Introduction
Quantum geometry is a differential geometry based on wave functions of quantum mechanics[1 , 2 , 3 ] . The key component is the quantum geometric tensor derived from quantum distance under infinitesimal momentum translation. Its real part gives the quantum metric, while its imaginary part gives the Berry curvature. It is well established that the Hall conductivity is related to the integral of the Berry curvature, whose relation is known as the Thouless-Kohmoto-Nightingale-Nijs (TKNN) formula[4 , 5 ] . On the other hand, quantum metric is related to various observables such as optical absorption[6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 ] , nonlinear conductivity[20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 ] and bulk photovoltaic effects[6 , 9 , 33 , 34 ] . Quantum geometric tensor is observed experimentally[35 , 36 ] . Recently, quantum geometry is generalized to the Zeeman quantum geometry by introducing spin translation in quantum distance. It is related to electromagnetic responses.
Ferromagnets are useful for magnetic memory and spintronics devices. However, there is a limitation of integration and fast dynamics due to the stray field. On the other hand, there is no such limitation in antiferromagnets. However, it is hard to readout the spin direction of antiferromagnet due to net zero magnetization. d d d -wave magnets are antiferromagnets whose electronic band structure has the d d d -wave splitting[37 , 38 , 39 , 40 , 41 ] . It is prominent that spin current can be generated by applying electric field or thermal gradient without using the spin-orbit interaction[37 , 38 , 39 , 40 , 41 ] . The notion of the d d d -wave magnets is generalized to altermagnets[42 , 43 ] including g g g -wave and i i i -wave altermagnets in two and three dimensions. They break time-reversal symmetry. In addition, p p p -wave[44 , 45 ] and f f f -wave magnets are also recognized, where time-reversal symmetry is preserved. They are summarized in the X X X -wave magnets[46 ] with X = p , d , f , g , i X = p, d, f, g, i X = p , d , f , g , i because they share universal features[47 ] irrespective of the presence or the absence of time-reversal symmetry.
In this paper, we review recent progress of quantum geometry and the X X X -wave magnets. We derive compact analytical formulas of the Zeeman quantum geometry based only on the two-band Hamiltonian without using eigenfunctions. We also generalize the Zeeman quantum geometry to multiband systems. Electromagnetic cross responses of the X X X -wave magnet with the Rashba interaction is also studied. We also review quantum geometry defined for non-Hermitian systems and density matrices. We note that there are some recent review articles on quantum geometry[48 , 49 , 50 , 51 ] and altermagnets[43 , 52 , 53 , 54 , 55 , 56 , 57 , 58 ] .
This paper is composed as follows. In § 2 , we construct quantum geometry starting from the fidelity of the wave function, which measures the similarity of the two wave functions. The quantum distance is defined based on the fidelity. By expanding the quantum distance with an infinitesimal momentum translation, we obtain the quantum geometric tensor. Its real part gives the quantum metric, while its imaginary part gives the Berry curvature. The integration of the Berry curvature leads to the Chern number, which is an integer characterizing a topological insulator. By making a Fourier transformation of the wave function from the momentum space to the real space, we obtain the Wannier function. It is shown that the quantum metric is related to the fluctuation of the position. We show an equality for the quantum geometry, which gives a lower bound determined by the Chern number. We summarize simple formulas to obtain the quantum metric and the Berry curvature directly from the two-band Hamiltonian without using information about the eigenfunctions. The quantum geometry is generalized to N N N -fold degenerate multiband systems.
In Sec. III, the TKNN formula is derived, which relates the Hall conductivity to the Chern number. It explains the quantization of the Hall conductivity for a topological insulator. We apply these results to the simplest Dirac system in two dimensions. Especially, optical dichroism is shown, where the optical absorption is selectively occurred depending on the chirality of the Dirac system. The sum rule which relates the optical conductivity and the quantum metric is derived. Bulk photovoltaic effects including the injection current and the shift current are related to the quantum metric. In addition, the second-order nonlinear conductivity is shown to be related to the quantum metric.
In § 4 , quantum geometry is generalized to include the spin degrees of freedom, which is called Zeemann quantum geometry. The relation between the Zeemann quantum geometry and electromagnetic cross responses are clarified. Simple formulas obtaining the Zeeman quantum geometric tensor for the two-band systems are derived based only on the Hamiltonian without using knowledge of the eigenfunctions. Their results are applied to the Rashba system. The Zeeman quantum geometry is generalized to N N N -fold degenerate systems.
In § 5 , quantum geometry is generalized to non-Hermitian systems. Especially, the quantum metric and the Berry curvature are obtained for the two-band system. They are studied in the Dirac system with a complex mass term.
In Sec. VI, quantum geometry is generalized to density matrices, which is known as the quantum information geometry. Especially, quantum distance is related to quantum Fisher information. The Clamér-Rao inequality is explained, which gives the lower bound of the covariance of the physical observable by the inverse of the quantum Fisher information. It is shown that the quantum Fisher information is reduced to the classical Fisher information for a pure state. Quantum geometry at thermal equilibrium is derived, and it is shown that the Uhlmann curvature is reduced to the Berry curvature at low temperature.
In Sec. VII, the notion of the X X X -wave magnets is introduced, which includes the d d d -wave, g g g -wave and i i i -wave altermagnets as well as the p p p -wave and f f f -wave magnets. Their symmetry properties are summarized. The quantum geometric properties are also calculated.
In Sec. VIII, transport properties of the X X X -wave magnets are summarized. The spin current generation by electric field and temperature gradient is analytically studied. It is shown that only the d d d -wave altermagnet has a linear response for the spin current. The analytic formula for tunneling magnetoresistance is also studied. Finally, we study a coupled system to the X X X -wave magnet and the Rashba system. It is shown that it is impossible to detect the Neel vector of the X X X -wave magnet by anomalous Hall conductivity based on the two-band model. Universal formula for the planar Hall conductivity for the X X X -wave magnets is presented. Electromagnetic responses of the X X X -wave magnets are also studied.
In Sec. IX, we analytically derive Landau levels for the p p p -wave magnet by using an analogy of the coherent state and for the d d d -wave altermagnet by using an analogy of the squeezed state.
In § 11 , we analytically derive the Friedel oscillation of the spatial profile of the local density states.
Sec. XI is devoted to summaries, discussions and outlooks.
This paper contains both reviews and original results. Most of the parts are review. The original ones are as follows. Two-band formulas for the Zeeman quantum geometry in § 4 .B. Non-Abelian Zeeman quantum geometry in § 4 .D. Zeeman quantum geometry induced cross responses in § 4 .E and F. Landau levels for the p p p -wave magnet in Sec. IX.A. The parts where the results are not original but give some new interpretations are as follows: Spin current generation and spin Nernst effect are derived analytically in Sec. VIII.A. The Landau levels for the d d d -wave altermagnet are derived by using the analogy of the coherent state in Sec. IX.A. The Friedel oscillation is discussed for the X X X -wave magnet in § 11 .
Quantum geometry
Quantum geometry is a differential geometry based on the wave functions. The basic notion is the fidelity, which measures the similarity of two wave functions. It acts as the distance in the context of the differential geometry. Starting from the fidelity with an infinitesimal translation in the momentum space, we derive the quantum geometric tensor. Its real part gives the quantum metric, while its imaginary part gives the Berry curvature. The latter is related to the anomalous Hall effect and topological insulators via the Chern number.
Quantum distance and quantum geometric tensor
The fidelity of the wave function ∣ ψ n ( k ) ⟩ |\psi_n(\mathbf{k})\rangle ∣ ψ n ( k )⟩ is defined by
F n ( k , k ′ ) ≡ ∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ ∣ , (1) F_n(\mathbf{k},\mathbf{k}') \equiv|\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle|, \tag*{(1)} F n ( k , k ′ ) ≡ ∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ ∣ , (1)
where the wave function is orthonormalized, ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ = δ ( k , k ′ ) \langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle=\delta(\mathbf{k},\mathbf{k}') ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ = δ ( k , k ′ ) . It satisfies 0 ≤ F n ( k , k ′ ) ≤ 1 0 \le F_n(\mathbf{k},\mathbf{k}') \le1 0 ≤ F n ( k , k ′ ) ≤ 1 . The Hilbert-Schmidt distance is defined by
s H S ( k , k ′ ) ≡ 1 − F n ( k , k ′ ) 2 , (2) s_{\mathrm{HS}}(\mathbf{k},\mathbf{k}') \equiv\sqrt{1-F_n(\mathbf{k},\mathbf{k}')^2}, \tag*{(2)} s HS ( k , k ′ ) ≡ 1 − F n ( k , k ′ ) 2 , (2)
where s H S ( k , k ′ ) = 0 s_{\mathrm{HS}}(\mathbf{k},\mathbf{k}')=0 s HS ( k , k ′ ) = 0 when the two wave functions are identical ∣ ψ n ( k ) ⟩ = ∣ ψ n ( k ′ ) ⟩ |\psi_n(\mathbf{k})\rangle=|\psi_n(\mathbf{k}')\rangle ∣ ψ n ( k )⟩ = ∣ ψ n ( k ′ )⟩ , while s H S ( k , k ′ ) = 1 s_{\mathrm{HS}}(\mathbf{k},\mathbf{k}')=1 s HS ( k , k ′ ) = 1 when they are orthogonal ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ = 0 \langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle=0 ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ = 0 .
The quantum geometric tensor F n μ ν F_n^{\mu\nu} F n μν is defined as follows. We start with the quantum distance d S H S d_{\mathrm{SHS}} d SHS for the infinitesimal translation d k d\mathbf{k} d k of the momentum as [1 , 2 , 59 ]
d S H S ( k ) = 1 − F n ( k , k + d k ) 2 , (3) d_{\mathrm{SHS}}(\mathbf{k})=\sqrt{1-F_n(\mathbf{k},\mathbf{k}+d\mathbf{k})^2}, \tag*{(3)} d SHS ( k ) = 1 − F n ( k , k + d k ) 2 , (3)
with
F n ( k , k + d k ) = ∣ ⟨ ψ n ( k ) ∣ U d k ∣ ψ n ( k ) ⟩ ∣ , (4) F_n(\mathbf{k},\mathbf{k}+d\mathbf{k})=|\langle\psi_n(\mathbf{k})|U_{d\mathbf{k}}|\psi_n(\mathbf{k})\rangle|, \tag*{(4)} F n ( k , k + d k ) = ∣ ⟨ ψ n ( k ) ∣ U d k ∣ ψ n ( k )⟩ ∣ , (4)
where
U d k ≡ e − i d k ⋅ r (5) U_{d\mathbf{k}}\equiv e^{-i d\mathbf{k}\cdot\mathbf{r}} \tag*{(5)} U d k ≡ e − i d k ⋅ r (5)
is the generator of the infinitesimal momentum translation d k d\mathbf{k} d k and r μ ≡ i ∂ / ∂ k μ r_\mu\equiv i\partial/\partial k_\mu r μ ≡ i ∂ / ∂ k μ is the position operator. The quantum distance for the infinitesimal momentum is expanded in terms of the quantum geometric tensor F n μ ν \mathcal{F}_{n}^{\mu\nu} F n μν as
( d s H S ) 2 = ∑ μ ν F n μ ν d k μ d k ν , (6) (ds_{\mathrm{HS}})^2=\sum_{\mu\nu}\mathcal{F}_{n}^{\mu\nu}\,dk_\mu dk_\nu, \tag*{(6)} ( d s HS ) 2 = μν ∑ F n μν d k μ d k ν , (6)
where μ , ν \mu,\nu μ , ν stand for x , y x,y x , y , and
F n μ ν ( k ) = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ) ∣ ∂ k ν ψ n ( k ) ⟩ , (7) \mathcal{F}_{n}^{\mu\nu}(\mathbf{k})=\langle\partial_{k_\mu}\psi_n(\mathbf{k})|(1-|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|)|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle, \tag*{(7)} F n μν ( k ) = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ) ∣ ∂ k ν ψ n ( k )⟩ , (7)
with ∂ k μ ≡ ∂ ∂ k μ \partial_{k_\mu}\equiv\frac{\partial}{\partial k_\mu} ∂ k μ ≡ ∂ k μ ∂ . It is called the Fubini-Study metric[60 , 61 ] . It is rewritten as
F n μ ν ( k ) = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P n ( k ) ) ∣ ∂ k ν ψ n ( k ) ⟩ , (8) \mathcal{F}_{n}^{\mu\nu}(\mathbf{k})=\langle\partial_{k_\mu}\psi_n(\mathbf{k})|(1-P_n(\mathbf{k}))|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle, \tag*{(8)} F n μν ( k ) = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P n ( k )) ∣ ∂ k ν ψ n ( k )⟩ , (8)
where we have defined the projection operator,
P n ( k ) ≡ ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ , (9) P_n(\mathbf{k})\equiv|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|, \tag*{(9)} P n ( k ) ≡ ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ , (9)
satisfying the idempotency condition P ( k ) 2 = P ( k ) P(\mathbf{k})^2=P(\mathbf{k}) P ( k ) 2 = P ( k ) . The presence of the projection operator is understood as follows. ∣ ∂ k ν ψ n ( k ) ⟩ |\partial_{k_\nu}\psi_n(\mathbf{k})\rangle ∣ ∂ k ν ψ n ( k )⟩ is generally mapped out of the band n n n , where the projection operator restricts ∣ ∂ k ν ψ n ( k ) ⟩ |\partial_{k_\nu}\psi_n(\mathbf{k})\rangle ∣ ∂ k ν ψ n ( k )⟩ to the band n n n . F n μ ν \mathcal{F}_n^{\mu\nu} F n μν is Hermitian, ( F n μ ν ) ∗ = F n ν μ (\mathcal{F}_n^{\mu\nu})^*=\mathcal{F}_n^{\nu\mu} ( F n μν ) ∗ = F n ν μ . The derivation of Eq. (7) is shown in Appendix A.
The quantum metric is the real part of the quantum geometric tensor,
g n μ ν = Re F n μ ν = F n μ ν ( k ) + ( F n μ ν ( k ) ) ∗ 2 , (10) g_n^{\mu\nu}=\operatorname{Re}\mathcal{F}_n^{\mu\nu}=\frac{\mathcal{F}_n^{\mu\nu}(\mathbf{k})+(\mathcal{F}_n^{\mu\nu}(\mathbf{k}))^*}{2}, \tag*{(10)} g n μν = Re F n μν = 2 F n μν ( k ) + ( F n μν ( k ) ) ∗ , (10)
while the Berry curvature is the imaginary part of the quantum geometric tensor,
Ω n μ ν = 2 Im F n μ ν = i ( F n μ ν ( k ) − ( F n μ ν ( k ) ) ∗ ) . (11) \Omega_n^{\mu\nu}=2\operatorname{Im}\mathcal{F}_n^{\mu\nu}=i\left(\mathcal{F}_n^{\mu\nu}(\mathbf{k})-(\mathcal{F}_n^{\mu\nu}(\mathbf{k}))^*\right). \tag*{(11)} Ω n μν = 2 Im F n μν = i ( F n μν ( k ) − ( F n μν ( k ) ) ∗ ) . (11)
There are relations
g n μ ν = g n ν μ , Ω n μ ν = − Ω n ν μ . (12) g_n^{\mu\nu}=g_n^{\nu\mu},\qquad\Omega_n^{\mu\nu}=-\Omega_n^{\nu\mu}. \tag*{(12)} g n μν = g n ν μ , Ω n μν = − Ω n ν μ . (12)
With the use of them, the quantum geometric tensor is decomposed into the real and imaginary parts,
F n μ ν = g n μ ν − i 2 Ω n μ ν . (13) \mathcal{F}_n^{\mu\nu}=g_n^{\mu\nu}-\frac{i}{2}\Omega_n^{\mu\nu}. \tag*{(13)} F n μν = g n μν − 2 i Ω n μν . (13)
The Berry curvature does not contribute to the quantum distance
( d s H S ) 2 = ∑ μ ν g n μ ν d k μ d k ν , (14) (ds_{\mathrm{HS}})^2=\sum_{\mu\nu}g_n^{\mu\nu}dk_\mu dk_\nu, \tag*{(14)} ( d s HS ) 2 = μν ∑ g n μν d k μ d k ν , (14)
because of the symmetry d k μ d k ν = d k ν d k μ dk_\mu dk_\nu=dk_\nu dk_\mu d k μ d k ν = d k ν d k μ .
By introducing the Wilczek-Zee connection[62 ] ,
a n m μ ( k ) ≡ i ⟨ ψ n ( k ) ∣ ∂ k μ ∣ ψ m ( k ) ⟩ , (15) a_{nm}^{\mu}(\mathbf{k})\equiv i\langle\psi_n(\mathbf{k})|\partial_{k_\mu}|\psi_m(\mathbf{k})\rangle, \tag*{(15)} a nm μ ( k ) ≡ i ⟨ ψ n ( k ) ∣ ∂ k μ ∣ ψ m ( k )⟩ , (15)
the quantum geometric tensor is rewritten as
F n μ ν ( k ) = ∑ m ≠ n ⟨ ∂ k μ ψ n ( k ) ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ = ∑ m ≠ n ( i ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) ∗ i ⟨ ψ m ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ = ∑ m ≠ n a m n μ ∗ ( k ) a m n ν ( k ) . (16) \begin{aligned}
\mathcal{F}_n^{\mu\nu}(\mathbf{k})
&=\sum_{m\ne n}\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\\
&=\sum_{m\ne n}\left(i\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)^*i\langle\psi_m(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\\
&=\sum_{m\ne n}a_{mn}^{\mu*}(\mathbf{k})a_{mn}^{\nu}(\mathbf{k}). \tag*{(16)}
\end{aligned} F n μν ( k ) = m = n ∑ ⟨ ∂ k μ ψ n ( k ) ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ ∂ k ν ψ n ( k )⟩ = m = n ∑ ( i ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) ∗ i ⟨ ψ m ( k ) ∣ ∂ k ν ψ n ( k )⟩ = m = n ∑ a mn μ ∗ ( k ) a mn ν ( k ) . (16)
The Wilczek-Zee connection is Hermitian,
a n m μ ∗ ( k ) = a m n μ ( k ) , (17) a_{nm}^{\mu*}(\mathbf{k})=a_{mn}^{\mu}(\mathbf{k}), \tag*{(17)} a nm μ ∗ ( k ) = a mn μ ( k ) , (17)
because
i ∂ ∂ k μ ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ = i ⟨ ∂ k μ ψ m ( k ) ∣ ψ n ( k ) ⟩ + i ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ = ( − i ⟨ ψ n ( k ) ∣ ∂ k μ ψ m ( k ) ⟩ ) ∗ + i ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ = − a n m μ ∗ ( k ) + a m n μ ( k ) = 0. (18) \begin{aligned}
i\frac{\partial}{\partial k_\mu}\langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle
&=i\langle\partial_{k_\mu}\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle+i\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&=\left(-i\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_m(\mathbf{k})\rangle\right)^*+i\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&=-a_{nm}^{\mu*}(\mathbf{k})+a_{mn}^{\mu}(\mathbf{k})=0. \tag*{(18)}
\end{aligned} i ∂ k μ ∂ ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ = i ⟨ ∂ k μ ψ m ( k ) ∣ ψ n ( k )⟩ + i ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ = ( − i ⟨ ψ n ( k ) ∣ ∂ k μ ψ m ( k )⟩ ) ∗ + i ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ = − a nm μ ∗ ( k ) + a mn μ ( k ) = 0. (18)
Then, the quantum geometric tensor is rewritten as
F n μ ν ( k ) = ∑ m ≠ n a n m μ ( k ) a m n ν ( k ) (19) \mathcal{F}_n^{\mu\nu}(\mathbf{k})=\sum_{m\ne n}a_{nm}^{\mu}(\mathbf{k})a_{mn}^{\nu}(\mathbf{k}) \tag*{(19)} F n μν ( k ) = m = n ∑ a nm μ ( k ) a mn ν ( k ) (19)
in terms of the Wilczek-Zee connection.
Berry connection, Berry curvature and Chern number
It follows that Ω n x x ( k ) = Ω n y y ( k ) = 0 \Omega_n^{xx}(\mathbf{k})=\Omega_n^{yy}(\mathbf{k})=0 Ω n xx ( k ) = Ω n y y ( k ) = 0 from the antisymmetric property Eq. (12) . Then, the nontrivial contributions are Ω n x y ( k ) \Omega_n^{xy}(\mathbf{k}) Ω n x y ( k ) and Ω n y x ( k ) \Omega_n^{yx}(\mathbf{k}) Ω n y x ( k ) . The Berry curvature (11) reads
Ω n x y ( k ) = i ⟨ ∂ k x ψ n ( k ) ∣ 1 − P ( k ) ∣ ∂ k y ψ n ( k ) ⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ 1 − P ( k ) ∣ ∂ k x ψ n ( k ) ⟩ = i ⟨ ∂ k x ψ n ( k ) ∣ ∂ k y ψ n ( k ) ⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ ∂ k x ψ n ( k ) ⟩ − i ⟨ ∂ k x ψ n ( k ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ∂ k y ψ n ( k ) ⟩ + i ⟨ ∂ k y ψ n ( k ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ∂ k x ψ n ( k ) ⟩ = i ⟨ ∂ k x ψ n ( k ) ∣ ∂ k y ψ n ( k ) ⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ ∂ k x ψ n ( k ) ⟩ + i a n x ∗ ( k ) a n y ( k ) − i a n y ∗ ( k ) a n x ( k ) , (20) \begin{aligned}
\Omega_n^{xy}(\mathbf{k})
&=i\langle\partial_{k_x}\psi_n(\mathbf{k})|1-P(\mathbf{k})|\partial_{k_y}\psi_n(\mathbf{k})\rangle\\
&\quad-i\langle\partial_{k_y}\psi_n(\mathbf{k})|1-P(\mathbf{k})|\partial_{k_x}\psi_n(\mathbf{k})\rangle\\
&=i\langle\partial_{k_x}\psi_n(\mathbf{k})|\partial_{k_y}\psi_n(\mathbf{k})\rangle-i\langle\partial_{k_y}\psi_n(\mathbf{k})|\partial_{k_x}\psi_n(\mathbf{k})\rangle\\
&\quad-i\langle\partial_{k_x}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\partial_{k_y}\psi_n(\mathbf{k})\rangle\\
&\quad+i\langle\partial_{k_y}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\partial_{k_x}\psi_n(\mathbf{k})\rangle\\
&=i\langle\partial_{k_x}\psi_n(\mathbf{k})|\partial_{k_y}\psi_n(\mathbf{k})\rangle-i\langle\partial_{k_y}\psi_n(\mathbf{k})|\partial_{k_x}\psi_n(\mathbf{k})\rangle\\
&\quad+ia_n^{x*}(\mathbf{k})a_n^y(\mathbf{k})-ia_n^{y*}(\mathbf{k})a_n^x(\mathbf{k}), \tag*{(20)}
\end{aligned} Ω n x y ( k ) = i ⟨ ∂ k x ψ n ( k ) ∣1 − P ( k ) ∣ ∂ k y ψ n ( k )⟩ − i ⟨ ∂ k y ψ n ( k ) ∣1 − P ( k ) ∣ ∂ k x ψ n ( k )⟩ = i ⟨ ∂ k x ψ n ( k ) ∣ ∂ k y ψ n ( k )⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ ∂ k x ψ n ( k )⟩ − i ⟨ ∂ k x ψ n ( k ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ∂ k y ψ n ( k )⟩ + i ⟨ ∂ k y ψ n ( k ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ∂ k x ψ n ( k )⟩ = i ⟨ ∂ k x ψ n ( k ) ∣ ∂ k y ψ n ( k )⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ ∂ k x ψ n ( k )⟩ + i a n x ∗ ( k ) a n y ( k ) − i a n y ∗ ( k ) a n x ( k ) , (20)
where we have defined the Berry connection
a n μ ( k ) ≡ a n n μ ( k ) , (21) a_n^\mu(\mathbf{k})\equiv a_{nn}^{\mu}(\mathbf{k}), \tag*{(21)} a n μ ( k ) ≡ a nn μ ( k ) , (21)
which is real, a n μ ∗ ( k ) = a n μ ( k ) a_n^{\mu*}(\mathbf{k})=a_n^\mu(\mathbf{k}) a n μ ∗ ( k ) = a n μ ( k ) .
Then, the Berry curvature (20) is rewritten as
Ω n x y ( k ) = i ⟨ ∂ k x ψ n ( k ) ∣ ∂ k y ψ n ( k ) ⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ ∂ k x ψ n ( k ) ⟩ = − 2 Im ⟨ ∂ ψ n ( k ) ∂ k x | ∂ ψ n ( k ) ∂ k y ⟩ = ∂ a n y ( k ) ∂ k x − ∂ a n x ( k ) ∂ k y = [ ∇ × a n ( k ) ] z . (22) \begin{aligned}
\Omega_n^{xy}(\mathbf{k})
&=i\langle\partial_{k_x}\psi_n(\mathbf{k})|\partial_{k_y}\psi_n(\mathbf{k})\rangle-i\langle\partial_{k_y}\psi_n(\mathbf{k})|\partial_{k_x}\psi_n(\mathbf{k})\rangle\\
&=-2\operatorname{Im}\left\langle\frac{\partial\psi_n(\mathbf{k})}{\partial k_x}\middle|\frac{\partial\psi_n(\mathbf{k})}{\partial k_y}\right\rangle\\
&=\frac{\partial a_n^y(\mathbf{k})}{\partial k_x}-\frac{\partial a_n^x(\mathbf{k})}{\partial k_y}=[\nabla\times\mathbf{a}_n(\mathbf{k})]_z. \tag*{(22)}
\end{aligned} Ω n x y ( k ) = i ⟨ ∂ k x ψ n ( k ) ∣ ∂ k y ψ n ( k )⟩ − i ⟨ ∂ k y ψ n ( k ) ∣ ∂ k x ψ n ( k )⟩ = − 2 Im ⟨ ∂ k x ∂ ψ n ( k ) ∂ k y ∂ ψ n ( k ) ⟩ = ∂ k x ∂ a n y ( k ) − ∂ k y ∂ a n x ( k ) = [ ∇ × a n ( k ) ] z . (22)
Its integral over the whole Brillouin zone is quantized
1 2 π ∫ d k Ω n x y ( k ) = 1 2 π ∮ d k μ [ ∇ × Ω n x y ( k ) ] = 1 2 π ∮ d k μ a n μ ( k ) = C n , (23) \frac{1}{2\pi}\int d\mathbf{k}\,\Omega_n^{xy}(\mathbf{k})=\frac{1}{2\pi}\oint d\mathbf{k}_\mu[\nabla\times\mathbf{\Omega}_n^{xy}(\mathbf{k})]
=\frac{1}{2\pi}\oint d\mathbf{k}_\mu a_n^\mu(\mathbf{k})=C_n, \tag*{(23)} 2 π 1 ∫ d k Ω n x y ( k ) = 2 π 1 ∮ d k μ [ ∇ × Ω n x y ( k )] = 2 π 1 ∮ d k μ a n μ ( k ) = C n , (23)
where C n C_n C n is the Chern number taking an integer. We go on to prove that C n C_n C n is an integer [63 ] . The formula (23) is interpreted as the integral of the Berry curvature Ω n x y ( k ) \Omega_n^{xy}(\mathbf{k}) Ω n x y ( k ) over the first Brillouin zone, which is the Chern number. Using the Stokes theorem, this can be rewritten as a contour integration along the boundary of the Brillouin zone,
C n = 1 2 π ∫ d k Ω n x y ( k ) = 1 2 π ∮ d k μ a n μ ( k ) . (24) C_n=\frac{1}{2\pi}\int d\mathbf{k}\,\Omega_n^{xy}(\mathbf{k})=\frac{1}{2\pi}\oint dk_\mu a_n^\mu(\mathbf{k}). \tag*{(24)} C n = 2 π 1 ∫ d k Ω n x y ( k ) = 2 π 1 ∮ d k μ a n μ ( k ) . (24)
Since it is a periodic system, it follows that C n = 0 C_n=0 C n = 0 if a n μ ( k ) a_n^\mu(\mathbf{k}) a n μ ( k ) is a regular function. However, the gauge potential a n μ ( k ) a_n^\mu(\mathbf{k}) a n μ ( k ) can be singular though the magnetic field Ω n x y ( k ) \Omega_n^{xy}(\mathbf{k}) Ω n x y ( k ) is regular. In this case it is necessary to choose the contour integration to avoid these singular points. Then, C n C_n C n counts the number of singularities, following the argument familiar in the theory of Dirac monopoles.
Wannier function and polarization
The Wannier function is defined by
∣ w n ( r ) ⟩ ≡ 1 2 π ∫ B Z d k e − i k ⋅ r ∣ ψ n ( k ) ⟩ , (25) |w_n(\mathbf{r})\rangle\equiv\frac{1}{2\pi}\int_{\mathrm{BZ}}d\mathbf{k}e^{-i\mathbf{k}\cdot\mathbf{r}}|\psi_n(\mathbf{k})\rangle, \tag*{(25)} ∣ w n ( r )⟩ ≡ 2 π 1 ∫ BZ d k e − i k ⋅ r ∣ ψ n ( k )⟩ , (25)
or
∣ ψ n ( k ) ⟩ = ∑ r e i k ⋅ r ∣ w n ( r ) ⟩ . (26) |\psi_n(\mathbf{k})\rangle=\sum_{\mathbf{r}}e^{i\mathbf{k}\cdot\mathbf{r}}|w_n(\mathbf{r})\rangle. \tag*{(26)} ∣ ψ n ( k )⟩ = r ∑ e i k ⋅ r ∣ w n ( r )⟩ . (26)
Then, the Berry connection is rewritten as
a n μ ( k ) = i ⟨ ψ n ( k ) ∣ d d k μ ∣ ψ n ( k ) ⟩ = ∑ r , r ′ ⟨ w n ( r ′ ) ∣ e − i k ⋅ r ′ d d k μ e i k ⋅ r ∣ w n ( r ) ⟩ = ∑ r , r ′ ⟨ w n ( r ′ ) ∣ e i k ⋅ ( r − r ′ ) r μ ∣ w n ( r ) ⟩ . (27) \begin{aligned}
a_n^\mu(\mathbf{k})=i\langle\psi_n(\mathbf{k})|\frac{d}{dk_\mu}|\psi_n(\mathbf{k})\rangle\\
&=\sum_{\mathbf{r},\mathbf{r}'}\langle w_n(\mathbf{r}')|e^{-i\mathbf{k}\cdot\mathbf{r}'}\frac{d}{dk_\mu}e^{i\mathbf{k}\cdot\mathbf{r}}|w_n(\mathbf{r})\rangle\\
&=\sum_{\mathbf{r},\mathbf{r}'}\langle w_n(\mathbf{r}')|e^{i\mathbf{k}\cdot(\mathbf{r}-\mathbf{r}')}r_\mu|w_n(\mathbf{r})\rangle. \tag*{(27)}
\end{aligned} a n μ ( k ) = i ⟨ ψ n ( k ) ∣ d k μ d ∣ ψ n ( k )⟩ = r , r ′ ∑ ⟨ w n ( r ′ ) ∣ e − i k ⋅ r ′ d k μ d e i k ⋅ r ∣ w n ( r )⟩ = r , r ′ ∑ ⟨ w n ( r ′ ) ∣ e i k ⋅ ( r − r ′ ) r μ ∣ w n ( r )⟩ . (27)
We integrate it over the Brillouin zone and obtain
1 2 π ∫ B Z a n μ ( k ) d k = ∑ r , r ′ ⟨ w n ( r ′ ) ∣ ∫ B Z e i k ⋅ ( r − r ′ ) d k r μ ∣ w n ( r ) ⟩ = ∑ r , r ′ ⟨ w n ( r ′ ) ∣ δ ( r − r ′ ) r μ ∣ w n ( r ) ⟩ = ∑ r ⟨ w n ( r ) ∣ r μ ∣ w n ( r ) ⟩ = ⟨ r μ ⟩ . (28) \begin{aligned}
\frac{1}{2\pi}\int_{\mathrm{BZ}}a_n^\mu(\mathbf{k})d\mathbf{k} \\
&=\sum_{\mathbf{r},\mathbf{r}'}\langle w_n(\mathbf{r}')|\int_{\mathrm{BZ}}e^{i\mathbf{k}\cdot(\mathbf{r}-\mathbf{r}')}d\mathbf{k}r_\mu|w_n(\mathbf{r})\rangle\\
&=\sum_{\mathbf{r},\mathbf{r}'}\langle w_n(\mathbf{r}')|\delta(\mathbf{r}-\mathbf{r}')r_\mu|w_n(\mathbf{r})\rangle\\
&=\sum_{\mathbf{r}}\langle w_n(\mathbf{r})|r_\mu|w_n(\mathbf{r})\rangle=\langle r_\mu\rangle. \tag*{(28)}
\end{aligned} 2 π 1 ∫ BZ a n μ ( k ) d k = r , r ′ ∑ ⟨ w n ( r ′ ) ∣ ∫ BZ e i k ⋅ ( r − r ′ ) d k r μ ∣ w n ( r )⟩ = r , r ′ ∑ ⟨ w n ( r ′ ) ∣ δ ( r − r ′ ) r μ ∣ w n ( r )⟩ = r ∑ ⟨ w n ( r ) ∣ r μ ∣ w n ( r )⟩ = ⟨ r μ ⟩ . (28)
It is the expectation value of the position r μ r_\mu r μ , which represents the polarization. The quantum metric is related to the fluctuation of the polarization [64 ] ,
1 2 π ∫ B Z g n μ μ ( k ) d k = ∑ r ⟨ w n ( r ) ∣ r μ 2 ∣ w n ( r ) ⟩ − ⟨ w n ( r ) ∣ r μ ∣ w n ( r ) ⟩ 2 = ⟨ r μ 2 ⟩ − ⟨ r μ ⟩ 2 . (29) \begin{aligned}
\frac{1}{2\pi}\int_{\mathrm{BZ}}g_n^{\mu\mu}(\mathbf{k})d\mathbf{k} \\
&=\sum_{\mathbf{r}}\langle w_n(\mathbf{r})|r_\mu^2|w_n(\mathbf{r})\rangle-\langle w_n(\mathbf{r})|r_\mu|w_n(\mathbf{r})\rangle^2 \\
&=\langle r_\mu^2\rangle-\langle r_\mu\rangle^2. \tag*{(29)}
\end{aligned} 2 π 1 ∫ BZ g n μμ ( k ) d k = r ∑ ⟨ w n ( r ) ∣ r μ 2 ∣ w n ( r )⟩ − ⟨ w n ( r ) ∣ r μ ∣ w n ( r ) ⟩ 2 = ⟨ r μ 2 ⟩ − ⟨ r μ ⟩ 2 . (29)
Inequality
There is an inequality for the quantum metric and the Berry curvature in two dimensions [65 , 66 ] ,
1 2 Tr g n μ ν ( k ) ≥ det g n μ ν ( k ) ≥ ∣ Ω n x y ( k ) ∣ 2 , (30) \frac{1}{2}\operatorname{Tr}g_n^{\mu\nu}(\mathbf{k})\geq\sqrt{\det g_n^{\mu\nu}(\mathbf{k})}\geq\frac{|\Omega_n^{xy}(\mathbf{k})|}{2}, \tag*{(30)} 2 1 Tr g n μν ( k ) ≥ det g n μν ( k ) ≥ 2 ∣ Ω n x y ( k ) ∣ , (30)
where the trace and the determinant are taken over μ \mu μ and ν \nu ν . The first inequality is proved by using the arithmetic geometric mean,
g + + g − 2 ≥ g + g − , (31) \frac{g_++g_-}{2}\geq\sqrt{g_+g_-}, \tag*{(31)} 2 g + + g − ≥ g + g − , (31)
where g ± g_\pm g ± is the diagonal component of the diagonalized g n μ ν ( k ) g_n^{\mu\nu}(\mathbf{k}) g n μν ( k ) ,
g n μ ν ( k ) = ( g + 0 0 g − ) , (32) g_n^{\mu\nu}(\mathbf{k})=
\begin{pmatrix}
g_+ & 0\\
0 & g_-
\end{pmatrix}, \tag*{(32)} g n μν ( k ) = ( g + 0 0 g − ) , (32)
with
g ± ≡ g n x x + g n y y ± ( g n x x − g n y y ) 2 + 4 ( g n x y ) 2 2 . (33) g_\pm\equiv\frac{g_n^{xx}+g_n^{yy}\pm\sqrt{(g_n^{xx}-g_n^{yy})^2+4(g_n^{xy})^2}}{2}. \tag*{(33)} g ± ≡ 2 g n xx + g n y y ± ( g n xx − g n y y ) 2 + 4 ( g n x y ) 2 . (33)
The second inequality is proved by using the non-negativity of F n μ ν ( k ) \mathcal{F}_n^{\mu\nu}(\mathbf{k}) F n μν ( k ) ,
det F n μ ν ( k ) ≥ 0 , (34) \det\mathcal{F}_n^{\mu\nu}(\mathbf{k})\geq0, \tag*{(34)} det F n μν ( k ) ≥ 0 , (34)
which is derived from the non-negativity of the quadratic form in Eq. (6) . It is equivalent to
det g n μ ν ( k ) ≥ ( Ω n x y 2 ) 2 , (35) \det g_n^{\mu\nu}(\mathbf{k})\geq\left(\frac{\Omega_n^{xy}}{2}\right)^2, \tag*{(35)} det g n μν ( k ) ≥ ( 2 Ω n x y ) 2 , (35)
where we have used
F n μ ν = ( g n x x g n x y − i Ω n x y / 2 g n x y + i Ω n x y / 2 g n y y ) , (36) \mathcal{F}_n^{\mu\nu}=
\begin{pmatrix}
g_n^{xx} & g_n^{xy}-i\Omega_n^{xy}/2\\
g_n^{xy}+i\Omega_n^{xy}/2 & g_n^{yy}
\end{pmatrix}, \tag*{(36)} F n μν = ( g n xx g n x y + i Ω n x y /2 g n x y − i Ω n x y /2 g n y y ) , (36)
which is derived from Eqs. (10) , (11) and (12) . By integrating the inequality (30) over the whole Brillouin zone, we obtain [65 , 66 , 67 , 68 ]
1 2 π ∫ d k Tr g n μ ν ( k ) ≥ 1 2 π ∫ d k det g n μ ν ( k ) ≥ 1 2 π ∫ d k ∣ Ω n x y ( k ) ∣ ≥ 1 2 π ∫ d k Ω n x y ( k ) = C n . (37) \frac{1}{2\pi}\int d\mathbf{k}\operatorname{Tr}g_n^{\mu\nu}(\mathbf{k})
\geq\frac{1}{2\pi}\int d\mathbf{k}\sqrt{\det g_n^{\mu\nu}(\mathbf{k})}
\geq\frac{1}{2\pi}\int d\mathbf{k}|\Omega_n^{xy}(\mathbf{k})|
\geq\frac{1}{2\pi}\int d\mathbf{k}\Omega_n^{xy}(\mathbf{k})=C_n. \tag*{(37)} 2 π 1 ∫ d k Tr g n μν ( k ) ≥ 2 π 1 ∫ d k det g n μν ( k ) ≥ 2 π 1 ∫ d k ∣ Ω n x y ( k ) ∣ ≥ 2 π 1 ∫ d k Ω n x y ( k ) = C n . (37)
This gives a lower bound for ∫ d k Tr g n μ ν ( k ) \int d\mathbf{k}\operatorname{Tr}g_n^{\mu\nu}(\mathbf{k}) ∫ d k Tr g n μν ( k ) , when the system is topological, C n ≥ 1 C_n\geq1 C n ≥ 1 . There are some applications to this inequality [12 , 69 , 70 ] . The quantum volume is defined by the Riemann volume as [67 ]
vol g ≡ 1 2 π ∫ d k det g n μ ν ( k ) . (38) \operatorname{vol}_g\equiv\frac{1}{2\pi}\int d\mathbf{k}\sqrt{\det g_n^{\mu\nu}(\mathbf{k})}. \tag*{(38)} vol g ≡ 2 π 1 ∫ d k det g n μν ( k ) . (38)
The quantum weight is defined by[71 ]
K μ ν = 1 2 π ∫ d k g n μ ν ( k ) (39) K^{\mu\nu}=\frac{1}{2\pi}\int dk\,g^{\mu\nu}_{n}(\mathbf{k}) \tag*{(39)} K μν = 2 π 1 ∫ d k g n μν ( k ) (39)
which measures quantum fluctuation in the center of mass. It is experimentally determined from X-ray scattering.
Quantum Geometry for two-band systems
We consider a two-band system with the index n = ± n=\pm n = ± . The Hamiltonian is generally given by
H ( k ) = h 0 ( k ) + σ ⋅ h ( k ) , (40) H(\mathbf{k})=h_{0}(\mathbf{k})+\boldsymbol{\sigma}\cdot\mathbf{h}(\mathbf{k}), \tag*{(40)} H ( k ) = h 0 ( k ) + σ ⋅ h ( k ) , (40)
where h ( k ) \mathbf{h}(\mathbf{k}) h ( k ) is parametrized as the normalized vector n ( k ) \mathbf{n}(\mathbf{k}) n ( k ) ,
n ( k ) ≡ h ( k ) ∣ h ( k ) ∣ = ( sin θ cos ϕ , sin θ sin ϕ , cos θ ) , (41) \mathbf{n}(\mathbf{k})\equiv\frac{\mathbf{h}(\mathbf{k})}{\lvert\mathbf{h}(\mathbf{k})\rvert}=(\sin\theta\cos\phi,\sin\theta\sin\phi,\cos\theta), \tag*{(41)} n ( k ) ≡ ∣ h ( k )∣ h ( k ) = ( sin θ cos ϕ , sin θ sin ϕ , cos θ ) , (41)
and σ j \sigma_{j} σ j is the Pauli matrix with j = x , y , z j=x,y,z j = x , y , z . In this case, there are simple forms in terms of n ( k ) \mathbf{n}(\mathbf{k}) n ( k ) .
The energy is given by
ε ± = h 0 ( k ) ± ∣ h ( k ) ∣ . (42) \varepsilon_{\pm}=h_{0}(\mathbf{k})\pm\lvert\mathbf{h}(\mathbf{k})\rvert. \tag*{(42)} ε ± = h 0 ( k ) ± ∣ h ( k )∣ . (42)
The eigenfunctions ψ ± ( k ) \psi_{\pm}(\mathbf{k}) ψ ± ( k ) of the Hamiltonian (40) corresponding to the energy ε ± \varepsilon_{\pm} ε ± are given by
ψ + ( k ) = e i α ( k ) ( e − i ϕ ( k ) cos θ ( k ) 2 sin θ ( k ) 2 ) , (43) \psi_{+}(\mathbf{k})=e^{i\alpha(\mathbf{k})}
\begin{pmatrix}
e^{-i\phi(\mathbf{k})}\cos\frac{\theta(\mathbf{k})}{2}\\
\sin\frac{\theta(\mathbf{k})}{2}
\end{pmatrix}, \tag*{(43)} ψ + ( k ) = e i α ( k ) ( e − i ϕ ( k ) cos 2 θ ( k ) sin 2 θ ( k ) ) , (43)
ψ − ( k ) = e i α ( k ) ( − e − i ϕ ( k ) sin θ ( k ) 2 cos θ ( k ) 2 ) , (44) \psi_{-}(\mathbf{k})=e^{i\alpha(\mathbf{k})}
\begin{pmatrix}
-e^{-i\phi(\mathbf{k})}\sin\frac{\theta(\mathbf{k})}{2}\\
\cos\frac{\theta(\mathbf{k})}{2}
\end{pmatrix}, \tag*{(44)} ψ − ( k ) = e i α ( k ) ( − e − i ϕ ( k ) sin 2 θ ( k ) cos 2 θ ( k ) ) , (44)
where α ( k ) \alpha(\mathbf{k}) α ( k ) is a real function representing a gauge degrees of freedom. Then, the Wilczek-Zee connection (15) is calculated as
α + − μ ( k ) = − i ⟨ ψ + ( k ) ∣ ∂ k μ ψ − ( k ) ⟩ = − i ∂ k μ θ 2 + ∂ k μ ϕ 2 sin θ , (45) \alpha_{+-}^{\mu}(\mathbf{k})=-i\langle\psi_{+}(\mathbf{k})\mid\partial_{k_{\mu}}\psi_{-}(\mathbf{k})\rangle
=-i\frac{\partial_{k_{\mu}}\theta}{2}+\frac{\partial_{k_{\mu}}\phi}{2}\sin\theta, \tag*{(45)} α +− μ ( k ) = − i ⟨ ψ + ( k ) ∣ ∂ k μ ψ − ( k )⟩ = − i 2 ∂ k μ θ + 2 ∂ k μ ϕ sin θ , (45)
and
α − + μ ( k ) = − i ⟨ ψ − ( k ) ∣ ∂ k μ ψ + ( k ) ⟩ = − i ∂ k μ θ 2 + ∂ k μ ϕ 2 sin θ , (46) \alpha_{-+}^{\mu}(\mathbf{k})=-i\langle\psi_{-}(\mathbf{k})\mid\partial_{k_{\mu}}\psi_{+}(\mathbf{k})\rangle
=-i\frac{\partial_{k_{\mu}}\theta}{2}+\frac{\partial_{k_{\mu}}\phi}{2}\sin\theta, \tag*{(46)} α −+ μ ( k ) = − i ⟨ ψ − ( k ) ∣ ∂ k μ ψ + ( k )⟩ = − i 2 ∂ k μ θ + 2 ∂ k μ ϕ sin θ , (46)
irrespective of α \alpha α .
The quantum metric is explicitly given by[12 , 15 , 17 , 18 , 72 , 73 , 74 , 75 ]
g ± μ ν ( k ) = ± 1 4 ( ∂ k μ n ) ⋅ ( ∂ k ν n ) = ± 1 4 ( ∂ k μ θ ∂ k ν θ + sin 2 θ ∂ k μ ϕ ∂ k ν ϕ ) (47) g_{\pm}^{\mu\nu}(\mathbf{k})=\pm\frac{1}{4}(\partial_{k_{\mu}}\mathbf{n})\cdot(\partial_{k_{\nu}}\mathbf{n})
=\pm\frac{1}{4}(\partial_{k_{\mu}}\theta\partial_{k_{\nu}}\theta+\sin^{2}\theta\partial_{k_{\mu}}\phi\partial_{k_{\nu}}\phi) \tag*{(47)} g ± μν ( k ) = ± 4 1 ( ∂ k μ n ) ⋅ ( ∂ k ν n ) = ± 4 1 ( ∂ k μ θ ∂ k ν θ + sin 2 θ ∂ k μ ϕ ∂ k ν ϕ ) (47)
with the normalized vector (41) . Two or three components of n \mathbf{n} n should be nonzero for nonzero g μ ν g^{\mu\nu} g μν . It is understood as follows: If only one component n z ( k ) n_{z}(\mathbf{k}) n z ( k ) is nonzero, it is given by n z ( k ) = h z ( k ) / ∣ h z ( k ) ∣ n_{z}(\mathbf{k})=h_{z}(\mathbf{k})/\lvert h_{z}(\mathbf{k})\rvert n z ( k ) = h z ( k ) / ∣ h z ( k )∣ , whose derivative ∂ k μ n \partial_{k_{\mu}}\mathbf{n} ∂ k μ n is singular.
The Berry curvature is explicitly given by[73 , 75 , 79 , 80 , 81 ]
Ω ± x y ( k ) = ∓ 1 2 n ⋅ ( ∂ k x n × ∂ k y n ) = ± 1 2 sin θ ( ∂ k x ϕ ∂ k y θ − ∂ k x θ ∂ k y ϕ ) . (48) \Omega_{\pm}^{xy}(\mathbf{k})=\mp\frac{1}{2}\mathbf{n}\cdot(\partial_{k_{x}}\mathbf{n}\times\partial_{k_{y}}\mathbf{n})
=\pm\frac{1}{2}\sin\theta\left(\partial_{k_{x}}\phi\partial_{k_{y}}\theta-\partial_{k_{x}}\theta\partial_{k_{y}}\phi\right). \tag*{(48)} Ω ± x y ( k ) = ∓ 2 1 n ⋅ ( ∂ k x n × ∂ k y n ) = ± 2 1 sin θ ( ∂ k x ϕ ∂ k y θ − ∂ k x θ ∂ k y ϕ ) . (48)
It is a solid angle of the vector n \mathbf{n} n . Hence, three components of n \mathbf{n} n should be nonzero for nonzero Ω ± x y \Omega_{\pm}^{xy} Ω ± x y .
Analogy of the theory of general relativity
The quantum metric and the Berry curvature are mainly studied in the current stage. In addition to them, there are following quantities that can be defined in the analogy of the theory of general relativity[15 , 76 ] . They are the Christoffel symbol
Γ μ ν λ = 1 2 g λ σ ( ∂ μ g ν σ + ∂ ν g σ μ − ∂ σ g μ ν ) , (49) \Gamma_{\mu\nu}^{\lambda}=\frac{1}{2}g^{\lambda\sigma}\left(\partial_{\mu}g_{\nu\sigma}+\partial_{\nu}g_{\sigma\mu}-\partial_{\sigma}g_{\mu\nu}\right), \tag*{(49)} Γ μν λ = 2 1 g λσ ( ∂ μ g ν σ + ∂ ν g σ μ − ∂ σ g μν ) , (49)
the Riemann tensor
R σ μ ν ρ = ∂ μ Γ ν σ ρ − ∂ ν Γ μ σ ρ + Γ μ λ ρ Γ ν σ λ − Γ ν λ ρ Γ μ σ λ , (50) R_{\sigma\mu\nu}^{\rho}=\partial_{\mu}\Gamma_{\nu\sigma}^{\rho}-\partial_{\nu}\Gamma_{\mu\sigma}^{\rho}+\Gamma_{\mu\lambda}^{\rho}\Gamma_{\nu\sigma}^{\lambda}-\Gamma_{\nu\lambda}^{\rho}\Gamma_{\mu\sigma}^{\lambda}, \tag*{(50)} R σ μν ρ = ∂ μ Γ ν σ ρ − ∂ ν Γ μ σ ρ + Γ μ λ ρ Γ ν σ λ − Γ ν λ ρ Γ μ σ λ , (50)
the Ricci tensor
R μ ν = R μ λ ν λ , (51) R_{\mu\nu}=R_{\mu\lambda\nu}^{\lambda}, \tag*{(51)} R μν = R μ λ ν λ , (51)
the Ricci scalar
R = g μ ν R ν μ , (52) R=g^{\mu\nu}R_{\nu\mu}, \tag*{(52)} R = g μν R ν μ , (52)
and the Einstein tensor
G μ ν ≡ R μ ν − 1 2 R g μ ν . (53) G_{\mu\nu}\equiv R_{\mu\nu}-\frac{1}{2}Rg_{\mu\nu}. \tag*{(53)} G μν ≡ R μν − 2 1 R g μν . (53)
Especially, the Christoffel symbol is given by
Γ μ ν λ = 1 4 ∂ λ n ⋅ ∂ μ ∂ ν n (54) \Gamma_{\mu\nu}^{\lambda}=\frac{1}{4}\partial_{\lambda}\mathbf{n}\cdot\partial_{\mu}\partial_{\nu}\mathbf{n} \tag*{(54)} Γ μν λ = 4 1 ∂ λ n ⋅ ∂ μ ∂ ν n (54)
for the two-band system[77 ] . Furthermore, the Riemann tensor reads
R σ μ ν ρ = 1 4 [ ∂ μ ∂ ρ n ⋅ ∂ ν ∂ σ n + ∂ ρ n ⋅ ∂ μ ∂ ν ∂ σ n − ∂ ν ∂ ρ n ⋅ ∂ μ ∂ σ n − ∂ ρ n ⋅ ∂ μ ∂ ν ∂ σ n ] + 1 16 ( ∂ ρ n ⋅ ∂ μ ∂ λ n ) ( ∂ λ n ⋅ ∂ ν ∂ σ n ) − 1 16 ( ∂ ρ n ⋅ ∂ ν ∂ λ n ) ( ∂ λ n ⋅ ∂ μ ∂ σ n ) , (55) R_{\sigma\mu\nu}^{\rho}=\frac{1}{4}\left[\partial_{\mu}\partial_{\rho}\mathbf{n}\cdot\partial_{\nu}\partial_{\sigma}\mathbf{n}+\partial_{\rho}\mathbf{n}\cdot\partial_{\mu}\partial_{\nu}\partial_{\sigma}\mathbf{n}-\partial_{\nu}\partial_{\rho}\mathbf{n}\cdot\partial_{\mu}\partial_{\sigma}\mathbf{n}-\partial_{\rho}\mathbf{n}\cdot\partial_{\mu}\partial_{\nu}\partial_{\sigma}\mathbf{n}\right]
+\frac{1}{16}(\partial_{\rho}\mathbf{n}\cdot\partial_{\mu}\partial_{\lambda}\mathbf{n})(\partial_{\lambda}\mathbf{n}\cdot\partial_{\nu}\partial_{\sigma}\mathbf{n})
-\frac{1}{16}(\partial_{\rho}\mathbf{n}\cdot\partial_{\nu}\partial_{\lambda}\mathbf{n})(\partial_{\lambda}\mathbf{n}\cdot\partial_{\mu}\partial_{\sigma}\mathbf{n}), \tag*{(55)} R σ μν ρ = 4 1 [ ∂ μ ∂ ρ n ⋅ ∂ ν ∂ σ n + ∂ ρ n ⋅ ∂ μ ∂ ν ∂ σ n − ∂ ν ∂ ρ n ⋅ ∂ μ ∂ σ n − ∂ ρ n ⋅ ∂ μ ∂ ν ∂ σ n ] + 16 1 ( ∂ ρ n ⋅ ∂ μ ∂ λ n ) ( ∂ λ n ⋅ ∂ ν ∂ σ n ) − 16 1 ( ∂ ρ n ⋅ ∂ ν ∂ λ n ) ( ∂ λ n ⋅ ∂ μ ∂ σ n ) , (55)
and the Ricci tensor reads
R μ ν = 1 4 ( ∂ λ ∂ λ n ⋅ ∂ ν ∂ μ n − ∂ ν ∂ λ n ⋅ ∂ λ ∂ μ n ) + 1 16 ( ∂ λ n ⋅ ∂ λ ∂ λ n ) ( ∂ λ n ⋅ ∂ ν ∂ μ n ) − 1 16 ( ∂ λ n ⋅ ∂ ν ∂ λ n ) ( ∂ λ n ⋅ ∂ λ ∂ μ n ) . (56) R_{\mu\nu}=\frac{1}{4}\left(\partial_{\lambda}\partial_{\lambda}\mathbf{n}\cdot\partial_{\nu}\partial_{\mu}\mathbf{n}-\partial_{\nu}\partial_{\lambda}\mathbf{n}\cdot\partial_{\lambda}\partial_{\mu}\mathbf{n}\right)
+\frac{1}{16}(\partial_{\lambda}\mathbf{n}\cdot\partial_{\lambda}\partial_{\lambda}\mathbf{n})(\partial_{\lambda}\mathbf{n}\cdot\partial_{\nu}\partial_{\mu}\mathbf{n})
-\frac{1}{16}(\partial_{\lambda}\mathbf{n}\cdot\partial_{\nu}\partial_{\lambda}\mathbf{n})(\partial_{\lambda}\mathbf{n}\cdot\partial_{\lambda}\partial_{\mu}\mathbf{n}). \tag*{(56)} R μν = 4 1 ( ∂ λ ∂ λ n ⋅ ∂ ν ∂ μ n − ∂ ν ∂ λ n ⋅ ∂ λ ∂ μ n ) + 16 1 ( ∂ λ n ⋅ ∂ λ ∂ λ n ) ( ∂ λ n ⋅ ∂ ν ∂ μ n ) − 16 1 ( ∂ λ n ⋅ ∂ ν ∂ λ n ) ( ∂ λ n ⋅ ∂ λ ∂ μ n ) . (56)
Non-Abelian quantum geometry
So far, we have studied the case where the target band is a single band. If the target bands are N N N -fold degenerate, we consider the N N N -fold degenerate wave function,
∣ ψ ( k ) ⟩ = ∑ n = 1 N c n ∣ ψ n ( k ) ⟩ , (57) |\psi(\mathbf{k})\rangle= \sum_{n=1}^{N} c_n|\psi_n(\mathbf{k})\rangle, \tag*{(57)} ∣ ψ ( k )⟩ = n = 1 ∑ N c n ∣ ψ n ( k )⟩ , (57)
satisfying the normalization condition, ∑ n = 1 N ∣ c n ∣ 2 = 1 \sum_{n=1}^{N}|c_n|^2=1 ∑ n = 1 N ∣ c n ∣ 2 = 1 . It is necessary to generalize quantum geometry to the non-Abelian quantum geometry[3 ] . The fidelity of the wave functions ∣ ψ ( k ) ⟩ |\psi(\mathbf{k})\rangle ∣ ψ ( k )⟩ and ∣ ψ ( k + d k ) ⟩ |\psi(\mathbf{k}+d\mathbf{k})\rangle ∣ ψ ( k + d k )⟩ is defined by
F ( k , k + d k ) ≡ ∣ ⟨ ψ ( k ) ∣ U d k ∣ ψ ( k ) ⟩ ∣ (58) F(\mathbf{k},\mathbf{k}+d\mathbf{k})\equiv|\langle\psi(\mathbf{k})|U_{d\mathbf{k}}|\psi(\mathbf{k})\rangle| \tag*{(58)} F ( k , k + d k ) ≡ ∣ ⟨ ψ ( k ) ∣ U d k ∣ ψ ( k )⟩ ∣ (58)
with U d k U_{d\mathbf{k}} U d k given by Eq. (5) . The quantum distance is
d S H S ≡ 1 − F ( k , k + d k ) 2 (59) d_{\mathrm{SHS}}\equiv\sqrt{1-F(\mathbf{k},\mathbf{k}+d\mathbf{k})^2} \tag*{(59)} d SHS ≡ 1 − F ( k , k + d k ) 2 (59)
The non-Abelian quantum geometric tensor is given by
( d S H S ) 2 ≡ ∑ μ ν ∑ n m F n m μ ν d k μ d k ν (60) (d_{\mathrm{SHS}})^2\equiv\sum_{\mu\nu}\sum_{nm}\mathcal{F}_{nm}^{\mu\nu}dk_\mu dk_\nu \tag*{(60)} ( d SHS ) 2 ≡ μν ∑ nm ∑ F nm μν d k μ d k ν (60)
with
F n m μ ν = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k ) ) ∣ ∂ k ν ψ m ( k ) ⟩ (61) \mathcal{F}_{nm}^{\mu\nu}=\langle\partial_{k_\mu}\psi_n(\mathbf{k})|(1-P(\mathbf{k}))|\partial_{k_\nu}\psi_m(\mathbf{k})\rangle \tag*{(61)} F nm μν = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k )) ∣ ∂ k ν ψ m ( k )⟩ (61)
where P ( k ) P(\mathbf{k}) P ( k ) is the projection operator onto the target bands,
P ( k ) ≡ ∑ n = 1 N ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ , (62) P(\mathbf{k})\equiv\sum_{n=1}^{N}|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|, \tag*{(62)} P ( k ) ≡ n = 1 ∑ N ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ , (62)
which satisfies the idempotence relation P ( k ) 2 = P ( k ) P(\mathbf{k})^2=P(\mathbf{k}) P ( k ) 2 = P ( k ) . The derivation is shown in Appendix B.
The quantum metric is the real part of the quantum geometric tensor,
g n m μ ν ≡ Re F n m μ ν = F m n μ ν ( k ) + ( F m n μ ν ( k ) ) ∗ 2 , (63) g_{nm}^{\mu\nu}\equiv\operatorname{Re}\mathcal{F}_{nm}^{\mu\nu}=\frac{\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k})+(\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k}))^*}{2}, \tag*{(63)} g nm μν ≡ Re F nm μν = 2 F mn μν ( k ) + ( F mn μν ( k ) ) ∗ , (63)
while the Berry curvature is the imaginary part of the quantum geometric tensor,
Ω n m μ ν ≡ 2 Im F n m μ ν = i ( F m n μ ν ( k ) − ( F m n μ ν ( k ) ) ∗ ) . (64) \Omega_{nm}^{\mu\nu}\equiv2\operatorname{Im}\mathcal{F}_{nm}^{\mu\nu}=i\left(\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k})-(\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k}))^*\right). \tag*{(64)} Ω nm μν ≡ 2 Im F nm μν = i ( F mn μν ( k ) − ( F mn μν ( k ) ) ∗ ) . (64)
It is further calculated as
Ω n m μ ν = i ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k ) ) ∣ ∂ k ν ψ m ( k ) ⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ( 1 − P ( k ) ) ∣ ∂ k μ ψ n ( k ) ⟩ = i ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ m ( k ) ⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ − i ∑ n ′ ≠ n , m ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ′ ( k ) ⟩ ⟨ ψ n ′ ( k ) ∣ ∂ k ν ψ m ( k ) ⟩ + i ∑ n ′ ≠ n , m ⟨ ∂ k ν ψ m ( k ) ∣ ψ n ′ ( k ) ⟩ ⟨ ψ n ′ ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ = i ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ m ( k ) ⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ + ∑ n ′ ≠ n , m ( i a n ′ n μ ∗ a n ′ m ν − i a n ′ m ν ∗ a n ′ n μ ) . (65) \begin{aligned}
\Omega_{nm}^{\mu\nu}
&=i\langle\partial_{k_\mu}\psi_n(\mathbf{k})|(1-P(\mathbf{k}))|\partial_{k_\nu}\psi_m(\mathbf{k})\rangle\\
&\quad-i\langle\partial_{k_\nu}\psi_m(\mathbf{k})|(1-P(\mathbf{k}))|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&=i\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_m(\mathbf{k})\rangle-i\langle\partial_{k_\nu}\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&\quad-i\sum_{n'\ne n,m}\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_{n'}(\mathbf{k})\rangle\langle\psi_{n'}(\mathbf{k})|\partial_{k_\nu}\psi_m(\mathbf{k})\rangle\\
&\quad+i\sum_{n'\ne n,m}\langle\partial_{k_\nu}\psi_m(\mathbf{k})|\psi_{n'}(\mathbf{k})\rangle\langle\psi_{n'}(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&=i\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_m(\mathbf{k})\rangle-i\langle\partial_{k_\nu}\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&\quad+\sum_{n'\ne n,m}\left(ia_{n'n}^{\mu*}a_{n'm}^{\nu}-ia_{n'm}^{\nu*}a_{n'n}^{\mu}\right). \tag*{(65)}
\end{aligned} Ω nm μν = i ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k )) ∣ ∂ k ν ψ m ( k )⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ( 1 − P ( k )) ∣ ∂ k μ ψ n ( k )⟩ = i ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ m ( k )⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ − i n ′ = n , m ∑ ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ′ ( k )⟩ ⟨ ψ n ′ ( k ) ∣ ∂ k ν ψ m ( k )⟩ + i n ′ = n , m ∑ ⟨ ∂ k ν ψ m ( k ) ∣ ψ n ′ ( k )⟩ ⟨ ψ n ′ ( k ) ∣ ∂ k μ ψ n ( k )⟩ = i ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ m ( k )⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ + n ′ = n , m ∑ ( i a n ′ n μ ∗ a n ′ m ν − i a n ′ m ν ∗ a n ′ n μ ) . (65)
With the use of the Hermiticity condition of the Berry connection (17) , we have
Ω n m μ ν = i ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ m ( k ) ⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ − ∑ n ′ ≠ n , m i a n n ′ μ a n ′ m ν + ∑ n ′ ≠ n , m i a m n ′ ν a n ′ n μ = ∇ × a n m μ − ∑ n ′ ≠ n , m i [ a n n ′ μ , a n ′ m ν ] . (66) \begin{aligned}
\Omega_{nm}^{\mu\nu}
&=i\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_m(\mathbf{k})\rangle-i\langle\partial_{k_\nu}\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&\quad-\sum_{n'\ne n,m}ia_{nn'}^\mu a_{n'm}^\nu+\sum_{n'\ne n,m}ia_{mn'}^\nu a_{n'n}^\mu\\
&=\nabla\times a_{nm}^\mu-\sum_{n'\ne n,m}i[a_{nn'}^\mu,a_{n'm}^\nu]. \tag*{(66)}
\end{aligned} Ω nm μν = i ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ m ( k )⟩ − i ⟨ ∂ k ν ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ − n ′ = n , m ∑ i a n n ′ μ a n ′ m ν + n ′ = n , m ∑ i a m n ′ ν a n ′ n μ = ∇ × a nm μ − n ′ = n , m ∑ i [ a n n ′ μ , a n ′ m ν ] . (66)
There is an additional term − i [ a n n ′ μ , a n ′ m ν ] -i[a_{nn'}^\mu,a_{n'm}^\nu] − i [ a n n ′ μ , a n ′ m ν ] in the Wilczek-Zee connection comparing with the Berry connection.
Quantum geometry in condensed matter physics
Quantum geometric properties are observable in condensed matter physics. First, we review that the Berry curvature is observable by the Hall conductance. The Hall conductance is quantized for an insulator, which is well described by the Chern number. On the other hand, the quantum metric appears in the optical absorption, the bulk photovoltaic effect and nonlinear conductivities.
The Berry curvature is related to the Hall conductivity. Especially, the Hall conductivity is quantized in an insulator and is proportional to the Chern number, which is known as the TKNN formula[4 ] .
According to the Kubo formula, the Hall conductance is given by
σ x y = − i ℏ e 2 ∫ d k ∑ n ≠ m f ( ε n ( k ) ) ( ε n ( k ) − ε m ( k ) ) 2 [ ⟨ ψ n ( k ) ∣ v x ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ v y ∣ ψ n ( k ) ⟩ − ⟨ ψ n ( k ) ∣ v y ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ v x ∣ ψ n ( k ) ⟩ ] . (67) \sigma_{xy}=-i\hbar e^2\int d\mathbf{k}\sum_{n\ne m}\frac{f(\varepsilon_n(\mathbf{k}))}{(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k}))^2}\left[\langle\psi_n(\mathbf{k})|v_x|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|v_y|\psi_n(\mathbf{k})\rangle-\langle\psi_n(\mathbf{k})|v_y|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|v_x|\psi_n(\mathbf{k})\rangle\right]. \tag*{(67)} σ x y = − i ℏ e 2 ∫ d k n = m ∑ ( ε n ( k ) − ε m ( k ) ) 2 f ( ε n ( k )) [ ⟨ ψ n ( k ) ∣ v x ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ v y ∣ ψ n ( k )⟩ − ⟨ ψ n ( k ) ∣ v y ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ v x ∣ ψ n ( k )⟩ ] . (67)
where f ( ε n ( k ) ) = 1 / ( exp [ ( ε n ( k ) − μ ) / ( k B T ) + 1 ] ) f(\varepsilon_n(\mathbf{k}))=1/(\exp[(\varepsilon_n(\mathbf{k})-\mu)/(k_{\mathrm{B}}T)+1]) f ( ε n ( k )) = 1/ ( exp [( ε n ( k ) − μ ) / ( k B T ) + 1 ]) is the Fermi distribution function and μ \mu μ is the chemical potential. By using the Hellmann-Feynman theorem
⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k ) ⟩ = 1 ℏ ∂ k μ ε n ( k ) ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ + 1 ℏ ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ (68) \langle\psi_m(\mathbf{k})|v_\mu|\psi_n(\mathbf{k})\rangle=\frac{1}{\hbar}\partial_{k_\mu}\varepsilon_n(\mathbf{k})\langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\frac{1}{\hbar}(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k}))\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle \tag*{(68)} ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k )⟩ = ℏ 1 ∂ k μ ε n ( k ) ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ + ℏ 1 ( ε n ( k ) − ε m ( k )) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ (68)
for the states subject to the orthonormalization condition, ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ = δ m n \langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle=\delta_{mn} ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ = δ mn , the Hall conductance is rewritten as
σ x y = − i e 2 ℏ ∫ d k ∑ n , m f ( ε n ( k ) ) [ ⟨ ψ n ( k ) | ∂ k x ψ m ( k ) ⟩ ⟨ ψ m ( k ) | ∂ k y ψ n ( k ) ⟩ − ⟨ ψ n ( k ) | ∂ k y ψ m ( k ) ⟩ ⟨ ψ m ( k ) | ∂ k x ψ n ( k ) ⟩ ] . (69) \sigma_{xy}=-\frac{ie^2}{\hbar}\int d\mathbf{k}\sum_{n,m} f\left(\varepsilon_n(\mathbf{k})\right)
\left[\left\langle\psi_n(\mathbf{k})\middle|\partial_{k_x}\psi_m(\mathbf{k})\right\rangle\left\langle\psi_m(\mathbf{k})\middle|\partial_{k_y}\psi_n(\mathbf{k})\right\rangle
-\left\langle\psi_n(\mathbf{k})\middle|\partial_{k_y}\psi_m(\mathbf{k})\right\rangle\left\langle\psi_m(\mathbf{k})\middle|\partial_{k_x}\psi_n(\mathbf{k})\right\rangle\right]. \tag*{(69)} σ x y = − ℏ i e 2 ∫ d k n , m ∑ f ( ε n ( k ) ) [ ⟨ ψ n ( k ) ∣ ∂ k x ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∂ k y ψ n ( k ) ⟩ − ⟨ ψ n ( k ) ∂ k y ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k x ψ n ( k ) ⟩ ] . (69)
It should be noticed that the sum can be extended to include the states with n = m n=m n = m in this formula, since
⟨ ψ n ( k ) | ∂ k x ψ n ( k ) ⟩ ⟨ ψ n ( k ) | ∂ k y ψ n ( k ) ⟩ = ⟨ ψ n ( k ) | ∂ k y ψ n ( k ) ⟩ ⟨ ψ n ( k ) | ∂ k x ψ n ( k ) ⟩ . (70) \left\langle\psi_n(\mathbf{k})\middle|\partial_{k_x}\psi_n(\mathbf{k})\right\rangle\left\langle\psi_n(\mathbf{k})\middle|\partial_{k_y}\psi_n(\mathbf{k})\right\rangle
=\left\langle\psi_n(\mathbf{k})\middle|\partial_{k_y}\psi_n(\mathbf{k})\right\rangle\left\langle\psi_n(\mathbf{k})\middle|\partial_{k_x}\psi_n(\mathbf{k})\right\rangle. \tag*{(70)} ⟨ ψ n ( k ) ∣ ∂ k x ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∂ k y ψ n ( k ) ⟩ = ⟨ ψ n ( k ) ∂ k y ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ∂ k x ψ n ( k ) ⟩ . (70)
The proof of the Hellmann–Feynman theorem is shown in Appendix C.
Making the use of the relation
⟨ ψ n ( k ) | ∂ k x ψ m ( k ) ⟩ + ⟨ ∂ k x ψ n ( k ) | ψ m ( k ) ⟩ = ∂ k x ⟨ ψ m ( k ) | ψ n ( k ) ⟩ = 0 , (71) \left\langle\psi_n(\mathbf{k})\middle|\partial_{k_x}\psi_m(\mathbf{k})\right\rangle+\left\langle\partial_{k_x}\psi_n(\mathbf{k})\middle|\psi_m(\mathbf{k})\right\rangle
=\partial_{k_x}\left\langle\psi_m(\mathbf{k})\middle|\psi_n(\mathbf{k})\right\rangle=0, \tag*{(71)} ⟨ ψ n ( k ) ∣ ∂ k x ψ m ( k ) ⟩ + ⟨ ∂ k x ψ n ( k ) ∣ ψ m ( k ) ⟩ = ∂ k x ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ = 0 , (71)
the Hall conductance is rewritten as
σ x y = − i e 2 ℏ ∫ d k ∑ n , m f ( ε ( k ) ) [ ⟨ ∂ k x ψ n ( k ) | ψ m ( k ) ⟩ ⟨ ψ m ( k ) | ∂ k y ψ n ( k ) ⟩ − ⟨ ∂ k y ψ n ( k ) | ψ m ( k ) ⟩ ⟨ ψ m ( k ) | ∂ k x ψ n ( k ) ⟩ ] . (72) \sigma_{xy}=-\frac{ie^2}{\hbar}\int d\mathbf{k}\sum_{n,m}f\left(\varepsilon(\mathbf{k})\right)
\left[\left\langle\partial_{k_x}\psi_n(\mathbf{k})\middle|\psi_m(\mathbf{k})\right\rangle\left\langle\psi_m(\mathbf{k})\middle|\partial_{k_y}\psi_n(\mathbf{k})\right\rangle
-\left\langle\partial_{k_y}\psi_n(\mathbf{k})\middle|\psi_m(\mathbf{k})\right\rangle\left\langle\psi_m(\mathbf{k})\middle|\partial_{k_x}\psi_n(\mathbf{k})\right\rangle\right]. \tag*{(72)} σ x y = − ℏ i e 2 ∫ d k n , m ∑ f ( ε ( k ) ) [ ⟨ ∂ k x ψ n ( k ) ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∂ k y ψ n ( k ) ⟩ − ⟨ ∂ k y ψ n ( k ) ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k x ψ n ( k ) ⟩ ] . (72)
Using the completeness condition ∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ = 1 \sum_m\lvert\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})\rvert=1 ∑ m ∣ ψ m ( k )⟩ ⟨ ψ m ( k )∣ = 1 , we obtain
σ x y = i e 2 ℏ ∫ d k ∑ n f ( ε n ( k ) ) [ ⟨ ∂ k x ψ n ( k ) | ∂ k y ψ n ( k ) ⟩ − ⟨ ∂ k y ψ n ( k ) | ∂ k x ψ n ( k ) ⟩ ] . (73) \sigma_{xy}=\frac{ie^2}{\hbar}\int d\mathbf{k}\sum_n f\left(\varepsilon_n(\mathbf{k})\right)\left[\left\langle\partial_{k_x}\psi_n(\mathbf{k})\middle|\partial_{k_y}\psi_n(\mathbf{k})\right\rangle-\left\langle\partial_{k_y}\psi_n(\mathbf{k})\middle|\partial_{k_x}\psi_n(\mathbf{k})\right\rangle\right]. \tag*{(73)} σ x y = ℏ i e 2 ∫ d k n ∑ f ( ε n ( k ) ) [ ⟨ ∂ k x ψ n ( k ) ∂ k y ψ n ( k ) ⟩ − ⟨ ∂ k y ψ n ( k ) ∂ k x ψ n ( k ) ⟩ ] . (73)
It is rewritten as
σ x y = e 2 2 π ℏ ∑ n ∫ d k f ( ε n ( k ) ) Ω n x y ( k ) , (74) \sigma_{xy}=\frac{e^2}{2\pi\hbar}\sum_n\int d\mathbf{k}\,f\left(\varepsilon_n(\mathbf{k})\right)\Omega_n^{xy}(\mathbf{k}), \tag*{(74)} σ x y = 2 π ℏ e 2 n ∑ ∫ d k f ( ε n ( k ) ) Ω n x y ( k ) , (74)
where
Ω n x y ( k ) = i [ ⟨ ∂ k x ψ n ( k ) | ∂ k y ψ n ( k ) ⟩ − ⟨ ∂ k y ψ n ( k ) | ∂ k x ψ n ( k ) ⟩ ] . (75) \Omega_n^{xy}(\mathbf{k})=i\left[\left\langle\partial_{k_x}\psi_n(\mathbf{k})\middle|\partial_{k_y}\psi_n(\mathbf{k})\right\rangle-\left\langle\partial_{k_y}\psi_n(\mathbf{k})\middle|\partial_{k_x}\psi_n(\mathbf{k})\right\rangle\right]. \tag*{(75)} Ω n x y ( k ) = i [ ⟨ ∂ k x ψ n ( k ) ∂ k y ψ n ( k ) ⟩ − ⟨ ∂ k y ψ n ( k ) ∂ k x ψ n ( k ) ⟩ ] . (75)
The conductance is the sum of the contributions from various bands indexed by n n n . It is convenient to define a "gauge potential" in the momentum space for each band index n n n by
a n μ ( k ) = − i ⟨ ψ n ( k ) | ∂ k μ ψ n ( k ) ⟩ , (76) a_n^\mu(\mathbf{k})=-i\left\langle\psi_n(\mathbf{k})\middle|\partial_{k_\mu}\psi_n(\mathbf{k})\right\rangle, \tag*{(76)} a n μ ( k ) = − i ⟨ ψ n ( k ) ∂ k μ ψ n ( k ) ⟩ , (76)
which is the Berry connection. We may rewrite (75) as
Ω n x y ( k ) = ∂ k x a n y ( k ) − ∂ k y a n x ( k ) . (77) \Omega_n^{xy}(\mathbf{k})=\partial_{k_x}a_n^y(\mathbf{k})-\partial_{k_y}a_n^x(\mathbf{k}). \tag*{(77)} Ω n x y ( k ) = ∂ k x a n y ( k ) − ∂ k y a n x ( k ) . (77)
This is the "magnetic field", which is the Berry curvature. We consider the zero-temperature limit, where the Fermi distribution function becomes a step function, f ( ε n ( k ) ) = θ ( μ − ε n ) f\left(\varepsilon_n(\mathbf{k})\right)=\theta(\mu-\varepsilon_n) f ( ε n ( k ) ) = θ ( μ − ε n ) . The conductance (74) is simplified as
σ x y = e 2 h ∑ n = 1 N C n , (78) \sigma_{xy}=\frac{e^2}{h}\sum_{n=1}^{N}C_n, \tag*{(78)} σ x y = h e 2 n = 1 ∑ N C n , (78)
because the integration is taken below the Fermi energy.
Dirac system
We apply the above formula to the Dirac Hamiltonian[?],
H ξ = ℏ v F ( ξ k x σ x + k y σ y ) + m σ z , (79) H^\xi=\hbar v_F\left(\xi k_x\sigma_x+k_y\sigma_y\right)+m\sigma_z, \tag*{(79)} H ξ = ℏ v F ( ξ k x σ x + k y σ y ) + m σ z , (79)
where ξ = ± 1 \xi=\pm1 ξ = ± 1 is the valley degrees of freedom. It describes the low-energy theory of graphene. It is expressed as H ξ = h ⋅ σ H^\xi=\mathbf{h}\cdot\boldsymbol{\sigma} H ξ = h ⋅ σ with
n ( k ) = 1 ( ℏ v F k ) 2 + m 2 ( ξ ℏ v F k x , ℏ v F k y , m ) (80) \mathbf{n}(\mathbf{k})=\frac{1}{\sqrt{(\hbar v_F k)^2+m^2}}\left(\xi\hbar v_F k_x,\hbar v_F k_y,m\right) \tag*{(80)} n ( k ) = ( ℏ v F k ) 2 + m 2 1 ( ξ ℏ v F k x , ℏ v F k y , m ) (80)
in the polar coordinate k = ( k , ϕ ) \mathbf{k}=(k,\phi) k = ( k , ϕ ) . The quantum metrices are calculated as
g ± x x ( k ) = ∓ ( ℏ v F ) 2 ( ( ℏ v F k y ) 2 + m 2 ) ( ( ℏ v F k ) 2 + m 2 ) 2 , (81) g_{\pm}^{xx}(\mathbf{k})=\mp\frac{(\hbar v_F)^2\left((\hbar v_F k_y)^2+m^2\right)}{\left((\hbar v_F k)^2+m^2\right)^2}, \tag*{(81)} g ± xx ( k ) = ∓ ( ( ℏ v F k ) 2 + m 2 ) 2 ( ℏ v F ) 2 ( ( ℏ v F k y ) 2 + m 2 ) , (81)
g ± y y ( k ) = ∓ ( ℏ v F ) 2 ( ( ℏ v F k x ) 2 + m 2 ) ( ( ℏ v F k ) 2 + m 2 ) 2 , (82) g_{\pm}^{yy}(\mathbf{k})=\mp\frac{(\hbar v_F)^2\left((\hbar v_F k_x)^2+m^2\right)}{\left((\hbar v_F k)^2+m^2\right)^2}, \tag*{(82)} g ± y y ( k ) = ∓ ( ( ℏ v F k ) 2 + m 2 ) 2 ( ℏ v F ) 2 ( ( ℏ v F k x ) 2 + m 2 ) , (82)
g ± x y ( k ) = ± ( ℏ v F ) 2 ( ℏ v F k x ) ( ℏ v F k y ) ( ( ℏ v F k ) 2 + m 2 ) 2 , (83) g_{\pm}^{xy}(\mathbf{k})=\pm\frac{(\hbar v_F)^2(\hbar v_F k_x)(\hbar v_F k_y)}{\left((\hbar v_F k)^2+m^2\right)^2}, \tag*{(83)} g ± x y ( k ) = ± ( ( ℏ v F k ) 2 + m 2 ) 2 ( ℏ v F ) 2 ( ℏ v F k x ) ( ℏ v F k y ) , (83)
while the Berry curvature is calculated as
Ω ± x y ( k ) = ± ξ m ( ℏ v F ) 2 ( ( ℏ v F k ) 2 + m 2 ) 3 / 2 . (84) \Omega_{\pm}^{xy}(\mathbf{k})=\pm\xi\frac{m(\hbar v_F)^2}{\left((\hbar v_F k)^2+m^2\right)^{3/2}}. \tag*{(84)} Ω ± x y ( k ) = ± ξ ( ( ℏ v F k ) 2 + m 2 ) 3/2 m ( ℏ v F ) 2 . (84)
When the chemical potential is within the gap μ < ∣ m ∣ \mu<\lvert m\rvert μ < ∣ m ∣ , we may calculate the Chern number explicitly as
C ± ξ = 1 2 π ∫ 0 ∞ 2 π k d k Ω ± x y ( k ) = − ξ 2 m ∣ m ∣ 3 ∫ 0 ∞ d k ( ℏ v F ) 2 2 ( ( ℏ v F / m ) 2 k 2 + 1 ) 3 / 2 = − ξ 4 m ∣ m ∣ ∫ 0 ∞ d x 1 ( x + 1 ) 3 / 2 = − ξ 4 m ∣ m ∣ ∫ 1 ∞ d y y − 3 / 2 = ξ 2 m ∣ m ∣ y − 1 / 2 ∣ 1 ∞ = − ξ 2 sgn ( m ) . (85) C^\xi_{\pm}=\frac{1}{2\pi}\int_0^\infty2\pi k\,dk\,\Omega_{\pm}^{xy}(\mathbf{k})
=-\frac{\xi}{2}\frac{m}{\lvert m\rvert^3}\int_0^\infty dk\frac{(\hbar v_F)^2}{2\left((\hbar v_F/m)^2k^2+1\right)^{3/2}}
=-\frac{\xi}{4}\frac{m}{\lvert m\rvert}\int_0^\infty dx\frac{1}{(x+1)^{3/2}}
=-\frac{\xi}{4}\frac{m}{\lvert m\rvert}\int_1^\infty dy\,y^{-3/2}
=\left.\frac{\xi}{2}\frac{m}{\lvert m\rvert}y^{-1/2}\right|_1^\infty=-\frac{\xi}{2}\operatorname{sgn}(m). \tag*{(85)} C ± ξ = 2 π 1 ∫ 0 ∞ 2 π k d k Ω ± x y ( k ) = − 2 ξ ∣ m ∣ 3 m ∫ 0 ∞ d k 2 ( ( ℏ v F / m ) 2 k 2 + 1 ) 3/2 ( ℏ v F ) 2 = − 4 ξ ∣ m ∣ m ∫ 0 ∞ d x ( x + 1 ) 3/2 1 = − 4 ξ ∣ m ∣ m ∫ 1 ∞ d y y − 3/2 = 2 ξ ∣ m ∣ m y − 1/2 1 ∞ = − 2 ξ sgn ( m ) . (85)
When the chemical potential is among the gap, ∣ μ ∣ < ∣ m ∣ \lvert\mu\rvert<\lvert m\rvert ∣ μ ∣ < ∣ m ∣ , the Chern number is already calculated as in (85) . When the chemical potential is below or above the gap, ∣ μ ∣ > ∣ m ∣ \lvert\mu\rvert>\lvert m\rvert ∣ μ ∣ > ∣ m ∣ , the integration is taken in the region for k > k F k>k_F k > k F with
k F = 1 ℏ v F μ 2 − m 2 . (86) k_F=\frac{1}{\hbar v_F}\sqrt{\mu^2-m^2}. \tag*{(86)} k F = ℏ v F 1 μ 2 − m 2 . (86)
which is the solution of ε ( k ) = μ \varepsilon(\boldsymbol{k})=\mu ε ( k ) = μ . The Hall conductivity is calculated as
σ x y ξ e 2 / h = 1 2 π ∫ k F ∞ 2 π k d k Ω ξ ( k ) = − ξ 2 m ∣ m ∣ 3 ∫ k F 2 ∞ d 2 k ( ℏ v F ) 2 2 ( ( ℏ v F / m ) 2 k 2 + 1 ) 3 / 2 = − ξ 2 m ∣ m ∣ ∫ μ 2 / m 2 − 1 ∞ d x 1 2 ( x + 1 ) 3 / 2 = − ξ 4 m ∣ m ∣ ∫ μ 2 / m 2 ∞ d y y − 3 / 2 = ξ 2 m ∣ m ∣ y − 1 / 2 ∣ μ 2 / m 2 ∞ = − ξ 2 ∣ m ∣ ∣ μ ∣ sgn ( m ) . (87) \begin{aligned}
\frac{\sigma_{xy}^{\xi}}{e^2/h}=\frac{1}{2\pi}\int_{k_{\mathrm{F}}}^{\infty}2\pi k\,\mathrm{d}k\,\Omega^{\xi}(\boldsymbol{k})
=-\frac{\xi}{2}\frac{m}{|m|^3}\int_{k_{\mathrm{F}}^2}^{\infty}\mathrm{d}^2k\,\frac{(\hbar v_{\mathrm{F}})^2}{2\left((\hbar v_{\mathrm{F}}/m)^2k^2+1\right)^{3/2}} \\
&=-\frac{\xi}{2}\frac{m}{|m|}\int_{\mu^2/m^2-1}^{\infty}\mathrm{d}x\,\frac{1}{2(x+1)^{3/2}} \\
&=-\frac{\xi}{4}\frac{m}{|m|}\int_{\mu^2/m^2}^{\infty}\mathrm{d}y\,y^{-3/2} \\
&=\frac{\xi}{2}\frac{m}{|m|}y^{-1/2}\bigg|_{\mu^2/m^2}^{\infty}
=-\frac{\xi}{2}\frac{|m|}{|\mu|}\operatorname{sgn}(m). \tag*{(87)}
\end{aligned} e 2 / h σ x y ξ = 2 π 1 ∫ k F ∞ 2 π k d k Ω ξ ( k ) = − 2 ξ ∣ m ∣ 3 m ∫ k F 2 ∞ d 2 k 2 ( ( ℏ v F / m ) 2 k 2 + 1 ) 3/2 ( ℏ v F ) 2 = − 2 ξ ∣ m ∣ m ∫ μ 2 / m 2 − 1 ∞ d x 2 ( x + 1 ) 3/2 1 = − 4 ξ ∣ m ∣ m ∫ μ 2 / m 2 ∞ d y y − 3/2 = 2 ξ ∣ m ∣ m y − 1/2 μ 2 / m 2 ∞ = − 2 ξ ∣ μ ∣ ∣ m ∣ sgn ( m ) . (87)
The TKNN formula dictates that the Hall conductance is proportional to the Chern number. The Hall conductance is given by σ x y = ∑ ξ σ x y ξ \sigma_{xy}=\sum_{\xi}\sigma_{xy}^{\xi} σ x y = ∑ ξ σ x y ξ with
σ x y ξ = { − ξ 2 e 2 h sgn ( m ) for ∣ μ ∣ < ∣ m ∣ , − ξ 2 ∣ m ∣ ∣ μ ∣ e 2 h sgn ( m ) for ∣ μ ∣ > ∣ m ∣ . (88) \sigma_{xy}^{\xi}=
\begin{cases}
-\dfrac{\xi}{2}\dfrac{e^2}{h}\operatorname{sgn}(m) & \text{for }|\mu|<|m|,\\
-\dfrac{\xi}{2}\dfrac{|m|}{|\mu|}\dfrac{e^2}{h}\operatorname{sgn}(m) & \text{for }|\mu|>|m|.
\end{cases} \tag*{(88)} σ x y ξ = ⎩ ⎨ ⎧ − 2 ξ h e 2 sgn ( m ) − 2 ξ ∣ μ ∣ ∣ m ∣ h e 2 sgn ( m ) for ∣ μ ∣ < ∣ m ∣ , for ∣ μ ∣ > ∣ m ∣. (88)
as a function of the chemical potential. It is quantized between the bulk band gap, ∣ μ ∣ < ∣ m ∣ |\mu|<|m| ∣ μ ∣ < ∣ m ∣ , which is known as the quantum anomalous Hall effect. On the other hand, the Hall conductance is anti-proportional to the chemical potential outside the band gap, ∣ μ ∣ > ∣ m ∣ |\mu|>|m| ∣ μ ∣ > ∣ m ∣ . Eq. (88) is shown in Figure 1 .
Figure 1. (a) The energy spectrum of the massive Dirac fermion with the mass m m m . (b) Hall conductance as a function of the chemical potential for the massive Dirac Hamiltonian. (c) The temperature dependence of the Hall conductance at the Fermi energy μ = 0 \mu= 0 μ = 0 .
We have
Tr g n μ ν ( k ) = ( ℏ v F ) 2 2 m 2 + ( ℏ v F k ) 2 2 ( ( ℏ v F k ) 2 + m 2 ) 2 . (89) \operatorname{Tr}g_{n}^{\mu\nu}(\boldsymbol{k})=(\hbar v_{\mathrm{F}})^2\frac{2m^2+(\hbar v_{\mathrm{F}}k)^2}{2\left((\hbar v_{\mathrm{F}}k)^2+m^2\right)^2}. \tag*{(89)} Tr g n μν ( k ) = ( ℏ v F ) 2 2 ( ( ℏ v F k ) 2 + m 2 ) 2 2 m 2 + ( ℏ v F k ) 2 . (89)
and
det g n μ ν ( k ) = ∣ m ∣ ( ℏ v F ) 2 2 ( ( ℏ v F k ) 2 + m 2 ) 3 / 2 . (90) \sqrt{\det g_{n}^{\mu\nu}(\boldsymbol{k})}=\frac{|m|(\hbar v_{\mathrm{F}})^2}{2\left((\hbar v_{\mathrm{F}}k)^2+m^2\right)^{3/2}}. \tag*{(90)} det g n μν ( k ) = 2 ( ( ℏ v F k ) 2 + m 2 ) 3/2 ∣ m ∣ ( ℏ v F ) 2 . (90)
Especially, there is a relation
2 det g n μ ν ( k ) = ∣ Ω ± x y ( k ) ∣ . (91) 2\sqrt{\det g_{n}^{\mu\nu}(\boldsymbol{k})}=|\Omega_{\pm}^{xy}(\boldsymbol{k})|. \tag*{(91)} 2 det g n μν ( k ) = ∣ Ω ± x y ( k ) ∣. (91)
We can check the inequality (30) is satisfied. The quantum volume is calculated as
∫ d k det g n μ ν ( k ) = 1 8 ( ℏ v F m ) 2 . (92) \int\mathrm{d}k\sqrt{\det g_{n}^{\mu\nu}(\boldsymbol{k})}=\frac{1}{8}\left(\frac{\hbar v_{\mathrm{F}}}{m}\right)^2. \tag*{(92)} ∫ d k det g n μν ( k ) = 8 1 ( m ℏ v F ) 2 . (92)
A comment is in order. The Dirac system gives a half-integer of the Chern number, which may contradict the fact that the Chern number is an integer. This problem is solved by the Nielsen-Ninomiya theorem, which shows that the number of Dirac cones must be even for the system defined for the Brillouin zone. Then, the total Chern number is always an integer, which is relevant in realistic materials.
Optical absorption and elliptic dichroism
We study optical inter-band transitions from the state ∣ ψ − ( k ) ⟩ |\psi_{-}(\boldsymbol{k})\rangle ∣ ψ − ( k )⟩ in the valence band to the state ∣ ψ + ( k ) ⟩ |\psi_{+}(\boldsymbol{k})\rangle ∣ ψ + ( k )⟩ in the conduction band. We apply a beam of elliptical polarized light perpendicular onto the sample, where the corresponding electromagnetic potential is given by A ( t ) = ( A x sin ω t , A y cos ω t ) \boldsymbol{A}(t)=(A_x\sin\omega t,A_y\cos\omega t) A ( t ) = ( A x sin ω t , A y cos ω t ) . The electromagnetic potential is introduced into the Hamiltonian by way of the minimal substitution, that is, by replacing the momentum k j k_j k j with the covariant momentum P j ≡ ℏ k j + e A j P_j\equiv\hbar k_j+eA_j P j ≡ ℏ k j + e A j .
Optical absorption is given by the optical conductivity
σ ( ω ; ϑ ) = σ 0 ∫ d k f − ( k ) − f + ( k ) ε − − ε + ∣ P ϑ ( k ) ∣ 2 × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] . (93) \sigma(\omega;\vartheta)=\sigma_0\int\mathrm{d}k\,\frac{f_{-}(\boldsymbol{k})-f_{+}(\boldsymbol{k})}{\varepsilon_{-}-\varepsilon_{+}}|P_{\vartheta}(\boldsymbol{k})|^2\times\delta\left[\varepsilon_{+}(\boldsymbol{k})-\varepsilon_{-}(\boldsymbol{k})-\hbar\omega\right]. \tag*{(93)} σ ( ω ; ϑ ) = σ 0 ∫ d k ε − − ε + f − ( k ) − f + ( k ) ∣ P ϑ ( k ) ∣ 2 × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] . (93)
where we have set
σ 0 ≡ π e 2 ( 2 π ) 2 . (94) \sigma_0\equiv\frac{\pi e^2}{(2\pi)^2}. \tag*{(94)} σ 0 ≡ ( 2 π ) 2 π e 2 . (94)
We start with the optical matrix element between the initial and final states in the photo-emission process given by[19 , 82 , 83 , 84 , 85 , 86 ]
P i ( k ) ≡ ℏ ⟨ ψ + ( k ) ∣ v μ ∣ ψ − ( k ) ⟩ , (95) P_i(\boldsymbol{k})\equiv\hbar\langle\psi_{+}(\boldsymbol{k})|v_{\mu}|\psi_{-}(\boldsymbol{k})\rangle, \tag*{(95)} P i ( k ) ≡ ℏ ⟨ ψ + ( k ) ∣ v μ ∣ ψ − ( k )⟩ , (95)
with the velocity
v μ = 1 ℏ ∂ H ( k ) ∂ k μ . (96) v_{\mu}=\frac{1}{\hbar}\frac{\partial H(\boldsymbol{k})}{\partial k_{\mu}}. \tag*{(96)} v μ = ℏ 1 ∂ k μ ∂ H ( k ) . (96)
The optical matrix element of the elliptic polarization is given by
P ϑ ( k ) = P x ( k ) cos ϑ + i P y ( k ) sin ϑ , (97) P_{\vartheta}(\boldsymbol{k})=P_x(\boldsymbol{k})\cos\vartheta+iP_y(\boldsymbol{k})\sin\vartheta, \tag*{(97)} P ϑ ( k ) = P x ( k ) cos ϑ + i P y ( k ) sin ϑ , (97)
where ϑ \vartheta ϑ is the ellipticity of the injected beam, with 0 < ϑ < π 0<\vartheta<\pi 0 < ϑ < π for the right polarization and − π < ϑ < 0 -\pi<\vartheta<0 − π < ϑ < 0 for the left polarization. P π / 4 ( k ) P_{\pi/4}(\boldsymbol{k}) P π /4 ( k ) corresponds to the right circularly polarized light, and P − π / 4 ( k ) P_{-\pi/4}(\boldsymbol{k}) P − π /4 ( k ) corresponds to the left circularly polarized light.
In the following, we assume that the temperature is absolutely zero. The optical matrix element for the elliptic polarized light is expanded as
∣ P ϑ ( k ) ∣ 2 = ∣ P x ( k ) ∣ 2 cos 2 ϑ + ∣ P y ( k ) ∣ 2 sin 2 ϑ + i [ P y ∗ ( k ) P x ( k ) − P x ∗ ( k ) P y ( k ) ] sin ϑ cos ϑ . (98) \left|P_{\vartheta}(\mathbf{k})\right|^{2}
= \left|P_x(\mathbf{k})\right|^{2}\cos^{2}\vartheta+\left|P_y(\mathbf{k})\right|^{2}\sin^{2}\vartheta
+i\left[P_y^{*}(\mathbf{k})P_x(\mathbf{k})-P_x^{*}(\mathbf{k})P_y(\mathbf{k})\right]\sin\vartheta\cos\vartheta. \tag*{(98)} ∣ P ϑ ( k ) ∣ 2 = ∣ P x ( k ) ∣ 2 cos 2 ϑ + ∣ P y ( k ) ∣ 2 sin 2 ϑ + i [ P y ∗ ( k ) P x ( k ) − P x ∗ ( k ) P y ( k ) ] sin ϑ cos ϑ . (98)
By using the Hellmann-Feynman theorem
⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k ) ⟩ = 1 ℏ ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ (99) \langle\psi_m(\mathbf{k})|v_\mu|\psi_n(\mathbf{k})\rangle
=\frac{1}{\hbar}\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle \tag*{(99)} ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k )⟩ = ℏ 1 ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ (99)
for m ≠ n m\ne n m = n , we have
∣ P μ ( k ) ∣ 2 ( ε n ( k ) − ε m ( k ) ) 2 = ℏ 2 ⟨ ψ − ( k ) ∣ v μ ∣ ψ + ( k ) ⟩ ⟨ ψ + ( k ) ∣ v μ ∣ ψ − ( k ) ⟩ = − Δ 2 ( k ) ⟨ ∂ k μ ψ − ( k ) ∣ ψ + ( k ) ⟩ ⟨ ψ + ( k ) ∣ ∂ k μ ψ − ( k ) ⟩ = − Δ 2 ( k ) g − μ μ ( k ) , (100) \frac{\left|P_\mu(\mathbf{k})\right|^2}{\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)^2}
=\hbar^2\langle\psi_-(\mathbf{k})|v_\mu|\psi_+(\mathbf{k})\rangle\langle\psi_+(\mathbf{k})|v_\mu|\psi_-(\mathbf{k})\rangle
=-\Delta^2(\mathbf{k})\langle\partial_{k_\mu}\psi_-(\mathbf{k})|\psi_+(\mathbf{k})\rangle\langle\psi_+(\mathbf{k})|\partial_{k_\mu}\psi_-(\mathbf{k})\rangle
=-\Delta^2(\mathbf{k})g_-^{\mu\mu}(\mathbf{k}), \tag*{(100)} ( ε n ( k ) − ε m ( k ) ) 2 ∣ P μ ( k ) ∣ 2 = ℏ 2 ⟨ ψ − ( k ) ∣ v μ ∣ ψ + ( k )⟩ ⟨ ψ + ( k ) ∣ v μ ∣ ψ − ( k )⟩ = − Δ 2 ( k ) ⟨ ∂ k μ ψ − ( k ) ∣ ψ + ( k )⟩ ⟨ ψ + ( k ) ∣ ∂ k μ ψ − ( k )⟩ = − Δ 2 ( k ) g − μμ ( k ) , (100)
with μ = x , y \mu=x,y μ = x , y , and
i [ P y ∗ ( k ) P x ( k ) − P x ∗ ( k ) P y ( k ) ] ( ε n ( k ) − ε m ( k ) ) 2 = i Δ 2 ( k ) [ ⟨ ψ − ( k ) ∣ v y ∣ ψ + ( k ) ⟩ ⟨ ψ + ( k ) ∣ v x ∣ ψ − ( k ) ⟩ − ⟨ ψ − ( k ) ∣ v x ∣ ψ + ( k ) ⟩ ⟨ ψ + ( k ) ∣ v y ∣ ψ − ( k ) ⟩ ] = i Δ 2 ( k ) [ F − + x y ( k ) − F − + y x ( k ) ] = Δ 2 ( k ) Ω − x y ( k ) . (101) \frac{i\left[P_y^{*}(\mathbf{k})P_x(\mathbf{k})-P_x^{*}(\mathbf{k})P_y(\mathbf{k})\right]}{\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)^2}
=i\Delta^2(\mathbf{k})\left[\langle\psi_-(\mathbf{k})|v_y|\psi_+(\mathbf{k})\rangle\langle\psi_+(\mathbf{k})|v_x|\psi_-(\mathbf{k})\rangle-\langle\psi_-(\mathbf{k})|v_x|\psi_+(\mathbf{k})\rangle\langle\psi_+(\mathbf{k})|v_y|\psi_-(\mathbf{k})\rangle\right]
=i\Delta^2(\mathbf{k})\left[\mathcal{F}_{-+}^{xy}(\mathbf{k})-\mathcal{F}_{-+}^{yx}(\mathbf{k})\right]
=\Delta^2(\mathbf{k})\Omega_-^{xy}(\mathbf{k}). \tag*{(101)} ( ε n ( k ) − ε m ( k ) ) 2 i [ P y ∗ ( k ) P x ( k ) − P x ∗ ( k ) P y ( k ) ] = i Δ 2 ( k ) [ ⟨ ψ − ( k ) ∣ v y ∣ ψ + ( k )⟩ ⟨ ψ + ( k ) ∣ v x ∣ ψ − ( k )⟩ − ⟨ ψ − ( k ) ∣ v x ∣ ψ + ( k )⟩ ⟨ ψ + ( k ) ∣ v y ∣ ψ − ( k )⟩ ] = i Δ 2 ( k ) [ F −+ x y ( k ) − F −+ y x ( k ) ] = Δ 2 ( k ) Ω − x y ( k ) . (101)
Then, the optical conductivity is rewritten by using the components of the quantum geometric tensor[19 ] as
σ ( ω ; ϑ ) = ℏ ω σ 0 ∫ d k f ( k ) G ( k ; ϑ ) δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] , (102) \sigma(\omega;\vartheta)=\hbar\omega\sigma_0\int d\mathbf{k}f(\mathbf{k})G(\mathbf{k};\vartheta)\delta\left[\varepsilon_+(\mathbf{k})-\varepsilon_-(\mathbf{k})-\hbar\omega\right], \tag*{(102)} σ ( ω ; ϑ ) = ℏ ω σ 0 ∫ d k f ( k ) G ( k ; ϑ ) δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] , (102)
with
G ( k ; ϑ ) ≡ g − x x ( k ) cos 2 ϑ + g − y y ( k ) sin 2 ϑ + Ω − x y ( k ) sin ϑ cos ϑ . (103) G(\mathbf{k};\vartheta)\equiv g_-^{xx}(\mathbf{k})\cos^2\vartheta+g_-^{yy}(\mathbf{k})\sin^2\vartheta+\Omega_-^{xy}(\mathbf{k})\sin\vartheta\cos\vartheta. \tag*{(103)} G ( k ; ϑ ) ≡ g − xx ( k ) cos 2 ϑ + g − y y ( k ) sin 2 ϑ + Ω − x y ( k ) sin ϑ cos ϑ . (103)
We study the optical absorption of the Dirac Hamiltonian (79) at the band edge ℏ ω = ε + ( 0 ) − ε − ( 0 ) \hbar\omega=\varepsilon_+(0)-\varepsilon_-(0) ℏ ω = ε + ( 0 ) − ε − ( 0 ) . It is simply given by
σ ( ω ; ϑ ) ℏ ω σ 0 = G ( 0 ; ϑ ) = g − x x ( 0 ) cos 2 ϑ + g − y y ( 0 ) sin 2 ϑ + Ω − x y ( 0 ) sin ϑ cos ϑ = ( ℏ v F ) 2 m 2 ( 1 − ξ 2 sin 2 ϑ ) , (104) \frac{\sigma(\omega;\vartheta)}{\hbar\omega\sigma_0}=G(0;\vartheta)
=g_-^{xx}(0)\cos^2\vartheta+g_-^{yy}(0)\sin^2\vartheta+\Omega_-^{xy}(0)\sin\vartheta\cos\vartheta
=\frac{(\hbar v_F)^2}{m^2}\left(1-\frac{\xi}{2}\sin2\vartheta\right), \tag*{(104)} ℏ ω σ 0 σ ( ω ; ϑ ) = G ( 0 ; ϑ ) = g − xx ( 0 ) cos 2 ϑ + g − y y ( 0 ) sin 2 ϑ + Ω − x y ( 0 ) sin ϑ cos ϑ = m 2 ( ℏ v F ) 2 ( 1 − 2 ξ sin 2 ϑ ) , (104)
where we have used quantum metrices
g − x x ( 0 ) = g − y y ( 0 ) = ( ℏ v F ) 2 m 2 , (105) g_-^{xx}(0)=g_-^{yy}(0)=\frac{(\hbar v_F)^2}{m^2}, \tag*{(105)} g − xx ( 0 ) = g − y y ( 0 ) = m 2 ( ℏ v F ) 2 , (105)
and the Berry curvature
Ω − x y ( 0 ) = − ξ ( ℏ v F ) 2 m 2 . (106) \Omega_-^{xy}(0)=-\xi\frac{(\hbar v_F)^2}{m^2}. \tag*{(106)} Ω − x y ( 0 ) = − ξ m 2 ( ℏ v F ) 2 . (106)
It shows elliptic dichroism, where there is no absorption
σ ( ω ; ϑ ) = 0 (107) \sigma(\omega;\vartheta)=0 \tag*{(107)} σ ( ω ; ϑ ) = 0 (107)
for ϑ = ξ π / 4 \vartheta=\xi\pi/4 ϑ = ξ π /4 . Namely, the right or left circularly polarized right is selectively absorbed depending on ξ \xi ξ .
Sum rule
The optical conductivity is given by
σ D μ ν ( ω ; ϑ ) = σ 0 ∫ d k f − ( k ) − f + ( k ) ε − − ε + ⟨ ψ − ( k ) ∣ v μ ∣ ψ + ( k ) ⟩ ⟨ ψ + ( k ) ∣ v ν ∣ ψ − ( k ) ⟩ × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] . (108) \sigma_{\mathrm{D}}^{\mu\nu}(\omega;\vartheta)=\sigma_0\int d\mathbf{k}\frac{f_-(\mathbf{k})-f_+(\mathbf{k})}{\varepsilon_--\varepsilon_+}\langle\psi_-(\mathbf{k})|v_\mu|\psi_+(\mathbf{k})\rangle\langle\psi_+(\mathbf{k})|v_\nu|\psi_-(\mathbf{k})\rangle
\times\delta\left[\varepsilon_+(\mathbf{k})-\varepsilon_-(\mathbf{k})-\hbar\omega\right]. \tag*{(108)} σ D μν ( ω ; ϑ ) = σ 0 ∫ d k ε − − ε + f − ( k ) − f + ( k ) ⟨ ψ − ( k ) ∣ v μ ∣ ψ + ( k )⟩ ⟨ ψ + ( k ) ∣ v ν ∣ ψ − ( k )⟩ × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] . (108)
By using the Hellmann-Feynman theorem, it is rewritten as
σ D μ ν ( ω ; ϑ ) = ω σ 0 ∫ d k f − + ( k ) a − + μ ( k ) a + − ν ( k ) × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] , (109) \sigma_{\mathrm{D}}^{\mu\nu}(\omega;\vartheta)=\omega\sigma_0\int d\mathbf{k}f_{-+}(\mathbf{k})a_{-+}^{\mu}(\mathbf{k})a_{+-}^{\nu}(\mathbf{k})
\times\delta\left[\varepsilon_+(\mathbf{k})-\varepsilon_-(\mathbf{k})-\hbar\omega\right], \tag*{(109)} σ D μν ( ω ; ϑ ) = ω σ 0 ∫ d k f −+ ( k ) a −+ μ ( k ) a +− ν ( k ) × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] , (109)
where f − + ( k ) ≡ f − ( k ) − f + ( k ) f_{-+}(\mathbf{k})\equiv f_-(\mathbf{k})-f_+(\mathbf{k}) f −+ ( k ) ≡ f − ( k ) − f + ( k ) . With the use of Eq. (19) , the optical conductivity is related to the quantum geometric tensor,
σ D μ ν ( ω ; ϑ ) = ω ℏ σ 0 ∫ d k f − + ( k ) F μ ν × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] . (110) \sigma_{\mathrm{D}}^{\mu\nu}(\omega;\vartheta)=\frac{\omega}{\hbar}\sigma_0\int d\mathbf{k}f_{-+}(\mathbf{k})\mathcal{F}^{\mu\nu}
\times\delta\left[\varepsilon_+(\mathbf{k})-\varepsilon_-(\mathbf{k})-\hbar\omega\right]. \tag*{(110)} σ D μν ( ω ; ϑ ) = ℏ ω σ 0 ∫ d k f −+ ( k ) F μν × δ [ ε + ( k ) − ε − ( k ) − ℏ ω ] . (110)
Then, we arrive at a relation between the quantum geometric tensor and the conductivity[10 , 12 ]
∫ 0 ∞ d ω ω σ D μ ν ( ω ) = π e 2 ℏ ( 2 π ) 2 ∫ F − μ ν d k , (111) \int_0^\infty\frac{d\omega}{\omega}\sigma_{\mathrm{D}}^{\mu\nu}(\omega)=\frac{\pi e^2}{\hbar(2\pi)^2}\int\mathcal{F}^{\mu\nu}_{-}\,d\mathbf{k}, \tag*{(111)} ∫ 0 ∞ ω d ω σ D μν ( ω ) = ℏ ( 2 π ) 2 π e 2 ∫ F − μν d k , (111)
where σ D μ ν ( ω ) \sigma_{\mathrm{D}}^{\mu\nu}(\omega) σ D μν ( ω ) denotes the dissipative (absorptive) component of the optical conductivity.
The real part reads[10 ]
∫ 0 ∞ d ω ω Re σ D μ ν ( ω ) = π e 2 ℏ ( 2 π ) 2 ∫ g μ ν d k . (112) \int_0^\infty\frac{d\omega}{\omega}\operatorname{Re}\sigma_{\mathrm{D}}^{\mu\nu}(\omega)=\frac{\pi e^2}{\hbar(2\pi)^2}\int g^{\mu\nu}\,d\mathbf{k}. \tag*{(112)} ∫ 0 ∞ ω d ω Re σ D μν ( ω ) = ℏ ( 2 π ) 2 π e 2 ∫ g μν d k . (112)
The imaginary part is related to the Chern number,
∫ 0 ∞ d ω ω Im σ D μ ν ( ω ) = π e 2 ℏ ( 2 π ) 2 ∫ Ω x y 2 d k = e 2 2 ℏ C − , (113) \int_0^\infty\frac{d\omega}{\omega}\operatorname{Im}\sigma_{\mathrm{D}}^{\mu\nu}(\omega)=\frac{\pi e^2}{\hbar(2\pi)^2}\int\frac{\Omega^{xy}}{2}\,d\mathbf{k}=\frac{e^2}{2\hbar}C_-, \tag*{(113)} ∫ 0 ∞ ω d ω Im σ D μν ( ω ) = ℏ ( 2 π ) 2 π e 2 ∫ 2 Ω x y d k = 2ℏ e 2 C − , (113)
where we have used the TKNN formula
σ D x y ( 0 ) ≡ σ x y = e 2 ℏ C − , (114) \sigma_{\mathrm{D}}^{xy}(0)\equiv\sigma_{xy}=\frac{e^2}{\hbar}C_-, \tag*{(114)} σ D x y ( 0 ) ≡ σ x y = ℏ e 2 C − , (114)
and the Kramers-Kronig relation. This relates the real and imaginary parts of the optical Hall conductivity
∫ 0 ∞ d ω ω Im σ D μ ν ( ω ) = π 2 σ x y ( 0 ) . (115) \int_0^\infty\frac{d\omega}{\omega}\operatorname{Im}\sigma_{\mathrm{D}}^{\mu\nu}(\omega)=\frac{\pi}{2}\sigma_{xy}(0). \tag*{(115)} ∫ 0 ∞ ω d ω Im σ D μν ( ω ) = 2 π σ x y ( 0 ) . (115)
Bulk photovoltaic effects
Photovoltaic currents are generated under photoirradiation, which will be useful for solar cell technologies. A p p p -n n n junction presents a conventional way to generate photocurrent. On the other hand, the bulk photovoltaic photocurrent generation presents another way without using a junction[87 , 88 , 89 , 90 , 91 ] . The injection current[6 , 9 , 33 , 89 , 90 , 92 , 93 , 94 , 95 ] and the shift current[6 , 9 , 33 , 88 , 89 , 90 , 92 , 93 , 94 , 96 , 97 , 98 , 99 , 100 , 101 ] are prominent, which are second order bulk photovoltaic currents.
The current density J μ J^\mu J μ along the μ \mu μ direction induced by the applied electric field E x E_x E x along the x x x direction is expanded in a power series of E x E_x E x as
J μ = ∑ ℓ = 1 ∞ σ μ ; x ℓ E x ℓ = ∑ ℓ = 1 ∞ J μ ; x ℓ , (116) J^\mu= \sum_{\ell=1}^{\infty}\sigma^{\mu;x^\ell}E_x^\ell= \sum_{\ell=1}^{\infty}J^{\mu;x^\ell}, \tag*{(116)} J μ = ℓ = 1 ∑ ∞ σ μ ; x ℓ E x ℓ = ℓ = 1 ∑ ∞ J μ ; x ℓ , (116)
where μ = x , y , z \mu=x,y,z μ = x , y , z . We refer to J μ ; x ℓ ≡ σ μ ; x ℓ E x ℓ J^{\mu;x^\ell}\equiv\sigma^{\mu;x^\ell}E_x^\ell J μ ; x ℓ ≡ σ μ ; x ℓ E x ℓ as the ℓ \ell ℓ -th order current, where σ μ ; x ℓ \sigma^{\mu;x^\ell} σ μ ; x ℓ is the ℓ \ell ℓ -th order conductivity. If there is inversion symmetry in the system, the even order conductivities with even ℓ \ell ℓ are prohibited, because
J μ ↦ − J μ , E x ↦ − E x (117) J^\mu\mapsto-J^\mu,\qquad E_x \mapsto-E_x \tag*{(117)} J μ ↦ − J μ , E x ↦ − E x (117)
under inversion symmetry.
The second order response has a form
J μ ; x 2 ( ω 1 + ω 2 ) = σ μ ; x 2 ( ω 1 + ω 2 ; ω 1 , ω 2 ) E x ( ω 1 ) E x ( ω 2 ) . (118) J^{\mu;x^2}(\omega_1+\omega_2)=\sigma^{\mu;x^2}(\omega_1+\omega_2;\omega_1,\omega_2)E_x(\omega_1)E_x(\omega_2). \tag*{(118)} J μ ; x 2 ( ω 1 + ω 2 ) = σ μ ; x 2 ( ω 1 + ω 2 ; ω 1 , ω 2 ) E x ( ω 1 ) E x ( ω 2 ) . (118)
We investigate the direct current generation,
J μ ; x 2 ( 0 ) = σ μ ; x 2 ( 0 ; ω , − ω ) E x ( ω ) E x ( − ω ) . (119) J^{\mu;x^2}(0)=\sigma^{\mu;x^2}(0;\omega,-\omega)E_x(\omega)E_x(-\omega). \tag*{(119)} J μ ; x 2 ( 0 ) = σ μ ; x 2 ( 0 ; ω , − ω ) E x ( ω ) E x ( − ω ) . (119)
In the following, we use the abbreviation J μ ; x 2 ≡ J μ ; x 2 ( 0 ) J^{\mu;x^2}\equiv J^{\mu;x^2}(0) J μ ; x 2 ≡ J μ ; x 2 ( 0 ) and σ μ ; x 2 ( ω ) ≡ σ μ ; x 2 ( 0 ; ω , − ω ) \sigma^{\mu;x^2}(\omega)\equiv\sigma^{\mu;x^2}(0;\omega,-\omega) σ μ ; x 2 ( ω ) ≡ σ μ ; x 2 ( 0 ; ω , − ω ) .
The conductivity of the injection current is in general given by the formula[6 , 9 , 33 , 89 , 90 , 92 , 93 , 94 , 95 ]
σ i n j e c t μ ; x 2 = − τ 2 π e 3 ℏ 2 ∑ n , m ∫ d k f n m Δ m n μ a n m x a m n x δ ( ω m n − ω ) , (120) \sigma_{\mathrm{inject}}^{\mu;x^2}=-\tau\frac{2\pi e^3}{\hbar^2}\sum_{n,m}\int dk f_{nm}\Delta_{mn}^{\mu}a_{nm}^{x}a_{mn}^{x}\delta(\omega_{mn}-\omega), \tag*{(120)} σ inject μ ; x 2 = − τ ℏ 2 2 π e 3 n , m ∑ ∫ d k f nm Δ mn μ a nm x a mn x δ ( ω mn − ω ) , (120)
where τ \tau τ is the relaxation time, f n m = f n − f m f_{nm}=f_n-f_m f nm = f n − f m with the Fermi distribution function f n f_n f n for the band n n n , a n m x a_{nm}^{x} a nm x is the Wilczek-Zee connection, ω n m ≡ ( ε n − ε m ) / ℏ \omega_{nm}\equiv(\varepsilon_n-\varepsilon_m)/\hbar ω nm ≡ ( ε n − ε m ) /ℏ , and
Δ m n μ = v m μ − v n μ (121) \Delta_{mn}^{\mu}=v_m^\mu-v_n^\mu \tag*{(121)} Δ mn μ = v m μ − v n μ (121)
is the difference of the velocities defined by
v m μ = 1 ℏ ⟨ m ∣ ∂ k μ H ∣ m ⟩ (122) v_m^\mu=\frac{1}{\hbar}\langle m|\partial_{k_\mu}H|m\rangle \tag*{(122)} v m μ = ℏ 1 ⟨ m ∣ ∂ k μ H ∣ m ⟩ (122)
with the Hamiltonian H H H . The injection current is induced when the velocities are different (Δ m n μ ≠ 0 \Delta_{mn}^{\mu}\ne0 Δ mn μ = 0 ) between the conduction band n n n and the valence band m m m along the μ \mu μ direction. The derivation is shown in Appendix D 1.
The shift current is in general given by the formula[6 , 9 , 33 , 89 , 90 , 92 , 93 , 94 , 96 , 97 , 98 , 99 , 100 , 101 ]
σ s h i f t μ ; x 2 = − π e 3 ℏ 2 ∑ n , m ∫ d k f n m R m n μ , x a n m x a m n x δ ( ω m n − ω ) , (123) \sigma_{\mathrm{shift}}^{\mu;x^2}=-\frac{\pi e^3}{\hbar^2}\sum_{n,m}\int dk f_{nm}R_{mn}^{\mu,x}a_{nm}^{x}a_{mn}^{x}\delta(\omega_{mn}-\omega), \tag*{(123)} σ shift μ ; x 2 = − ℏ 2 π e 3 n , m ∑ ∫ d k f nm R mn μ , x a nm x a mn x δ ( ω mn − ω ) , (123)
where
R m n μ , x = a m m x − a n n x + i ∂ k μ log a m n x (124) R_{mn}^{\mu,x}=a_{mm}^{x}-a_{nn}^{x}+i\partial_{k_\mu}\log a_{mn}^{x} \tag*{(124)} R mn μ , x = a mm x − a nn x + i ∂ k μ log a mn x (124)
is the shift vector[90 ] . The shift vector is gauge invariant although the Wilczek-Zee connection is not gauge invariant. The shift vector describes the difference of the mean position of the Wannier function between two bands m m m and n n n . The integrand in Eq. (124) is rewritten as
R m n μ , x a n m x a m n x = i a m n x a n m , μ x , (125) R_{mn}^{\mu,x}a_{nm}^{x}a_{mn}^{x}=ia_{mn}^{x}a_{nm,\mu}^{x}, \tag*{(125)} R mn μ , x a nm x a mn x = i a mn x a nm , μ x , (125)
where we have defined the covariant derivative,
∇ k μ a n m x ≡ a n m , μ x ≡ ∂ a n m x ∂ k μ − i a n m x ( a n n x − a m m x ) . (126) \nabla_{k_\mu}a_{nm}^{x}\equiv a_{nm,\mu}^{x}\equiv\frac{\partial a_{nm}^{x}}{\partial k_\mu}-ia_{nm}^{x}\left(a_{nn}^{x}-a_{mm}^{x}\right). \tag*{(126)} ∇ k μ a nm x ≡ a nm , μ x ≡ ∂ k μ ∂ a nm x − i a nm x ( a nn x − a mm x ) . (126)
The shift current is induced when the mean positions are different (R m n μ , x ≠ 0 R_{mn}^{\mu,x}\ne0 R mn μ , x = 0 ) between the conduction band n n n and the valence band m m m . The derivation is shown in Appendix D 2.
In the following, we concentrate on the longitudinal conductivities by setting μ = x \mu=x μ = x .
The injection current is rewritten in terms of the quantum metric[6 , 9 , 34 ] as
σ i n j e c t x ; x 2 = − τ 2 π e 3 ℏ 2 ∫ d k f + − Δ + − x g − x x δ ( ω + − − ω ) . (127) \sigma_{\mathrm{inject}}^{x;x^2}=-\tau\frac{2\pi e^3}{\hbar^2}\int dk f_{+-}\Delta_{+-}^{x}g_{-}^{xx}\delta(\omega_{+-}-\omega). \tag*{(127)} σ inject x ; x 2 = − τ ℏ 2 2 π e 3 ∫ d k f +− Δ +− x g − xx δ ( ω +− − ω ) . (127)
while the shift current is rewritten in terms of the quantum metric[33 , 34 ] as
σ s h i f t x ; x 2 = − π e 3 ℏ 2 V ∫ d k f + − R + − x , x g − x x δ ( ω + − − ω ) . (128) \sigma_{\mathrm{shift}}^{x;x^2}=-\frac{\pi e^3}{\hbar^2}V\int dk f_{+-}R_{+-}^{x,x}g_{-}^{xx}\delta(\omega_{+-}-\omega). \tag*{(128)} σ shift x ; x 2 = − ℏ 2 π e 3 V ∫ d k f +− R +− x , x g − xx δ ( ω +− − ω ) . (128)
Nonlinear conductivity
The second-order nonlinear conductivity σ μ ν ; ρ \sigma^{\mu\nu;\rho} σ μν ; ρ is expanded in terms of the electron relaxation time τ \tau τ as[27 ]
σ μ ν ; ρ = σ N L D r u d e μ ν ; ρ + σ D i p o l e μ ν ; ρ + σ M e t r i c μ ν ; ρ , (129) \sigma^{\mu\nu;\rho}=\sigma_{\mathrm{NLDrude}}^{\mu\nu;\rho}+\sigma_{\mathrm{Dipole}}^{\mu\nu;\rho}+\sigma_{\mathrm{Metric}}^{\mu\nu;\rho}, \tag*{(129)} σ μν ; ρ = σ NLDrude μν ; ρ + σ Dipole μν ; ρ + σ Metric μν ; ρ , (129)
where
σ M e t r i c μ ν ; ρ ∝ τ 0 , σ D i p o l e μ ν ; ρ ∝ τ , σ N L D r u d e μ ν ; ρ ∝ τ 2 . (130) \sigma_{\mathrm{Metric}}^{\mu\nu;\rho}\propto\tau^0,\qquad\sigma_{\mathrm{Dipole}}^{\mu\nu;\rho}\propto\tau,\qquad\sigma_{\mathrm{NLDrude}}^{\mu\nu;\rho}\propto\tau^2. \tag*{(130)} σ Metric μν ; ρ ∝ τ 0 , σ Dipole μν ; ρ ∝ τ , σ NLDrude μν ; ρ ∝ τ 2 . (130)
First, σ N L D r u d e μ ν ; ρ \sigma_{\mathrm{NLDrude}}^{\mu\nu;\rho} σ NLDrude μν ; ρ is the nonlinear Drude conductivity[102 , 103 ] ,
σ N L D r u d e μ ν ; ρ = − e 3 τ 2 ℏ 3 ∑ n ∫ d k f n ∂ 3 ε n ∂ k μ ∂ k ν ∂ k ρ . (131) \sigma_{\mathrm{NLDrude}}^{\mu\nu;\rho}=-\frac{e^3\tau^2}{\hbar^3}\sum_n\int dk f_n\frac{\partial^3\varepsilon_n}{\partial k_\mu\partial k_\nu\partial k_\rho}. \tag*{(131)} σ NLDrude μν ; ρ = − ℏ 3 e 3 τ 2 n ∑ ∫ d k f n ∂ k μ ∂ k ν ∂ k ρ ∂ 3 ε n . (131)
It is also an extrinsic nonlinear conductivity.
Second, σ D i p o l e μ ν ; ρ \sigma_{\mathrm{Dipole}}^{\mu\nu;\rho} σ Dipole μν ; ρ is the nonlinear transverse conductivity induced by the Berry curvature dipole (BCD)[20 ] ,
σ D i p o l e μ ν ; ρ = − e 3 τ ℏ 2 ∑ n ∫ d k f n ( ∂ Ω n ν ρ ∂ k μ + ∂ Ω n μ ρ ∂ k ν ) . (132) \sigma_{\mathrm{Dipole}}^{\mu\nu;\rho}=-\frac{e^{3}\tau}{\hbar^{2}}\sum_{n}\int d\mathbf{k}f_{n}\left(\frac{\partial\Omega_{n}^{\nu\rho}}{\partial k_{\mu}}+\frac{\partial\Omega_{n}^{\mu\rho}}{\partial k_{\nu}}\right). \tag*{(132)} σ Dipole μν ; ρ = − ℏ 2 e 3 τ n ∑ ∫ d k f n ( ∂ k μ ∂ Ω n ν ρ + ∂ k ν ∂ Ω n μ ρ ) . (132)
It is an extrinsic nonlinear conductivity, since it vanishes as τ → 0 \tau\to0 τ → 0 .
Third, only the term σ M e t r i c μ ν ; ρ \sigma_{\mathrm{Metric}}^{\mu\nu;\rho} σ Metric μν ; ρ survives in the dirty limit τ → 0 \tau\to0 τ → 0 , which is the intrinsic nonlinear conductivity. It is the quantum-metric induced nonlinear conductivity. There are still debates[104 , 105 ] on the coefficients of the quantum-metric induced nonlinear conductivity.
(i) The Luttinger-Kohn approach[27 ] gives
σ M e t r i c μ ν ; ρ = − e 3 ℏ ∑ n ∫ d k f n ( 2 ∂ G n μ ν ∂ k ρ − 1 2 ( ∂ G n ν ρ ∂ k μ + ∂ G n μ ρ ∂ k ν ) ) , (133) \sigma_{\mathrm{Metric}}^{\mu\nu;\rho}=-\frac{e^{3}}{\hbar}\sum_{n}\int d\mathbf{k}f_{n}\left(2\frac{\partial G_{n}^{\mu\nu}}{\partial k_{\rho}}-\frac{1}{2}\left(\frac{\partial G_{n}^{\nu\rho}}{\partial k_{\mu}}+\frac{\partial G_{n}^{\mu\rho}}{\partial k_{\nu}}\right)\right), \tag*{(133)} σ Metric μν ; ρ = − ℏ e 3 n ∑ ∫ d k f n ( 2 ∂ k ρ ∂ G n μν − 2 1 ( ∂ k μ ∂ G n ν ρ + ∂ k ν ∂ G n μ ρ ) ) , (133)
where G n a b G_{n}^{ab} G n ab is the band–energy normalized quantum metric or the Berry connection polarizability. It is given as[22 , 24 , 27 , 106 , 107 , 108 ]
G n μ ν = 2 Re ∑ m ≠ n a n m μ ( k ) a m n ν ( k ) ε n ( k ) − ε m ( k ) . (134) G_{n}^{\mu\nu}=2\operatorname{Re}\sum_{m\ne n}\frac{a_{nm}^{\mu}(\mathbf{k})a_{mn}^{\nu}(\mathbf{k})}{\varepsilon_{n}(\mathbf{k})-\varepsilon_{m}(\mathbf{k})}. \tag*{(134)} G n μν = 2 Re m = n ∑ ε n ( k ) − ε m ( k ) a nm μ ( k ) a mn ν ( k ) . (134)
(ii) The wave packet dynamics approach[107 ] gives
σ M e t r i c μ ν ; ρ = − e 3 ℏ ∑ n ∫ d k f n ( ∂ G n μ ν ∂ k ρ − 1 2 ( ∂ G n ν ρ ∂ k μ + ∂ G n μ ρ ∂ k ν ) ) . (135) \sigma_{\mathrm{Metric}}^{\mu\nu;\rho}=-\frac{e^{3}}{\hbar}\sum_{n}\int d\mathbf{k}f_{n}\left(\frac{\partial G_{n}^{\mu\nu}}{\partial k_{\rho}}-\frac{1}{2}\left(\frac{\partial G_{n}^{\nu\rho}}{\partial k_{\mu}}+\frac{\partial G_{n}^{\mu\rho}}{\partial k_{\nu}}\right)\right). \tag*{(135)} σ Metric μν ; ρ = − ℏ e 3 n ∑ ∫ d k f n ( ∂ k ρ ∂ G n μν − 2 1 ( ∂ k μ ∂ G n ν ρ + ∂ k ν ∂ G n μ ρ ) ) . (135)
(iii) The quantum kinetics approach[24 ] gives
σ M e t r i c μ ν ; ρ = − e 3 ℏ ∑ n ∫ d k f n ( 1 2 ∂ G n μ ν ∂ k ρ − ( ∂ G n ν ρ ∂ k μ + ∂ G n μ ρ ∂ k ν ) ) . (136) \sigma_{\mathrm{Metric}}^{\mu\nu;\rho}=-\frac{e^{3}}{\hbar}\sum_{n}\int d\mathbf{k}f_{n}\left(\frac{1}{2}\frac{\partial G_{n}^{\mu\nu}}{\partial k_{\rho}}-\left(\frac{\partial G_{n}^{\nu\rho}}{\partial k_{\mu}}+\frac{\partial G_{n}^{\mu\rho}}{\partial k_{\nu}}\right)\right). \tag*{(136)} σ Metric μν ; ρ = − ℏ e 3 n ∑ ∫ d k f n ( 2 1 ∂ k ρ ∂ G n μν − ( ∂ k μ ∂ G n ν ρ + ∂ k ν ∂ G n μ ρ ) ) . (136)
(iv) The intrinsic Ohmic conductivity[109 ] gives
σ M e t r i c μ ν ; ρ = e 3 ℏ ∑ n ∫ d k f n ( ∂ G n μ ν ∂ k ρ + ∂ G n ν ρ ∂ k μ + ∂ G n μ ρ ∂ k ν ) . (137) \sigma_{\mathrm{Metric}}^{\mu\nu;\rho}=\frac{e^{3}}{\hbar}\sum_{n}\int d\mathbf{k}f_{n}\left(\frac{\partial G_{n}^{\mu\nu}}{\partial k_{\rho}}+\frac{\partial G_{n}^{\nu\rho}}{\partial k_{\mu}}+\frac{\partial G_{n}^{\mu\rho}}{\partial k_{\nu}}\right). \tag*{(137)} σ Metric μν ; ρ = ℏ e 3 n ∑ ∫ d k f n ( ∂ k ρ ∂ G n μν + ∂ k μ ∂ G n ν ρ + ∂ k ν ∂ G n μ ρ ) . (137)
Zeeman Quantum geometry for momentum and spin
So far, we have considered quantum geometry for momentum translation. However, there is also a spin degree of freedom for electrons. It is important to study the effect of spin rotation in the context of spintronics. This is achieved by generalizing quantum geometry to Zeeman quantum geometry[110 , 111 , 112 , 113 ] , where local spin rotation between two adjacent wave functions is also taken into account.
The Zeeman quantum geometric tensor g n m μ ν g_{nm}^{\mu\nu} g nm μν is defined by the quantum distance d S H S d_{\mathrm{SHS}} d SHS for the infinitesimal translation d k d\mathbf{k} d k of the momentum and infinitesimal spin rotation d θ d\boldsymbol{\theta} d θ as[110 , 111 , 112 , 113 ]
d S H S ( k ) ≡ 1 − ∣ ⟨ ψ n ( k ) ∣ U d θ U d k ∣ ψ n ( k ) ⟩ ∣ 2 , (138) d_{\mathrm{SHS}}(\mathbf{k})\equiv\sqrt{1-\left|\langle\psi_{n}(\mathbf{k})|U_{d\theta}U_{d\mathbf{k}}|\psi_{n}(\mathbf{k})\rangle\right|^{2}}, \tag*{(138)} d SHS ( k ) ≡ 1 − ∣ ⟨ ψ n ( k ) ∣ U d θ U d k ∣ ψ n ( k )⟩ ∣ 2 , (138)
where
U d θ ≡ e − i 2 d θ ⋅ σ , U d k ≡ e − i d k ⋅ r (139) U_{d\theta}\equiv e^{-\frac{i}{2}d\boldsymbol{\theta}\cdot\boldsymbol{\sigma}},\qquad U_{d\mathbf{k}}\equiv e^{-i d\mathbf{k}\cdot\mathbf{r}} \tag*{(139)} U d θ ≡ e − 2 i d θ ⋅ σ , U d k ≡ e − i d k ⋅ r (139)
are the generators of the spin angular momentum d θ d\boldsymbol{\theta} d θ and the momentum d k d\mathbf{k} d k . The Zeeman quantum geometric tensor z n m μ ν z_{nm}^{\mu\nu} z nm μν and the spin quantum geometric tensor s n m μ ν s_{nm}^{\mu\nu} s nm μν are given by
( d S H S ) 2 = ∑ μ ν ∑ n ≠ m ( F n m μ ν d k μ d k ν + s n m μ ν 4 d θ μ d θ ν + z n m μ ν + z m n μ ν 2 d k μ d θ ν ) . (140) (d_{\mathrm{SHS}})^{2}=\sum_{\mu\nu}\sum_{n\ne m}\left(F_{nm}^{\mu\nu}dk_{\mu}dk_{\nu}+\frac{s_{nm}^{\mu\nu}}{4}d\theta_{\mu}d\theta_{\nu}+\frac{z_{nm}^{\mu\nu}+z_{mn}^{\mu\nu}}{2}dk_{\mu}d\theta_{\nu}\right). \tag*{(140)} ( d SHS ) 2 = μν ∑ n = m ∑ ( F nm μν d k μ d k ν + 4 s nm μν d θ μ d θ ν + 2 z nm μν + z mn μν d k μ d θ ν ) . (140)
where
F n m μ ν ≡ a n m μ a m n ν (141) F_{nm}^{\mu\nu}\equiv a_{nm}^{\mu}a_{mn}^{\nu} \tag*{(141)} F nm μν ≡ a nm μ a mn ν (141)
is the quantum geometric tensor,
s n m μ ν ≡ s m n μ s m n ν (142) s_{nm}^{\mu\nu}\equiv s_{mn}^{\mu}s_{mn}^{\nu} \tag*{(142)} s nm μν ≡ s mn μ s mn ν (142)
is the spin quantum geometric tensor, and
z n m μ ν ≡ a n m μ s m n ν (143) z_{nm}^{\mu\nu}\equiv a_{nm}^{\mu}s_{mn}^{\nu} \tag*{(143)} z nm μν ≡ a nm μ s mn ν (143)
is the Zeeman quantum geometric tensor, where we have defined the expectation value of the spin operator,
s n m μ ≡ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k ) ⟩ . (144) s_{nm}^{\mu}\equiv\langle\psi_{n}(\mathbf{k})|\sigma_{\mu}|\psi_{n}(\mathbf{k})\rangle. \tag*{(144)} s nm μ ≡ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k )⟩ . (144)
The derivation from Eq. (140) ~Eq. (143) is shown in Appendix E.
As in the case of the quantum metric and the Berry curvature, the Zeeman quantum metric Q n m μ ν Q_{nm}^{\mu\nu} Q nm μν is defined by the real part of the Zeeman quantum geometric tensor,
Q n m μ ν ≡ Re z n m μ ν = ( a n m μ s m n ν + a m n μ s n m ν ) / 2 , (145) Q_{nm}^{\mu\nu}\equiv\operatorname{Re}z_{nm}^{\mu\nu}=\left(a_{nm}^{\mu}s_{mn}^{\nu}+a_{mn}^{\mu}s_{nm}^{\nu}\right)/2, \tag*{(145)} Q nm μν ≡ Re z nm μν = ( a nm μ s mn ν + a mn μ s nm ν ) /2 , (145)
and the Zeeman Berry curvature Z n m μ ν Z_{nm}^{\mu\nu} Z nm μν by the imaginary part of the Zeeman quantum geometric tensor,
Z n m μ ν ≡ 2 Im z n m μ ν = i ( a n m μ s m n ν − a m n μ s n m ν ) . (146) Z_{nm}^{\mu\nu}\equiv2\operatorname{Im}z_{nm}^{\mu\nu}=i\left(a_{nm}^{\mu}s_{mn}^{\nu}-a_{mn}^{\mu}s_{nm}^{\nu}\right). \tag*{(146)} Z nm μν ≡ 2 Im z nm μν = i ( a nm μ s mn ν − a mn μ s nm ν ) . (146)
We also define the spin quantum metric S n m μ ν S_{nm}^{\mu\nu} S nm μν by the real part of the spin quantum geometric tensor,
S n m μ ν ≡ Re s n m μ ν = ( s n m μ s m n ν + s m n μ s n m ν ) / 2 , (147) S_{nm}^{\mu\nu}\equiv\operatorname{Re}s_{nm}^{\mu\nu}=\left(s_{nm}^{\mu}s_{mn}^{\nu}+s_{mn}^{\mu}s_{nm}^{\nu}\right)/2, \tag*{(147)} S nm μν ≡ Re s nm μν = ( s nm μ s mn ν + s mn μ s nm ν ) /2 , (147)
and the spin Berry curvature A n m μ ν A_{nm}^{\mu\nu} A nm μν by
A n m μ ν ≡ 2 Im s n m μ ν = i ( s n m μ s m n ν − s m n μ s n m ν ) . (148) A_{nm}^{\mu\nu}\equiv2\operatorname{Im}s_{nm}^{\mu\nu}=i\left(s_{nm}^{\mu}s_{mn}^{\nu}-s_{mn}^{\mu}s_{nm}^{\nu}\right). \tag*{(148)} A nm μν ≡ 2 Im s nm μν = i ( s nm μ s mn ν − s mn μ s nm ν ) . (148)
Responses originated from the Zeeman geometry
In the linear response theory, we obtain the response of the current J μ ; ν J^{\mu;\nu} J μ ; ν or the spin polarization S μ ; ν S^{\mu;\nu} S μ ; ν by applying electric field E ν E^\nu E ν or magnetic field B ν B^\nu B ν .
We apply alternating electric field E ν ( t ) E^\nu(t) E ν ( t ) and alternating magnetic field B ν ( t ) B^\nu(t) B ν ( t ) ,
E ν ( t ) = 1 2 E ν ∑ ω 1 = ± ω e − i ω 1 t , (149) E^\nu(t)=\frac{1}{2}E^\nu\sum_{\omega_1=\pm\omega}e^{-i\omega_1t}, \tag*{(149)} E ν ( t ) = 2 1 E ν ω 1 = ± ω ∑ e − i ω 1 t , (149)
B ν ( t ) = 1 2 B ν ∑ ω 1 = ± ω e − i ω 1 t . (150) B^\nu(t)=\frac{1}{2}B^\nu\sum_{\omega_1=\pm\omega}e^{-i\omega_1t}. \tag*{(150)} B ν ( t ) = 2 1 B ν ω 1 = ± ω ∑ e − i ω 1 t . (150)
The Hamiltonian for the external fields is given by
H 1 = E ν ( t ) a ν − g μ B B ν ( t ) σ ν . (151) H_1=E^\nu(t)a^\nu-g\mu_{\mathrm{B}}B^\nu(t)\sigma^\nu. \tag*{(151)} H 1 = E ν ( t ) a ν − g μ B B ν ( t ) σ ν . (151)
We solve the quantum Liouville equation
i ∂ ρ ∂ t = [ H ′ , ρ ] (152) i\frac{\partial\rho}{\partial t}=[H^\prime,\rho] \tag*{(152)} i ∂ t ∂ ρ = [ H ′ , ρ ] (152)
with H ′ = H + H 1 H^\prime=H+H_1 H ′ = H + H 1 , where H H H is the non-perturbated Hamiltonian. The first-order solution of the density matrix is given by[111 ]
ρ m n μ ; ν = 1 2 ∑ ω 1 = ± ω [ i δ m n ∂ k μ f m + f n m a m n μ ] E ν e − i ω 1 t ℏ ω 1 − ε m n + i η − g μ B 2 ∑ ω 1 = ± ω f n m s m n μ B ν e − i ω 1 t ℏ ω 1 − ε m n + i η , (153) \rho_{mn}^{\mu;\nu}=\frac{1}{2}\sum_{\omega_1=\pm\omega}\left[i\delta_{mn}\partial_{k_\mu}f_m+f_{nm}a_{mn}^{\mu}\right]\frac{E^\nu e^{-i\omega_1t}}{\hbar\omega_1-\varepsilon_{mn}+i\eta}
-\frac{g\mu_{\mathrm{B}}}{2}\sum_{\omega_1=\pm\omega}\frac{f_{nm}s_{mn}^{\mu}B^\nu e^{-i\omega_1t}}{\hbar\omega_1-\varepsilon_{mn}+i\eta}, \tag*{(153)} ρ mn μ ; ν = 2 1 ω 1 = ± ω ∑ [ i δ mn ∂ k μ f m + f nm a mn μ ] ℏ ω 1 − ε mn + i η E ν e − i ω 1 t − 2 g μ B ω 1 = ± ω ∑ ℏ ω 1 − ε mn + i η f nm s mn μ B ν e − i ω 1 t , (153)
where f n m ≡ f n − f m f_{nm}\equiv f_n-f_m f nm ≡ f n − f m and ε m n ≡ ε m − ε n \varepsilon_{mn}\equiv\varepsilon_m-\varepsilon_n ε mn ≡ ε m − ε n .
In the linear response theory, the current J μ ; ν J^{\mu;\nu} J μ ; ν is given by
J μ ; ν = ∑ n m ∫ d k v n m μ ρ m n μ ; ν , (154) J^{\mu;\nu}=\sum_{nm}\int dk\,v_{nm}^{\mu}\rho_{mn}^{\mu;\nu}, \tag*{(154)} J μ ; ν = nm ∑ ∫ d k v nm μ ρ mn μ ; ν , (154)
where v n m μ v_{nm}^{\mu} v nm μ is the velocity operator. By using the Hellmann–Feynman theorem, it is rewritten as
v n m μ = i ε n m a n m μ . (155) v_{nm}^{\mu}=i\varepsilon_{nm}a_{nm}^{\mu}. \tag*{(155)} v nm μ = i ε nm a nm μ . (155)
On the other hand, the spin S μ ; ν S^{\mu;\nu} S μ ; ν polarization is given by
S μ ; ν = ∑ n m ∫ d k s n m μ ρ m n μ ; ν . (156) S^{\mu;\nu}=\sum_{nm}\int dk\,s_{nm}^{\mu}\rho_{mn}^{\mu;\nu}. \tag*{(156)} S μ ; ν = nm ∑ ∫ d k s nm μ ρ mn μ ; ν . (156)
There are four types of responses, J μ ; ν / E ν J^{\mu;\nu}/E^\nu J μ ; ν / E ν , J μ ; ν / B ν J^{\mu;\nu}/B^\nu J μ ; ν / B ν , S μ ; ν / E ν S^{\mu;\nu}/E^\nu S μ ; ν / E ν and S μ ; ν / B ν S^{\mu;\nu}/B^\nu S μ ; ν / B ν , where J μ ; ν / E ν J^{\mu;\nu}/E^\nu J μ ; ν / E ν and S μ ; ν / B ν S^{\mu;\nu}/B^\nu S μ ; ν / B ν are direct responses, while J μ ; ν / B ν J^{\mu;\nu}/B^\nu J μ ; ν / B ν and S μ ; ν / E ν S^{\mu;\nu}/E^\nu S μ ; ν / E ν are cross responses.
(i) The current J μ ; ν J^{\mu;\nu} J μ ; ν is induced by electric field E ν E^\nu E ν as[110 ]
J μ ; ν E ν = ∫ d k ∑ n > m f n m [ Ω n m μ ν cos ω t + 2 g n m μ ν ω sin ω t ε m n ] . (157) \frac{J^{\mu;\nu}}{E^\nu}=\int dk\sum_{n>m}f_{nm}\left[\Omega_{nm}^{\mu\nu}\cos\omega t+2g_{nm}^{\mu\nu}\frac{\omega\sin\omega t}{\varepsilon_{mn}}\right]. \tag*{(157)} E ν J μ ; ν = ∫ d k n > m ∑ f nm [ Ω nm μν cos ω t + 2 g nm μν ε mn ω sin ω t ] . (157)
The first term represents the Hall current generated by the Berry curvature Ω n m μ ν \Omega_{nm}^{\mu\nu} Ω nm μν . The quantum metric g n m μ ν g_{nm}^{\mu\nu} g nm μν in the second term appears as an oscillating response[114 ] .
(ii) The current J μ ; ν J^{\mu;\nu} J μ ; ν is induced by magnetic field B ν B^\nu B ν as[111 ]
J μ ; ν B ν = g μ B ∫ d k ∑ n > m f n m [ Z n m μ ν cos ω t + 2 Q n m μ ν ω sin ω t ε m n ] . (158) \frac{J^{\mu;\nu}}{B^\nu}=g\mu_{\mathrm{B}}\int dk\sum_{n>m}f_{nm}\left[Z_{nm}^{\mu\nu}\cos\omega t+2Q_{nm}^{\mu\nu}\frac{\omega\sin\omega t}{\varepsilon_{mn}}\right]. \tag*{(158)} B ν J μ ; ν = g μ B ∫ d k n > m ∑ f nm [ Z nm μν cos ω t + 2 Q nm μν ε mn ω sin ω t ] . (158)
The Zeeman quantum geometric tensor Z n m μ ν Z_{nm}^{\mu\nu} Z nm μν contributes to the cross response between the magnetic field B ν B^\nu B ν and the current J μ ; ν J^{\mu;\nu} J μ ; ν , where J μ ; ν J^{\mu;\nu} J μ ; ν is called the intrinsic gyrotropic magnetic current J μ ; ν J^{\mu;\nu} J μ ; ν . The Zeeman quantum metric Q n m μ ν Q_{nm}^{\mu\nu} Q nm μν in the second term appears as an oscillating response.
(iii) Spin polarization S μ ; ν S^{\mu;\nu} S μ ; ν is induced by electric field E ν E^\nu E ν as[110 ]
S μ ; ν E ν = − ∫ d k ∑ n > m f n m [ 2 Q n m μ ν ε m n cos ω t + Z n m μ ν ω sin ω t ε m n 2 ] . (159) \frac{S^{\mu;\nu}}{E^\nu}=-\int dk\sum_{n>m}f_{nm}\left[\frac{2Q_{nm}^{\mu\nu}}{\varepsilon_{mn}}\cos\omega t+Z_{nm}^{\mu\nu}\frac{\omega\sin\omega t}{\varepsilon_{mn}^{2}}\right]. \tag*{(159)} E ν S μ ; ν = − ∫ d k n > m ∑ f nm [ ε mn 2 Q nm μν cos ω t + Z nm μν ε mn 2 ω sin ω t ] . (159)
The Zeeman quantum metric Q n m μ ν Q_{nm}^{\mu\nu} Q nm μν contributes to the cross response between the electric field E ν E^\nu E ν and the spin polarization S μ ; ν S^{\mu;\nu} S μ ; ν . The Zeeman quantum geometric tensor Z n m μ ν Z_{nm}^{\mu\nu} Z nm μν in the second term appears as an oscillating response.
(iv) Spin polarization S μ ; ν S^{\mu;\nu} S μ ; ν is also induced by magnetic field B ν B^\nu B ν as
S μ ; ν B ν = g μ B ∫ d k ∑ n > m f n m [ S n m μ ν ε m n cos ω t + 2 A n m μ ν ω sin ω t ε m n 2 ] . (160) \frac{S^{\mu;\nu}}{B^\nu}=g\mu_{\mathrm{B}}\int dk\sum_{n>m}f_{nm}\left[\frac{S_{nm}^{\mu\nu}}{\varepsilon_{mn}}\cos\omega t+2A_{nm}^{\mu\nu}\frac{\omega\sin\omega t}{\varepsilon_{mn}^{2}}\right]. \tag*{(160)} B ν S μ ; ν = g μ B ∫ d k n > m ∑ f nm [ ε mn S nm μν cos ω t + 2 A nm μν ε mn 2 ω sin ω t ] . (160)
It is interesting that there may be off-diagonal response if S n m μ ν S_{nm}^{\mu\nu} S nm μν or A n m μ ν A_{nm}^{\mu\nu} A nm μν has off-diagonal components. Detailed derivations of Eqs. (157) , (158) , (159) and (160) are given in Appendix F.
We may call the first terms in Eqs. (157) , (158) , (159) and (160) the static terms and the second terms the dynamic terms, because the second terms vanish for ω = 0 \omega=0 ω = 0 .
Zeeman Quantum Geometry for two-band systems
We consider the two-band system (40) with the index n = ± n=\pm n = ± . The simple relations for the Zeeman quantum metric are derived as
Z + − μ ν = − Z − + μ ν = ∂ n ν ∂ k μ . (161) Z_{+-}^{\mu\nu}=-Z_{-+}^{\mu\nu}=\frac{\partial n_\nu}{\partial k_\mu}. \tag*{(161)} Z +− μν = − Z −+ μν = ∂ k μ ∂ n ν . (161)
Two or three components of n \mathbf{n} n should be nonzero for nonzero Z + − μ ν Z^{\mu\nu}_{+-} Z +− μν . In general, we have Z + − μ ν ≠ Z + − ν μ Z^{\mu\nu}_{+-} \ne Z^{\nu\mu}_{+-} Z +− μν = Z +− ν μ .
The Zeeman Berry curvature is derived as
Q + − μ ν = 1 2 ∑ ρ σ ε ν ρ σ n ρ ∂ n σ ∂ k μ , (162) Q^{\mu\nu}_{+-}=\frac{1}{2}\sum_{\rho\sigma}\varepsilon_{\nu\rho\sigma}n_{\rho}\frac{\partial n_{\sigma}}{\partial k_{\mu}}, \tag*{(162)} Q +− μν = 2 1 ρ σ ∑ ε ν ρ σ n ρ ∂ k μ ∂ n σ , (162)
Two or three components of n \mathbf{n} n should be nonzero for nonzero Q + − μ ν Q^{\mu\nu}_{+-} Q +− μν .
The spin quantum metric is
S + − μ ν = δ μ ν − n μ n ν , (163) S^{\mu\nu}_{+-}=\delta_{\mu\nu}-n_{\mu}n_{\nu}, \tag*{(163)} S +− μν = δ μν − n μ n ν , (163)
while the spin Berry curvature is
A + − μ ν = − 2 ∑ ρ ε μ ν ρ n ρ . (164) \mathcal{A}^{\mu\nu}_{+-}=-2\sum_{\rho}\varepsilon_{\mu\nu\rho}n_{\rho}. \tag*{(164)} A +− μν = − 2 ρ ∑ ε μν ρ n ρ . (164)
They are defined even when only a single component of n \mathbf{n} n is nonzero.
Rashba system
As an example of the two-band system, we consider the Rashba system described by
H ( k ) = ℏ 2 k 2 2 m + λ ( k × σ ) z + B σ z , (165) H(\mathbf{k})=\frac{\hbar^{2}k^{2}}{2m}+\lambda(\mathbf{k}\times\boldsymbol{\sigma})_{z}+B\sigma_{z}, \tag*{(165)} H ( k ) = 2 m ℏ 2 k 2 + λ ( k × σ ) z + B σ z , (165)
where λ \lambda λ is the strength of the Rashba interaction and B B B is static magnetic field applied perpendicularly to the plane.
The energy is given by
E χ = ℏ 2 k 2 2 m + χ λ 2 k 2 + B 2 (166) E_{\chi}=\frac{\hbar^{2}k^{2}}{2m}+\chi\sqrt{\lambda^{2}k^{2}+B^{2}} \tag*{(166)} E χ = 2 m ℏ 2 k 2 + χ λ 2 k 2 + B 2 (166)
for the lower band with χ = − 1 \chi=-1 χ = − 1 and for the upper band with χ = 1 \chi=1 χ = 1 . The Fermi surface is given by
k η χ = m μ + m λ 2 + η m 2 λ 4 + 2 μ m λ 2 + B 2 , (167) k_{\eta}^{\chi}=\sqrt{m}\sqrt{\mu+m\lambda^{2}+\eta\sqrt{m^{2}\lambda^{4}+2\mu m\lambda^{2}+B^{2}}}, \tag*{(167)} k η χ = m μ + m λ 2 + η m 2 λ 4 + 2 μ m λ 2 + B 2 , (167)
where η = − 1 \eta=-1 η = − 1 describes the inner Fermi surface and η = 1 \eta=1 η = 1 describes the outer Fermi surface.
The Berry curvature is given by
Ω x y = − B λ 2 2 ( λ 2 k 2 + B 2 ) 3 / 2 , (168) \Omega^{xy}=-\frac{B\lambda^{2}}{2(\lambda^{2}k^{2}+B^{2})^{3/2}}, \tag*{(168)} Ω x y = − 2 ( λ 2 k 2 + B 2 ) 3/2 B λ 2 , (168)
while the quantum metrices are given by
g + − x x = λ 2 ( λ 2 k y 2 + B 2 ) 2 ( λ 2 k 2 + B 2 ) 2 , (169) g^{xx}_{+-}=\frac{\lambda^{2}(\lambda^{2}k_{y}^{2}+B^{2})}{2(\lambda^{2}k^{2}+B^{2})^{2}}, \tag*{(169)} g +− xx = 2 ( λ 2 k 2 + B 2 ) 2 λ 2 ( λ 2 k y 2 + B 2 ) , (169)
g + − x y = g + − y x = − λ 4 k x k y 2 ( λ 2 k 2 + B 2 ) 2 , (170) g^{xy}_{+-}=g^{yx}_{+-}=-\frac{\lambda^{4}k_{x}k_{y}}{2(\lambda^{2}k^{2}+B^{2})^{2}}, \tag*{(170)} g +− x y = g +− y x = − 2 ( λ 2 k 2 + B 2 ) 2 λ 4 k x k y , (170)
g + − y y = λ 2 ( λ 2 k x 2 + B 2 ) 2 ( λ 2 k 2 + B 2 ) 2 . (171) g^{yy}_{+-}=\frac{\lambda^{2}(\lambda^{2}k_{x}^{2}+B^{2})}{2(\lambda^{2}k^{2}+B^{2})^{2}}. \tag*{(171)} g +− y y = 2 ( λ 2 k 2 + B 2 ) 2 λ 2 ( λ 2 k x 2 + B 2 ) . (171)
There is no contribution from the quantum metric g + − x y g^{xy}_{+-} g +− x y in the Hall current (157) because g + − x y ∝ sin 2 ϕ g^{xy}_{+-}\propto\sin2\phi g +− x y ∝ sin 2 ϕ leads to ∫ g + − x y d ϕ = 0 \int g^{xy}_{+-}d\phi=0 ∫ g +− x y d ϕ = 0 .
The Zeeman Berry curvatures are calculated as
Z + − x x = ∂ n x ∂ k x = λ 3 k x k y ( λ 2 k 2 + B 2 ) 3 / 2 , (172) Z^{xx}_{+-}=\frac{\partial n_{x}}{\partial k_{x}}=\frac{\lambda^{3}k_{x}k_{y}}{(\lambda^{2}k^{2}+B^{2})^{3/2}}, \tag*{(172)} Z +− xx = ∂ k x ∂ n x = ( λ 2 k 2 + B 2 ) 3/2 λ 3 k x k y , (172)
Z + − x y = ∂ n y ∂ k x = λ ( λ 2 k y 2 + B 2 ) ( λ 2 k 2 + B 2 ) 3 / 2 , (173) Z^{xy}_{+-}=\frac{\partial n_{y}}{\partial k_{x}}=\frac{\lambda(\lambda^{2}k_{y}^{2}+B^{2})}{(\lambda^{2}k^{2}+B^{2})^{3/2}}, \tag*{(173)} Z +− x y = ∂ k x ∂ n y = ( λ 2 k 2 + B 2 ) 3/2 λ ( λ 2 k y 2 + B 2 ) , (173)
Z + − y x = ∂ n x ∂ k y = − λ ( λ 2 k x 2 + B 2 ) ( λ 2 k 2 + B 2 ) 3 / 2 , (174) Z^{yx}_{+-}=\frac{\partial n_{x}}{\partial k_{y}}=-\frac{\lambda(\lambda^{2}k_{x}^{2}+B^{2})}{(\lambda^{2}k^{2}+B^{2})^{3/2}}, \tag*{(174)} Z +− y x = ∂ k y ∂ n x = − ( λ 2 k 2 + B 2 ) 3/2 λ ( λ 2 k x 2 + B 2 ) , (174)
Z + − x z = ∂ n z ∂ k x = − λ 2 B k x ( λ 2 k 2 + B 2 ) 3 / 2 , (175) Z^{xz}_{+-}=\frac{\partial n_{z}}{\partial k_{x}}=-\frac{\lambda^{2}Bk_{x}}{(\lambda^{2}k^{2}+B^{2})^{3/2}}, \tag*{(175)} Z +− x z = ∂ k x ∂ n z = − ( λ 2 k 2 + B 2 ) 3/2 λ 2 B k x , (175)
Z + − y z = ∂ n z ∂ k y = − λ 2 B k y ( λ 2 k 2 + B 2 ) 3 / 2 . (176) Z^{yz}_{+-}=\frac{\partial n_{z}}{\partial k_{y}}=-\frac{\lambda^{2}Bk_{y}}{(\lambda^{2}k^{2}+B^{2})^{3/2}}. \tag*{(176)} Z +− y z = ∂ k y ∂ n z = − ( λ 2 k 2 + B 2 ) 3/2 λ 2 B k y . (176)
Only the Zeeman Berry curvature Z + − x y Z^{xy}_{+-} Z +− x y contributes to a nonzero response because we have
∫ Z + − x x d ϕ = ∫ Z + − x z d ϕ = ∫ Z + − y z d ϕ = 0. (177) \int Z^{xx}_{+-}d\phi=\int Z^{xz}_{+-}d\phi=\int Z^{yz}_{+-}d\phi=0. \tag*{(177)} ∫ Z +− xx d ϕ = ∫ Z +− x z d ϕ = ∫ Z +− y z d ϕ = 0. (177)
We note that Z + − x x Z^{xx}_{+-} Z +− xx and Z + − x y Z^{xy}_{+-} Z +− x y are singular at k = 0 k=0 k = 0 , while Z + − x z Z^{xz}_{+-} Z +− x z and Z + − y z Z^{yz}_{+-} Z +− y z are zero for B = 0 B=0 B = 0 .
The static intrinsic gyrotropic magnetic current at zero temperature is calculated as
J x ; y B y = g μ B ∫ d k ∑ n > m f n m Z n m x y = g μ B ∫ k η x k η x k d k d ϕ λ ( λ 2 k y 2 + B 2 ) ( λ 2 k 2 + B 2 ) 3 / 2 = g μ B ∑ η = ± 1 η π λ ( k η x ) 2 λ 2 ( k η x ) 2 + B 2 . (178) \begin{aligned}
\frac{J^{x;y}}{B^{y}}=g\mu_{\mathrm{B}}\int dk\sum_{n>m}f_{nm}Z^{xy}_{nm} \\
&=g\mu_{\mathrm{B}}\int_{k_{\eta}^{x}}^{k_{\eta}^{x}}kdkd\phi\frac{\lambda(\lambda^{2}k_{y}^{2}+B^{2})}{(\lambda^{2}k^{2}+B^{2})^{3/2}} \\
&=g\mu_{\mathrm{B}}\sum_{\eta=\pm1}\frac{\eta\pi\lambda(k_{\eta}^{x})^{2}}{\sqrt{\lambda^{2}(k_{\eta}^{x})^{2}+B^{2}}}. \tag*{(178)}
\end{aligned} B y J x ; y = g μ B ∫ d k n > m ∑ f nm Z nm x y = g μ B ∫ k η x k η x k d k d ϕ ( λ 2 k 2 + B 2 ) 3/2 λ ( λ 2 k y 2 + B 2 ) = g μ B η = ± 1 ∑ λ 2 ( k η x ) 2 + B 2 η π λ ( k η x ) 2 . (178)
The electric-filed induced spin polarization at zero temperature is calculated as
S μ ; ν E ν = − ∫ d k f + − Z + − ν μ ω ϵ 2 2 sin ( ω t ) = ∑ η = ± 1 η π λ ( k η x ) 2 + 4 B 2 ( λ 2 ( k η x ) 2 + B 2 ) 3 / 2 ω sin ( ω t ) . (179) \begin{aligned}
\frac{S^{\mu;\nu}}{E^{\nu}}=-\int dkf_{+-}Z^{\nu\mu}_{+-}\frac{\omega}{\epsilon_{2}^{2}}\sin(\omega t) \\
&=\sum_{\eta=\pm1}\eta\pi\frac{\lambda(k_{\eta}^{x})^{2}+4B^{2}}{(\lambda^{2}(k_{\eta}^{x})^{2}+B^{2})^{3/2}}\omega\sin(\omega t). \tag*{(179)}
\end{aligned} E ν S μ ; ν = − ∫ d k f +− Z +− ν μ ϵ 2 2 ω sin ( ω t ) = η = ± 1 ∑ η π ( λ 2 ( k η x ) 2 + B 2 ) 3/2 λ ( k η x ) 2 + 4 B 2 ω sin ( ω t ) . (179)
The Zeeman quantum metrices are calculated as
Q + − x x = Q + − y y = − B λ 2 ( λ 2 k 2 + B 2 ) , (180) Q^{xx}_{+-}=Q^{yy}_{+-}=-\frac{B\lambda}{2(\lambda^{2}k^{2}+B^{2})}, \tag*{(180)} Q +− xx = Q +− y y = − 2 ( λ 2 k 2 + B 2 ) B λ , (180)
Q + − x y = Q + − y x = 0 , (181) Q^{xy}_{+-}=Q^{yx}_{+-}=0, \tag*{(181)} Q +− x y = Q +− y x = 0 , (181)
Q + − x z = − λ 2 k y 2 ( λ 2 k 2 + B 2 ) . (182) Q^{xz}_{+-}=-\frac{\lambda^{2}k_{y}}{2(\lambda^{2}k^{2}+B^{2})}. \tag*{(182)} Q +− x z = − 2 ( λ 2 k 2 + B 2 ) λ 2 k y . (182)
Only the Zeeman quantum metric Q + − x x Q^{xx}_{+-} Q +− xx contributes to a nonzero response because
∫ Q + − x z d ϕ = 0. (183) \int Q^{xz}_{+-}d\phi=0. \tag*{(183)} ∫ Q +− x z d ϕ = 0. (183)
The static spin polarization is induced by electric field as
S x ; x E x = − ∫ d k f + − 2 Q + − x x ϵ + − = − ∑ η = ± 1 η π B λ λ 2 ( k η x ) 2 + B 2 . (184) \frac{S^{x;x}}{E^x}=-\int dk f_{+-}\frac{2Q^{xx}_{+-}}{\epsilon_{+-}}
=-\sum_{\eta=\pm1}\frac{\eta\pi B}{\lambda\sqrt{\lambda^2(k_\eta^x)^2+B^2}}. \tag*{(184)} E x S x ; x = − ∫ d k f +− ϵ +− 2 Q +− xx = − η = ± 1 ∑ λ λ 2 ( k η x ) 2 + B 2 η π B . (184)
The dynamic intrinsic gyrotropic magnetic current is obtained as
J x ; x B x = g μ B ∫ d k ∑ n > m f n m 2 Q n m x x ω ϵ m n sin ω t = g μ B ∑ η = ± 1 η 2 π B λ λ 2 ( k η x ) 2 + B 2 ω sin ω t . (185) \frac{J^{x;x}}{B^x}=g\mu_{\mathrm{B}}\int dk\sum_{n>m}f_{nm}2Q^{xx}_{nm}\frac{\omega}{\epsilon_{mn}}\sin\omega t
=g\mu_{\mathrm{B}}\sum_{\eta=\pm1}\frac{\eta2\pi B}{\lambda\sqrt{\lambda^2(k_\eta^x)^2+B^2}}\omega\sin\omega t. \tag*{(185)} B x J x ; x = g μ B ∫ d k n > m ∑ f nm 2 Q nm xx ϵ mn ω sin ω t = g μ B η = ± 1 ∑ λ λ 2 ( k η x ) 2 + B 2 η 2 π B ω sin ω t . (185)
The spin Berry curvatures are calculated as
A + − x y = − 2 n z = − 2 B λ 2 k 2 + B 2 , (186) \mathcal{A}^{xy}_{+-}=-2n_z=-\frac{2B}{\sqrt{\lambda^2k^2+B^2}}, \tag*{(186)} A +− x y = − 2 n z = − λ 2 k 2 + B 2 2 B , (186)
A + − y z = − 2 n x = 2 λ k y λ 2 k 2 + B 2 , (187) \mathcal{A}^{yz}_{+-}=-2n_x=\frac{2\lambda k_y}{\sqrt{\lambda^2k^2+B^2}}, \tag*{(187)} A +− y z = − 2 n x = λ 2 k 2 + B 2 2 λ k y , (187)
A + − z x = − 2 n y = − 2 λ k x λ 2 k 2 + B 2 , (188) \mathcal{A}^{zx}_{+-}=-2n_y=-\frac{2\lambda k_x}{\sqrt{\lambda^2k^2+B^2}}, \tag*{(188)} A +− z x = − 2 n y = − λ 2 k 2 + B 2 2 λ k x , (188)
among which there is only nonzero contribution from A + − x y \mathcal{A}^{xy}_{+-} A +− x y because
∫ d ϕ A + − y z = ∫ d ϕ A + − z x = 0. (189) \int d\phi\mathcal{A}^{yz}_{+-}=\int d\phi\mathcal{A}^{zx}_{+-}=0. \tag*{(189)} ∫ d ϕ A +− y z = ∫ d ϕ A +− z x = 0. (189)
Spin polarization is also induced by magnetic field as
S x ; y B y = g μ B ∫ d k f + − 2 A + − x y ω ϵ + − 2 sin ω t = g μ B ∑ η = ± 1 η 4 π B λ 2 λ 2 ( k η x ) 2 + B 2 . (190) \frac{S^{x;y}}{B^y}=g\mu_{\mathrm{B}}\int dk f_{+-}2\mathcal{A}^{xy}_{+-}\frac{\omega}{\epsilon_{+-}^2}\sin\omega t
=g\mu_{\mathrm{B}}\sum_{\eta=\pm1}\frac{\eta4\pi B}{\lambda^2\sqrt{\lambda^2(k_\eta^x)^2+B^2}}. \tag*{(190)} B y S x ; y = g μ B ∫ d k f +− 2 A +− x y ϵ +− 2 ω sin ω t = g μ B η = ± 1 ∑ λ 2 λ 2 ( k η x ) 2 + B 2 η 4 π B . (190)
The diagonal spin quantum metrices are calculated as
S + − x x = λ 2 k x 2 + B 2 λ 2 k 2 + B 2 , (191) S^{xx}_{+-}=\frac{\lambda^2k_x^2+B^2}{\lambda^2k^2+B^2}, \tag*{(191)} S +− xx = λ 2 k 2 + B 2 λ 2 k x 2 + B 2 , (191)
S + − y y = λ 2 k y 2 + B 2 λ 2 k 2 + B 2 , (192) S^{yy}_{+-}=\frac{\lambda^2k_y^2+B^2}{\lambda^2k^2+B^2}, \tag*{(192)} S +− y y = λ 2 k 2 + B 2 λ 2 k y 2 + B 2 , (192)
S + − z z = λ 2 k 2 λ 2 k 2 + B 2 , (193) S^{zz}_{+-}=\frac{\lambda^2k^2}{\lambda^2k^2+B^2}, \tag*{(193)} S +− z z = λ 2 k 2 + B 2 λ 2 k 2 , (193)
all of which contribute to nonzero responses. The off-diagonal spin quantum metrices are calculated as
S + − x y = 2 λ 2 k x k y λ 2 k 2 + B 2 , (194) S^{xy}_{+-}=\frac{2\lambda^2k_xk_y}{\lambda^2k^2+B^2}, \tag*{(194)} S +− x y = λ 2 k 2 + B 2 2 λ 2 k x k y , (194)
S + − y z = − 2 B λ k x λ 2 k 2 + B 2 , (195) S^{yz}_{+-}=-\frac{2B\lambda k_x}{\lambda^2k^2+B^2}, \tag*{(195)} S +− y z = − λ 2 k 2 + B 2 2 B λ k x , (195)
S + − z x = 2 B λ k y λ 2 k 2 + B 2 . (196) S^{zx}_{+-}=\frac{2B\lambda k_y}{\lambda^2k^2+B^2}. \tag*{(196)} S +− z x = λ 2 k 2 + B 2 2 B λ k y . (196)
all of which do not contribute to the off-diagonal responses because
∫ S + − x y d ϕ = ∫ S + − y z d ϕ = ∫ S + − z x d ϕ = 0. (197) \int S^{xy}_{+-}d\phi=\int S^{yz}_{+-}d\phi=\int S^{zx}_{+-}d\phi=0. \tag*{(197)} ∫ S +− x y d ϕ = ∫ S +− y z d ϕ = ∫ S +− z x d ϕ = 0. (197)
Non-Abelian Zeeman quantum geometry
We generalize the Zeeman quantum geometry to the multiband systems. The Zeeman quantum geometric tensor g n m μ ν g^{\mu\nu}_{nm} g nm μν is defined by the quantum distance d s H S ds_{\mathrm{HS}} d s HS for the infinitesimal translation d k dk d k of the momentum as
( d s H S ) 2 = 1 − ∣ ⟨ ψ ( k ) ∣ U d θ U d k ∣ ψ ( k ) ⟩ ∣ 2 . (198) (ds_{\mathrm{HS}})^2=\sqrt{1-\left|\langle\psi(\boldsymbol{k})|U_{d\theta}U_{dk}|\psi(\boldsymbol{k})\rangle\right|^2}. \tag*{(198)} ( d s HS ) 2 = 1 − ∣ ⟨ ψ ( k ) ∣ U d θ U d k ∣ ψ ( k )⟩ ∣ 2 . (198)
where ∣ ψ ( k ) ⟩ |\psi(\boldsymbol{k})\rangle ∣ ψ ( k )⟩ is given by Eq. (57) , U d θ U_{d\theta} U d θ and U d k U_{dk} U d k are given by Eq. (139) . The Zeeman quantum geometric tensor is given by
( d s H S ) 2 = ∑ μ ν ∑ n ≠ m g n m μ ν d k μ d k ν + s n m μ ν 4 d θ μ d θ ν + z n m μ ν + z m n μ ν 2 d k μ d θ ν . (199) (ds_{\mathrm{HS}})^2=\sum_{\mu\nu}\sum_{n\ne m}g^{\mu\nu}_{nm}dk_\mu dk_\nu+\frac{s^{\mu\nu}_{nm}}{4}d\theta_\mu d\theta_\nu+\frac{z^{\mu\nu}_{nm}+z^{\mu\nu}_{mn}}{2}dk_\mu d\theta_\nu. \tag*{(199)} ( d s HS ) 2 = μν ∑ n = m ∑ g nm μν d k μ d k ν + 4 s nm μν d θ μ d θ ν + 2 z nm μν + z mn μν d k μ d θ ν . (199)
where
g n m μ ν = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k ) ) ∣ ∂ k ν ψ m ( k ) ⟩ (200) g^{\mu\nu}_{nm}=\langle\partial_{k_\mu}\psi_n(\boldsymbol{k})|(1-P(\boldsymbol{k}))|\partial_{k_\nu}\psi_m(\boldsymbol{k})\rangle \tag*{(200)} g nm μν = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k )) ∣ ∂ k ν ψ m ( k )⟩ (200)
is the quantum metric,
s n m μ ν = ⟨ ψ n ( k ) ∣ σ μ ( 1 − P ( k ) ) σ ν ∣ ψ m ( k ) ⟩ (201) s^{\mu\nu}_{nm}=\langle\psi_n(\boldsymbol{k})|\sigma_\mu(1-P(\boldsymbol{k}))\sigma_\nu|\psi_m(\boldsymbol{k})\rangle \tag*{(201)} s nm μν = ⟨ ψ n ( k ) ∣ σ μ ( 1 − P ( k )) σ ν ∣ ψ m ( k )⟩ (201)
is the spin geometric tensor, and
z n m μ ν = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k ) ) σ ν ∣ ψ m ( k ) ⟩ (202) z^{\mu\nu}_{nm}=\langle\partial_{k_\mu}\psi_n(\boldsymbol{k})|(1-P(\boldsymbol{k}))\sigma_\nu|\psi_m(\boldsymbol{k})\rangle \tag*{(202)} z nm μν = ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k )) σ ν ∣ ψ m ( k )⟩ (202)
is the Zeeman geometric tensor. The derivations are shown in Appendix Appendix G.
Quantum geometry for non-Hermitian systems
Open quantum system and non-Hermitian Hamiltonian
Non-Hermitian systems attract much attention. It is required that the Hamiltonian is Hermitian in quantum mechanics. However, it becomes non-Hermitian if we consider an open quantum system. An open quantum system is described by the Lindblad equation for the density matrix ρ \rho ρ ,
d ρ d t = − i ℏ [ H , ρ ] + ∑ k γ k ( L k ρ L k † − 1 2 { L k † L k , ρ } ) . (203) \frac{d\rho}{dt}=-\frac{i}{\hbar}[H,\rho]+\sum_k\gamma_k\left(L_k\rho L_k^\dagger-\frac{1}{2}\left\{L_k^\dagger L_k,\rho\right\}\right). \tag*{(203)} d t d ρ = − ℏ i [ H , ρ ] + k ∑ γ k ( L k ρ L k † − 2 1 { L k † L k , ρ } ) . (203)
where L k L_k L k is the Lindblad operator and γ \gamma γ is the strength of the dissipation. The Lindblad equation is rewritten in the form of
d ρ d t = − i ℏ ( H e f f ρ − ρ H e f f † ) + ∑ k γ k L k ρ L k † , (204) \frac{d\rho}{dt}=-\frac{i}{\hbar}\left(H_{\mathrm{eff}}\rho-\rho H_{\mathrm{eff}}^\dagger\right)+\sum_k\gamma_kL_k\rho L_k^\dagger, \tag*{(204)} d t d ρ = − ℏ i ( H eff ρ − ρ H eff † ) + k ∑ γ k L k ρ L k † , (204)
where we have introduced a non-Hermitian Hamiltonian by
H e f f ≡ H − i ℏ 2 ∑ k γ γ k L k † L k . (205) H_{\mathrm{eff}}\equiv H-\frac{i\hbar}{2}\sum_{k\gamma}\gamma_kL_k^\dagger L_k. \tag*{(205)} H eff ≡ H − 2 i ℏ k γ ∑ γ k L k † L k . (205)
Non-Hermitian quantum geometry
Quantum geometry is generalized to non-Hermitian systems[75 , 115 , 116 , 117 , 118 ] , where H † ≠ H H^{\dagger} \ne H H † = H . There are right and left eigenfunctions and eigenvalues,
H = ε n R ∣ ψ n R ( k ) ⟩ , ⟨ ψ n L ( k ) ∣ H = ⟨ ψ n L ( k ) ∣ ε n L . (206) H=\varepsilon_n^{\mathrm{R}}\left|\psi_n^{\mathrm{R}}(\mathbf{k})\right\rangle,\qquad\left\langle\psi_n^{\mathrm{L}}(\mathbf{k})\right|H=\left\langle\psi_n^{\mathrm{L}}(\mathbf{k})\right|\varepsilon_n^{\mathrm{L}}. \tag*{(206)} H = ε n R ψ n R ( k ) ⟩ , ⟨ ψ n L ( k ) H = ⟨ ψ n L ( k ) ε n L . (206)
where
( ∣ ψ n R ( k ) ⟩ ) † ≠ ⟨ ψ n L ( k ) ∣ . \left(\left|\psi_n^{\mathrm{R}}(\mathbf{k})\right\rangle\right)^{\dagger}\ne\left\langle\psi_n^{\mathrm{L}}(\mathbf{k})\right|. ( ψ n R ( k ) ⟩ ) † = ⟨ ψ n L ( k ) .
in general. However, it is straightforward to generalize quantum geometry to non-Hermitian case. The eigen functions are orthonormalized,
⟨ ψ n L ( k ) | ψ n R ( k ′ ) ⟩ = δ ( k , k ′ ) . (207) \left\langle\psi_n^{\mathrm{L}}(\mathbf{k})\middle|\psi_n^{\mathrm{R}}(\mathbf{k}')\right\rangle=\delta(\mathbf{k},\mathbf{k}'). \tag*{(207)} ⟨ ψ n L ( k ) ψ n R ( k ′ ) ⟩ = δ ( k , k ′ ) . (207)
If the target bands are N N N -fold degenerate, the right and left eigenfunctions are given by
∣ ψ R ( k ) ⟩ = ∑ n = 1 N c n R ∣ ψ n R ( k ) ⟩ , (208) \left|\psi^{\mathrm{R}}(\mathbf{k})\right\rangle=\sum_{n=1}^{N}c_n^{\mathrm{R}}\left|\psi_n^{\mathrm{R}}(\mathbf{k})\right\rangle, \tag*{(208)} ψ R ( k ) ⟩ = n = 1 ∑ N c n R ψ n R ( k ) ⟩ , (208)
⟨ ψ L ( k ) ∣ = ∑ n = 1 N c n L ⟨ ψ n L ( k ) ∣ . (209) \left\langle\psi^{\mathrm{L}}(\mathbf{k})\right|=\sum_{n=1}^{N}c_n^{\mathrm{L}}\left\langle\psi_n^{\mathrm{L}}(\mathbf{k})\right|. \tag*{(209)} ⟨ ψ L ( k ) = n = 1 ∑ N c n L ⟨ ψ n L ( k ) . (209)
The fidelity is defined by
F ( k , k ′ ) = ⟨ ψ L ( k ) | ψ R ( k ′ ) ⟩ ⟨ ψ L ( k ′ ) | ψ R ( k ) ⟩ . (210) F(\mathbf{k},\mathbf{k}')=\sqrt{\left\langle\psi^{\mathrm{L}}(\mathbf{k})\middle|\psi^{\mathrm{R}}(\mathbf{k}')\right\rangle\left\langle\psi^{\mathrm{L}}(\mathbf{k}')\middle|\psi^{\mathrm{R}}(\mathbf{k})\right\rangle}. \tag*{(210)} F ( k , k ′ ) = ⟨ ψ L ( k ) ∣ ψ R ( k ′ ) ⟩ ⟨ ψ L ( k ′ ) ∣ ψ R ( k ) ⟩ . (210)
The Hilbert-Schmidt distance is defined by
s H S ( k , k ′ ) = 1 − F ( k , k ′ ) 2 . (211) s_{\mathrm{HS}}(\mathbf{k},\mathbf{k}')=\sqrt{1-F(\mathbf{k},\mathbf{k}')^{2}}. \tag*{(211)} s HS ( k , k ′ ) = 1 − F ( k , k ′ ) 2 . (211)
The quantum distance with the infinitesimal momentum translation d k d\mathbf{k} d k is given by
d s H S ( k ) = 1 − F ( k , k + d k ) 2 (212) d s_{\mathrm{HS}}(\mathbf{k})=\sqrt{1-F(\mathbf{k},\mathbf{k}+d\mathbf{k})^{2}} \tag*{(212)} d s HS ( k ) = 1 − F ( k , k + d k ) 2 (212)
with
F ( k ) = ⟨ ψ L ( k ) | U d k | ψ R ( k ) ⟩ ⟨ ψ L ( k ) | U d k − 1 | ψ R ( k ) ⟩ , (213) F(\mathbf{k})=\sqrt{\left\langle\psi^{\mathrm{L}}(\mathbf{k})\middle|U_{d\mathbf{k}}\middle|\psi^{\mathrm{R}}(\mathbf{k})\right\rangle\left\langle\psi^{\mathrm{L}}(\mathbf{k})\middle|U_{d\mathbf{k}}^{-1}\middle|\psi^{\mathrm{R}}(\mathbf{k})\right\rangle}, \tag*{(213)} F ( k ) = ⟨ ψ L ( k ) ∣ U d k ∣ ψ R ( k ) ⟩ ⟨ ψ L ( k ) U d k − 1 ψ R ( k ) ⟩ , (213)
where U d k U_{d\mathbf{k}} U d k is given by Eq. (5) . The quantum distance for the infinitesimal momentum is expanded in terms of the quantum geometric tensor F n m μ ν \mathcal{F}_{nm}^{\mu\nu} F nm μν as
( d s H S ) 2 = ∑ μ ν ∑ n m F n m μ ν d k μ d k ν , (214) (d s_{\mathrm{HS}})^{2}=\sum_{\mu\nu}\sum_{nm}\mathcal{F}_{nm}^{\mu\nu}d k_{\mu}d k_{\nu}, \tag*{(214)} ( d s HS ) 2 = μν ∑ nm ∑ F nm μν d k μ d k ν , (214)
The quantum geometric tensor is
F n m μ ν ( k ) = ⟨ ∂ k μ ψ n L ( k ) | ( 1 − P ( k ) ) | ∂ k ν ψ m R ( k ) ⟩ (215) \mathcal{F}_{nm}^{\mu\nu}(\mathbf{k})=\left\langle\partial_{k_{\mu}}\psi_n^{\mathrm{L}}(\mathbf{k})\middle|\left(1-P(\mathbf{k})\right)\middle|\partial_{k_{\nu}}\psi_m^{\mathrm{R}}(\mathbf{k})\right\rangle \tag*{(215)} F nm μν ( k ) = ⟨ ∂ k μ ψ n L ( k ) ( 1 − P ( k ) ) ∂ k ν ψ m R ( k ) ⟩ (215)
with the projection operator
P ( k ) ≡ ∑ n = 1 N ∣ ψ n R ( k ) ⟩ ⟨ ψ n L ( k ) ∣ . (216) P(\mathbf{k})\equiv\sum_{n=1}^{N}\left|\psi_n^{\mathrm{R}}(\mathbf{k})\right\rangle\left\langle\psi_n^{\mathrm{L}}(\mathbf{k})\right|. \tag*{(216)} P ( k ) ≡ n = 1 ∑ N ψ n R ( k ) ⟩ ⟨ ψ n L ( k ) . (216)
The quantum metric is the real part of the quantum geometric tensor,
g n m μ ν ≡ Re F n m μ ν = F m n μ ν ( k ) + ( F m n μ ν ( k ) ) ∗ 2 , (217) g_{nm}^{\mu\nu}\equiv\operatorname{Re}\mathcal{F}_{nm}^{\mu\nu}=\frac{\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k})+\left(\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k})\right)^{*}}{2}, \tag*{(217)} g nm μν ≡ Re F nm μν = 2 F mn μν ( k ) + ( F mn μν ( k ) ) ∗ , (217)
while the Berry curvature is the imaginary part of the quantum geometric tensor,
Ω n m μ ν ≡ 2 Im F n m μ ν = i ( F m n μ ν ( k ) − ( F m n μ ν ( k ) ) ∗ ) . (218) \Omega_{nm}^{\mu\nu}\equiv2\operatorname{Im}\mathcal{F}_{nm}^{\mu\nu}=i\left(\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k})-\left(\mathcal{F}_{mn}^{\mu\nu}(\mathbf{k})\right)^{*}\right). \tag*{(218)} Ω nm μν ≡ 2 Im F nm μν = i ( F mn μν ( k ) − ( F mn μν ( k ) ) ∗ ) . (218)
The Wilczek-Zee connection is given by
a n m μ ( k ) ≡ − i ⟨ ψ n L | ∂ k μ | ψ m R ⟩ , (219) a_{nm}^{\mu}(\mathbf{k})\equiv-i\left\langle\psi_n^{\mathrm{L}}\middle|\partial_{k_{\mu}}\middle|\psi_m^{\mathrm{R}}\right\rangle, \tag*{(219)} a nm μ ( k ) ≡ − i ⟨ ψ n L ∂ k μ ψ m R ⟩ , (219)
while the Berry curvature is given by[120 , 122 , 123 , 124 , 126 ]
Ω n m μ ν ( k ) = ∇ × a n m ( k ) − i ∑ n ′ [ a n n ′ μ , a n ′ m ν ] . (220) \Omega_{nm}^{\mu\nu}(\mathbf{k})=\nabla\times\mathbf{a}_{nm}(\mathbf{k})-i\sum_{n'}\left[a_{nn'}^{\mu},a_{n'm}^{\nu}\right]. \tag*{(220)} Ω nm μν ( k ) = ∇ × a nm ( k ) − i n ′ ∑ [ a n n ′ μ , a n ′ m ν ] . (220)
We note that there are some other generalizations to non-Hermitian systems[75 , 115 , 116 , 117 , 118 , 119 ] .
Two-band systems
We consider a two-band system with the index n = ± n=\pm n = ± . The Hamiltonian is generally given by
H ( k ) = h 0 ( k ) + σ ⋅ h ( k ) , (221) H(\mathbf{k})=h_0(\mathbf{k})+\boldsymbol{\sigma}\cdot\mathbf{h}(\mathbf{k}), \tag*{(221)} H ( k ) = h 0 ( k ) + σ ⋅ h ( k ) , (221)
where
h 0 ( k ) = h 0 R e ( k ) + i h 0 I m ( k ) , (222) h_0(\mathbf{k})=h_{0\mathrm{Re}}(\mathbf{k})+i h_{0\mathrm{Im}}(\mathbf{k}), \tag*{(222)} h 0 ( k ) = h 0 Re ( k ) + i h 0 Im ( k ) , (222)
h ( k ) = h R e ( k ) + i h I m ( k ) , (223) \mathbf{h}(\mathbf{k})=\mathbf{h}_{\mathrm{Re}}(\mathbf{k})+i\mathbf{h}_{\mathrm{Im}}(\mathbf{k}), \tag*{(223)} h ( k ) = h Re ( k ) + i h Im ( k ) , (223)
where h 0 ( k ) h_0(\mathbf{k}) h 0 ( k ) and h ( k ) \mathbf{h}(\mathbf{k}) h ( k ) are complex functions with h 0 R e h_{0\mathrm{Re}} h 0 Re , h 0 I m h_{0\mathrm{Im}} h 0 Im , h R e ( k ) \mathbf{h}_{\mathrm{Re}}(\mathbf{k}) h Re ( k ) and h I m ( k ) \mathbf{h}_{\mathrm{Im}}(\mathbf{k}) h Im ( k ) being real functions. The energy is given by
ε ± = h 0 ( k ) ± h ( k ) 2 , (224) \varepsilon_{\pm}=h_0(\mathbf{k})\pm\sqrt{\mathbf{h}(\mathbf{k})^2}, \tag*{(224)} ε ± = h 0 ( k ) ± h ( k ) 2 , (224)
where
h ( k ) 2 ≡ h x 2 ( k ) + h y 2 ( k ) + h z 2 ( k ) . (225) \mathbf{h}(\mathbf{k})^2\equiv h_x^2(\mathbf{k})+h_y^2(\mathbf{k})+h_z^2(\mathbf{k}). \tag*{(225)} h ( k ) 2 ≡ h x 2 ( k ) + h y 2 ( k ) + h z 2 ( k ) . (225)
Its right and left eigenvalues are given by[121 ]
∣ ψ ± R ⟩ = 1 2 ε ± ( ε ± − h z ) ( h x − i h y , ε ± − h z ) T , (226) \left|\psi_{\pm}^{\mathrm{R}}\right\rangle=\frac{1}{\sqrt{2\varepsilon_{\pm}(\varepsilon_{\pm}-h_z)}}\left(h_x-i h_y,\varepsilon_{\pm}-h_z\right)^{T}, \tag*{(226)} ψ ± R ⟩ = 2 ε ± ( ε ± − h z ) 1 ( h x − i h y , ε ± − h z ) T , (226)
⟨ ψ ± L ∣ = 1 2 ε ± ( ε ± − h z ) ( h x + i h y , ε ± − h z ) , (227) \left\langle\psi_{\pm}^{\mathrm{L}}\right|=\frac{1}{\sqrt{2\varepsilon_{\pm}(\varepsilon_{\pm}-h_z)}}\left(h_x+i h_y,\varepsilon_{\pm}-h_z\right), \tag*{(227)} ⟨ ψ ± L = 2 ε ± ( ε ± − h z ) 1 ( h x + i h y , ε ± − h z ) , (227)
which satisfy the biorthogal condition, ⟨ ψ ± L | ψ ± R ⟩ = 1 \left\langle\psi_{\pm}^{\mathrm{L}}\middle|\psi_{\pm}^{\mathrm{R}}\right\rangle=1 ⟨ ψ ± L ψ ± R ⟩ = 1 . The non-Hermitian Berry connection is calculated as[121 ]
a ± μ = h x ∂ k μ h y − h y ∂ k μ h x ε ± ( ε ± − h z ) . (228) a_{\pm}^{\mu}=\frac{h_x\partial_{k_{\mu}}h_y-h_y\partial_{k_{\mu}}h_x}{\varepsilon_{\pm}(\varepsilon_{\pm}-h_z)}. \tag*{(228)} a ± μ = ε ± ( ε ± − h z ) h x ∂ k μ h y − h y ∂ k μ h x . (228)
The non-Hermitian Berry curvature reads[127 ]
Ω ± ( k ) = ∇ × a ± ( k ) = 1 2 ε ± 3 / 2 ϵ μ ν ρ h μ ∂ k x h ν ∂ k y h ρ . (229) \Omega_{\pm}(\mathbf{k})=\nabla\times\mathbf{a}_{\pm}(\mathbf{k})=\frac{1}{2\varepsilon_{\pm}^{3/2}}\epsilon_{\mu\nu\rho}h_{\mu}\partial_{k_x}h_{\nu}\partial_{k_y}h_{\rho}. \tag*{(229)} Ω ± ( k ) = ∇ × a ± ( k ) = 2 ε ± 3/2 1 ϵ μν ρ h μ ∂ k x h ν ∂ k y h ρ . (229)
It is further simplified as
Ω ± ( k ) = ∓ 1 2 n ⋅ ( ∂ k x n × ∂ k y n ) (230) \Omega_{\pm}(\mathbf{k}) = \mp\frac{1}{2}\mathbf{n}\cdot\left(\partial_{k_x}\mathbf{n}\times\partial_{k_y}\mathbf{n}\right) \tag*{(230)} Ω ± ( k ) = ∓ 2 1 n ⋅ ( ∂ k x n × ∂ k y n ) (230)
with
n = h ( k ) h ( k ) 2 . (231) \mathbf{n}=\frac{\mathbf{h}(\mathbf{k})}{\sqrt{\mathbf{h}(\mathbf{k})^2}}. \tag*{(231)} n = h ( k ) 2 h ( k ) . (231)
It is identical to Eq. (48) for the Hermitian system. Note that
h ( k ) 2 ≠ ∣ h ( k ) ∣ (232) \sqrt{\mathbf{h}(\mathbf{k})^2}\ne|\mathbf{h}(\mathbf{k})| \tag*{(232)} h ( k ) 2 = ∣ h ( k ) ∣ (232)
for non-Hermitian systems. The non-Hermitian Chern number is defined by[122 , 125 ]
C ± = 1 2 π ∫ Ω ± ( k ) d 2 k , (233) C_{\pm}=\frac{1}{2\pi}\int\Omega_{\pm}(\mathbf{k})\,d^2k, \tag*{(233)} C ± = 2 π 1 ∫ Ω ± ( k ) d 2 k , (233)
where the integration is over the Brillouin zone.
We can check
g ± μ ν ( k ) = ± 1 2 ( ∂ k μ n ) ⋅ ( ∂ k ν n ) (234) g_{\pm}^{\mu\nu}(\mathbf{k})=\pm\frac{1}{2}\left(\partial_{k_\mu}\mathbf{n}\right)\cdot\left(\partial_{k_\nu}\mathbf{n}\right) \tag*{(234)} g ± μν ( k ) = ± 2 1 ( ∂ k μ n ) ⋅ ( ∂ k ν n ) (234)
even for the non-Hermitian systems. It is identical to Eq. (47) .
Dirac system with a complex mass
We consider a Dirac Hamiltonian with a complex mass m = B + i γ m=B+i\gamma m = B + iγ , whose Hamiltonian is given by
H ( k ) = λ ( k × σ ) z + ( B + i γ ) σ z , (235) H(\mathbf{k})=\lambda(\mathbf{k}\times\boldsymbol{\sigma})_z+(B+i\gamma)\sigma_z, \tag*{(235)} H ( k ) = λ ( k × σ ) z + ( B + iγ ) σ z , (235)
where γ \gamma γ is real. The Berry curvature (229) is calculated as
Ω ± ( k ) = ± ( m + i γ ) λ 2 2 ( λ 2 k 2 + ( m + i γ ) 2 ) 3 / 2 , (236) \Omega_{\pm}(\mathbf{k})=\pm\frac{(m+i\gamma)\lambda^2}{2\left(\lambda^2k^2+(m+i\gamma)^2\right)^{3/2}}, \tag*{(236)} Ω ± ( k ) = ± 2 ( λ 2 k 2 + ( m + iγ ) 2 ) 3/2 ( m + iγ ) λ 2 , (236)
which leads to the Chern number[128 ]
C ± = ± m + i γ 2 ∣ m + i γ ∣ , (237) C_{\pm}=\pm\frac{m+i\gamma}{2|m+i\gamma|}, \tag*{(237)} C ± = ± 2∣ m + iγ ∣ m + iγ , (237)
where ∣ C ± ∣ = 1 / 2 |C_{\pm}|=1/2 ∣ C ± ∣ = 1/2 .
The quantum metrices are calculated as
g ± x x ( k ) = ∓ λ 2 ( ( λ k y ) 2 + ( m + i γ ) 2 ) ( ( λ k ) 2 + ( m + i γ ) 2 ) 2 , (238) g_{\pm}^{xx}(\mathbf{k})=\mp\frac{\lambda^2\left((\lambda k_y)^2+(m+i\gamma)^2\right)}{\left((\lambda k)^2+(m+i\gamma)^2\right)^2}, \tag*{(238)} g ± xx ( k ) = ∓ ( ( λk ) 2 + ( m + iγ ) 2 ) 2 λ 2 ( ( λ k y ) 2 + ( m + iγ ) 2 ) , (238)
g ± y y ( k ) = ∓ λ 2 ( ( λ k x ) 2 + ( m + i γ ) 2 ) ( ( λ k ) 2 + ( m + i γ ) 2 ) 2 , (239) g_{\pm}^{yy}(\mathbf{k})=\mp\frac{\lambda^2\left((\lambda k_x)^2+(m+i\gamma)^2\right)}{\left((\lambda k)^2+(m+i\gamma)^2\right)^2}, \tag*{(239)} g ± y y ( k ) = ∓ ( ( λk ) 2 + ( m + iγ ) 2 ) 2 λ 2 ( ( λ k x ) 2 + ( m + iγ ) 2 ) , (239)
g ± x y ( k ) = ± λ 2 ( λ k x ) ( λ k y ) ( ( λ k ) 2 + ( m + i γ ) 2 ) 2 , (240) g_{\pm}^{xy}(\mathbf{k})=\pm\frac{\lambda^2(\lambda k_x)(\lambda k_y)}{\left((\lambda k)^2+(m+i\gamma)^2\right)^2}, \tag*{(240)} g ± x y ( k ) = ± ( ( λk ) 2 + ( m + iγ ) 2 ) 2 λ 2 ( λ k x ) ( λ k y ) , (240)
The quantum volume is calculated as
∫ d k det g ± μ ν ( k ) = 1 8 ( λ m + i γ ) 2 . (241) \int d k\sqrt{\det g_{\pm}^{\mu\nu}(\mathbf{k})}=\frac{1}{8}\left(\frac{\lambda}{m+i\gamma}\right)^2. \tag*{(241)} ∫ d k det g ± μν ( k ) = 8 1 ( m + iγ λ ) 2 . (241)
It is negative for pure imaginary mass m = 0 m=0 m = 0 and γ ≠ 0 \gamma\ne0 γ = 0 .
Uhlmann quantum geometry for density matrix
So far, quantum geometry is constructed for the wave function. It means that it is constructed for pure states. On the other hand, mixed states are important for quantum information, finite temperature system and open quantum systems. They are described by the density matrix
ρ ( k ) = ∑ n = 1 N p n ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ , (242) \rho(\mathbf{k})=\sum_{n=1}^{N}p_n\left|\psi_n(\mathbf{k})\right\rangle\left\langle\psi_n(\mathbf{k})\right|, \tag*{(242)} ρ ( k ) = n = 1 ∑ N p n ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ , (242)
where p n = exp ( − ε n / k B T ) p_n=\exp(-\varepsilon_n/k_BT) p n = exp ( − ε n / k B T ) . Quantum geometry for it is constructed by Uhlmann[129 , 130 , 131 , 132 , 133 , 141 , 144 ] [143]. Starting from the fidelity of the density matrix, we naturally obtain quantum Fisher information, which gives the lower boundary of the quantum fluctuation. Hence, quantum geometry for density matrices is called quantum information geometry.
We start with the Uhlmann fidelity[137 , 138 ] for the density matrix defined by
F ( k , k ′ ) = Tr ρ ( k ) ρ ( k ′ ) ρ ( k ) . (243) F(\mathbf{k},\mathbf{k}')=\operatorname{Tr}\sqrt{\sqrt{\rho(\mathbf{k})}\rho(\mathbf{k}')\sqrt{\rho(\mathbf{k})}}. \tag*{(243)} F ( k , k ′ ) = Tr ρ ( k ) ρ ( k ′ ) ρ ( k ) . (243)
It is symmetric
F ( k , k ′ ) = F ( k ′ , k ) , (244) F(\mathbf{k},\mathbf{k}')=F(\mathbf{k}',\mathbf{k}), \tag*{(244)} F ( k , k ′ ) = F ( k ′ , k ) , (244)
because ρ ( k ) ρ ( k ′ ) ρ ( k ) \sqrt{\rho(\mathbf{k})}\rho(\mathbf{k}')\sqrt{\rho(\mathbf{k})} ρ ( k ) ρ ( k ′ ) ρ ( k ) and ρ ( k ′ ) ρ ( k ) ρ ( k ′ ) \sqrt{\rho(\mathbf{k}')}\rho(\mathbf{k})\sqrt{\rho(\mathbf{k}')} ρ ( k ′ ) ρ ( k ) ρ ( k ′ ) have the same eigenvalues. It is bounded as
0 ≤ F ( k , k ′ ) ≤ 1 , (245) 0\le F(\mathbf{k},\mathbf{k}')\le1, \tag*{(245)} 0 ≤ F ( k , k ′ ) ≤ 1 , (245)
where
F ( k , k ) = 1 , (246) F(\mathbf{k},\mathbf{k})=1, \tag*{(246)} F ( k , k ) = 1 , (246)
because
Tr ρ ( k ) = 1. (247) \operatorname{Tr}\rho(\mathbf{k})=1. \tag*{(247)} Tr ρ ( k ) = 1. (247)
It reduces to the fidelity (1) for the pure states,
F ( k , k ′ ) = Tr ρ ( k ) ρ ( k ′ ) = Tr ∣ ψ ( k ) ⟩ ⟨ ψ ( k ) ∣ ψ ( k ′ ) ⟩ ⟨ ψ ( k ′ ) ∣ = ⟨ ψ ( k ) ∣ ψ ( k ′ ) ⟩ ⟨ ψ ( k ′ ) ∣ ψ ( k ) ∣ = ∣ ⟨ ψ ( k ) ∣ ψ ( k ′ ) ⟩ ∣ . (248) \begin{aligned}
F(\mathbf{k},\mathbf{k}')=\operatorname{Tr}\sqrt{\rho(\mathbf{k})\rho(\mathbf{k}')} \\
&=\operatorname{Tr}\sqrt{|\psi(\mathbf{k})\rangle\langle\psi(\mathbf{k})|\psi(\mathbf{k}')\rangle\langle\psi(\mathbf{k}')|} \\
&=\sqrt{\langle\psi(\mathbf{k})|\psi(\mathbf{k}')\rangle\langle\psi(\mathbf{k}')|\psi(\mathbf{k})|} \\
&=|\langle\psi(\mathbf{k})|\psi(\mathbf{k}')\rangle|. \tag*{(248)}
\end{aligned} F ( k , k ′ ) = Tr ρ ( k ) ρ ( k ′ ) = Tr ∣ ψ ( k )⟩ ⟨ ψ ( k ) ∣ ψ ( k ′ )⟩ ⟨ ψ ( k ′ ) ∣ = ⟨ ψ ( k ) ∣ ψ ( k ′ )⟩ ⟨ ψ ( k ′ ) ∣ ψ ( k ) ∣ = ∣ ⟨ ψ ( k ) ∣ ψ ( k ′ )⟩ ∣. (248)
With the use of the Uhlmann fidelity (243) , the Bures distance[139 ] is defined by
s B = 1 − F ( k , k ′ ) 2 . (249) s_B=\sqrt{1-F(\mathbf{k},\mathbf{k}')^2}. \tag*{(249)} s B = 1 − F ( k , k ′ ) 2 . (249)
We make a purification of the density matrix with the use of the amplitude W W W satisfying
ρ = W W † , (250) \rho=WW^\dagger, \tag*{(250)} ρ = W W † , (250)
where the amplitude W W W is given by
W = ( p 1 ∣ ψ 1 ( k ) ⟩ , p 2 ∣ ψ 2 ( k ) ⟩ , ⋯ , p N ∣ ψ N ( k ) ⟩ ) , (251) W=\left(\sqrt{p_1}\lvert\psi_1(\mathbf{k})\rangle,\sqrt{p_2}\lvert\psi_2(\mathbf{k})\rangle,\cdots,\sqrt{p_N}\lvert\psi_N(\mathbf{k})\rangle\right), \tag*{(251)} W = ( p 1 ∣ ψ 1 ( k )⟩ , p 2 ∣ ψ 2 ( k )⟩ , ⋯ , p N ∣ ψ N ( k )⟩ ) , (251)
W † = ( p 1 ⟨ ψ 1 ( k ) ∣ p 2 ⟨ ψ 2 ( k ) ∣ ⋮ p N ⟨ ψ N ( k ) ∣ ) . (252) W^\dagger=
\begin{pmatrix}
\sqrt{p_1}\langle\psi_1(\mathbf{k})\rvert\\
\sqrt{p_2}\langle\psi_2(\mathbf{k})\rvert\\
\vdots\\
\sqrt{p_N}\langle\psi_N(\mathbf{k})\rvert
\end{pmatrix}. \tag*{(252)} W † = p 1 ⟨ ψ 1 ( k )∣ p 2 ⟨ ψ 2 ( k )∣ ⋮ p N ⟨ ψ N ( k )∣ . (252)
Hence, purification is always possible using the spectral decomposition. It is identical to represent
ρ = Tr E ∣ Ψ ⟩ ⟨ Ψ ∣ (253) \rho=\operatorname{Tr}_{E}\lvert\Psi\rangle\langle\Psi\rvert \tag*{(253)} ρ = Tr E ∣ Ψ ⟩ ⟨ Ψ ∣ (253)
for the extended Hamiltonian
H e x t ≡ H ⊗ H E , (254) H_{\mathrm{ext}}\equiv H\otimes H_E, \tag*{(254)} H ext ≡ H ⊗ H E , (254)
by attaching an environment Hamiltonian H E H_E H E , where the trace is taken over the environment. It means that the mixed states are represented by a pure state ∣ Ψ ⟩ \lvert\Psi\rangle ∣ Ψ ⟩ if we prepare a large system described by H e x t H_{\mathrm{ext}} H ext . Comparing Eq. (250) and Eq. (253) , W W W and W † W^\dagger W † correspond to ∣ Ψ ⟩ \lvert\Psi\rangle ∣ Ψ ⟩ and ⟨ Ψ ∣ \langle\Psi\rvert ⟨ Ψ ∣ , respectively.
There is a gauge degrees of freedom U ′ U' U ′ in W W W as
W = ρ U ′ , (255) W=\sqrt{\rho}U', \tag*{(255)} W = ρ U ′ , (255)
where U U U is a unitary matrix. By using the purification, the Uhlmann fidelity is rewritten as
F ( k , k ′ ) = Tr W † ( k ) W ( k ′ ) W † ( k ′ ) W ( k ) = Tr ∣ W † ( k ) W ( k ′ ) ∣ . (256) F(\mathbf{k},\mathbf{k}')=\operatorname{Tr}\sqrt{W^\dagger(\mathbf{k})W(\mathbf{k}')W^\dagger(\mathbf{k}')W(\mathbf{k})}
=\operatorname{Tr}\left\lvert W^\dagger(\mathbf{k})W(\mathbf{k}')\right\rvert. \tag*{(256)} F ( k , k ′ ) = Tr W † ( k ) W ( k ′ ) W † ( k ′ ) W ( k ) = Tr W † ( k ) W ( k ′ ) . (256)
The Bures distance (249) is rewritten as
s B 2 = 2 ( 1 − Tr ∣ W † ( k ) W ( k ′ ) ∣ ) = 2 Tr ( W ( k ′ ) − W ( k ) ) ( W ( k ′ ) − W ( k ) ) † , (257) s_{\mathrm{B}}^2=2\left(1-\operatorname{Tr}\left\lvert W^\dagger(\mathbf{k})W(\mathbf{k}')\right\rvert\right)
=2\operatorname{Tr}\left(W(\mathbf{k}')-W(\mathbf{k})\right)\left(W(\mathbf{k}')-W(\mathbf{k})\right)^\dagger, \tag*{(257)} s B 2 = 2 ( 1 − Tr W † ( k ) W ( k ′ ) ) = 2 Tr ( W ( k ′ ) − W ( k ) ) ( W ( k ′ ) − W ( k ) ) † , (257)
where we have used the normalization condition for the density matrix,
Tr ρ ( k ) = Tr W ( k ) W † ( k ) = 1 , (258) \operatorname{Tr}\rho(\mathbf{k})=\operatorname{Tr}W(\mathbf{k})W^\dagger(\mathbf{k})=1, \tag*{(258)} Tr ρ ( k ) = Tr W ( k ) W † ( k ) = 1 , (258)
and the Uhlmann parallel transport condition[129 ]
W † ( k ) W ( k ′ ) = W † ( k ′ ) W ( k ) > 0. (259) W^\dagger(\mathbf{k})W(\mathbf{k}')=W^\dagger(\mathbf{k}')W(\mathbf{k})>0. \tag*{(259)} W † ( k ) W ( k ′ ) = W † ( k ′ ) W ( k ) > 0. (259)
The Bures distance for the infinitesimal distance d k d\mathbf{k} d k is given by setting k ′ = k + d k \mathbf{k}'=\mathbf{k}+d\mathbf{k} k ′ = k + d k as
( d s B ) 2 = 2 Tr d W d W † = 2 Tr ∣ d W ∣ 2 . (260) (ds_{\mathrm{B}})^2=2\operatorname{Tr}dW\,dW^\dagger=2\operatorname{Tr}\lvert dW\rvert^2. \tag*{(260)} ( d s B ) 2 = 2 Tr d W d W † = 2 Tr ∣ d W ∣ 2 . (260)
The differential form of the Uhlmann parallel transport condition (259) is given by
W † d W = ( d W † ) W . (261) W^\dagger dW=(dW^\dagger)W. \tag*{(261)} W † d W = ( d W † ) W . (261)
It is fulfilled by the ansatz
d W = 1 2 L W , L † = L . (262) dW=\frac{1}{2}\mathcal{L}W,\qquad\mathcal{L}^\dagger=\mathcal{L}. \tag*{(262)} d W = 2 1 L W , L † = L . (262)
Indeed, the left and right hand sides of Eq. (261) are given by
W † d W = 1 2 W † L W , (263) W^\dagger dW=\frac{1}{2}W^\dagger\mathcal{L}W, \tag*{(263)} W † d W = 2 1 W † L W , (263)
( d W † ) W = 1 2 ( L W ) † W = 1 2 W † L † W = 1 2 W † L W , (264) (dW^\dagger)W=\frac{1}{2}(\mathcal{L}W)^\dagger W=\frac{1}{2}W^\dagger\mathcal{L}^\dagger W=\frac{1}{2}W^\dagger\mathcal{L}W, \tag*{(264)} ( d W † ) W = 2 1 ( L W ) † W = 2 1 W † L † W = 2 1 W † L W , (264)
and identical with the use of Eq. (262) . L \mathcal{L} L is called the symmetric logarithmic derivative (SLD)[140 , 145 , 146 ] .
Then, the Bures distance (260) reads
( d s B ) 2 = 2 Tr ∣ d W ∣ 2 = 2 Tr ∣ 1 2 L W ∣ 2 = 1 2 Tr L W W † L = 1 2 Tr L ρ L = 1 2 Tr ρ L 2 = 1 2 ⟨ L 2 ⟩ . (265) (ds_{\mathrm{B}})^2=2\operatorname{Tr}\lvert dW\rvert^2=2\operatorname{Tr}\left\lvert\frac{1}{2}\mathcal{L}W\right\rvert^2
=\frac{1}{2}\operatorname{Tr}\mathcal{L}WW^\dagger\mathcal{L}
=\frac{1}{2}\operatorname{Tr}\mathcal{L}\rho\mathcal{L}
=\frac{1}{2}\operatorname{Tr}\rho\mathcal{L}^2=\frac{1}{2}\langle\mathcal{L}^2\rangle. \tag*{(265)} ( d s B ) 2 = 2 Tr ∣ d W ∣ 2 = 2 Tr 2 1 L W 2 = 2 1 Tr L W W † L = 2 1 Tr L ρ L = 2 1 Tr ρ L 2 = 2 1 ⟨ L 2 ⟩ . (265)
By using one form of the SLD,
L = L μ d k μ , (266) \mathcal{L}=\mathcal{L}^{\mu}dk_{\mu}, \tag*{(266)} L = L μ d k μ , (266)
we obtain the Uhlmann quantum geometric tensor F U μ ν \mathcal{F}_{\mathrm{U}}^{\mu\nu} F U μν as
( d s B ) 2 = Tr [ ρ ∑ μ ν L μ L ν d k μ d k ν ] ≡ ∑ μ ν F U μ ν d k μ d k ν (267) (ds_{\mathrm{B}})^2=\operatorname{Tr}\left[\rho\sum_{\mu\nu}\mathcal{L}^{\mu}\mathcal{L}^{\nu}dk_{\mu}dk_{\nu}\right]\equiv\sum_{\mu\nu}\mathcal{F}_{\mathrm{U}}^{\mu\nu}dk_{\mu}dk_{\nu} \tag*{(267)} ( d s B ) 2 = Tr [ ρ μν ∑ L μ L ν d k μ d k ν ] ≡ μν ∑ F U μν d k μ d k ν (267)
with
F U μ ν ≡ Tr [ ρ L μ L ν ] = ⟨ L μ L ν ⟩ . (268) \mathcal{F}_{\mathrm{U}}^{\mu\nu}\equiv\operatorname{Tr}[\rho\mathcal{L}^{\mu}\mathcal{L}^{\nu}]=\langle\mathcal{L}^{\mu}\mathcal{L}^{\nu}\rangle. \tag*{(268)} F U μν ≡ Tr [ ρ L μ L ν ] = ⟨ L μ L ν ⟩ . (268)
The Uhlmann quantum geometric tensor F U μ ν \mathcal{F}_{\mathrm{U}}^{\mu\nu} F U μν is decomposed as
F U μ ν = F Q F i s h e r μ ν + i U ˉ μ ν (269) \mathcal{F}_{\mathrm{U}}^{\mu\nu}=\mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}+i\bar{U}^{\mu\nu} \tag*{(269)} F U μν = F QFisher μν + i U ˉ μν (269)
with the SLD quantum Fisher information[140 , 142 ] [143]
F Q F i s h e r μ ν ≡ 1 2 Tr [ ρ { L μ , L ν } ] = 1 2 ⟨ { L μ , L ν } ⟩ , (270) \mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}\equiv\frac{1}{2}\operatorname{Tr}\left[\rho\{\mathcal{L}^{\mu},\mathcal{L}^{\nu}\}\right]=\frac{1}{2}\langle\{\mathcal{L}^{\mu},\mathcal{L}^{\nu}\}\rangle, \tag*{(270)} F QFisher μν ≡ 2 1 Tr [ ρ { L μ , L ν } ] = 2 1 ⟨{ L μ , L ν }⟩ , (270)
and the mean Uhlmann curvature[133 ] [143]
U ˉ μ ν ≡ − i 2 Tr [ ρ [ L μ , L ν ] ] = − i 2 ⟨ [ L μ , L ν ] ⟩ . (271) \bar{U}^{\mu\nu}\equiv-\frac{i}{2}\operatorname{Tr}\left[\rho[\mathcal{L}^{\mu},\mathcal{L}^{\nu}]\right]=-\frac{i}{2}\langle[\mathcal{L}^{\mu},\mathcal{L}^{\nu}]\rangle. \tag*{(271)} U ˉ μν ≡ − 2 i Tr [ ρ [ L μ , L ν ] ] = − 2 i ⟨[ L μ , L ν ]⟩ . (271)
We note that
U μ ν ≡ − i 2 [ L μ , L ν ] (272) U^{\mu\nu}\equiv-\frac{i}{2}[\mathcal{L}^{\mu},\mathcal{L}^{\nu}] \tag*{(272)} U μν ≡ − 2 i [ L μ , L ν ] (272)
is the Uhlmann curvature, which satisfies
U ˉ μ ν = Tr [ ρ U μ ν ] = ⟨ U μ ν ⟩ . (273) \bar{U}^{\mu\nu}=\operatorname{Tr}[\rho U^{\mu\nu}]=\langle U^{\mu\nu}\rangle. \tag*{(273)} U ˉ μν = Tr [ ρ U μν ] = ⟨ U μν ⟩ . (273)
It follows from Eq. (250) that
d ρ = d ( W W † ) = ( d W ) W † + W d W † = 1 2 L W W † + W 1 2 ( L W ) † = 1 2 ( L ρ + ρ L ) = 1 2 { L , ρ } . (274) d\rho=d\left(WW^\dagger\right)=(dW)W^\dagger+WdW^\dagger
=\frac{1}{2}\mathcal{L}WW^\dagger+W\frac{1}{2}\left(\mathcal{L}W\right)^\dagger
=\frac{1}{2}\left(\mathcal{L}\rho+\rho\mathcal{L}\right)=\frac{1}{2}\{\mathcal{L},\rho\}. \tag*{(274)} d ρ = d ( W W † ) = ( d W ) W † + W d W † = 2 1 L W W † + W 2 1 ( L W ) † = 2 1 ( L ρ + ρ L ) = 2 1 { L , ρ } . (274)
By using Eq. (242) and Eq. (274) , the SLD L μ \mathcal{L}^{\mu} L μ is explicitly given by [134 , 135 , 136 , 140 ] [143]
L μ = ∑ n ∂ ln p n ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ + 2 i ∑ n ≠ m p m − p n p m + p n a n m μ ∣ ψ n ⟩ ⟨ ψ m ∣ . (275) \mathcal{L}^{\mu}=\sum_n\frac{\partial\ln p_n}{\partial k_\mu}|\psi_n\rangle\langle\psi_n|+2i\sum_{n\ne m}\frac{p_m-p_n}{p_m+p_n}a^\mu_{nm}|\psi_n\rangle\langle\psi_m|. \tag*{(275)} L μ = n ∑ ∂ k μ ∂ ln p n ∣ ψ n ⟩ ⟨ ψ n ∣ + 2 i n = m ∑ p m + p n p m − p n a nm μ ∣ ψ n ⟩ ⟨ ψ m ∣. (275)
The derivation is shown in Appendix § 1 .
The symmetric logarithmic derivative is decomposed into the classical part L C μ \mathcal{L}^{\mu}_{C} L C μ and the quantum part L Q μ \mathcal{L}^{\mu}_{Q} L Q μ as [143]
L μ = L C μ + L Q μ , (276) \mathcal{L}^{\mu}=\mathcal{L}^{\mu}_{C}+\mathcal{L}^{\mu}_{Q}, \tag*{(276)} L μ = L C μ + L Q μ , (276)
L C μ = ∑ n ∂ ln p n ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ , (277) \mathcal{L}^{\mu}_{C}=\sum_n\frac{\partial\ln p_n}{\partial k_\mu}|\psi_n\rangle\langle\psi_n|, \tag*{(277)} L C μ = n ∑ ∂ k μ ∂ ln p n ∣ ψ n ⟩ ⟨ ψ n ∣ , (277)
L Q μ = 2 i ∑ n ≠ m p m − p n p m + p n a n m μ ∣ ψ n ⟩ ⟨ ψ m ∣ . (278) \mathcal{L}^{\mu}_{Q}=2i\sum_{n\ne m}\frac{p_m-p_n}{p_m+p_n}a^\mu_{nm}|\psi_n\rangle\langle\psi_m|. \tag*{(278)} L Q μ = 2 i n = m ∑ p m + p n p m − p n a nm μ ∣ ψ n ⟩ ⟨ ψ m ∣. (278)
When the SLD L μ \mathcal{L}^{\mu} L μ commutes with the density matrix ρ \rho ρ ,
[ L μ , ρ ] = 0 , (279) [\mathcal{L}^{\mu},\rho]=0, \tag*{(279)} [ L μ , ρ ] = 0 , (279)
we have L Q μ = 0 \mathcal{L}^{\mu}_{Q}=0 L Q μ = 0 and L μ = L C μ \mathcal{L}^{\mu}=\mathcal{L}^{\mu}_{C} L μ = L C μ . Then, we obtain the classical Fisher information based on the classical part of the SLD as
1 2 Tr [ ρ { L C μ , L C ν } ] = 1 2 ∑ n = 1 N p n ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ( ∑ n ′ ∂ ln p n ′ ∂ k μ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ ∑ n ′ ′ ∂ ln p n ′ ′ ∂ k ν ∣ ψ n ′ ′ ⟩ ⟨ ψ n ′ ′ ∣ + ∑ n ′ ′ ∂ ln p n ′ ′ ∂ k ν ∣ ψ n ′ ′ ⟩ ⟨ ψ n ′ ′ ∣ ∑ n ′ ∂ ln p n ′ ∂ k μ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ ) = ∑ n , n ′ , n ′ ′ = 1 N p n ∂ ln p n ′ ∂ k μ ∂ ln p n ′ ′ ∂ k ν δ n n ′ δ n ′ n ′ ′ = ∑ n = 1 N p n ∂ ln p n ∂ k μ ∂ ln p n ∂ k ν ≡ F C F i s h e r μ ν . (280) \frac{1}{2}\operatorname{Tr}\left[\rho\{\mathcal{L}^{\mu}_{C},\mathcal{L}^{\nu}_{C}\}\right]
=\frac{1}{2}\sum_{n=1}^{N}p_n|\psi_n(\boldsymbol{k})\rangle\langle\psi_n(\boldsymbol{k})|
\left(\sum_{n'}\frac{\partial\ln p_{n'}}{\partial k_\mu}|\psi_{n'}\rangle\langle\psi_{n'}|\sum_{n''}\frac{\partial\ln p_{n''}}{\partial k_\nu}|\psi_{n''}\rangle\langle\psi_{n''}|
+\sum_{n''}\frac{\partial\ln p_{n''}}{\partial k_\nu}|\psi_{n''}\rangle\langle\psi_{n''}|\sum_{n'}\frac{\partial\ln p_{n'}}{\partial k_\mu}|\psi_{n'}\rangle\langle\psi_{n'}|\right)
=\sum_{n,n',n''=1}^{N}p_n\frac{\partial\ln p_{n'}}{\partial k_\mu}\frac{\partial\ln p_{n''}}{\partial k_\nu}\delta_{nn'}\delta_{n'n''}
=\sum_{n=1}^{N}p_n\frac{\partial\ln p_n}{\partial k_\mu}\frac{\partial\ln p_n}{\partial k_\nu}\equiv\mathcal{F}^{\mu\nu}_{\mathrm{CFisher}}. \tag*{(280)} 2 1 Tr [ ρ { L C μ , L C ν } ] = 2 1 n = 1 ∑ N p n ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ( n ′ ∑ ∂ k μ ∂ ln p n ′ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ n ′′ ∑ ∂ k ν ∂ ln p n ′′ ∣ ψ n ′′ ⟩ ⟨ ψ n ′′ ∣ + n ′′ ∑ ∂ k ν ∂ ln p n ′′ ∣ ψ n ′′ ⟩ ⟨ ψ n ′′ ∣ n ′ ∑ ∂ k μ ∂ ln p n ′ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ ) = n , n ′ , n ′′ = 1 ∑ N p n ∂ k μ ∂ ln p n ′ ∂ k ν ∂ ln p n ′′ δ n n ′ δ n ′ n ′′ = n = 1 ∑ N p n ∂ k μ ∂ ln p n ∂ k ν ∂ ln p n ≡ F CFisher μν . (280)
On the other hand, the classical part of the mean Uhlmann curvature is zero,
U ˉ μ ν ≡ − i 2 Tr [ ρ [ L C μ , L C ν ] ] = 0. (281) \bar{U}^{\mu\nu}\equiv-\frac{i}{2}\operatorname{Tr}\left[\rho[\mathcal{L}^{\mu}_{C},\mathcal{L}^{\nu}_{C}]\right]=0. \tag*{(281)} U ˉ μν ≡ − 2 i Tr [ ρ [ L C μ , L C ν ] ] = 0. (281)
Quantum Cramér-Rao inequality
There is an inequality known as the Quantum Cramér-Rao inequality [142 , 145 , 146 ] ,
F C F i s h e r ≤ F Q F i s h e r , (282) \mathcal{F}_{\mathrm{CFisher}}\leq\mathcal{F}_{\mathrm{QFisher}}, \tag*{(282)} F CFisher ≤ F QFisher , (282)
where we have defined
F C F i s h e r = ∑ μ ν a μ F C F i s h e r μ ν a ν , (283) \mathcal{F}_{\mathrm{CFisher}}=\sum_{\mu\nu}a_\mu\mathcal{F}^{\mu\nu}_{\mathrm{CFisher}}a_\nu, \tag*{(283)} F CFisher = μν ∑ a μ F CFisher μν a ν , (283)
F Q F i s h e r = ∑ μ ν a μ F Q F i s h e r μ ν a ν (284) \mathcal{F}_{\mathrm{QFisher}}=\sum_{\mu\nu}a_\mu\mathcal{F}^{\mu\nu}_{\mathrm{QFisher}}a_\nu \tag*{(284)} F QFisher = μν ∑ a μ F QFisher μν a ν (284)
for an arbitrary set of parameters a μ a_\mu a μ . The proof is shown in Appendix § 1 .2.
On the other hand, the classical Cramér-Rao inequality states that the covariance is bounded by the inverse of the classical Fisher information matrix,
∑ μ ν a μ p n ( k μ − ⟨ k μ ⟩ ) ( k ν − ⟨ k ν ⟩ ) a ν ≥ 1 N ∑ μ ν a μ ( F C F i s h e r − 1 ) μ ν a ν . (285) \sum_{\mu\nu}a_\mu p_n(k_\mu-\langle k_\mu\rangle)(k_\nu-\langle k_\nu\rangle)a_\nu\geq\frac{1}{N}\sum_{\mu\nu}a_\mu\left(\mathcal{F}^{-1}_{\mathrm{CFisher}}\right)^{\mu\nu}a_\nu. \tag*{(285)} μν ∑ a μ p n ( k μ − ⟨ k μ ⟩) ( k ν − ⟨ k ν ⟩) a ν ≥ N 1 μν ∑ a μ ( F CFisher − 1 ) μν a ν . (285)
Combining the quantum and classical Cramér-Rao inequalities, the lower bound of the covariance is determined by the inverse of the quantum Fisher information.
By inserting Eq. (266) to Eq. (274) , we obtain
d ρ = 1 2 { L , ρ } = 1 2 { L μ d k μ , ρ } , (286) d\rho=\frac{1}{2}\{\mathcal{L},\rho\}=\frac{1}{2}\{\mathcal{L}^{\mu}dk_\mu,\rho\}, \tag*{(286)} d ρ = 2 1 { L , ρ } = 2 1 { L μ d k μ , ρ } , (286)
which leads to
∂ ρ ∂ k μ = 1 2 { L μ , ρ } . (287) \frac{\partial\rho}{\partial k_\mu}=\frac{1}{2}\{\mathcal{L}^{\mu},\rho\}. \tag*{(287)} ∂ k μ ∂ ρ = 2 1 { L μ , ρ } . (287)
We show that the quantum Fisher information reproduces the quantum metric when the density matrix ρ \rho ρ describes a pure state ∣ ψ n ⟩ |\psi_n\rangle ∣ ψ n ⟩ . By using the relation ρ 2 = ρ \rho^2=\rho ρ 2 = ρ , we have
∂ ρ ∂ k μ = ∂ ρ 2 ∂ k μ = ρ ∂ ρ ∂ k μ + ∂ ρ ∂ k μ ρ = { ∂ ρ ∂ k μ , ρ } . (288) \frac{\partial\rho}{\partial k_\mu}=\frac{\partial\rho^2}{\partial k_\mu}=\rho\frac{\partial\rho}{\partial k_\mu}+\frac{\partial\rho}{\partial k_\mu}\rho=\left\{\frac{\partial\rho}{\partial k_\mu},\rho\right\}. \tag*{(288)} ∂ k μ ∂ ρ = ∂ k μ ∂ ρ 2 = ρ ∂ k μ ∂ ρ + ∂ k μ ∂ ρ ρ = { ∂ k μ ∂ ρ , ρ } . (288)
Comparing it with Eq. (287) , we have
L μ = 2 ∂ ρ ∂ k μ . (289) \mathcal{L}^{\mu}=2\frac{\partial\rho}{\partial k_\mu}. \tag*{(289)} L μ = 2 ∂ k μ ∂ ρ . (289)
Then, the quantum Fisher information is rewritten as
F Q F i s h e r μ ν = 2 Tr [ ρ { ∂ ρ ∂ k μ , ∂ ρ ∂ k ν } ] = 2 ⟨ { ∂ ρ ∂ k μ , ∂ ρ ∂ k ν } ⟩ . (290) \mathcal{F}^{\mu\nu}_{\mathrm{QFisher}}=2\operatorname{Tr}\left[\rho\left\{\frac{\partial\rho}{\partial k_\mu},\frac{\partial\rho}{\partial k_\nu}\right\}\right]=2\left\langle\left\{\frac{\partial\rho}{\partial k_\mu},\frac{\partial\rho}{\partial k_\nu}\right\}\right\rangle. \tag*{(290)} F QFisher μν = 2 Tr [ ρ { ∂ k μ ∂ ρ , ∂ k ν ∂ ρ } ] = 2 ⟨ { ∂ k μ ∂ ρ , ∂ k ν ∂ ρ } ⟩ . (290)
By inserting
ρ = ∣ ψ n ⟩ ⟨ ψ n ∣ , (291) \rho= \lvert\psi_n\rangle\langle\psi_n\rvert, \tag*{(291)} ρ = ∣ ψ n ⟩ ⟨ ψ n ∣ , (291)
the quantum Fisher information is reduced to the quantum metric
F Q F i s h e r μ ν = 2 Tr ( ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν − ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ − ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ) = 2 ( F n μ ν + F n ν μ ) = 4 g n μ ν . (292) \begin{aligned}
\mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}=2\operatorname{Tr}\left(\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}-\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\lvert\psi_n\rangle\langle\psi_n\rvert\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}+\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}-\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\lvert\psi_n\rangle\langle\psi_n\rvert\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\right) \\
&=2\left(\mathcal{F}_n^{\mu\nu}+\mathcal{F}_n^{\nu\mu}\right)=4g_n^{\mu\nu}. \tag*{(292)}
\end{aligned} F QFisher μν = 2 Tr ( ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ − ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ − ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ) = 2 ( F n μν + F n ν μ ) = 4 g n μν . (292)
while the mean Uhlmann curvature is reduced to the Berry curvature
U ‾ μ ν = − 2 i Tr ( ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν − ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν − ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ + ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ) = − 2 i ( F n μ ν + F n ν μ ) = i 4 Ω n μ ν . (293) \begin{aligned}
\overline{\mathcal{U}}^{\mu\nu}=-2i\operatorname{Tr}\left(\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}-\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\lvert\psi_n\rangle\langle\psi_n\rvert\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}-\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}+\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\lvert\psi_n\rangle\langle\psi_n\rvert\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\right) \\
&=-2i\left(\mathcal{F}_n^{\mu\nu}+\mathcal{F}_n^{\nu\mu}\right)=i4\Omega_n^{\mu\nu}. \tag*{(293)}
\end{aligned} U μν = − 2 i Tr ( ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ − ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ − ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ + ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ) = − 2 i ( F n μν + F n ν μ ) = i 4 Ω n μν . (293)
Then, the quantum Fisher information is reduced to the quantum metric, while the mean Uhlmann curvature is reduced to the Berry curvature. Thus, the Uhlmann quantum geometry is a generalization of quantum geometry. The derivation of Eq. (292) is shown in Appendix § 1 3.
Fluctuation-dissipation theorem
There is a fluctuation-dissipation theorem[143],
− 1 2 ℏ g B u r e s μ ν = tanh 2 ℏ ω 2 k B T 1 − e − ℏ ω / k B T χ μ ν ( ℏ ω ) 2 , (294) -\frac{1}{2\hbar}g_{\mathrm{Bures}}^{\mu\nu}=\frac{\tanh^2\frac{\hbar\omega}{2k_{\mathrm{B}}T}}{1-e^{-\hbar\omega/k_{\mathrm{B}}T}}\frac{\chi^{\mu\nu}}{(\hbar\omega)^2}, \tag*{(294)} − 2ℏ 1 g Bures μν = 1 − e − ℏ ω / k B T tanh 2 2 k B T ℏ ω ( ℏ ω ) 2 χ μν , (294)
where χ μ ν \chi^{\mu\nu} χ μν is the susceptibility. It is reduced to Eq. (111) for the pure state.
Quantum geometry at thermal equilibrium
At a thermal equilibrium, the probability is given by
p n = exp [ − β ε n ] , (295) p_n=\exp[-\beta\varepsilon_n], \tag*{(295)} p n = exp [ − β ε n ] , (295)
where β ≡ 1 / k B T \beta\equiv1/k_{\mathrm{B}}T β ≡ 1/ k B T is the inverse temperature. We consider a two-band system with n = ± n=\pm n = ± . The SLD operator (275) is given by
L μ = − β ∑ n = ± ∂ ε n ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ − 2 i tanh Δ ε 2 ( a + − μ ∣ ψ − ⟩ ⟨ ψ + ∣ − a − + μ ∣ ψ + ⟩ ⟨ ψ − ∣ ) (296) \mathcal{L}^{\mu}=-\beta\sum_{n=\pm}\frac{\partial\varepsilon_n}{\partial k_\mu}\lvert\psi_n\rangle\langle\psi_n\rvert-2i\tanh\frac{\Delta\varepsilon}{2}\left(a^\mu_{+-}\lvert\psi_-\rangle\langle\psi_+\rvert-a^\mu_{-+}\lvert\psi_+\rangle\langle\psi_-\rvert\right) \tag*{(296)} L μ = − β n = ± ∑ ∂ k μ ∂ ε n ∣ ψ n ⟩ ⟨ ψ n ∣ − 2 i tanh 2 Δ ε ( a +− μ ∣ ψ − ⟩ ⟨ ψ + ∣ − a −+ μ ∣ ψ + ⟩ ⟨ ψ − ∣ ) (296)
with
Δ ε = ε + − ε − , (297) \Delta\varepsilon=\varepsilon_+-\varepsilon_-, \tag*{(297)} Δ ε = ε + − ε − , (297)
where we have used
e − β ε + − e − β ε − e − β ε + + e − β ε − = − tanh β Δ ε 2 . (298) \frac{e^{-\beta\varepsilon_+}-e^{-\beta\varepsilon_-}}{e^{-\beta\varepsilon_+}+e^{-\beta\varepsilon_-}}=-\tanh\frac{\beta\Delta\varepsilon}{2}. \tag*{(298)} e − β ε + + e − β ε − e − β ε + − e − β ε − = − tanh 2 β Δ ε . (298)
It is rewritten as
L μ = Ψ † ( − β ∂ ε + ∂ k μ 2 i tanh β Δ ε 2 a + − μ − 2 i tanh β Δ ε 2 a − + μ − β ∂ ε − ∂ k μ ) Ψ (299) \mathcal{L}^{\mu}=\Psi^\dagger\begin{pmatrix}-\beta\frac{\partial\varepsilon_+}{\partial k_\mu}&2i\tanh\frac{\beta\Delta\varepsilon}{2}a^\mu_{+-}\\-2i\tanh\frac{\beta\Delta\varepsilon}{2}a^\mu_{-+}&-\beta\frac{\partial\varepsilon_-}{\partial k_\mu}\end{pmatrix}\Psi \tag*{(299)} L μ = Ψ † ( − β ∂ k μ ∂ ε + − 2 i tanh 2 β Δ ε a −+ μ 2 i tanh 2 β Δ ε a +− μ − β ∂ k μ ∂ ε − ) Ψ (299)
with
Ψ = ( ⟨ ψ + ∣ ⟨ ψ − ∣ ) . (300) \Psi=\begin{pmatrix}\langle\psi_+\rvert\\\langle\psi_-\rvert\end{pmatrix}. \tag*{(300)} Ψ = ( ⟨ ψ + ∣ ⟨ ψ − ∣ ) . (300)
The Uhlmann curvature is calculated as
U μ ν = − i 2 [ L μ , L ν ] = 2 tanh β Δ ε 2 ( U + + U + − U + − − U + + ) (301) \mathcal{U}^{\mu\nu}=-\frac{i}{2}\left[\mathcal{L}^{\mu},\mathcal{L}^{\nu}\right]=2\tanh\frac{\beta\Delta\varepsilon}{2}\begin{pmatrix}\mathcal{U}^{++}&\mathcal{U}^{+-}\\\mathcal{U}^{+-}&-\mathcal{U}^{++}\end{pmatrix} \tag*{(301)} U μν = − 2 i [ L μ , L ν ] = 2 tanh 2 β Δ ε ( U ++ U +− U +− − U ++ ) (301)
with
U + + = Ω μ ν tanh β Δ ε 2 , (302) \mathcal{U}^{++}=\Omega^{\mu\nu}\tanh\frac{\beta\Delta\varepsilon}{2}, \tag*{(302)} U ++ = Ω μν tanh 2 β Δ ε , (302)
U + − = ∂ ε + ∂ k μ a + − ν − ∂ ε + ∂ k ν a + − μ . (303) \mathcal{U}^{+-}=\frac{\partial\varepsilon_+}{\partial k_\mu}a^\nu_{+-}-\frac{\partial\varepsilon_+}{\partial k_\nu}a^\mu_{+-}. \tag*{(303)} U +− = ∂ k μ ∂ ε + a +− ν − ∂ k ν ∂ ε + a +− μ . (303)
U + + \mathcal{U}^{++} U ++ recovers the Berry curvature at the low temperature limit, β → 0 \beta\to0 β → 0 ,
U + + = − lim β → 0 2 Ω μ ν tanh 2 β Δ ε 2 = − 2 Ω μ ν . (304) \mathcal{U}^{++}=-\lim_{\beta\to0}2\Omega^{\mu\nu}\tanh^2\frac{\beta\Delta\varepsilon}{2}=-2\Omega^{\mu\nu}. \tag*{(304)} U ++ = − β → 0 lim 2 Ω μν tanh 2 2 β Δ ε = − 2 Ω μν . (304)
We define the quantum Fisher information density by
F Q F i s h e r μ ν ≡ 1 2 { L μ , L ν } , (305) \mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}\equiv\frac{1}{2}\{\mathcal{L}^{\mu},\mathcal{L}^{\nu}\}, \tag*{(305)} F QFisher μν ≡ 2 1 { L μ , L ν } , (305)
which satisfies
F Q F i s h e r μ ν ≡ Tr [ ρ F Q F i s h e r μ ν ] = ⟨ F Q F i s h e r μ ν ⟩ . (306) \mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}\equiv\operatorname{Tr}\left[\rho\mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}\right]=\langle\mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}\rangle. \tag*{(306)} F QFisher μν ≡ Tr [ ρ F QFisher μν ] = ⟨ F QFisher μν ⟩ . (306)
It is explicitly given by
F Q F i s h e r μ ν = ( ∂ ε + ∂ k μ ∂ ε + ∂ k ν + 4 g μ ν tanh 2 β Δ ε 2 ) I 2 , (307) \mathcal{F}_{\mathrm{QFisher}}^{\mu\nu}=\left(\frac{\partial\varepsilon_+}{\partial k_\mu}\frac{\partial\varepsilon_+}{\partial k_\nu}+4g^{\mu\nu}\tanh^2\frac{\beta\Delta\varepsilon}{2}\right)I_2, \tag*{(307)} F QFisher μν = ( ∂ k μ ∂ ε + ∂ k ν ∂ ε + + 4 g μν tanh 2 2 β Δ ε ) I 2 , (307)
where I 2 I_2 I 2 is the 2 by 2 identity matrix.
X-wave magnets
Fermi surface symmetry
The Fermi surface of electrons coupled with a ferromagnet is known to have the s s s -wave symmetry as shown in Figure 2 (a). Recently proposed altermagnets generalize it to Fermi surfaces possessing higher-wave symmetries[42 , 43 ] . The Fermi surface has 0, 1, 2, 3, 4 and 6 nodes for the s s s -wave, p p p -wave, d d d -wave, f f f -wave, g g g -wave and i i i -wave symmetry, respectively. The d d d -wave altermagnet has the Fermi surface with the d d d -wave symmetry as shown in Figure 2 (cFigure 1 . In the similar way, the Fermi surface of the g g g -wave altermagnet is shown in Figure 2 (eFigure 1 and that of the i i i -wave altermagnet is shown in Figure 2 (fFigure 1 . Altermagnets break time-reversal symmetry. On the other hand, the p p p -wave magnet preserves time-reversal symmetry. Its Fermi surface has the p p p -wave symmetry as shown in Figure 2 (bFigure 1 . In a similar way, the f f f -wave magnet has the Fermi surface with the f f f -wave symmetry as shown in Figure 2 (dFigure 1 . We call them the X-wave magnets with X = p , d , f , g , i X=p,d,f,g,i X = p , d , f , g , i . We note that there are no h h h -wave magnets because of the incompatibility between the five-fold rotational symmetry and the lattice symmetry.
Figure 2. Fermi surfaces in two and three dimensions. (a1), (a3) s s s -wave magnet; (b1), (b3) p p p -wave magnet; (c1), (c3) d d d -wave altermagnet; (c2) d ′ d' d ′ -wave altermagnet; (d1), (d3) f f f -wave magnet; (d2) f ′ f' f ′ -wave magnet; (e1), (e3) g g g -wave altermagnet; (e2) g ′ g' g ′ -wave altermagnet; ((f1) and (f3) i i i -wave altermagnet. Red (blue) curves indicate up (down)-spin Fermi surfaces.
The simplest expressions on magnetic terms with higher symmetries in two dimensions are summarized as follows. The X-wave magnet is characterized by a function f X 2 D ( k ) f_X^{2\mathrm{D}}(\mathbf{k}) f X 2 D ( k ) , which reads[42 , 43 , 46 , 47 , 147 ] ,
f s 2 D ( k ) = 1 , (308) f_s^{2\mathrm{D}}(\mathbf{k})=1, \tag*{(308)} f s 2 D ( k ) = 1 , (308)
f p 2 D ( k ) = a k x = a k cos ϕ , (309) f_p^{2\mathrm{D}}(\mathbf{k})=ak_x=ak\cos\phi, \tag*{(309)} f p 2 D ( k ) = a k x = ak cos ϕ , (309)
f d 2 D ( k ) = 2 a 2 k x k y = a 2 k 2 sin 2 ϕ , (310) f_d^{2\mathrm{D}}(\mathbf{k})=2a^2k_xk_y=a^2k^2\sin2\phi, \tag*{(310)} f d 2 D ( k ) = 2 a 2 k x k y = a 2 k 2 sin 2 ϕ , (310)
f f 2 D ( k ) = a 3 k x ( k x 2 − 3 k y 2 ) = a 3 k 3 cos 3 ϕ , (311) f_f^{2\mathrm{D}}(\mathbf{k})=a^3k_x\left(k_x^2-3k_y^2\right)=a^3k^3\cos3\phi, \tag*{(311)} f f 2 D ( k ) = a 3 k x ( k x 2 − 3 k y 2 ) = a 3 k 3 cos 3 ϕ , (311)
f g 2 D ( k ) = 4 a 4 k x k y ( k x 2 − k y 2 ) = a 4 k 4 sin 4 ϕ , (312) f_g^{2\mathrm{D}}(\mathbf{k})=4a^4k_xk_y\left(k_x^2-k_y^2\right)=a^4k^4\sin4\phi, \tag*{(312)} f g 2 D ( k ) = 4 a 4 k x k y ( k x 2 − k y 2 ) = a 4 k 4 sin 4 ϕ , (312)
f i 2 D ( k ) = 2 a 6 k x k y ( 3 k x 2 − k y 2 ) ( k y 2 − 3 k x 2 ) = a 6 k 6 sin 6 ϕ , (313) f_i^{2\mathrm{D}}(\mathbf{k})=2a^6k_xk_y\left(3k_x^2-k_y^2\right)\left(k_y^2-3k_x^2\right)=a^6k^6\sin6\phi, \tag*{(313)} f i 2 D ( k ) = 2 a 6 k x k y ( 3 k x 2 − k y 2 ) ( k y 2 − 3 k x 2 ) = a 6 k 6 sin 6 ϕ , (313)
where k x = k cos ϕ k_x=k\cos\phi k x = k cos ϕ , k y = k sin ϕ k_y=k\sin\phi k y = k sin ϕ . We note that the d d d -wave altermagnet described by the function f d 2 D ( k ) f_d^{2\mathrm{D}}(\mathbf{k}) f d 2 D ( k ) is commonly called the d x y d_{xy} d x y -wave altermagnet. The X-wave magnet has N X N_X N X nodes in the band structure, where N X = 1 , 2 , 3 , 4 , 6 N_X=1,2,3,4,6 N X = 1 , 2 , 3 , 4 , 6 for X = p , d , f , g , i X=p,d,f,g,i X = p , d , f , g , i , respectively.
There is another type of the X ′ X' X ′ -wave magnet in two dimensions characterized by a function f X ′ 2 D ( k ) f_{X'}^{2\mathrm{D}}(\mathbf{k}) f X ′ 2 D ( k ) such that
f p ′ 2 D ( k ) = a k y = a k sin ϕ , (314) f_{p'}^{2\mathrm{D}}(\mathbf{k})=ak_y=ak\sin\phi, \tag*{(314)} f p ′ 2 D ( k ) = a k y = ak sin ϕ , (314)
f d ′ 2 D ( k ) = a 2 ( k x 2 − k y 2 ) = a 2 k 2 cos 2 ϕ , (315) f_{d'}^{2\mathrm{D}}(\mathbf{k})=a^2\left(k_x^2-k_y^2\right)=a^2k^2\cos2\phi, \tag*{(315)} f d ′ 2 D ( k ) = a 2 ( k x 2 − k y 2 ) = a 2 k 2 cos 2 ϕ , (315)
f f ′ 2 D ( k ) = a 3 k x ( 3 k x 2 − k y 2 ) = a 3 k 3 sin 3 ϕ , (316) f_{f'}^{2\mathrm{D}}(\mathbf{k})=a^3k_x\left(3k_x^2-k_y^2\right)=a^3k^3\sin3\phi, \tag*{(316)} f f ′ 2 D ( k ) = a 3 k x ( 3 k x 2 − k y 2 ) = a 3 k 3 sin 3 ϕ , (316)
f g ′ 2 D ( k ) = a 4 ( k x 2 − k y 2 − 2 k x k y ) ( k x 2 − k y 2 + 2 k x k y ) = a 4 k 4 cos 4 ϕ , (317) \begin{aligned}
f_{g'}^{2\mathrm{D}}(\mathbf{k})&=a^4\left(k_x^2-k_y^2-2k_xk_y\right)\left(k_x^2-k_y^2+2k_xk_y\right)\\
&=a^4k^4\cos4\phi, \tag*{(317)}
\end{aligned} f g ′ 2 D ( k ) = a 4 ( k x 2 − k y 2 − 2 k x k y ) ( k x 2 − k y 2 + 2 k x k y ) = a 4 k 4 cos 4 ϕ , (317)
f i ′ 2 D ( k ) = 2 a 6 ( k x 2 − k y 2 ) ( k x 4 + k y 4 − 14 k x 2 k y 2 ) = a 6 k 6 cos 6 ϕ , (318) \begin{aligned}
f_{i'}^{2\mathrm{D}}(\mathbf{k})&=2a^6\left(k_x^2-k_y^2\right)\left(k_x^4+k_y^4-14k_x^2k_y^2\right)\\
&=a^6k^6\cos6\phi, \tag*{(318)}
\end{aligned} f i ′ 2 D ( k ) = 2 a 6 ( k x 2 − k y 2 ) ( k x 4 + k y 4 − 14 k x 2 k y 2 ) = a 6 k 6 cos 6 ϕ , (318)
where X ′ = p ′ , d ′ , f ′ , g ′ , i ′ X'=p',d',f',g',i' X ′ = p ′ , d ′ , f ′ , g ′ , i ′ . We note that the d ′ d' d ′ -wave altermagnet described by the function f d ′ 2 D ( k ) f_{d'}^{2\mathrm{D}}(\mathbf{k}) f d ′ 2 D ( k ) is commonly called the d x 2 − y 2 d_{x^2-y^2} d x 2 − y 2 -wave altermagnet, which is obtained by rotating the d x y d_{xy} d x y altermagnet by 45 degrees.
The X-wave magnet and the X ′ X' X ′ -wave magnet are summarized by the function in the form of
f X , X ′ 2 D ( k ) = ( a k ) N X sin N X ϕ , (319) f_{X,X'}^{2\mathrm{D}}(\mathbf{k})=(ak)^{N_X}\sin N_X\phi, \tag*{(319)} f X , X ′ 2 D ( k ) = ( ak ) N X sin N X ϕ , (319)
or
f X , X ′ 2 D ( k ) = ( a k ) N X cos N X ϕ . (320) f_{X,X'}^{2\mathrm{D}}(\mathbf{k})=(ak)^{N_X}\cos N_X\phi. \tag*{(320)} f X , X ′ 2 D ( k ) = ( ak ) N X cos N X ϕ . (320)
The magnetic terms of the X-wave magnet in three dimensions are given by
f d 3 D ( k ) = a 2 k z ( k x + k y ) , (321) f_d^{3\mathrm{D}}(\mathbf{k})=a^2k_z\left(k_x+k_y\right), \tag*{(321)} f d 3 D ( k ) = a 2 k z ( k x + k y ) , (321)
f f 3 D ( k ) = 2 a 3 k x k y k z = a 2 k 2 cos θ sin 2 ϕ , (322) f_f^{3\mathrm{D}}(\mathbf{k})=2a^3k_xk_yk_z=a^2k^2\cos\theta\sin2\phi, \tag*{(322)} f f 3 D ( k ) = 2 a 3 k x k y k z = a 2 k 2 cos θ sin 2 ϕ , (322)
f g 3 D ( k ) = a 4 k z k x ( k x 2 − 3 k y 2 ) = a 3 k 3 cos θ cos 3 ϕ , (323) f_g^{3\mathrm{D}}(\mathbf{k})=a^4k_zk_x\left(k_x^2-3k_y^2\right)=a^3k^3\cos\theta\cos3\phi, \tag*{(323)} f g 3 D ( k ) = a 4 k z k x ( k x 2 − 3 k y 2 ) = a 3 k 3 cos θ cos 3 ϕ , (323)
f i 3 D ( k ) = a 6 ( k x 2 − k y 2 ) ( k y 2 − k z 2 ) ( k z 2 − k x 2 ) . (324) f_i^{3\mathrm{D}}(\mathbf{k})=a^6\left(k_x^2-k_y^2\right)\left(k_y^2-k_z^2\right)\left(k_z^2-k_x^2\right). \tag*{(324)} f i 3 D ( k ) = a 6 ( k x 2 − k y 2 ) ( k y 2 − k z 2 ) ( k z 2 − k x 2 ) . (324)
There are relations between the X-wave magnet between two and three dimensions,
f f 3 D ( k ) = a k z f d 2 D ( k ) , (325) f_f^{3\mathrm{D}}(\mathbf{k})=ak_zf_d^{2\mathrm{D}}(\mathbf{k}), \tag*{(325)} f f 3 D ( k ) = a k z f d 2 D ( k ) , (325)
f g 3 D ( k ) = a k z f f 2 D ( k ) . (326) f_g^{3\mathrm{D}}(\mathbf{k})=ak_zf_f^{2\mathrm{D}}(\mathbf{k}). \tag*{(326)} f g 3 D ( k ) = a k z f f 2 D ( k ) . (326)
There are also X ′ X' X ′ -wave magnets in three dimensions[148 ] ,
f d ′ 3 D ( k ) = a 2 k y ( k x + k z ) , (327) f_{d'}^{3\mathrm{D}}(\mathbf{k})=a^2k_y\left(k_x+k_z\right), \tag*{(327)} f d ′ 3 D ( k ) = a 2 k y ( k x + k z ) , (327)
f f ′ 3 D ( k ) = a 3 k z ( k x 2 − k y 2 ) = a 2 k 2 cos θ cos 2 ϕ , (328) f_{f'}^{3\mathrm{D}}(\mathbf{k})=a^3k_z\left(k_x^2-k_y^2\right)=a^2k^2\cos\theta\cos2\phi, \tag*{(328)} f f ′ 3 D ( k ) = a 3 k z ( k x 2 − k y 2 ) = a 2 k 2 cos θ cos 2 ϕ , (328)
f g ′ 3 D ( k ) = a 4 k z k y ( 3 k x 2 − k y 2 ) = a 3 k 3 cos θ sin 3 ϕ . (329) f_{g'}^{3\mathrm{D}}(\mathbf{k})=a^4k_zk_y\left(3k_x^2-k_y^2\right)=a^3k^3\cos\theta\sin3\phi. \tag*{(329)} f g ′ 3 D ( k ) = a 4 k z k y ( 3 k x 2 − k y 2 ) = a 3 k 3 cos θ sin 3 ϕ . (329)
We set a = 1 a=1 a = 1 in the following.
Model Hamiltonian
We consider the two-band Hamiltonian described[46 ] by
H ( k ) = ℏ 2 k 2 2 m + J f X ( k ) n ⋅ σ + λ ( k x σ y − k y σ x ) + B ⋅ σ . (330) H(\mathbf{k})=\frac{\hbar^2k^2}{2m}+Jf_X(\mathbf{k})\,\mathbf{n}\cdot\boldsymbol{\sigma}+\lambda\left(k_x\sigma_y-k_y\sigma_x\right)+\mathbf{B}\cdot\boldsymbol{\sigma}. \tag*{(330)} H ( k ) = 2 m ℏ 2 k 2 + J f X ( k ) n ⋅ σ + λ ( k x σ y − k y σ x ) + B ⋅ σ . (330)
The first term represents the kinetic energy, making the system metallic. The second term represents the X-wave term with the X-wave function f X ( k ) f_X(\mathbf{k}) f X ( k ) , where n \mathbf{n} n is the direction of the spin-splitting of the band structure, and J J J is the coupling constant induced by the X-wave magnet. The third term represents the Rashba interaction introduced by making an interface between the X-wave magnet and the substrate, where λ \lambda λ is the magnitude of the Rashba interaction. The fourth term is the magnetic field term. The Rashba interaction is introduced by placing an altermagnet on the substrate[31 , 42 , 43 , 149 , 150 , 151 , 152 , 153 ] .
Symmtery
We summarize symmetry properties of the X-wave magnets[58 ] .
Spin diagonal case
We consider the Hamiltonian (330) by setting λ = 0 \lambda=0 λ = 0 , n = ( 0 , 0 , 1 ) \mathbf{n}=(0,0,1) n = ( 0 , 0 , 1 ) and B = 0 \mathbf{B}=0 B = 0 , where the spin is a good quantum number, σ z = s = ± 1 \sigma_z=s=\pm1 σ z = s = ± 1 . The Hamiltonian is diagonal with respect to the spin σ \sigma σ , where s = ± 1 s=\pm1 s = ± 1 . Let us use s = ↑ ↓ s=\uparrow\downarrow s =↑↓ within indices and s = ± 1 s=\pm1 s = ± 1 in equations.
The energy is given by
ε s = ℏ 2 k 2 2 m + s f X . (331) \varepsilon_s=\frac{\hbar^2k^2}{2m}+sf_X. \tag*{(331)} ε s = 2 m ℏ 2 k 2 + s f X . (331)
Time-reversal symmetry is defined by
T ϵ s ( k ) T − 1 = ϵ − s ( − k ) . (332) T\epsilon_s(\mathbf{k})T^{-1}=\epsilon_{-s}(-\mathbf{k}). \tag*{(332)} T ϵ s ( k ) T − 1 = ϵ − s ( − k ) . (332)
Inversion symmetry is defined by
P ϵ s ( k ) P − 1 = ϵ s ( − k ) . (333) P\epsilon_s(\mathbf{k})P^{-1}=\epsilon_s(-\mathbf{k}). \tag*{(333)} P ϵ s ( k ) P − 1 = ϵ s ( − k ) . (333)
When there are both symmetries, we have
P T ϵ s ( k ) ( P T ) − 1 = ϵ − s ( k ) . (334) PT\epsilon_s(\mathbf{k})(PT)^{-1}=\epsilon_{-s}(\mathbf{k}). \tag*{(334)} P T ϵ s ( k ) ( P T ) − 1 = ϵ − s ( k ) . (334)
Hence, the band with up and down spins are degenerate perfectly. In this sense, the breaking of time-reversal symmetry or inversion symmetry is necessary for spin-splitting band structure. In the d d d -wave, the g g g -wave and the i i i -wave altermagnets, time-reversal symmetry is broken but inversion symmetry is preserved. On the other hand, in the p p p -wave and the f f f -wave magnets, inversion symmetry is broken but time-reversal symmetry is preserved. Hence, the spin splitting occurs in all X X X -wave magnets. These properties are summarized in the following table,
Table 1.
Spin nondiagonal case
In the presence of the Rashba interaction, the system is not diagonal with respect to spin σ z \sigma_z σ z . In this case, time-reversal symmetry is defined by
T H ( k ) T − 1 = H ( − k ) (335) TH(\mathbf{k})T^{-1}=H(-\mathbf{k}) \tag*{(335)} T H ( k ) T − 1 = H ( − k ) (335)
with the time-reversal symmetry operator T = i σ y K T=i\sigma_yK T = i σ y K , where K K K is the complex conjugation operator. Thus, T T T is an antiunitary operator. Time-reversal symmetry is broken
T H ( k ) T − 1 = − H ( − k ) (336) TH(\mathbf{k})T^{-1}=-H(-\mathbf{k}) \tag*{(336)} T H ( k ) T − 1 = − H ( − k ) (336)
for the d d d -wave, the g g g -wave and the i i i -wave altermagnets. On the other hand, time-reversal symmetry is preserved for the p p p -wave and the f f f -wave magnets.
Inversion symmetry is defined by
P H ( k ) P − 1 = H ( − k ) (337) P H(\mathbf{k}) P^{-1} = H(-\mathbf{k}) \tag*{(337)} P H ( k ) P − 1 = H ( − k ) (337)
with the inversion symmetry operator P = σ z P = \sigma_z P = σ z , which is a unitary operator. Inversion symmetry is preserved for s s s -wave, d d d -wave and g g g -wave magnets. On the other hand, it is broken
P H ( k ) P − 1 = − H ( − k ) (338) P H(\mathbf{k}) P^{-1} = -H(-\mathbf{k}) \tag*{(338)} P H ( k ) P − 1 = − H ( − k ) (338)
for p p p -wave, f f f -wave and i i i -wave magnets.
PT symmetry is a combination operator of inversion symmetry and time-reversal symmetry. If there is PT symmetry
P T H ( k ) ( P T ) − 1 = H ( k ) , (339) P T H(\mathbf{k}) (P T)^{-1} = H(\mathbf{k}), \tag*{(339)} P T H ( k ) ( P T ) − 1 = H ( k ) , (339)
the bands are two-fold degenerate. All X X X -wave magnets break PT symmetry, and hence, the band structure is spin split.
The X X X -wave symmetric magnets with N X N_X N X are characterized by the spin group, which a combination group of spatial group and spin. The 2 π N X \frac{2\pi}{N_X} N X 2 π -rotation along the z z z axis of the momentum is given by
R z ( π N X ) : ( k x k y ) ↦ ( cos π N X sin π N X − sin π N X cos π N X ) ( k x k y ) , (340) R_z\left(\frac{\pi}{N_X}\right):\begin{pmatrix} k_x \\ k_y \end{pmatrix}\mapsto
\begin{pmatrix}
\cos\frac{\pi}{N_X} & \sin\frac{\pi}{N_X} \\
-\sin\frac{\pi}{N_X} & \cos\frac{\pi}{N_X}
\end{pmatrix}
\begin{pmatrix} k_x \\ k_y \end{pmatrix}, \tag*{(340)} R z ( N X π ) : ( k x k y ) ↦ ( cos N X π − sin N X π sin N X π cos N X π ) ( k x k y ) , (340)
where N X = 1 , 2 , 3 , 4 , 6 N_X = 1, 2, 3, 4, 6 N X = 1 , 2 , 3 , 4 , 6 for X = p , d , f , g , i X = p, d, f, g, i X = p , d , f , g , i , respectively. The X X X -wave symmetric magnets with N X N_X N X in two dimensions have a combinational symmetry of 2 π N X \frac{2\pi}{N_X} N X 2 π -rotation and time-reversal symmetry
[ R z ( π N X ) T ] H ( k ) [ R z ( π N X ) T ] − 1 = H ( k ) . (341) \left[R_z\left(\frac{\pi}{N_X}\right)T\right]H(\mathbf{k})\left[R_z\left(\frac{\pi}{N_X}\right)T\right]^{-1}=H(\mathbf{k}). \tag*{(341)} [ R z ( N X π ) T ] H ( k ) [ R z ( N X π ) T ] − 1 = H ( k ) . (341)
The X X X -wave symmetric magnets in three dimensions also have a combinational symmetry of spatial group and time-reversal symmetry.
Quantum geometry of X-wave magnets
We take n = ( 0 , 0 , 1 ) \mathbf{n}=(0,0,1) n = ( 0 , 0 , 1 ) and B = ( 0 , 0 , B ) \mathbf{B}=(0,0,B) B = ( 0 , 0 , B ) in the Hamiltonian Eq. (330) . The Berry curvature is given by
Ω ± x y = ∓ λ 2 ( B + J f X − J k ∂ k f X ) 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 , (342) \Omega_{\pm}^{xy}=\mp\frac{\lambda^2(B+Jf_X-Jk\partial_k f_X)}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}, \tag*{(342)} Ω ± x y = ∓ 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 λ 2 ( B + J f X − J k ∂ k f X ) , (342)
while the quantum metrices are given by
g ± x x = ∓ λ 2 ( λ 2 k y 2 + ( B + J f X ) 2 ) 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 2 , (343) g_{\pm}^{xx}=\mp\frac{\lambda^2\left(\lambda^2 k_y^2+(B+Jf_X)^2\right)}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^2}, \tag*{(343)} g ± xx = ∓ 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 2 λ 2 ( λ 2 k y 2 + ( B + J f X ) 2 ) , (343)
g ± x y = g ± y x = ∓ − λ 4 k x k y 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 2 , (344) g_{\pm}^{xy}=g_{\pm}^{yx}=\mp\frac{-\lambda^4 k_x k_y}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^2}, \tag*{(344)} g ± x y = g ± y x = ∓ 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 2 − λ 4 k x k y , (344)
g ± y y = ∓ λ 2 ( λ 2 k x 2 + ( B + J f X ) 2 ) 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 2 . (345) g_{\pm}^{yy}=\mp\frac{\lambda^2\left(\lambda^2 k_x^2+(B+Jf_X)^2\right)}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^2}. \tag*{(345)} g ± y y = ∓ 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 2 λ 2 ( λ 2 k x 2 + ( B + J f X ) 2 ) . (345)
Zeeman quantum geometry of X-wave magnets
By inserting Eq. (330) to Eq. (161) , the Zeeman Berry curvatures are calculated as
Z + − x x = ∂ n x ∂ k x = λ k y λ 2 k x + ( B + J f X ) J ∂ k x f X ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 , (346) \mathcal{Z}_{+-}^{xx}=\frac{\partial n_x}{\partial k_x}=\lambda k_y\frac{\lambda^2 k_x+(B+Jf_X)J\partial_{k_x}f_X}{\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}, \tag*{(346)} Z +− xx = ∂ k x ∂ n x = λ k y ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 λ 2 k x + ( B + J f X ) J ∂ k x f X , (346)
Z + − y y = ∂ n y ∂ k y = − λ k x λ 2 k y + ( B + J f X ) J ∂ k y f X ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 , (347) \mathcal{Z}_{+-}^{yy}=\frac{\partial n_y}{\partial k_y}=-\lambda k_x\frac{\lambda^2 k_y+(B+Jf_X)J\partial_{k_y}f_X}{\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}, \tag*{(347)} Z +− y y = ∂ k y ∂ n y = − λ k x ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 λ 2 k y + ( B + J f X ) J ∂ k y f X , (347)
Z + − x y = ∂ n y ∂ k x = λ λ 2 k 2 + ( B + J f X ) 2 − λ k x 2 λ 2 k x + 2 ( B + f X ) J ∂ k x f X 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 , (348) \mathcal{Z}_{+-}^{xy}=\frac{\partial n_y}{\partial k_x}=\frac{\lambda}{\sqrt{\lambda^2 k^2+(B+Jf_X)^2}}-\lambda k_x\frac{2\lambda^2 k_x+2(B+f_X)J\partial_{k_x}f_X}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}, \tag*{(348)} Z +− x y = ∂ k x ∂ n y = λ 2 k 2 + ( B + J f X ) 2 λ − λ k x 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 2 λ 2 k x + 2 ( B + f X ) J ∂ k x f X , (348)
Z + − y x = ∂ n x ∂ k y = − λ λ 2 k 2 + ( B + J f X ) 2 + λ k y 2 λ 2 k y + 2 ( B + f X ) J ∂ k y f X 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 , (349) \mathcal{Z}_{+-}^{yx}=\frac{\partial n_x}{\partial k_y}=-\frac{\lambda}{\sqrt{\lambda^2 k^2+(B+Jf_X)^2}}+\lambda k_y\frac{2\lambda^2 k_y+2(B+f_X)J\partial_{k_y}f_X}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}, \tag*{(349)} Z +− y x = ∂ k y ∂ n x = − λ 2 k 2 + ( B + J f X ) 2 λ + λ k y 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 2 λ 2 k y + 2 ( B + f X ) J ∂ k y f X , (349)
Z + − z x = ∂ n z ∂ k x = λ 2 − k x ( B + J f X ) + J k 2 J ∂ k x f X 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 , (350) \mathcal{Z}_{+-}^{zx}=\frac{\partial n_z}{\partial k_x}=\lambda^2\frac{-k_x(B+Jf_X)+Jk^2J\partial_{k_x}f_X}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}, \tag*{(350)} Z +− z x = ∂ k x ∂ n z = λ 2 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 − k x ( B + J f X ) + J k 2 J ∂ k x f X , (350)
Z + − z y = ∂ n z ∂ k y = λ 2 − k y ( B + J f X ) + J k 2 J ∂ k y f X 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 . (351) \mathcal{Z}_{+-}^{zy}=\frac{\partial n_z}{\partial k_y}=\lambda^2\frac{-k_y(B+Jf_X)+Jk^2J\partial_{k_y}f_X}{2\left(\lambda^2 k^2+(B+Jf_X)^2\right)^{3/2}}. \tag*{(351)} Z +− z y = ∂ k y ∂ n z = λ 2 2 ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 − k y ( B + J f X ) + J k 2 J ∂ k y f X . (351)
By inserting Eq. (330) to Eq. (162) , the Zeeman quantum metrices are calculated as
Q + − x x = n y ∂ n z ∂ k x − n z ∂ n y ∂ k x 2 = − λ ( B + J f X − J k x ∂ k x f X ) 2 ( λ 2 k 2 + ( B + J f X ) 2 ) (352) Q^{xx}_{+-}=\frac{n_y\frac{\partial n_z}{\partial k_x}-n_z\frac{\partial n_y}{\partial k_x}}{2}=-\frac{\lambda\left(B+Jf_X-Jk_x\partial_{k_x}f_X\right)}{2\left(\lambda^2k^2+\left(B+Jf_X\right)^2\right)} \tag*{(352)} Q +− xx = 2 n y ∂ k x ∂ n z − n z ∂ k x ∂ n y = − 2 ( λ 2 k 2 + ( B + J f X ) 2 ) λ ( B + J f X − J k x ∂ k x f X ) (352)
Q + − y y = n z ∂ n x ∂ k y − n x ∂ n z ∂ k y 2 = − λ ( B + J f X − J k y ∂ k y f X ) 2 ( λ 2 k 2 + ( B + J f X ) 2 ) (353) Q^{yy}_{+-}=\frac{n_z\frac{\partial n_x}{\partial k_y}-n_x\frac{\partial n_z}{\partial k_y}}{2}=-\frac{\lambda\left(B+Jf_X-Jk_y\partial_{k_y}f_X\right)}{2\left(\lambda^2k^2+\left(B+Jf_X\right)^2\right)} \tag*{(353)} Q +− y y = 2 n z ∂ k y ∂ n x − n x ∂ k y ∂ n z = − 2 ( λ 2 k 2 + ( B + J f X ) 2 ) λ ( B + J f X − J k y ∂ k y f X ) (353)
Q + − x y = n z ∂ n y ∂ k x − n x ∂ n z ∂ k x 2 = λ k y J ∂ k x f X 2 ( λ 2 k 2 + ( B + J f X ) 2 ) (354) Q^{xy}_{+-}=\frac{n_z\frac{\partial n_y}{\partial k_x}-n_x\frac{\partial n_z}{\partial k_x}}{2}=\frac{\lambda k_yJ\partial_{k_x}f_X}{2\left(\lambda^2k^2+\left(B+Jf_X\right)^2\right)} \tag*{(354)} Q +− x y = 2 n z ∂ k x ∂ n y − n x ∂ k x ∂ n z = 2 ( λ 2 k 2 + ( B + J f X ) 2 ) λ k y J ∂ k x f X (354)
Q + − y x = n y ∂ n z ∂ k y − n z ∂ n y ∂ k y 2 = λ k x J ∂ k y f X 2 ( λ 2 k 2 + ( B + J f X ) 2 ) (355) Q^{yx}_{+-}=\frac{n_y\frac{\partial n_z}{\partial k_y}-n_z\frac{\partial n_y}{\partial k_y}}{2}=\frac{\lambda k_xJ\partial_{k_y}f_X}{2\left(\lambda^2k^2+\left(B+Jf_X\right)^2\right)} \tag*{(355)} Q +− y x = 2 n y ∂ k y ∂ n z − n z ∂ k y ∂ n y = 2 ( λ 2 k 2 + ( B + J f X ) 2 ) λ k x J ∂ k y f X (355)
Q + − x z = n x ∂ n y ∂ k x − n y ∂ n x ∂ k x 2 = − λ 2 k y 2 ( λ 2 k 2 + ( B + J f X ) 2 ) (356) Q^{xz}_{+-}=\frac{n_x\frac{\partial n_y}{\partial k_x}-n_y\frac{\partial n_x}{\partial k_x}}{2}=-\frac{\lambda^2k_y}{2\left(\lambda^2k^2+\left(B+Jf_X\right)^2\right)} \tag*{(356)} Q +− x z = 2 n x ∂ k x ∂ n y − n y ∂ k x ∂ n x = − 2 ( λ 2 k 2 + ( B + J f X ) 2 ) λ 2 k y (356)
Q + − y z = n x ∂ n y ∂ k y − n y ∂ n x ∂ k y 2 = λ 2 k x 2 ( λ 2 k 2 + ( B + J f X ) 2 ) (357) Q^{yz}_{+-}=\frac{n_x\frac{\partial n_y}{\partial k_y}-n_y\frac{\partial n_x}{\partial k_y}}{2}=\frac{\lambda^2k_x}{2\left(\lambda^2k^2+\left(B+Jf_X\right)^2\right)} \tag*{(357)} Q +− y z = 2 n x ∂ k y ∂ n y − n y ∂ k y ∂ n x = 2 ( λ 2 k 2 + ( B + J f X ) 2 ) λ 2 k x (357)
The diagonal spin quantum metrices are calculated as
S + − x x = λ 2 k x 2 + ( B + J f X ) 2 λ 2 k 2 + ( B + J f X ) 2 (358) S^{xx}_{+-}=\frac{\lambda^2k_x^2+\left(B+Jf_X\right)^2}{\lambda^2k^2+\left(B+Jf_X\right)^2} \tag*{(358)} S +− xx = λ 2 k 2 + ( B + J f X ) 2 λ 2 k x 2 + ( B + J f X ) 2 (358)
S + − y y = λ 2 k y 2 + ( B + J f X ) 2 λ 2 k 2 + ( B + J f X ) 2 (359) S^{yy}_{+-}=\frac{\lambda^2k_y^2+\left(B+Jf_X\right)^2}{\lambda^2k^2+\left(B+Jf_X\right)^2} \tag*{(359)} S +− y y = λ 2 k 2 + ( B + J f X ) 2 λ 2 k y 2 + ( B + J f X ) 2 (359)
S + − z z = λ 2 k 2 λ 2 k 2 + ( B + J f X ) 2 (360) S^{zz}_{+-}=\frac{\lambda^2k^2}{\lambda^2k^2+\left(B+Jf_X\right)^2} \tag*{(360)} S +− z z = λ 2 k 2 + ( B + J f X ) 2 λ 2 k 2 (360)
The off-diagonal spin quantum metrices are calculated as
S + − x y = λ 2 k x k y λ 2 k 2 + ( B + J f X ) 2 (361) S^{xy}_{+-}=\frac{\lambda^2k_xk_y}{\sqrt{\lambda^2k^2+\left(B+Jf_X\right)^2}} \tag*{(361)} S +− x y = λ 2 k 2 + ( B + J f X ) 2 λ 2 k x k y (361)
S + − y z = − ( B + J f X ) k x λ 2 k 2 + ( B + J f X ) 2 (362) S^{yz}_{+-}=-\frac{\left(B+Jf_X\right)k_x}{\sqrt{\lambda^2k^2+\left(B+Jf_X\right)^2}} \tag*{(362)} S +− y z = − λ 2 k 2 + ( B + J f X ) 2 ( B + J f X ) k x (362)
S + − z x = ( B + f X ) k y λ 2 k 2 + ( B + J f X ) 2 (363) S^{zx}_{+-}=\frac{\left(B+f_X\right)k_y}{\sqrt{\lambda^2k^2+\left(B+Jf_X\right)^2}} \tag*{(363)} S +− z x = λ 2 k 2 + ( B + J f X ) 2 ( B + f X ) k y (363)
The spin Berry curvatures are calculated as
A + − x y = − 2 B + J f X λ 2 k 2 + ( B + J f X ) 2 (364) \mathcal{A}^{xy}_{+-}=-2\frac{B+Jf_X}{\sqrt{\lambda^2k^2+\left(B+Jf_X\right)^2}} \tag*{(364)} A +− x y = − 2 λ 2 k 2 + ( B + J f X ) 2 B + J f X (364)
A + − y z = 2 λ k y λ 2 k 2 + ( B + J f X ) 2 (365) \mathcal{A}^{yz}_{+-}=\frac{2\lambda k_y}{\sqrt{\lambda^2k^2+\left(B+Jf_X\right)^2}} \tag*{(365)} A +− y z = λ 2 k 2 + ( B + J f X ) 2 2 λ k y (365)
A + − z x = − 2 λ k x λ 2 k 2 + ( B + J f X ) 2 (366) \mathcal{A}^{zx}_{+-}=-\frac{2\lambda k_x}{\sqrt{\lambda^2k^2+\left(B+Jf_X\right)^2}} \tag*{(366)} A +− z x = − λ 2 k 2 + ( B + J f X ) 2 2 λ k x (366)
Zeeman quantum geometry induced cross response
We study the X-wave magnet coupled with the Rashba interaction without applying magnetic field[112 ] , where we set J ≠ 0 J \ne0 J = 0 and B = 0 B=0 B = 0 . It is possible to determine whether there is a response by integrating the quantum geometric tensors Z + − μ ν Z^{\mu\nu}_{+-} Z +− μν and Q + − μ ν Q^{\mu\nu}_{+-} Q +− μν over the angle ϕ \phi ϕ . Most of them vanishes by integration over the angle ϕ \phi ϕ .
First, we study the spin polarization S μ ; ν S^{\mu;\nu} S μ ; ν induced by electric field E ν E^\nu E ν . It is determined by Q + − μ ν Q^{\mu\nu}_{+-} Q +− μν as in Eq. (159) .
S x ; x / E x S^{x;x}/E^x S x ; x / E x
In the absence of the X-wave magnet, there is no response in the absence of magnetic field, while there is nonzero response in the presence of magnetic field. Only in the d x 2 − y 2 d_{x^2-y^2} d x 2 − y 2 altermagnet, there emerges the diagonal spin polarization S μ ; μ S^{\mu;\mu} S μ ; μ by applying electric field E μ E^\mu E μ with μ = x , y \mu=x,y μ = x , y in the absence of magnetic field. Once magnetic field is turned on, there is nonzero response for all X-wave magnets.
S x ; y / E y S^{x;y}/E^y S x ; y / E y
In addition, only in the d x y d_{xy} d x y altermagnet, there emerges the off-diagonal spin polarization S x ; y S^{x;y} S x ; y (S y ; x S^{y;x} S y ; x ) by applying electric field E y E^y E y (E x E^x E x ) due to the contribution from the Zeeman quantum metric Z + − x y Z^{xy}_{+-} Z +− x y (Z + − y x Z^{yx}_{+-} Z +− y x ). Diagonal spin polarization S μ S^\mu S μ is induced by electric field E μ E^\mu E μ for all X-wave and X ′ X' X ′ -wave magnets, which originate from the contribution of the Rashba interaction.
Next, we study current J μ ; ν J^{\mu;\nu} J μ ; ν induced by magnetic field B ν B^\nu B ν . It is determined by Z + − μ ν Z^{\mu\nu}_{+-} Z +− μν as in Eq. (158) .
J x ; x / B x J^{x;x}/B^x J x ; x / B x
The diagonal current J μ ; μ J^{\mu;\mu} J μ ; μ is not induced by magnetic field B μ B^\mu B μ .
J x ; y / B y J^{x;y}/B^y J x ; y / B y
On the other hand, the off-diagonal current J x ; y J^{x;y} J x ; y (J y ; x J^{y;x} J y ; x ) is induced by magnetic field B y B^y B y (B x B^x B x ) when there is nonzero Rashba interaction irrespective of the presence of the X-wave magnet.
They are summarized in the following table. Detailed derivations are shown in Appendix.H,
Table 368.
Materials
We summarize materials realizing X-wave magnets[52 , 58 ] . Altermagnets have a collinear spin texture[42 , 43 ] , while p p p -wave magnets have a spiral spin texture[44 , 154 ] .
p-wave magnet
It was theoretically proposed that CeNiAsO is a p p p -wave magnet[45 ] and experimentally realized[155 ] . They were recently realized experimentally in[156 ] Gd3 _3 3 Ru4 _4 4 Al12 _{12} 12 and in[157 ] NiI2 _2 2 . It was also theoretically proposed that a p p p -wave magnet is realized in graphene by introducing spin nematic order[158 ] .
d-wave altermagnet
The d d d -wave magnet in two dimensions was theoretically proposed in organic materials[37 ] , perovskite materials[39 ] , and twisted magnetic Van der Waals bilayers[159 ] . The d d d -wave altermagnet in three dimensions is experimentally realized in RuO2 _2 2 [160 , 161 , 162 , 163 , 164 ] , Mn5 _5 5 Si3 _3 3 [165 ] , FeSb2 _2 2 [166 ] , KV2 _2 2 Se2 _2 2 [167 ] .
f-wave magnet
It was theoretically proposed that an f f f -wave magnet is theoretically proposed in Ba3 _3 3 MnNb2 _2 2 O9 _9 9 [168 ] , FePO4 _4 4 [44 ] and in graphene by introducing spin nematic order[158 ] .
g-wave altermagnet
A g g g -wave altermagnet in two-dimensions was theoretically proposed in twisted magnetic Van der Waals bilayers[159 ] . The Fermi surface splitting of the g g g -wave altermagnet in three dimensions is experimentally observed in MnTe[169 , 170 , 171 , 172 , 173 , 174 ] , CrSb[175 , 176 , 177 , 178 , 179 ] , V1 / 3 _{1/3} 1/3 NbS2 _2 2 [180 ] and FeS[181 ] .
i-wave altermagnet
An i i i -wave altermagnet in two dimensions was theoretically proposed in twisted magnetic Van der Waals bilayers[159 ] and in MnP(S,Se)3 _3 3 [182 ] .
Transport properties of X-wave magnets
Without Rashba interaction
Spin current generation
One of the key feature of the d d d -wave altermagnet is that spin current is generated without using the Rashba interaction[37 ] . We analytically show it.
The current is given by
j = − e ∫ d k f v , (367) \mathbf{j}=-e\int d\mathbf{k}f\mathbf{v}, \tag*{(367)} j = − e ∫ d k f v , (367)
where f f f is the Fermi distribution function in the presence of E \mathbf{E} E ; v \mathbf{v} v is the velocity,
v = 1 ℏ ∂ ε ∂ k , (368) \mathbf{v}=\frac{1}{\hbar}\frac{\partial\varepsilon}{\partial\mathbf{k}}, \tag*{(368)} v = ℏ 1 ∂ k ∂ ε , (368)
where ε \varepsilon ε is the energy of the Hamiltonian, and we have dropped the anomalous velocity term proportional to − e E × Ω / ℏ -e\mathbf{E}\times\mathbf{\Omega}/\hbar − e E × Ω /ℏ because the Berry curvature Ω \mathbf{\Omega} Ω is zero due to the single band condition.
We expand the current in terms of electric field E \mathbf{E} E as
j b = ∑ ℓ 1 , ℓ 2 = 0 ∞ σ ℓ 1 y ℓ 2 ; b ( E x ) ℓ 1 ( E y ) ℓ 2 . (369) j_b=\sum_{\ell_1,\ell_2=0}^{\infty}\sigma^{\ell_1 y\ell_2;b}(E_x)^{\ell_1}(E_y)^{\ell_2}. \tag*{(369)} j b = ℓ 1 , ℓ 2 = 0 ∑ ∞ σ ℓ 1 y ℓ 2 ; b ( E x ) ℓ 1 ( E y ) ℓ 2 . (369)
Then, the ( ℓ 1 + ℓ 2 ) (\ell_1+\ell_2) ( ℓ 1 + ℓ 2 ) -th order conductivity is defined by
σ ℓ 1 y ℓ 2 ; b = 1 ℓ 1 ! ℓ 2 ! ∂ ℓ 1 + ℓ 2 j b ∂ E x ℓ 1 ∂ E y ℓ 2 . (370) \sigma^{\ell_1 y\ell_2;b}=\frac{1}{\ell_1!\ell_2!}\frac{\partial^{\ell_1+\ell_2}j_b}{\partial E_x^{\ell_1}\partial E_y^{\ell_2}}. \tag*{(370)} σ ℓ 1 y ℓ 2 ; b = ℓ 1 ! ℓ 2 ! 1 ∂ E x ℓ 1 ∂ E y ℓ 2 ∂ ℓ 1 + ℓ 2 j b . (370)
The semi-classical Boltzmann equation in the presence of electric field E \mathbf{E} E is given by
∂ t f − e E ℏ ⋅ ∇ k f = − f − f ( 0 ) τ , (371) \partial_t f-\frac{e\mathbf{E}}{\hbar}\cdot\nabla_{\mathbf{k}}f=-\frac{f-f^{(0)}}{\tau}, \tag*{(371)} ∂ t f − ℏ e E ⋅ ∇ k f = − τ f − f ( 0 ) , (371)
where τ \tau τ is the relaxation time, and f ( 0 ) f^{(0)} f ( 0 ) is the Fermi distribution function at the equilibrium with the chemical potential μ \mu μ ,
f ( 0 ) = 1 / ( exp ( ε − μ ) + 1 ) . (372) f^{(0)}=1/(\exp(\varepsilon-\mu)+1). \tag*{(372)} f ( 0 ) = 1/ ( exp ( ε − μ ) + 1 ) . (372)
Corresponding to Eq. (369) , we expand the Fermi distribution in powers of E \mathbf{E} E ,
f = f ( 0 ) + f ( 1 ) + ⋯ . (373) f=f^{(0)}+f^{(1)}+\cdots. \tag*{(373)} f = f ( 0 ) + f ( 1 ) + ⋯ . (373)
The recursive solution gives[29 ]
f ( ℓ 1 + ℓ 2 ) = ( e / ℏ i ω + 1 / τ ) ℓ 1 + ℓ 2 ∂ ℓ 1 + ℓ 2 f ( 0 ) ∂ k x ℓ 1 ∂ k y ℓ 2 ( E x ) ℓ 1 ( E x ) ℓ 2 , (374) f^{(\ell_{1}+\ell_{2})}=\left(\frac{e/\hbar}{i\omega+1/\tau}\right)^{\ell_{1}+\ell_{2}}\frac{\partial^{\ell_{1}+\ell_{2}}f^{(0)}}{\partial k_{x}^{\ell_{1}}\partial k_{y}^{\ell_{2}}}(E_x)^{\ell_{1}}(E_x)^{\ell_{2}}, \tag*{(374)} f ( ℓ 1 + ℓ 2 ) = ( iω + 1/ τ e /ℏ ) ℓ 1 + ℓ 2 ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ ℓ 1 + ℓ 2 f ( 0 ) ( E x ) ℓ 1 ( E x ) ℓ 2 , (374)
where ω \omega ω is the frequency of the applied electric field E \mathbf{E} E . The current (367) is expanded as in Eq. (373) with
j b ( ℓ 1 + ℓ 2 ) = − e ∫ d k f ( ℓ 1 + ℓ 2 ) v b = − e ℏ ( e / ℏ i ω + 1 / τ ) ℓ 1 + ℓ 2 × ∫ d k ∂ ε ∂ k b ∂ ℓ 1 + ℓ 2 f ( 0 ) ∂ k x ℓ 1 ∂ k y ℓ 2 ( E x ) ℓ 1 ( E x ) ℓ 2 . (375) \begin{aligned}
j_b^{(\ell_{1}+\ell_{2})}=-e\int\mathrm{d}k f^{(\ell_{1}+\ell_{2})}v_b \\
&=-\frac{e}{\hbar}\left(\frac{e/\hbar}{i\omega+1/\tau}\right)^{\ell_{1}+\ell_{2}}\times\int\mathrm{d}k\frac{\partial\varepsilon}{\partial k_b}\frac{\partial^{\ell_{1}+\ell_{2}}f^{(0)}}{\partial k_x^{\ell_{1}}\partial k_y^{\ell_{2}}}(E_x)^{\ell_{1}}(E_x)^{\ell_{2}}. \tag*{(375)}
\end{aligned} j b ( ℓ 1 + ℓ 2 ) = − e ∫ d k f ( ℓ 1 + ℓ 2 ) v b = − ℏ e ( iω + 1/ τ e /ℏ ) ℓ 1 + ℓ 2 × ∫ d k ∂ k b ∂ ε ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ ℓ 1 + ℓ 2 f ( 0 ) ( E x ) ℓ 1 ( E x ) ℓ 2 . (375)
The ( ℓ 1 + ℓ 2 ) (\ell_{1}+\ell_{2}) ( ℓ 1 + ℓ 2 ) -th nonlinear Drude conductivity is defined by
σ x ℓ 1 y ℓ 2 ; b = ( − e / ℏ ) ℓ 1 + ℓ 2 + 1 ( i ω + 1 / τ ) ℓ 1 + ℓ 2 ∫ d k f ( 0 ) ∂ ℓ 1 + ℓ 2 + 1 ε ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ k b . (376) \sigma^{x^{\ell_{1}}y^{\ell_{2}};b}=\frac{(-e/\hbar)^{\ell_{1}+\ell_{2}+1}}{(i\omega+1/\tau)^{\ell_{1}+\ell_{2}}}\int\mathrm{d}k f^{(0)}\frac{\partial^{\ell_{1}+\ell_{2}+1}\varepsilon}{\partial k_x^{\ell_{1}}\partial k_y^{\ell_{2}}\partial k_b}. \tag*{(376)} σ x ℓ 1 y ℓ 2 ; b = ( iω + 1/ τ ) ℓ 1 + ℓ 2 ( − e /ℏ ) ℓ 1 + ℓ 2 + 1 ∫ d k f ( 0 ) ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ k b ∂ ℓ 1 + ℓ 2 + 1 ε . (376)
The static limit is obtained simply by setting ω = 0 \omega=0 ω = 0 in this equation.
We define the ℓ \ell ℓ -th order spin-dependent Drude conductivity for each spin s s s by the formula
σ s x ℓ 1 y ℓ 2 ; b = ( − e / ℏ ) ℓ 1 + ℓ 2 + 1 ( i ω + 1 / τ ) ℓ 1 + ℓ 2 ∫ d k f s ( 0 ) ∂ ℓ 1 + ℓ 2 + 1 ε s ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ k b . (377) \sigma_s^{x^{\ell_{1}}y^{\ell_{2}};b}=\frac{(-e/\hbar)^{\ell_{1}+\ell_{2}+1}}{(i\omega+1/\tau)^{\ell_{1}+\ell_{2}}}\int\mathrm{d}k f_s^{(0)}\frac{\partial^{\ell_{1}+\ell_{2}+1}\varepsilon_s}{\partial k_x^{\ell_{1}}\partial k_y^{\ell_{2}}\partial k_b}. \tag*{(377)} σ s x ℓ 1 y ℓ 2 ; b = ( iω + 1/ τ ) ℓ 1 + ℓ 2 ( − e /ℏ ) ℓ 1 + ℓ 2 + 1 ∫ d k f s ( 0 ) ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ k b ∂ ℓ 1 + ℓ 2 + 1 ε s . (377)
This formula is nontrivial only when
∂ ℓ 1 + ℓ 2 + 1 ε s ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ k b ≠ 0 , (378) \frac{\partial^{\ell_{1}+\ell_{2}+1}\varepsilon_s}{\partial k_x^{\ell_{1}}\partial k_y^{\ell_{2}}\partial k_b}\ne0, \tag*{(378)} ∂ k x ℓ 1 ∂ k y ℓ 2 ∂ k b ∂ ℓ 1 + ℓ 2 + 1 ε s = 0 , (378)
which leads to a conclusion that there is no ℓ \ell ℓ -th order nonlinear spin-Drude conductivity for ℓ ≥ ℓ 1 + ℓ 2 \ell\ge\ell_{1}+\ell_{2} ℓ ≥ ℓ 1 + ℓ 2 . It is necessary to calculate explicitly the ℓ \ell ℓ -th order nonlinear spin-Drude conductivity for ℓ = 0 , 1 , ⋯ , ℓ 1 + ℓ 2 − 1 \ell=0,1,\cdots,\ell_{1}+\ell_{2}-1 ℓ = 0 , 1 , ⋯ , ℓ 1 + ℓ 2 − 1 . In particular, the choice of ℓ = 0 \ell=0 ℓ = 0 and 1 1 1 yield the persistent spin current without electric field and the linear spin conductivity, respectively.
We define the ( ℓ 1 + ℓ 2 ) (\ell_{1}+\ell_{2}) ( ℓ 1 + ℓ 2 ) -th order nonlinear spin-Drude conductivity by
σ s p i n x ℓ 1 y ℓ 2 ; b = σ ↑ x ℓ 1 y ℓ 2 ; b − σ ↓ x ℓ 1 y ℓ 2 ; b 2 . (379) \sigma_{\mathrm{spin}}^{x^{\ell_{1}}y^{\ell_{2}};b}=\frac{\sigma_{\uparrow}^{x^{\ell_{1}}y^{\ell_{2}};b}-\sigma_{\downarrow}^{x^{\ell_{1}}y^{\ell_{2}};b}}{2}. \tag*{(379)} σ spin x ℓ 1 y ℓ 2 ; b = 2 σ ↑ x ℓ 1 y ℓ 2 ; b − σ ↓ x ℓ 1 y ℓ 2 ; b . (379)
On the other hand, the charge conductivity is given by
σ c h a r g e x ℓ 1 y ℓ 2 ; b = σ ↑ x ℓ 1 y ℓ 2 ; b + σ ↓ x ℓ 1 y ℓ 2 ; b . (380) \sigma_{\mathrm{charge}}^{x^{\ell_{1}}y^{\ell_{2}};b}=\sigma_{\uparrow}^{x^{\ell_{1}}y^{\ell_{2}};b}+\sigma_{\downarrow}^{x^{\ell_{1}}y^{\ell_{2}};b}. \tag*{(380)} σ charge x ℓ 1 y ℓ 2 ; b = σ ↑ x ℓ 1 y ℓ 2 ; b + σ ↓ x ℓ 1 y ℓ 2 ; b . (380)
Spin current is generated in linear response only for the d d d -wave altermagnet. The second-order nonlinear spin current is generated in the f f f -wave magnet, the third-order nonlinear spin current is generated in the g g g -wave altermagnet and the fifth-order nonlinear spin current is generated in the i i i -wave altermagnet[147 ] . The results are summarized in the following table.
Table 383.
Spin Nernst effects
The spin Nernst effect is an effect that spin current is generated perpendicularly to the thermal gradient. It is discussed in the d d d -wave altermagnet[37 ] . So far, there is no analytic result on it, which we derive.
Tts momentum derivative reads
∂ f ∂ k = ∂ ε ( k ) ∂ k ∂ f ∂ ε , (381) \frac{\partial f}{\partial\mathbf{k}}=\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}\frac{\partial f}{\partial\varepsilon}, \tag*{(381)} ∂ k ∂ f = ∂ k ∂ ε ( k ) ∂ ε ∂ f , (381)
while its spatial derivative reads
∂ f ∂ r = ∂ T ( r ) ∂ r ∂ f ∂ T . (382) \frac{\partial f}{\partial\mathbf{r}}=\frac{\partial T(\mathbf{r})}{\partial\mathbf{r}}\frac{\partial f}{\partial T}. \tag*{(382)} ∂ r ∂ f = ∂ r ∂ T ( r ) ∂ T ∂ f . (382)
We rewrite the Boltzmann equation as
d f d t = ∂ f ∂ t + ∂ k ∂ t ⋅ ∂ f ∂ k + ∂ r ∂ t ⋅ ∂ f ∂ r = ∂ k ∂ t ⋅ ( ∂ ε ( k ) ∂ k ∂ f ∂ ε ) + ∂ r ∂ t ⋅ ( ∂ T ( r ) ∂ r ∂ f ∂ T ) = − f − f ( 0 ) τ . (383) \begin{aligned}
\frac{\mathrm{d}f}{\mathrm{d}t}=\frac{\partial f}{\partial t}+\frac{\partial\mathbf{k}}{\partial t}\cdot\frac{\partial f}{\partial\mathbf{k}}+\frac{\partial\mathbf{r}}{\partial t}\cdot\frac{\partial f}{\partial\mathbf{r}} \\
&=\frac{\partial\mathbf{k}}{\partial t}\cdot\left(\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}\frac{\partial f}{\partial\varepsilon}\right)+\frac{\partial\mathbf{r}}{\partial t}\cdot\left(\frac{\partial T(\mathbf{r})}{\partial\mathbf{r}}\frac{\partial f}{\partial T}\right) \\
&=-\frac{f-f^{(0)}}{\tau}. \tag*{(383)}
\end{aligned} d t d f = ∂ t ∂ f + ∂ t ∂ k ⋅ ∂ k ∂ f + ∂ t ∂ r ⋅ ∂ r ∂ f = ∂ t ∂ k ⋅ ( ∂ k ∂ ε ( k ) ∂ ε ∂ f ) + ∂ t ∂ r ⋅ ( ∂ r ∂ T ( r ) ∂ T ∂ f ) = − τ f − f ( 0 ) . (383)
We use the kinetic equation
∂ r ∂ t = 1 ℏ ∂ ε ( k ) ∂ k , (384) \frac{\partial\mathbf{r}}{\partial t}=\frac{1}{\hbar}\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}, \tag*{(384)} ∂ t ∂ r = ℏ 1 ∂ k ∂ ε ( k ) , (384)
∂ k ∂ t = 0 , (385) \frac{\partial\mathbf{k}}{\partial t}=0, \tag*{(385)} ∂ t ∂ k = 0 , (385)
and obtain
1 ℏ ( ∂ ε ( k ) ∂ k ⋅ ∂ T ( r ) ∂ r ) ∂ f ∂ T = − f − f ( 0 ) τ . (386) \frac{1}{\hbar}\left(\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}\cdot\frac{\partial T(\mathbf{r})}{\partial\mathbf{r}}\right)\frac{\partial f}{\partial T}=-\frac{f-f^{(0)}}{\tau}. \tag*{(386)} ℏ 1 ( ∂ k ∂ ε ( k ) ⋅ ∂ r ∂ T ( r ) ) ∂ T ∂ f = − τ f − f ( 0 ) . (386)
The first order expansion reads
1 ℏ ( ∂ ε ( k ) ∂ k ⋅ ∂ T ( r ) ∂ r ) ∂ f ( 0 ) ∂ T = − f ( 1 ) τ , (387) \frac{1}{\hbar}\left(\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}\cdot\frac{\partial T(\mathbf{r})}{\partial\mathbf{r}}\right)\frac{\partial f^{(0)}}{\partial T}=-\frac{f^{(1)}}{\tau}, \tag*{(387)} ℏ 1 ( ∂ k ∂ ε ( k ) ⋅ ∂ r ∂ T ( r ) ) ∂ T ∂ f ( 0 ) = − τ f ( 1 ) , (387)
whose solution is given by
f ( 1 ) = − τ ℏ ( ∂ ε ( k ) ∂ k ⋅ ∂ T ( r ) ∂ r ) ∂ f ( 0 ) ∂ T . (388) f^{(1)}=-\frac{\tau}{\hbar}\left(\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}\cdot\frac{\partial T(\mathbf{r})}{\partial\mathbf{r}}\right)\frac{\partial f^{(0)}}{\partial T}. \tag*{(388)} f ( 1 ) = − ℏ τ ( ∂ k ∂ ε ( k ) ⋅ ∂ r ∂ T ( r ) ) ∂ T ∂ f ( 0 ) . (388)
The current driven by temperature gradient is given by
j b = − e ∫ d k f ( 1 ) v b = e τ ℏ 2 ∫ d k ( ∂ ε ( k ) ∂ k ⋅ ∂ T ( r ) ∂ r ) ∂ ε ∂ k b ∂ f ( 0 ) ∂ T . (389) \begin{aligned}
j_b=-e\int\mathrm{d}k f^{(1)}v_b \\
&=\frac{e\tau}{\hbar^2}\int\mathrm{d}k\left(\frac{\partial\varepsilon(\mathbf{k})}{\partial\mathbf{k}}\cdot\frac{\partial T(\mathbf{r})}{\partial\mathbf{r}}\right)\frac{\partial\varepsilon}{\partial k_b}\frac{\partial f^{(0)}}{\partial T}. \tag*{(389)}
\end{aligned} j b = − e ∫ d k f ( 1 ) v b = ℏ 2 e τ ∫ d k ( ∂ k ∂ ε ( k ) ⋅ ∂ r ∂ T ( r ) ) ∂ k b ∂ ε ∂ T ∂ f ( 0 ) . (389)
Especially, current j x j_x j x and j y j_y j y flow when the temperature gradient is along the x x x axis,
j x = e τ ∫ d k ( 1 ℏ ∂ ε ∂ k x ) 2 ∂ T ( r ) ∂ x ∂ f ( 0 ) ∂ T (390) j_x=e\tau\int\mathrm{d}k\left(\frac{1}{\hbar}\frac{\partial\varepsilon}{\partial k_x}\right)^2\frac{\partial T(\mathbf{r})}{\partial x}\frac{\partial f^{(0)}}{\partial T} \tag*{(390)} j x = e τ ∫ d k ( ℏ 1 ∂ k x ∂ ε ) 2 ∂ x ∂ T ( r ) ∂ T ∂ f ( 0 ) (390)
which is known as the Seebeck effect, and
j y = − e ∫ d k f ( 1 ) v y = e τ ℏ 2 ∫ d k ∂ ε ∂ k x ⋅ ∂ T ( r ) ∂ x ∂ ε ∂ k y ∂ f ( 0 ) ∂ T , (391) \begin{aligned}
j_y&=-e\int\mathrm{d}k f^{(1)}v_y\\
&=\frac{e\tau}{\hbar^2}\int\mathrm{d}k\frac{\partial\varepsilon}{\partial k_x}\cdot\frac{\partial T(\mathbf{r})}{\partial x}\frac{\partial\varepsilon}{\partial k_y}\frac{\partial f^{(0)}}{\partial T}, \tag*{(391)}
\end{aligned} j y = − e ∫ d k f ( 1 ) v y = ℏ 2 e τ ∫ d k ∂ k x ∂ ε ⋅ ∂ x ∂ T ( r ) ∂ k y ∂ ε ∂ T ∂ f ( 0 ) , (391)
which is known as the Nernst effect. We assume that the temperature gradient is linear,
T ( r ) = a x . (392) T(\mathbf{r})=ax. \tag*{(392)} T ( r ) = a x . (392)
When ε ( k ) − μ ≫ k B T ( r ) \varepsilon(\mathbf{k})-\mu\gg k_{\mathrm{B}}T(\mathbf{r}) ε ( k ) − μ ≫ k B T ( r ) , we approximate the Fermi distribution by the Boltzmann distribution,
f = 1 exp ( ε ( k ) − μ k B T ( r ) ) + 1 ≃ exp ( − ε ( k ) − μ k B T ( r ) ) . (393) f=\frac{1}{\exp\left(\frac{\varepsilon(\mathbf{k})-\mu}{k_{\mathrm{B}}T(\mathbf{r})}\right)+1}\simeq\exp\left(-\frac{\varepsilon(\mathbf{k})-\mu}{k_{\mathrm{B}}T(\mathbf{r})}\right). \tag*{(393)} f = exp ( k B T ( r ) ε ( k ) − μ ) + 1 1 ≃ exp ( − k B T ( r ) ε ( k ) − μ ) . (393)
Then, the spin Hall conductivity is obtained as
j s = e τ ℏ 2 8 e μ / k B T J m π ( 2 T − μ ) 1 − 4 J 2 m 2 (394) j_s=\frac{e\tau}{\hbar^2}\frac{8e^{\mu/k_{\mathrm{B}}T}Jm\pi(2T-\mu)}{\sqrt{1-4J^2m^2}} \tag*{(394)} j s = ℏ 2 e τ 1 − 4 J 2 m 2 8 e μ / k B T J mπ ( 2 T − μ ) (394)
for the d d d -wave altermagnet. It is proportional to the coupling constant J J J of the X X X -wave magnet in the Hamiltonian (330) . Hence, it is possible to detect the sign of J J J by the spin Nernst effect. On the other hand, there is no spin current generation for the other X X X -wave magnets.
Tunneling magnetoresistance
Magnetic tunneling junction is a most successful spintronic device[183 , 184 ] . It consists of the bilayer ferromagnets spaced by an insulator as shown in Figure 3 , where the resistance is low (large) if the directions of spins are identical (opposite). It is called the tunneling magnetoresistance (TMR).
Figure 3. Illustration of a bilayer magnetic tunneling junction made of magnets. (a) Parallel configuration, where the spin directions are identical at each lattice site between the two layers. (b) Antiparallel configuration, where the spin directions are opposite at each lattice site between the two layers. The green arrow indicates the tunneling current, which is larger in the parallel configuration.
The spin direction of the memory can be readout by using the TMR. Tunneling magnetoresistance is discussed in the d d d -wave altermagnet[43 , 185 , 186 , 187 , 188 , 189 , 190 , 191 ] and the p p p -wave magnet[192 ] .
The differential conductance G = d I / d V G=\mathrm{d}I/\mathrm{d}V G = d I / d V is calculated based on the Green function[192 ] ,
G 4 e π 3 = ∑ s = ± 1 ∑ k 1 , k 2 ∣ T k 1 , k 2 ∣ 2 Tr [ Im G s T ( 0 ; k 1 ) Im G s B ( 0 ; k 2 ) ] (395) \frac{G}{4e\pi^3}=\sum_{s=\pm1}\sum_{\mathbf{k}_1,\mathbf{k}_2}\lvert T_{\mathbf{k}_1,\mathbf{k}_2}\rvert^2\operatorname{Tr}\left[\operatorname{Im}G_s^{\mathrm{T}}(0;\mathbf{k}_1)\operatorname{Im}G_s^{\mathrm{B}}(0;\mathbf{k}_2)\right] \tag*{(395)} 4 e π 3 G = s = ± 1 ∑ k 1 , k 2 ∑ ∣ T k 1 , k 2 ∣ 2 Tr [ Im G s T ( 0 ; k 1 ) Im G s B ( 0 ; k 2 ) ] (395)
where G s T G_s^{\mathrm{T}} G s T (G s B G_s^{\mathrm{B}} G s B ) is the retarded Green function of the top (bottom) layer defined by
G s T ( ω ; k ) ≡ G s B ( ω ; k ) ≡ G s ( ω ; k ) (396) G_s^{\mathrm{T}}(\omega;\mathbf{k})\equiv G_s^{\mathrm{B}}(\omega;\mathbf{k})\equiv G_s(\omega;\mathbf{k}) \tag*{(396)} G s T ( ω ; k ) ≡ G s B ( ω ; k ) ≡ G s ( ω ; k ) (396)
for the parallel configuration, and
G s T ( ω ; k ) ≡ G s ( ω ; k ) , G s B ( ω ; k ) ≡ G − s ( ω ; k ) (397) G_s^{\mathrm{T}}(\omega;\mathbf{k})\equiv G_s(\omega;\mathbf{k}),\qquad G_s^{\mathrm{B}}(\omega;\mathbf{k})\equiv G_{-s}(\omega;\mathbf{k}) \tag*{(397)} G s T ( ω ; k ) ≡ G s ( ω ; k ) , G s B ( ω ; k ) ≡ G − s ( ω ; k ) (397)
for the antiparallel configuration, where we have defined
G s ( ω ; k ) ≡ 1 ℏ ω − ε s + i Γ (398) G_s(\omega;\mathbf{k})\equiv\frac{1}{\hbar\omega-\varepsilon_s+\mathrm{i}\Gamma} \tag*{(398)} G s ( ω ; k ) ≡ ℏ ω − ε s + i Γ 1 (398)
with the self-energy Γ \Gamma Γ and the energy ε s \varepsilon_s ε s . Im G s ( 0 ; k 1 ) \operatorname{Im}G_s(0;\mathbf{k}_1) Im G s ( 0 ; k 1 ) represents the density of states depending on the spin s s s , which is shown in Figure 4 (a) and (b).
Figure 4. (a) Spin density ⟨ S z ⟩ \langle S_z\rangle ⟨ S z ⟩ when J > 0 J>0 J > 0 , and (b) that when J < 0 J<0 J < 0 in one layer. Red (blue) color indicates up (down) spin. (c) Overlap O P \mathcal{O}_{\mathrm{P}} O P for the parallel configuration, and (d) overlap O A P \mathcal{O}_{\mathrm{AP}} O AP for the antiparallel configuration. Red color indicates a large overlap.
The differential conductivity is determined by the overlap of the density of states Im G s ( 0 ; k 1 ) \operatorname{Im}G_s(0;\mathbf{k}_1) Im G s ( 0 ; k 1 ) , which is shown in Figure 4 (c) and (d).
The differential conductivity for the parallel configuration is analytically given by[47 ]
lim Γ → 0 G P 4 e π 3 = 2 m π 2 ℏ 2 Γ (399) \lim_{\Gamma\to0}\frac{G_{\mathrm{P}}}{4e\pi^3}=\frac{2m\pi^2}{\hbar^2\Gamma} \tag*{(399)} Γ → 0 lim 4 e π 3 G P = ℏ 2 Γ 2 m π 2 (399)
while that for the antiparallel configuration is analytically given by
lim Γ → 0 G A P 4 e π 3 = m 2 μ N X π 2 ℏ ∣ J ∣ (400) \lim_{\Gamma\to0}\frac{G_{\mathrm{AP}}}{4e\pi^3}=\sqrt{\frac{m}{2\mu}}\frac{N_X\pi^2}{\hbar\lvert J\rvert} \tag*{(400)} Γ → 0 lim 4 e π 3 G AP = 2 μ m ℏ ∣ J ∣ N X π 2 (400)
where J J J is the strength of the X X X -wave magnet, Γ \Gamma Γ is the self-energy, μ \mu μ is the chemical potential, a a a is the lattice constant and m m m is the mass of electrons. Hence, the TMR ratio is given by
lim Γ → 0 T M R r a t i o ≡ lim Γ → 0 G P − G A P G A P = 2 a ∣ J ∣ 2 m μ N X ℏ Γ . (401) \lim_{\Gamma\to0}\mathrm{TMR}_{\mathrm{ratio}}\equiv\lim_{\Gamma\to0}\frac{G_{\mathrm{P}}-G_{\mathrm{AP}}}{G_{\mathrm{AP}}}=\frac{2a\lvert J\rvert\sqrt{2m\mu}}{N_X\hbar\Gamma}. \tag*{(401)} Γ → 0 lim TMR ratio ≡ Γ → 0 lim G AP G P − G AP = N X ℏΓ 2 a ∣ J ∣ 2 m μ . (401)
It is to be contrasted with the TMR ratio based on ferromagnets, where it is given by
lim Γ → 0 T M R r a t i o = 2 J 2 Γ 2 . (402) \lim_{\Gamma\to0}\mathrm{TMR}_{\mathrm{ratio}}=\frac{2J^2}{\Gamma^2}. \tag*{(402)} Γ → 0 lim TMR ratio = Γ 2 2 J 2 . (402)
Therefore, the TMR ratio is larger in ferromagnets for ∣ J ∣ > Γ \lvert J\rvert>\Gamma ∣ J ∣ > Γ . However, the X X X -wave magnets are expected to achieve high-speed and ultra-dense memory owing to the zero net magnetization.
With Rashba interaction
Anomalous Hall effects
One of the motivation of studying altermagnets is that the Néel vector can be readout by measuring anomalous Hall conductivity because time-reversal symmetry is broken[162 , 165 , 193 , 194 ] . Actually, it is verified by the density-functional theory and experiments.
The Berry curvature is obtained as
Ω X = − λ 2 ( B + J f X − J k ∂ k f X ) 2 ( ( B + J f X ) 2 + λ 2 k 2 ) 3 / 2 (403) \Omega_{\mathrm{X}}=-\frac{\lambda^{2}\left(B+Jf_{\mathrm{X}}-Jk\partial_k f_{\mathrm{X}}\right)}{2\left((B+Jf_{\mathrm{X}})^{2}+\lambda^{2}k^{2}\right)^{3/2}} \tag*{(403)} Ω X = − 2 ( ( B + J f X ) 2 + λ 2 k 2 ) 3/2 λ 2 ( B + J f X − J k ∂ k f X ) (403)
The Hall conductivity is calculated as
1 2 π ∫ Ω X d k = − 1 2 sgn B . (404) \frac{1}{2\pi}\int\Omega_{\mathrm{X}}\,\mathrm{d}k=-\frac{1}{2}\operatorname{sgn}B. \tag*{(404)} 2 π 1 ∫ Ω X d k = − 2 1 sgn B . (404)
It does not depend on J J J . Hence, it is impossible to detect the sign of J J J by measuring the anomalous Hall conductivity based on the two-band model. It is necessary to introduce the orbital degrees of freedom to the two-band model.
Planar Hall effects
The Hall effect is a prominent phenomenon in two-dimensional materials. A current flows into a direction perpendicular to the applied electric field, when the magnetic field is applied perpendicular to the plane, as illustrated in Figure 5 (a). Similarly, a current flows into a direction perpendicular to the applied electric field, when the magnetic field is applied parallel to the plane, as illustrated in Figure 5 (b). It is the planar Hall effect[195 , 196 , 197 , 198 , 199 , 200 , 201 ] . The planar Hall effect is discussed in d d d -wave altermagnets[202 , 203 ] .
Figure 5. Illustration for (a) Hall effect and (b) planar Hall effect. In the planar Hall effect, when the electric field is applied along the x x x axis and the magnetic field ( B cos Φ , B sin Φ , 0 ) (B\cos\Phi,B\sin\Phi,0) ( B cos Φ , B sin Φ , 0 ) is applied parallel to the system, the Hall current flows along the y y y axis. The Hall conductivity is predicted to be given by the formula (408) .
The planar Hall effect is discussed in the X X X -wave magnet[46 ] . The Dirac point shifts in the presence of the in-plane magnetic field. The shifted Dirac point is given by ( k x ′ , k y ′ ) = 0 (k'_x,k'_y)=0 ( k x ′ , k y ′ ) = 0 , where
k x ′ = k x + B y λ , k y ′ = k y − B x λ , (405) k'_x = k_x + \frac{B_y}{\lambda}, \qquad k'_y = k_y - \frac{B_x}{\lambda}, \tag*{(405)} k x ′ = k x + λ B y , k y ′ = k y − λ B x , (405)
by solving the equation
λ ( k x σ y − k y σ x ) + B x σ x + B y σ y = 0. (406) \lambda(k_x\sigma_y-k_y\sigma_x)+B_x\sigma_x+B_y\sigma_y=0. \tag*{(406)} λ ( k x σ y − k y σ x ) + B x σ x + B y σ y = 0. (406)
The Dirac gap is explicitly determined as
Δ = ( − 1 ) s X B N X λ N X sin N X Φ , (407) \Delta=(-1)^{s_X}\frac{B^{N_X}}{\lambda^{N_X}}\sin N_X\Phi, \tag*{(407)} Δ = ( − 1 ) s X λ N X B N X sin N X Φ , (407)
which leads to the Hall conductance
σ x y = e 2 h ( − 1 ) s X 2 sgn ( J B N X λ N X sin N X Φ ) , (408) \sigma_{xy}=\frac{e^2}{h}\frac{(-1)^{s_X}}{2}\operatorname{sgn}\left(J\frac{B^{N_X}}{\lambda^{N_X}}\sin N_X\Phi\right), \tag*{(408)} σ x y = h e 2 2 ( − 1 ) s X sgn ( J λ N X B N X sin N X Φ ) , (408)
where s X = 1 , 1 , − 1 , − 1 , 1 s_X=1,1,-1,-1,1 s X = 1 , 1 , − 1 , − 1 , 1 for X = p , d , f , g , i X=p,d,f,g,i X = p , d , f , g , i , respectively.
Quantum Hall effects
Landau levels
If we apply magnetic field perpendicular to the sample, Landau levels are formed. Landau levels are obtained for the d d d -wave altermagnet[204 , 205 ] . We apply a homogeneous magnetic field B = ∇ × A = ( 0 , 0 , − B ) \mathbf{B}=\nabla\times\mathbf{A}=(0,0,-B) B = ∇ × A = ( 0 , 0 , − B ) with B > 0 B>0 B > 0 along the z z z axis to the sample. The Hamiltonian under magnetic field is obtained by replacing the momentum k i k_i k i to the covariant momentum P i ≡ k i + e A i P_i\equiv k_i+eA_i P i ≡ k i + e A i . We introduce a pair of Landau-level ladder operators,
a ^ = ℓ B ( P x + i P y ) 2 ℏ , a ^ † = ℓ B ( P x − i P y ) 2 ℏ , (409) \hat{a}=\frac{\ell_B(P_x+iP_y)}{\sqrt{2}\hbar}, \qquad\hat{a}^{\dagger}=\frac{\ell_B(P_x-iP_y)}{\sqrt{2}\hbar}, \tag*{(409)} a ^ = 2 ℏ ℓ B ( P x + i P y ) , a ^ † = 2 ℏ ℓ B ( P x − i P y ) , (409)
satisfying the bosonic commutation relation [ a ^ , a ^ † ] = 1 [\hat{a},\hat{a}^{\dagger}]=1 [ a ^ , a ^ † ] = 1 , where ℓ B = ℏ / e B \ell_B=\sqrt{\hbar/eB} ℓ B = ℏ/ e B is the magnetic length. The inverse relations read
P x = ℏ 2 ℓ B ( a ^ + a ^ † ) , P y = ℏ i 2 ℓ B ( a ^ − a ^ † ) . (410) P_x=\frac{\hbar}{\sqrt{2}\ell_B}(\hat{a}+\hat{a}^{\dagger}), \qquad P_y=\frac{\hbar}{i\sqrt{2}\ell_B}(\hat{a}-\hat{a}^{\dagger}). \tag*{(410)} P x = 2 ℓ B ℏ ( a ^ + a ^ † ) , P y = i 2 ℓ B ℏ ( a ^ − a ^ † ) . (410)
The Hamiltonian for free electrons reads
H 0 = ℏ ω 0 a ^ † a ^ . (411) H_0=\hbar\omega_0\hat{a}^{\dagger}\hat{a}. \tag*{(411)} H 0 = ℏ ω 0 a ^ † a ^ . (411)
p-wave magnets and coherent states
By inserting Eq. (410) to Eq. (309) , the p p p -wave term under magnetic field reads
f p 2 D ( k ) = a ℏ 2 ℓ B ( a ^ + a ^ † ) . (412) f_p^{2\mathrm{D}}(\mathbf{k})=\frac{a\hbar}{\sqrt{2}\ell_B}(\hat{a}+\hat{a}^{\dagger}). \tag*{(412)} f p 2 D ( k ) = 2 ℓ B a ℏ ( a ^ + a ^ † ) . (412)
The Hamiltonian is identical to that of the coherent state
H ^ = ℏ ω 0 a ^ † a ^ + c ∗ a ^ + c a ^ † (413) \hat{H}=\hbar\omega_0\hat{a}^{\dagger}\hat{a}+c^*\hat{a}+c\hat{a}^{\dagger} \tag*{(413)} H ^ = ℏ ω 0 a ^ † a ^ + c ∗ a ^ + c a ^ † (413)
provided
c = c ∗ = a ℏ 2 ℓ B . (414) c=c^*=\frac{a\hbar}{\sqrt{2}\ell_B}. \tag*{(414)} c = c ∗ = 2 ℓ B a ℏ . (414)
It is diagonalized as
H ^ = ℏ ω b ^ † b ^ − c 2 ℏ ω , (415) \hat{H}=\hbar\omega\hat{b}^{\dagger}\hat{b}-\frac{c^2}{\hbar\omega}, \tag*{(415)} H ^ = ℏ ω b ^ † b ^ − ℏ ω c 2 , (415)
where we have introduced displaced operators
b ^ ≡ a ^ + c ℏ ω , b ^ † ≡ a ^ † + c ℏ ω . (416) \hat{b}\equiv\hat{a}+\frac{c}{\hbar\omega},\qquad\hat{b}^{\dagger}\equiv\hat{a}^{\dagger}+\frac{c}{\hbar\omega}. \tag*{(416)} b ^ ≡ a ^ + ℏ ω c , b ^ † ≡ a ^ † + ℏ ω c . (416)
The energy of the landau level is obtained as
H ^ = ℏ ω N − a 2 ℏ 2 ℓ B 2 ω , (417) \hat{H}=\hbar\omega N-\frac{a^2\hbar}{2\ell_B^2\omega}, \tag*{(417)} H ^ = ℏ ω N − 2 ℓ B 2 ω a 2 ℏ , (417)
while the eigenstate is determined as
∣ N ⟩ b ≡ 1 N ! ( b ^ † ) N ∣ 0 ⟩ . (418) |N\rangle_b\equiv\frac{1}{\sqrt{N!}}(\hat{b}^{\dagger})^N|0\rangle. \tag*{(418)} ∣ N ⟩ b ≡ N ! 1 ( b ^ † ) N ∣0 ⟩ . (418)
d-wave altermagnets and squeezed states
By inserting Eq. (410) to Eq. (310) , the d d d -wave term under magnetic field reads
f d 2 D ( k ) = ℏ 2 a 2 2 ℓ B 2 ( ( a ^ + a ^ † ) 2 + ( a ^ − a ^ † ) 2 ) (419) f_d^{2\mathrm{D}}(\mathbf{k})=\frac{\hbar^2a^2}{2\ell_B^2}\left((\hat{a}+\hat{a}^{\dagger})^2+(\hat{a}-\hat{a}^{\dagger})^2\right) \tag*{(419)} f d 2 D ( k ) = 2 ℓ B 2 ℏ 2 a 2 ( ( a ^ + a ^ † ) 2 + ( a ^ − a ^ † ) 2 ) (419)
= ℏ 2 a 2 ℓ B 2 ( ( a ^ † ) 2 + a ^ 2 ) . (420) =\frac{\hbar^2a^2}{\ell_B^2}\left((\hat{a}^{\dagger})^2+\hat{a}^2\right). \tag*{(420)} = ℓ B 2 ℏ 2 a 2 ( ( a ^ † ) 2 + a ^ 2 ) . (420)
The Hamiltonian is identical to that of the squeezed state,
H ^ ≡ ℏ ω 0 + c 2 ( a ^ † ) 2 + c ∗ 2 a ^ 2 (421) \hat{H}\equiv\hbar\omega_0+\frac{c}{2}(\hat{a}^{\dagger})^2+\frac{c^*}{2}\hat{a}^2 \tag*{(421)} H ^ ≡ ℏ ω 0 + 2 c ( a ^ † ) 2 + 2 c ∗ a ^ 2 (421)
provided
c = c ∗ = 2 ℏ 2 a 2 ℓ B 2 . (422) c=c^*=\frac{2\hbar^2a^2}{\ell_B^2}. \tag*{(422)} c = c ∗ = ℓ B 2 2 ℏ 2 a 2 . (422)
It is diagonalized as
H = ℏ ω b ^ † b ^ − ℏ ω 0 − ω 2 (423) H=\hbar\omega\hat{b}^{\dagger}\hat{b}-\hbar\frac{\omega_0-\omega}{2} \tag*{(423)} H = ℏ ω b ^ † b ^ − ℏ 2 ω 0 − ω (423)
with
ω = ω 0 2 − c 2 , (424) \omega=\sqrt{\omega_0^2-c^2}, \tag*{(424)} ω = ω 0 2 − c 2 , (424)
where we have made the Bogoliubov transformation of bosons
( b ^ b ^ † ) = ( cosh r − sinh r − sinh r cosh r ) ( a ^ a ^ † ) (425) \begin{pmatrix}\hat{b}\\\hat{b}^{\dagger}\end{pmatrix}
=
\begin{pmatrix}\cosh r&-\sinh r\\-\sinh r&\cosh r\end{pmatrix}
\begin{pmatrix}\hat{a}\\\hat{a}^{\dagger} \tag*{(425)}\end{pmatrix} ( b ^ b ^ † ) = ( cosh r − sinh r − sinh r cosh r ) ( a ^ a ^ † ) (425)
with
tanh 2 r = c ℏ ω . (426) \tanh2r=\frac{c}{\hbar\omega}. \tag*{(426)} tanh 2 r = ℏ ω c . (426)
The energy of the landau level is obtained as
E n = ℏ ω N − ℏ ω 0 − ω 2 , (427) E_n=\hbar\omega N-\hbar\frac{\omega_0-\omega}{2}, \tag*{(427)} E n = ℏ ω N − ℏ 2 ω 0 − ω , (427)
while the eigenstate is determined as
∣ N ⟩ b ≡ 1 N ! ( b ^ † ) N ∣ 0 ⟩ . (428) |N\rangle_b\equiv\frac{1}{\sqrt{N!}}(\hat{b}^{\dagger})^N|0\rangle. \tag*{(428)} ∣ N ⟩ b ≡ N ! 1 ( b ^ † ) N ∣0 ⟩ . (428)
X-wave magnets
In general, it is hard to exactly diagonalize the Hamiltonian except for the p p p -wave magnet and the d d d -wave altermagnet because the Hamiltonian contains more than quadratic order of the creation and annihilation operators. However, it is possible to numerically obtain the energy of the Landau levels by using the states
∣ N ⟩ = 1 N ! ( a ^ † ) N ∣ 0 ⟩ . (429) |N\rangle= \frac{1}{\sqrt{N!}}(\hat{a}^{\dagger})^{N}|0\rangle. \tag*{(429)} ∣ N ⟩ = N ! 1 ( a ^ † ) N ∣0 ⟩ . (429)
Magneto-optical conductivity
Optical absorption between Landau levels are known as magneto-optical conductivity, which is calculated by the Kubo formula[206 , 207 , 208 , 209 ] ,
σ μ ν ( ω ) = − i ℏ e 2 ∫ d k ∑ n ≠ m f ( ε n ( k ) ) − f ( ε m ( k ) ) ( ε n ( k ) − ε m ( k ) ) ( ε n ( k ) − ε m ( k ) + ℏ ω + i Γ ) × [ ⟨ ψ n ( k ) ∣ v μ ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ v ν ∣ ψ n ( k ) ⟩ ] (430) \sigma_{\mu\nu}(\omega)=-i\hbar e^{2}\int d\mathbf{k}\sum_{n\ne m}\frac{f(\varepsilon_n(\mathbf{k}))-f(\varepsilon_m(\mathbf{k}))}{(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k}))(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})+\hbar\omega+i\Gamma)}
\times[\langle\psi_n(\mathbf{k})|v_\mu|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|v_\nu|\psi_n(\mathbf{k})\rangle] \tag*{(430)} σ μν ( ω ) = − i ℏ e 2 ∫ d k n = m ∑ ( ε n ( k ) − ε m ( k )) ( ε n ( k ) − ε m ( k ) + ℏ ω + i Γ ) f ( ε n ( k )) − f ( ε m ( k )) × [⟨ ψ n ( k ) ∣ v μ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ v ν ∣ ψ n ( k )⟩] (430)
with the velocity operators
v x ≡ ∂ H ∂ P x , v y ≡ ∂ H ∂ P y . (431) v_x\equiv\frac{\partial H}{\partial P_x},\qquad v_y\equiv\frac{\partial H}{\partial P_y}. \tag*{(431)} v x ≡ ∂ P x ∂ H , v y ≡ ∂ P y ∂ H . (431)
Especially, the magneto-optical conductivity for the d d d -wave altermagnet is studied[204 ] .
p-wave magnets
We first study magneto-optical conductivity for the p p p -wave magnet. The velocity operators are explicitly obtained as
v x = ∂ H ∂ P x = ∂ b ^ † ∂ P x ∂ H ∂ b ^ † + ∂ b ^ ∂ P x ∂ H ∂ b ^ = ℓ B ω 2 ( b ^ + b ^ † ) , (432) v_x=\frac{\partial H}{\partial P_x}=\frac{\partial\hat{b}^{\dagger}}{\partial P_x}\frac{\partial H}{\partial\hat{b}^{\dagger}}+\frac{\partial\hat{b}}{\partial P_x}\frac{\partial H}{\partial\hat{b}}=\frac{\ell_B\omega}{\sqrt{2}}(\hat{b}+\hat{b}^{\dagger}), \tag*{(432)} v x = ∂ P x ∂ H = ∂ P x ∂ b ^ † ∂ b ^ † ∂ H + ∂ P x ∂ b ^ ∂ b ^ ∂ H = 2 ℓ B ω ( b ^ + b ^ † ) , (432)
and
v y = ∂ H ∂ P y = ∂ b ^ † ∂ P y ∂ H ∂ b ^ † + ∂ b ^ ∂ P y ∂ H ∂ b ^ = i ℓ B ω 2 ( − b ^ + b ^ † ) , (433) v_y=\frac{\partial H}{\partial P_y}=\frac{\partial\hat{b}^{\dagger}}{\partial P_y}\frac{\partial H}{\partial\hat{b}^{\dagger}}+\frac{\partial\hat{b}}{\partial P_y}\frac{\partial H}{\partial\hat{b}}=i\frac{\ell_B\omega}{\sqrt{2}}(-\hat{b}+\hat{b}^{\dagger}), \tag*{(433)} v y = ∂ P y ∂ H = ∂ P y ∂ b ^ † ∂ b ^ † ∂ H + ∂ P y ∂ b ^ ∂ b ^ ∂ H = i 2 ℓ B ω ( − b ^ + b ^ † ) , (433)
where we have used
b ^ = ℓ B 2 ℏ ( P x + i P y ) + c ℏ ω , (434) \hat{b}=\frac{\ell_B}{\sqrt{2}\hbar}(P_x+iP_y)+\frac{c}{\hbar\omega}, \tag*{(434)} b ^ = 2 ℏ ℓ B ( P x + i P y ) + ℏ ω c , (434)
b ^ † = ℓ B 2 ℏ ( P x − i P y ) + c ∗ ℏ ω , (435) \hat{b}^{\dagger}=\frac{\ell_B}{\sqrt{2}\hbar}(P_x-iP_y)+\frac{c^{*}}{\hbar\omega}, \tag*{(435)} b ^ † = 2 ℏ ℓ B ( P x − i P y ) + ℏ ω c ∗ , (435)
derived from Eq. (416) . The interband matrix elements ⟨ ψ n ( k ) ∣ v μ ∣ ψ m ( k ) ⟩ \langle\psi_n(\mathbf{k})|v_\mu|\psi_m(\mathbf{k})\rangle ⟨ ψ n ( k ) ∣ v μ ∣ ψ m ( k )⟩ and ⟨ ψ m ( k ) ∣ v ν ∣ ψ n ( k ) ⟩ \langle\psi_m(\mathbf{k})|v_\nu|\psi_n(\mathbf{k})\rangle ⟨ ψ m ( k ) ∣ v ν ∣ ψ n ( k )⟩ are nonzero for m = n ± 1 m=n\pm1 m = n ± 1 . Then, the optical absorption occurs between the adjacent Landau levels. There are nonzero Re σ x x \operatorname{Re}\sigma_{xx} Re σ xx , Re σ y y \operatorname{Re}\sigma_{yy} Re σ y y and Im σ x y \operatorname{Im}\sigma_{xy} Im σ x y because v x v_x v x is real and v y v_y v y is imaginary.
d-wave altermagnets
We next study magneto-optical conductivity for the d d d -wave altermagnet. The velocity operators are explicitly obtained as
v x = ∂ H ∂ P x = ∂ b ^ † ∂ P x ∂ H ∂ b ^ † + ∂ b ^ ∂ P x ∂ H ∂ b ^ = ℓ B ω 2 ( cosh r + sinh r ) ( b ^ + b ^ † ) , (436) v_x=\frac{\partial H}{\partial P_x}=\frac{\partial\hat{b}^{\dagger}}{\partial P_x}\frac{\partial H}{\partial\hat{b}^{\dagger}}+\frac{\partial\hat{b}}{\partial P_x}\frac{\partial H}{\partial\hat{b}}=\frac{\ell_B\omega}{\sqrt{2}(\cosh r+\sinh r)}(\hat{b}+\hat{b}^{\dagger}), \tag*{(436)} v x = ∂ P x ∂ H = ∂ P x ∂ b ^ † ∂ b ^ † ∂ H + ∂ P x ∂ b ^ ∂ b ^ ∂ H = 2 ( cosh r + sinh r ) ℓ B ω ( b ^ + b ^ † ) , (436)
and
v y = ∂ H ∂ P y = ∂ b ^ † ∂ P y ∂ H ∂ b ^ † + ∂ b ^ ∂ P y ∂ H ∂ b ^ = − i ℓ B ω 2 ( cosh r − sinh r ) ( b ^ − b ^ † ) (437) v_y=\frac{\partial H}{\partial P_y}=\frac{\partial\hat{b}^{\dagger}}{\partial P_y}\frac{\partial H}{\partial\hat{b}^{\dagger}}+\frac{\partial\hat{b}}{\partial P_y}\frac{\partial H}{\partial\hat{b}}=\frac{-i\ell_B\omega}{\sqrt{2}(\cosh r-\sinh r)}(\hat{b}-\hat{b}^{\dagger}) \tag*{(437)} v y = ∂ P y ∂ H = ∂ P y ∂ b ^ † ∂ b ^ † ∂ H + ∂ P y ∂ b ^ ∂ b ^ ∂ H = 2 ( cosh r − sinh r ) − i ℓ B ω ( b ^ − b ^ † ) (437)
where we have used
b ^ = ℓ B 2 ( P x cosh r + sinh r + i P y cosh r − sinh r ) , (438) \hat{b}=\frac{\ell_B}{\sqrt{2}}\left(\frac{P_x}{\cosh r+\sinh r}+\frac{iP_y}{\cosh r-\sinh r}\right), \tag*{(438)} b ^ = 2 ℓ B ( cosh r + sinh r P x + cosh r − sinh r i P y ) , (438)
b ^ † = ℓ B 2 ( P x cosh r + sinh r − i P y cosh r − sinh r ) , (439) \hat{b}^{\dagger}=\frac{\ell_B}{\sqrt{2}}\left(\frac{P_x}{\cosh r+\sinh r}-\frac{iP_y}{\cosh r-\sinh r}\right), \tag*{(439)} b ^ † = 2 ℓ B ( cosh r + sinh r P x − cosh r − sinh r i P y ) , (439)
derived from Eq. (425) . The interband matrix elements ⟨ ψ n ( k ) ∣ v μ ∣ ψ m ( k ) ⟩ \langle\psi_n(\mathbf{k})|v_\mu|\psi_m(\mathbf{k})\rangle ⟨ ψ n ( k ) ∣ v μ ∣ ψ m ( k )⟩ and ⟨ ψ m ( k ) ∣ v ν ∣ ψ n ( k ) ⟩ \langle\psi_m(\mathbf{k})|v_\nu|\psi_n(\mathbf{k})\rangle ⟨ ψ m ( k ) ∣ v ν ∣ ψ n ( k )⟩ are nonzero for m = n ± 1 m=n\pm1 m = n ± 1 . Then, the optical absorption occurs between the adjacent Landau levels. There are nonzero Re σ x x \operatorname{Re}\sigma_{xx} Re σ xx , Re σ y y \operatorname{Re}\sigma_{yy} Re σ y y and Im σ x y \operatorname{Im}\sigma_{xy} Im σ x y because v x v_x v x is real and v y v_y v y is imaginary.
Magnetic circular dichroism
Circular dichroism is a phenomenon that the optical absorption is different between the right-hand and left-hand polarized light,
σ ± = σ x x ± i σ x y . (440) \sigma_{\pm}=\sigma_{xx}\pm i\sigma_{xy}. \tag*{(440)} σ ± = σ xx ± i σ x y . (440)
Its real part is given by
Re σ ± = Re σ x x ∓ Im σ x y . (441) \operatorname{Re}\sigma_{\pm}=\operatorname{Re}\sigma_{xx}\mp\operatorname{Im}\sigma_{xy}. \tag*{(441)} Re σ ± = Re σ xx ∓ Im σ x y . (441)
By using it, the degree of the circular dichroism is evaluated by the quantity
I C D ≡ Re σ + − Re σ − Re σ + + Re σ − = − Im σ x y Re σ x x . (442) I_{\mathrm{CD}}\equiv\frac{\operatorname{Re}\sigma_{+}-\operatorname{Re}\sigma_{-}}{\operatorname{Re}\sigma_{+}+\operatorname{Re}\sigma_{-}}=-\frac{\operatorname{Im}\sigma_{xy}}{\operatorname{Re}\sigma_{xx}}. \tag*{(442)} I CD ≡ Re σ + + Re σ − Re σ + − Re σ − = − Re σ xx Im σ x y . (442)
Hence, circular dichroism occurs when Im σ x y ≠ 0 \operatorname{Im}\sigma_{xy}\ne0 Im σ x y = 0 . Indeed, it occurs in the case of the p p p -wave magnet and the d d d -wave altermagnet, where Im σ x y ≠ 0 \operatorname{Im}\sigma_{xy}\ne0 Im σ x y = 0 as we have seen.
Friedel oscillation
We study the effect of an impurity. Friedel oscillation is that the local the density of sates (LDOS) oscillates in the presence of the impurity. We show that this LDOS has the X-wave symmetry for the X-wave magnet. The energy-dependent LDOS is determined by
ρ ( ω , r ) = − 1 π Im G R ( ω , r , r ) , (443) \rho(\omega,\mathbf{r})=-\frac{1}{\pi}\operatorname{Im}G^{R}(\omega,\mathbf{r},\mathbf{r}), \tag*{(443)} ρ ( ω , r ) = − π 1 Im G R ( ω , r , r ) , (443)
where G ( ω , r , r ′ ) G(\omega,\mathbf{r},\mathbf{r}') G ( ω , r , r ′ ) is the real space representation of the retarded Green function in the presence of the impurity defined by
G R ( ω , r , r ′ ) ≡ ⟨ r ∣ 1 ℏ ω − H + i η ∣ r ′ ⟩ . (444) G^{R}(\omega,\mathbf{r},\mathbf{r}')\equiv\langle\mathbf{r}|\frac{1}{\hbar\omega-H+i\eta}|\mathbf{r}'\rangle. \tag*{(444)} G R ( ω , r , r ′ ) ≡ ⟨ r ∣ ℏ ω − H + i η 1 ∣ r ′ ⟩ . (444)
The Green function is obtained by using the Dyson equation
G R ( ω , r , r ′ ) = G 0 R ( ω , r , r ′ ) + ∫ d r 1 d r 2 G 0 R ( ω , r , r 1 ) T ( ω , r 1 , r 2 ) G 0 R ( ω , r 2 , r ′ ) . (445) \begin{aligned}
G^{R}(\omega,\mathbf{r},\mathbf{r}')&=G_{0}^{R}(\omega,\mathbf{r},\mathbf{r}')\\
&\quad+\int d\mathbf{r}_{1}d\mathbf{r}_{2}G_{0}^{R}(\omega,\mathbf{r},\mathbf{r}_{1})T(\omega,\mathbf{r}_{1},\mathbf{r}_{2})G_{0}^{R}(\omega,\mathbf{r}_{2},\mathbf{r}'). \tag*{(445)}
\end{aligned} G R ( ω , r , r ′ ) = G 0 R ( ω , r , r ′ ) + ∫ d r 1 d r 2 G 0 R ( ω , r , r 1 ) T ( ω , r 1 , r 2 ) G 0 R ( ω , r 2 , r ′ ) . (445)
with the T matrix
T ( ω , r , r 1 ) = V ( 1 − G 0 R ( ω , r , r 1 ) V ) − 1 . (446) T(\omega,\mathbf{r},\mathbf{r}_{1})=V\left(1-G_{0}^{R}(\omega,\mathbf{r},\mathbf{r}_{1})V\right)^{-1}. \tag*{(446)} T ( ω , r , r 1 ) = V ( 1 − G 0 R ( ω , r , r 1 ) V ) − 1 . (446)
We assume that the impurity is described by the delta function potential
( V 0 σ 0 + V s σ z ) δ ( r ) , (447) (V_{0}\sigma_{0}+V_{s}\sigma_{z})\delta(\mathbf{r}), \tag*{(447)} ( V 0 σ 0 + V s σ z ) δ ( r ) , (447)
where V 0 V_{0} V 0 describes a non-magnetic impurity and V s V_{s} V s describes a magnetic impurity. It is enough to study spin-dependent potential
( V 0 + V s s ) δ ( r ) (448) (V_{0}+V_{s}s)\delta(\mathbf{r}) \tag*{(448)} ( V 0 + V s s ) δ ( r ) (448)
with s = ± 1 s=\pm1 s = ± 1 because the spin operator is diagonal. In this case, the T matrix has a form
T ( ω , r , r 1 ) = T ( ω ) δ ( r ) δ ( r ′ ) , (449) T(\omega,\mathbf{r},\mathbf{r}_{1})=T(\omega)\delta(\mathbf{r})\delta(\mathbf{r}'), \tag*{(449)} T ( ω , r , r 1 ) = T ( ω ) δ ( r ) δ ( r ′ ) , (449)
which leads to the Dyson equation
G R ( ω , r , r ′ ) = G 0 R ( ω , r , r ′ ) + G 0 R ( ω , r , 0 ) T ( ω ) G 0 R ( ω , 0 , r ′ ) . (450) G^{R}(\omega,\mathbf{r},\mathbf{r}')=G_{0}^{R}(\omega,\mathbf{r},\mathbf{r}')+G_{0}^{R}(\omega,\mathbf{r},0)T(\omega)G_{0}^{R}(\omega,0,\mathbf{r}'). \tag*{(450)} G R ( ω , r , r ′ ) = G 0 R ( ω , r , r ′ ) + G 0 R ( ω , r , 0 ) T ( ω ) G 0 R ( ω , 0 , r ′ ) . (450)
For small V V V , the difference of the local density of states is simply given by
δ ρ = − 1 π Im [ G R ( ω , r , r ) − G 0 R ( ω , r , r ) ] = − 1 π Im [ V G 0 R ( ω , r , 0 ) G 0 R ( ω , 0 , r ′ ) ] 2 . (451) \begin{aligned}
\delta\rho&=-\frac{1}{\pi}\operatorname{Im}\left[G^{R}(\omega,\mathbf{r},\mathbf{r})-G_{0}^{R}(\omega,\mathbf{r},\mathbf{r})\right]\\
&=-\frac{1}{\pi}\operatorname{Im}\left[V G_{0}^{R}(\omega,\mathbf{r},0)G_{0}^{R}(\omega,0,\mathbf{r}')\right]^{2}. \tag*{(451)}
\end{aligned} δ ρ = − π 1 Im [ G R ( ω , r , r ) − G 0 R ( ω , r , r ) ] = − π 1 Im [ V G 0 R ( ω , r , 0 ) G 0 R ( ω , 0 , r ′ ) ] 2 . (451)
The real space Green function is obtained by the Fourier transformation of the momentum space Green function as
G 0 R ( ω , r , r ′ ) = G 0 R ( ω , r − r ′ , 0 ) ≡ 1 ( 2 π ) 2 ∫ e − i k ⋅ ( r − r ′ ) G 0 R ( ω , k ) d r , (452) G_{0}^{R}(\omega,\mathbf{r},\mathbf{r}')=G_{0}^{R}(\omega,\mathbf{r}-\mathbf{r}',0)\equiv\frac{1}{(2\pi)^{2}}\int e^{-i\mathbf{k}\cdot(\mathbf{r}-\mathbf{r}')}G_{0}^{R}(\omega,\mathbf{k})\,d\mathbf{r}, \tag*{(452)} G 0 R ( ω , r , r ′ ) = G 0 R ( ω , r − r ′ , 0 ) ≡ ( 2 π ) 2 1 ∫ e − i k ⋅ ( r − r ′ ) G 0 R ( ω , k ) d r , (452)
with
G 0 R ( ω , k ) ≡ 1 i ω − H + i η (453) G_{0}^{R}(\omega,\mathbf{k})\equiv\frac{1}{i\omega-H+i\eta} \tag*{(453)} G 0 R ( ω , k ) ≡ iω − H + i η 1 (453)
where we have used the translational symmetry of the Green function in the absence of the impurity.
Free electrons
First, we review the Friedel oscillation of free electrons. We derive the real space representation of the Green function in the presence of the impurity and determine the symmetry of the LDOS. The Green function for free electrons is given by
G 0 R ( ω , k ) = 1 ℏ 2 2 m ( k 0 2 − k 2 ) + i η , (454) G_{0}^{R}(\omega,\mathbf{k})=\frac{1}{\frac{\hbar^{2}}{2m}(k_{0}^{2}-k^{2})+i\eta}, \tag*{(454)} G 0 R ( ω , k ) = 2 m ℏ 2 ( k 0 2 − k 2 ) + i η 1 , (454)
where we have set
ω = ℏ 2 k 0 2 2 m . (455) \omega=\frac{\hbar^{2}k_{0}^{2}}{2m}. \tag*{(455)} ω = 2 m ℏ 2 k 0 2 . (455)
We make a Fourier transformation
G 0 R ( r ) = 1 ( 2 π ) 2 ∫ 0 ∞ k e i k r cos ( θ − φ ) ℏ 2 2 m ( k 0 2 − k 2 ) + i η d k d θ = 1 2 π ∫ 0 ∞ k J 0 ( k r ) ℏ 2 2 m ( k 0 2 − k 2 ) + i η d k , (456) \begin{aligned}
G_{0}^{R}(r)&=\frac{1}{(2\pi)^{2}}\int_{0}^{\infty}\frac{k e^{ikr\cos(\theta-\varphi)}}{\frac{\hbar^{2}}{2m}(k_{0}^{2}-k^{2})+i\eta}\,dkd\theta\\
&=\frac{1}{2\pi}\int_{0}^{\infty}\frac{kJ_{0}(kr)}{\frac{\hbar^{2}}{2m}(k_{0}^{2}-k^{2})+i\eta}\,dk, \tag*{(456)}
\end{aligned} G 0 R ( r ) = ( 2 π ) 2 1 ∫ 0 ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) + i η k e ik r c o s ( θ − φ ) d k d θ = 2 π 1 ∫ 0 ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) + i η k J 0 ( k r ) d k , (456)
where J 0 J_{0} J 0 is the Bessel function and we have used the formula
∫ 0 2 π e i k r cos θ d θ = 2 π J 0 ( k r ) . (457) \int_{0}^{2\pi}e^{ikr\cos\theta}\,d\theta=2\pi J_{0}(kr). \tag*{(457)} ∫ 0 2 π e ik r c o s θ d θ = 2 π J 0 ( k r ) . (457)
The Bessel function is decomposed as the Hankel function
J 0 ( k r ) = H 0 ( 1 ) ( k r ) + H 0 ( 2 ) ( k r ) 2 , (458) J_{0}(kr)=\frac{H_{0}^{(1)}(kr)+H_{0}^{(2)}(kr)}{2}, \tag*{(458)} J 0 ( k r ) = 2 H 0 ( 1 ) ( k r ) + H 0 ( 2 ) ( k r ) , (458)
where H 0 ( 1 ) ( k r ) H_{0}^{(1)}(kr) H 0 ( 1 ) ( k r ) (H 0 ( 2 ) ( k r ) H_{0}^{(2)}(kr) H 0 ( 2 ) ( k r ) ) is the Hankel function the first (second) kind. The asymptotic forms of the Hankel functions are given by
H 0 ( 1 ) ( k r ) ∼ 2 π k r e i ( k r − π 4 ) , (459) H_{0}^{(1)}(kr)\sim\sqrt{\frac{2}{\pi kr}}e^{i(kr-\frac{\pi}{4})}, \tag*{(459)} H 0 ( 1 ) ( k r ) ∼ π k r 2 e i ( k r − 4 π ) , (459)
H 0 ( 2 ) ( k r ) ∼ 2 π k r e − i ( k r − π 4 ) . (460) H_{0}^{(2)}(kr)\sim\sqrt{\frac{2}{\pi kr}}e^{-i(kr-\frac{\pi}{4})}. \tag*{(460)} H 0 ( 2 ) ( k r ) ∼ π k r 2 e − i ( k r − 4 π ) . (460)
We make a complex integral
G 0 ( r ) = 1 16 π ∫ − ∞ ∞ k ( H 0 ( 1 ) ( k r ) + H 0 ( 2 ) ( k r ) ) ℏ 2 2 m ( k 0 2 − k 2 ) + i η d k , (461) G_{0}(r)=\frac{1}{16\pi}\int_{-\infty}^{\infty}\frac{k\left(H_{0}^{(1)}(kr)+H_{0}^{(2)}(kr)\right)}{\frac{\hbar^{2}}{2m}(k_{0}^{2}-k^{2})+i\eta}\,dk, \tag*{(461)} G 0 ( r ) = 16 π 1 ∫ − ∞ ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) + i η k ( H 0 ( 1 ) ( k r ) + H 0 ( 2 ) ( k r ) ) d k , (461)
where poles exist at
k = ± k 0 + i η . (462) k = \pm k_0 + i\eta. \tag*{(462)} k = ± k 0 + i η . (462)
We take the upper half circle for H 0 ( 1 ) ( k r ) H_0^{(1)}(kr) H 0 ( 1 ) ( k r ) and the lower half circle for H 0 ( 2 ) ( k r ) H_0^{(2)}(kr) H 0 ( 2 ) ( k r ) so that the complex integral converges. Only the contribution from H 0 ( 1 ) ( k r ) H_0^{(1)}(kr) H 0 ( 1 ) ( k r ) is nonzero and the Green function is calculated as
G 0 ( r ) = 1 16 π ∫ − ∞ ∞ k H 0 ( 1 ) ( k r ) ℏ 2 2 m ( k 0 2 − k 2 ) + i η d k = i 8 ( H 0 ( 1 ) ( k 0 r ) + H 0 ( 1 ) ( − k 0 r ) ) = i 4 H 0 ( 1 ) ( k 0 r ) . (463) \begin{aligned}
G_0(r) &= \frac{1}{16\pi}\int_{-\infty}^{\infty}\frac{kH_0^{(1)}(kr)}{\frac{\hbar^2}{2m}(k_0^2-k^2)+i\eta}\,dk \\
&= \frac{i}{8}\left(H_0^{(1)}(k_0r)+H_0^{(1)}(-k_0r)\right) \\
&= \frac{i}{4}H_0^{(1)}(k_0r). \tag*{(463)}
\end{aligned} G 0 ( r ) = 16 π 1 ∫ − ∞ ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) + i η k H 0 ( 1 ) ( k r ) d k = 8 i ( H 0 ( 1 ) ( k 0 r ) + H 0 ( 1 ) ( − k 0 r ) ) = 4 i H 0 ( 1 ) ( k 0 r ) . (463)
Then, the LDOS has the s s s -wave symmetry.
p-wave magnets
Second, we study the Friedel oscillation of the p p p -wave magnet[211 ] . We calculate the real space representation of the Green function
G ( r ) = 1 ( 2 π ) 2 ∫ 0 ∞ k e i k r cos ( θ − ϕ ) ℏ 2 2 m ( k 0 2 − k 2 ) − s J k cos ϕ + i η d k d θ = 1 2 π ∫ 0 ∞ k J 0 ( k r ) ℏ 2 2 m ( k 0 2 − k 2 ) − s J k cos ϕ + i η d k . (464) \begin{aligned}
G(r) &= \frac{1}{(2\pi)^2}\int_0^\infty k\frac{e^{ikr\cos(\theta-\phi)}}{\frac{\hbar^2}{2m}(k_0^2-k^2)-sJk\cos\phi+i\eta}\,dkd\theta\\
&= \frac{1}{2\pi}\int_0^\infty\frac{kJ_0(kr)}{\frac{\hbar^2}{2m}(k_0^2-k^2)-sJk\cos\phi+i\eta}\,dk. \tag*{(464)}
\end{aligned} G ( r ) = ( 2 π ) 2 1 ∫ 0 ∞ k 2 m ℏ 2 ( k 0 2 − k 2 ) − s J k cos ϕ + i η e ik r c o s ( θ − ϕ ) d k d θ = 2 π 1 ∫ 0 ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) − s J k cos ϕ + i η k J 0 ( k r ) d k . (464)
Poles exist at
k p = − s J m cos ϕ ℏ 2 ± k 0 2 + J 2 m 2 ℏ 4 cos 2 ϕ + i η , (465) k_p=-\frac{sJm\cos\phi}{\hbar^2}\pm\sqrt{k_0^2+\frac{J^2m^2}{\hbar^4}\cos^2\phi+i\eta}, \tag*{(465)} k p = − ℏ 2 s J m cos ϕ ± k 0 2 + ℏ 4 J 2 m 2 cos 2 ϕ + i η , (465)
and we obtain
G ( r ) = i 4 H 0 ( 1 ) ( k p r ) . (466) G(r)=\frac{i}{4}H_0^{(1)}(k_pr). \tag*{(466)} G ( r ) = 4 i H 0 ( 1 ) ( k p r ) . (466)
Then, the LDOS has the p p p -wave symmetry and spin dependence.
d-wave altermagnets
Third, we study the Friedel oscillation of the d d d -wave altermagnet[210 , 211 ] . We calculate the real space representation of the Green function
G ( r ) = 1 ( 2 π ) 2 ∫ 0 ∞ k e i k r cos ( θ − ϕ ) k 0 2 − k 2 − s J k 2 cos ϕ + i η d k d θ = 1 2 π ∫ 0 ∞ k J 0 ( k r ) ℏ 2 2 m ( k 0 2 − k 2 ) − s J J k 2 cos ϕ + i η d k . (467) \begin{aligned}
G(r) &= \frac{1}{(2\pi)^2}\int_0^\infty k\frac{e^{ikr\cos(\theta-\phi)}}{k_0^2-k^2-sJk^2\cos\phi+i\eta}\,dkd\theta\\
&= \frac{1}{2\pi}\int_0^\infty\frac{kJ_0(kr)}{\frac{\hbar^2}{2m}(k_0^2-k^2)-s_JJk^2\cos\phi+i\eta}\,dk. \tag*{(467)}
\end{aligned} G ( r ) = ( 2 π ) 2 1 ∫ 0 ∞ k k 0 2 − k 2 − s J k 2 cos ϕ + i η e ik r c o s ( θ − ϕ ) d k d θ = 2 π 1 ∫ 0 ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) − s J J k 2 cos ϕ + i η k J 0 ( k r ) d k . (467)
Poles exist at
k d = ± k 0 1 + 2 m s J ℏ 2 cos ϕ + i η , (468) k_d=\pm\frac{k_0}{\sqrt{1+\frac{2msJ}{\hbar^2}\cos\phi}}+i\eta, \tag*{(468)} k d = ± 1 + ℏ 2 2 m s J cos ϕ k 0 + i η , (468)
and we obtain
G ( r ) = i 4 H 0 ( 1 ) ( k d r ) . (469) G(r)=\frac{i}{4}H_0^{(1)}(k_dr). \tag*{(469)} G ( r ) = 4 i H 0 ( 1 ) ( k d r ) . (469)
Then, the LDOS has the d d d -wave symmetry.
X-wave magnets
Finally, we study the Friedel oscillation of the X X X -wave magnet. We calculate the real space representation of the Green function
G ( r ) = 1 ( 2 π ) 2 ∫ 0 ∞ k e i k r cos ( θ − ϕ ) k 0 2 − k 2 − f X , X ′ 2 D ( k ) + i η d k d θ = 1 2 π ∫ 0 ∞ k J 0 ( k r ) ℏ 2 2 m ( k 0 2 − k 2 ) − f X , X ′ 2 D ( k ) + i η d k . (470) \begin{aligned}
G(r) &= \frac{1}{(2\pi)^2}\int_0^\infty k\frac{e^{ikr\cos(\theta-\phi)}}{k_0^2-k^2-f^{2D}_{X,X'}(k)+i\eta}\,dkd\theta\\
&= \frac{1}{2\pi}\int_0^\infty\frac{kJ_0(kr)}{\frac{\hbar^2}{2m}(k_0^2-k^2)-f^{2D}_{X,X'}(k)+i\eta}\,dk. \tag*{(470)}
\end{aligned} G ( r ) = ( 2 π ) 2 1 ∫ 0 ∞ k k 0 2 − k 2 − f X , X ′ 2 D ( k ) + i η e ik r c o s ( θ − ϕ ) d k d θ = 2 π 1 ∫ 0 ∞ 2 m ℏ 2 ( k 0 2 − k 2 ) − f X , X ′ 2 D ( k ) + i η k J 0 ( k r ) d k . (470)
The pole is determined by the solution of
ℏ 2 2 m ( k 0 2 − k X , X ′ 2 ) − f X , X ′ 2 D ( k X , X ′ ) + i η = 0 , (471) \frac{\hbar^2}{2m}(k_0^2-k_{X,X'}^2)-f^{2D}_{X,X'}(k_{X,X'})+i\eta=0, \tag*{(471)} 2 m ℏ 2 ( k 0 2 − k X , X ′ 2 ) − f X , X ′ 2 D ( k X , X ′ ) + i η = 0 , (471)
which has the X X X -wave symmetry although it is hard to obtain analytic formula except for the p p p -wave magnet and the d d d -wave altermagnet. By using it, we obtain
G ( r ) = i 4 H 0 ( 1 ) ( k X , X ′ r ) . (472) G(r)=\frac{i}{4}H_0^{(1)}(k_{X,X'}r). \tag*{(472)} G ( r ) = 4 i H 0 ( 1 ) ( k X , X ′ r ) . (472)
Then, the LDOS has the X X X -wave symmetry and spin dependence.
Summaries, discussions and outlooks
We reviewed recent progress on quantum geometry and X X X -wave magnets. We systematically formulated quantum geometry based on the quantum distance. Quantum geometry is generalized to the Zeemann quantum geometry, non-Hermitian quantum geometry and quantum information geometry. Then, we reviewed the X X X -wave magnets including altermagnets and p p p -wave magnets in a universal manner. Universal transport properties were discussed for the X X X -wave magnets.
In this paper, we have derived various analytic formulas based on the two-band Hamiltonian. However, actual materials are more complicated. For example, four-band models including the orbital degrees of freedom have been studied[45 , 148 ] . Nevertheless, if there are only two bands at the Fermi energy, the effective Hamiltonian is reduced to be a two-band Hamiltonian[32 ] . In addition, optical absorption is well described by the two-band Hamiltonian consisting of the valence and conduction bands. The two-band Hamiltonian reveals the basic structure of the relevant phenomena in these cases. It is an interesting problem to study based on more realistic models.
In passing, we give some outlooks. The general relativistic quantities such as the Christoffel symbols are defined based on the quantum metric as shown in § 2 F. However, it is not clear whether there are physical observable quantities such as nonlinear conductivities or nonlinear optical conductivities directly relating to them. Quantum geometry is mainly studied in the context of condensed matter physics. Applications of the notion of quantum geometry to quantum information and quantum computing are still rare, where quantum information geometry for density matrices will be useful. There will be developments in this direction. Zeeman quantum geometry will be useful for spintronics.
Magnonic properties of the X X X -wave magnets will also be interesting, which are mainly studied so far for d d d -wave altermagnets[212 , 213 ] and g g g -wave altermagnets[214 ] . The magnon spectrum is given by[215 ]
E ± ( k ) = v k + κ k 3 ± J k N X sin N X ϕ (473) E_{\pm}(\mathbf{k})=v k+\kappa k^{3}\pm J k^{N_X}\sin N_X\phi \tag*{(473)} E ± ( k ) = v k + κ k 3 ± J k N X sin N X ϕ (473)
for the X X X -wave magnet in general. With respect to material realization of the X X X -wave magnets, there are no experiments on the f f f -wave, g g g -wave and i i i -wave magnets in two dimensions and the f f f -wave and i i i -wave magnets in three dimensions. In this paper, we have mainly focused on electronic and optical properties coupled with the X X X -wave magnet. On the other hand, the control of the magnetism of the X X X -wave magnet is also important. It is a nontrivial problem to switch the direction of the spins of the X X X -wave magnet. In order to switch the direction of spins, multiferroic altermagnets[216 , 217 ] are promising because the spin direction is reversed by applying electric field. It is a fascinating problem both theoretically and experimentally to search universal physics in the viewpoint of the X X X -wave magnets.
Acknowledgements
The author is grateful to N. Nagaosa, M. Hirschberger, S. Okumura, Y. Motome, T. Morimoto, H. Seo, J. Wang, B. J. Yang and W. Chen for helpful discussions on the subject. This work is supported by CREST, JST (Grants No. JPMJCR20T2) and Grants-in-Aid for Scientific Research from MEXT KAKENHI (Grant No. 23H00171).
Appendices
Quantum distance and Abelian quantum geometric tensor
The Hilbert-Schmidt distance is defined by
( d s H S ) 2 ≡ 1 − ∣ ⟨ ψ n ( k ) ∣ ψ n ( k + d k ) ⟩ ∣ 2 . (474) (ds_{\mathrm{HS}})^2\equiv1-\left|\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}+d\mathbf{k})\rangle\right|^2. \tag*{(474)} ( d s HS ) 2 ≡ 1 − ∣ ⟨ ψ n ( k ) ∣ ψ n ( k + d k )⟩ ∣ 2 . (474)
The wave function is expanded as
∣ ψ n ( k + d k ) ⟩ = ∣ ψ n ( k ) ⟩ + ∣ ∂ k μ ψ n ( k ) ⟩ d k μ + 1 2 ∣ ∂ k μ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν + ⋯ . (475) |\psi_n(\mathbf{k}+d\mathbf{k})\rangle=|\psi_n(\mathbf{k})\rangle+|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle dk_\mu+\frac{1}{2}|\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})\rangle dk_\mu dk_\nu+\cdots. \tag*{(475)} ∣ ψ n ( k + d k )⟩ = ∣ ψ n ( k )⟩ + ∣ ∂ k μ ψ n ( k )⟩ d k μ + 2 1 ∣ ∂ k μ ∂ k ν ψ n ( k )⟩ d k μ d k ν + ⋯ . (475)
Quantum distance and non-Abelian quantum geometric tensor
The inner product reads
∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ ∣ 2 = ⟨ ψ n ( k ′ ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ = ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ + ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k ) ⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ + ( 1 2 ⟨ ∂ k μ ∂ k ν ψ n ( k ) ∣ ψ n ( k ) ⟩ + 1 2 ⟨ ψ n ( k ) ∣ ∂ k μ ∂ k ν ψ n ( k ) ⟩ + ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ + ⟨ ∂ k ν ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ d k ν . (476) \begin{aligned}
\left|\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle\right|^2
={}&\langle\psi_n(\mathbf{k}')|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle\\
={}&\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\\
&+\left(\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)dk_\mu\\
&+\left(\frac{1}{2}\langle\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\frac{1}{2}\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})\rangle+\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\right.\\
&\left.\qquad+\langle\partial_{k_\nu}\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)dk_\mu dk_\nu. \tag*{(476)}
\end{aligned} ∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ ∣ 2 = = ⟨ ψ n ( k ′ ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ + ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k )⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) d k μ + ( 2 1 ⟨ ∂ k μ ∂ k ν ψ n ( k ) ∣ ψ n ( k )⟩ + 2 1 ⟨ ψ n ( k ) ∣ ∂ k μ ∂ k ν ψ n ( k )⟩ + ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k )⟩ + ⟨ ∂ k ν ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) d k μ d k ν . (476)
The term proportional to d k μ dk_\mu d k μ is zero because
( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k ) ⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) = ∂ k μ ( ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ ) = 0 , (477) \left(\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)=\partial_{k_\mu}\left(\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\right)=0, \tag*{(477)} ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k )⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) = ∂ k μ ( ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ ) = 0 , (477)
where we have used the normalization condition ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ = 1 \langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle=1 ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ = 1 . By using the relation
∂ k μ ∂ k ν ( ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ ) = ⟨ ∂ k μ ∂ k ν ψ n ( k ) ∣ ψ n ( k ) ⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ∂ k ν ψ n ( k ) ⟩ + ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ + ⟨ ∂ k ν ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ , (478) \begin{aligned}
\partial_{k_\mu}\partial_{k_\nu}\left(\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\right)
={}&\langle\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\\
&+\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle+\langle\partial_{k_\nu}\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle, \tag*{(478)}
\end{aligned} ∂ k μ ∂ k ν ( ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ ) = ⟨ ∂ k μ ∂ k ν ψ n ( k ) ∣ ψ n ( k )⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ∂ k ν ψ n ( k )⟩ + ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k )⟩ + ⟨ ∂ k ν ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ , (478)
we have
1 2 ⟨ ∂ k μ ∂ k ν ψ n ( k ) ∣ ψ n ( k ) ⟩ + 1 2 ⟨ ψ n ( k ) ∣ ∂ k μ ∂ k ν ψ n ( k ) ⟩ = − 1 2 ( ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ + ⟨ ∂ k ν ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ d k ν = − ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν . (479) \begin{aligned}
&\frac{1}{2}\langle\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\frac{1}{2}\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\\
&=-\frac{1}{2}\left(\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle+\langle\partial_{k_\nu}\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)dk_\mu dk_\nu\\
&=-\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle dk_\mu dk_\nu. \tag*{(479)}
\end{aligned} 2 1 ⟨ ∂ k μ ∂ k ν ψ n ( k ) ∣ ψ n ( k )⟩ + 2 1 ⟨ ψ n ( k ) ∣ ∂ k μ ∂ k ν ψ n ( k )⟩ = − 2 1 ( ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k )⟩ + ⟨ ∂ k ν ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) d k μ d k ν = − ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k )⟩ d k μ d k ν . (479)
Then, the inner product is written as
∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ ∣ 2 = 1 + ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ − ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ ) d k μ d k ν = 1 + ⟨ ∂ k μ ψ n ( k ) ∣ ( ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ − 1 ) ∣ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν . (480) \begin{aligned}
\left|\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle\right|^2
={}&1+\left(\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\right.\\
&\left.-\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\right)dk_\mu dk_\nu\\
={}&1+\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\left(|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|-1\right)|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle dk_\mu dk_\nu. \tag*{(480)}
\end{aligned} ∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ ∣ 2 = = 1 + ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ∂ k ν ψ n ( k )⟩ − ⟨ ∂ k μ ψ n ( k ) ∣ ∂ k ν ψ n ( k )⟩ ) d k μ d k ν 1 + ⟨ ∂ k μ ψ n ( k ) ∣ ( ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ − 1 ) ∣ ∂ k ν ψ n ( k )⟩ d k μ d k ν . (480)
and the quantum distance reads
( d s H S ) 2 ≡ ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ) ∣ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν = F μ ν H d k μ d k ν . (481) \begin{aligned}
(ds_{\mathrm{HS}})^2
&\equiv\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\left(1-|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\right)|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle dk_\mu dk_\nu\\
&=\mathcal{F}^{\mathrm{H}}_{\mu\nu}dk_\mu dk_\nu. \tag*{(481)}
\end{aligned} ( d s HS ) 2 ≡ ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ) ∣ ∂ k ν ψ n ( k )⟩ d k μ d k ν = F μν H d k μ d k ν . (481)
This is Eq. (7) in the main text.
Appendix B: Quantum distance and non-Abelian quantum geometric tensor
We summarize non-Abelian quantum geometry for the N N N -fold degenerate wave functions[3 ] .
The Hilbert-Schmidt distance is defined by
( d H S ) 2 ≡ 1 − ∣ ⟨ u ( k ) ∣ u ( k + d k ) ⟩ ∣ 2 (482) (d_{\mathrm{HS}})^2 \equiv1-\left|\langle u(\mathbf{k})|u(\mathbf{k}+d\mathbf{k})\rangle\right|^2 \tag*{(482)} ( d HS ) 2 ≡ 1 − ∣ ⟨ u ( k ) ∣ u ( k + d k )⟩ ∣ 2 (482)
where ∣ u ( k ) ⟩ |u(\mathbf{k})\rangle ∣ u ( k )⟩ is the wave function of N N N -fold degenerate bands
∣ u ( k ) ⟩ = ∑ n = 1 N c n ∣ ψ n ( k ) ⟩ (483) |u(\mathbf{k})\rangle=\sum_{n=1}^{N}c_n|\psi_n(\mathbf{k})\rangle \tag*{(483)} ∣ u ( k )⟩ = n = 1 ∑ N c n ∣ ψ n ( k )⟩ (483)
where
∑ n = 1 N ∣ c n ∣ 2 = 1. (484) \sum_{n=1}^{N}|c_n|^2=1. \tag*{(484)} n = 1 ∑ N ∣ c n ∣ 2 = 1. (484)
The inner product is calculated as
∣ ⟨ u ( k ) ∣ u ( k + d k ) ⟩ ∣ 2 = ⟨ u ( k ) ∣ u ( k ) ⟩ + ( ⟨ ∂ k μ u ( k ) ∣ u ( k ) ⟩ + ⟨ u ( k ) ∣ ∂ k μ u ( k ) ⟩ ) d k μ + ( 1 2 ⟨ ∂ k μ ∂ k ν u ( k ) ∣ u ( k ) ⟩ + 1 2 ⟨ u ( k ) ∣ ∂ k μ ∂ k ν u ( k ) ⟩ + ⟨ ∂ k μ u ( k ) ∣ u ( k ) ⟩ ⟨ u ( k ) ∣ ∂ k ν u ( k ) ⟩ ) d k μ d k ν = 1 + ( − 1 2 ⟨ ∂ k μ u ( k ) ∣ ∂ k ν u ( k ) ⟩ − 1 2 ⟨ ∂ k ν u ( k ) ∣ ∂ k μ u ( k ) ⟩ + ⟨ ∂ k μ u ( k ) ∣ u ( k ) ⟩ ⟨ u ( k ) ∣ ∂ k ν u ( k ) ⟩ ) d k μ d k ν = 1 + ⟨ ∂ k μ u ( k ) ∣ ( ∣ u ( k ) ⟩ ⟨ u ( k ) ∣ − 1 ) ∣ ∂ k ν u ( k ) ⟩ d k μ d k ν . (485) \begin{aligned}
\left|\langle u(\mathbf{k})|u(\mathbf{k}+d\mathbf{k})\rangle\right|^2
={}&\langle u(\mathbf{k})|u(\mathbf{k})\rangle\\
&+\left(\langle\partial_{k_\mu}u(\mathbf{k})|u(\mathbf{k})\rangle+\langle u(\mathbf{k})|\partial_{k_\mu}u(\mathbf{k})\rangle\right)dk_\mu\\
&+\left(\frac{1}{2}\langle\partial_{k_\mu}\partial_{k_\nu}u(\mathbf{k})|u(\mathbf{k})\rangle+\frac{1}{2}\langle u(\mathbf{k})|\partial_{k_\mu}\partial_{k_\nu}u(\mathbf{k})\rangle+\langle\partial_{k_\mu}u(\mathbf{k})|u(\mathbf{k})\rangle\langle u(\mathbf{k})|\partial_{k_\nu}u(\mathbf{k})\rangle\right)dk_\mu dk_\nu\\
={}&1+\left(-\frac{1}{2}\langle\partial_{k_\mu}u(\mathbf{k})|\partial_{k_\nu}u(\mathbf{k})\rangle-\frac{1}{2}\langle\partial_{k_\nu}u(\mathbf{k})|\partial_{k_\mu}u(\mathbf{k})\rangle+\langle\partial_{k_\mu}u(\mathbf{k})|u(\mathbf{k})\rangle\langle u(\mathbf{k})|\partial_{k_\nu}u(\mathbf{k})\rangle\right)dk_\mu dk_\nu\\
={}&1+\langle\partial_{k_\mu}u(\mathbf{k})|\left(|u(\mathbf{k})\rangle\langle u(\mathbf{k})|-1\right)|\partial_{k_\nu}u(\mathbf{k})\rangle dk_\mu dk_\nu. \tag*{(485)}
\end{aligned} ∣ ⟨ u ( k ) ∣ u ( k + d k )⟩ ∣ 2 = = = ⟨ u ( k ) ∣ u ( k )⟩ + ( ⟨ ∂ k μ u ( k ) ∣ u ( k )⟩ + ⟨ u ( k ) ∣ ∂ k μ u ( k )⟩ ) d k μ + ( 2 1 ⟨ ∂ k μ ∂ k ν u ( k ) ∣ u ( k )⟩ + 2 1 ⟨ u ( k ) ∣ ∂ k μ ∂ k ν u ( k )⟩ + ⟨ ∂ k μ u ( k ) ∣ u ( k )⟩ ⟨ u ( k ) ∣ ∂ k ν u ( k )⟩ ) d k μ d k ν 1 + ( − 2 1 ⟨ ∂ k μ u ( k ) ∣ ∂ k ν u ( k )⟩ − 2 1 ⟨ ∂ k ν u ( k ) ∣ ∂ k μ u ( k )⟩ + ⟨ ∂ k μ u ( k ) ∣ u ( k )⟩ ⟨ u ( k ) ∣ ∂ k ν u ( k )⟩ ) d k μ d k ν 1 + ⟨ ∂ k μ u ( k ) ∣ ( ∣ u ( k )⟩ ⟨ u ( k ) ∣ − 1 ) ∣ ∂ k ν u ( k )⟩ d k μ d k ν . (485)
Then, we have
( d H S ) 2 = ⟨ ∂ k μ u ( k ) ∣ ( 1 − P ( k ) ) ∣ ∂ k ν u ( k ) ⟩ d k μ d k ν , (486) (d_{\mathrm{HS}})^2=\langle\partial_{k_\mu}u(\mathbf{k})|(1-P(\mathbf{k}))|\partial_{k_\nu}u(\mathbf{k})\rangle dk_\mu dk_\nu, \tag*{(486)} ( d HS ) 2 = ⟨ ∂ k μ u ( k ) ∣ ( 1 − P ( k )) ∣ ∂ k ν u ( k )⟩ d k μ d k ν , (486)
where P ( k ) ≡ ∣ u ( k ) ⟩ ⟨ u ( k ) ∣ P(\mathbf{k})\equiv|u(\mathbf{k})\rangle\langle u(\mathbf{k})| P ( k ) ≡ ∣ u ( k )⟩ ⟨ u ( k ) ∣ is the projection operator to the N N N -fold degenerate bands satisfying P ( k ) 2 = P ( k ) P(\mathbf{k})^2=P(\mathbf{k}) P ( k ) 2 = P ( k ) . By diagonalizing it, we obtain
P ( k ) ≡ ∑ n = 1 N ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ . (487) P(\mathbf{k})\equiv\sum_{n=1}^{N}|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|. \tag*{(487)} P ( k ) ≡ n = 1 ∑ N ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣. (487)
This is Eq.[61 ] in the main text.
Hellmann-Feynman theorem
By differentiating the eigenvalue equation
H ∣ ψ n ( k ) ⟩ = ε n ( k ) ∣ ψ n ( k ) ⟩ (488) H|\psi_n(\mathbf{k})\rangle=\varepsilon_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle \tag*{(488)} H ∣ ψ n ( k )⟩ = ε n ( k ) ∣ ψ n ( k )⟩ (488)
with respect to k μ k_\mu k μ , we obtain
∂ k μ H ( k ) ∣ ψ n ( k ) ⟩ + H ( k ) ∂ k μ ∣ ψ n ( k ) ⟩ = ∂ k μ ε n ( k ) ∣ ψ n ( k ) ⟩ + ε n ( k ) ∂ k μ ∣ ψ n ( k ) ⟩ . (489) \partial_{k_\mu}H(\mathbf{k})|\psi_n(\mathbf{k})\rangle+H(\mathbf{k})\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle=\partial_{k_\mu}\varepsilon_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\varepsilon_n(\mathbf{k})\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle. \tag*{(489)} ∂ k μ H ( k ) ∣ ψ n ( k )⟩ + H ( k ) ∂ k μ ∣ ψ n ( k )⟩ = ∂ k μ ε n ( k ) ∣ ψ n ( k )⟩ + ε n ( k ) ∂ k μ ∣ ψ n ( k )⟩ . (489)
Applying ⟨ ψ m ( k ) ∣ \langle\psi_m(\mathbf{k})| ⟨ ψ m ( k ) ∣ to the above equation, we obtain
⟨ ψ m ( k ) ∣ ∂ k μ H ( k ) ∣ ψ n ( k ) ⟩ + ⟨ ψ m ( k ) ∣ H ( k ) ∂ k μ ∣ ψ n ( k ) ⟩ = ⟨ ψ m ( k ) ∣ ∂ k μ ε n ( k ) ∣ ψ n ( k ) ⟩ + ⟨ ψ m ( k ) ∣ ε n ( k ) ∂ k μ ∣ ψ n ( k ) ⟩ . (490) \langle\psi_m(\mathbf{k})|\partial_{k_\mu}H(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\langle\psi_m(\mathbf{k})|H(\mathbf{k})\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle=\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\varepsilon_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\langle\psi_m(\mathbf{k})|\varepsilon_n(\mathbf{k})\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle. \tag*{(490)} ⟨ ψ m ( k ) ∣ ∂ k μ H ( k ) ∣ ψ n ( k )⟩ + ⟨ ψ m ( k ) ∣ H ( k ) ∂ k μ ∣ ψ n ( k )⟩ = ⟨ ψ m ( k ) ∣ ∂ k μ ε n ( k ) ∣ ψ n ( k )⟩ + ⟨ ψ m ( k ) ∣ ε n ( k ) ∂ k μ ∣ ψ n ( k )⟩ . (490)
Using (C1) and defining the velocity
v μ = ∂ H ( k ) ℏ ∂ k μ , (491) v_\mu=\frac{\partial H(\mathbf{k})}{\hbar\partial k_\mu}, \tag*{(491)} v μ = ℏ ∂ k μ ∂ H ( k ) , (491)
we obtain
ℏ ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k ) ⟩ + ⟨ ψ m ( k ) ∣ ε m ( k ) ∂ k μ ∣ ψ n ( k ) ⟩ = ⟨ ψ m ( k ) ∣ ∂ k μ ε n ( k ) ∣ ψ n ( k ) ⟩ + ε n ( k ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ , (492) \hbar\langle\psi_m(\mathbf{k})|v_\mu|\psi_n(\mathbf{k})\rangle+\langle\psi_m(\mathbf{k})|\varepsilon_m(\mathbf{k})\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle=\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\varepsilon_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\varepsilon_n(\mathbf{k})\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle, \tag*{(492)} ℏ ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k )⟩ + ⟨ ψ m ( k ) ∣ ε m ( k ) ∂ k μ ∣ ψ n ( k )⟩ = ⟨ ψ m ( k ) ∣ ∂ k μ ε n ( k ) ∣ ψ n ( k )⟩ + ε n ( k ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ , (492)
where we have used the eigenequation
⟨ ψ m ( k ) ∣ H ( k ) = ⟨ ψ m ( k ) ∣ ε m ( k ) . (493) \langle\psi_m(\mathbf{k})|H(\mathbf{k})=\langle\psi_m(\mathbf{k})|\varepsilon_m(\mathbf{k}). \tag*{(493)} ⟨ ψ m ( k ) ∣ H ( k ) = ⟨ ψ m ( k ) ∣ ε m ( k ) . (493)
This equation is rewritten as
⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k ) ⟩ = 1 ℏ ∂ k μ ε n ( k ) ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ + 1 ℏ ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ . (494) \langle\psi_m(\mathbf{k})|v_\mu|\psi_n(\mathbf{k})\rangle=\frac{1}{\hbar}\partial_{k_\mu}\varepsilon_n(\mathbf{k})\langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\frac{1}{\hbar}\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle. \tag*{(494)} ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k )⟩ = ℏ 1 ∂ k μ ε n ( k ) ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ + ℏ 1 ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ . (494)
It is Eq.[68 ] in the main text. Assuming that the states satisfy the orthonormalization condition ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ = δ m n \langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle=\delta_{mn} ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ = δ mn we find
⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k ) ⟩ = 1 ℏ ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ , for m ≠ n . (495) \langle\psi_m(\mathbf{k})|v_\mu|\psi_n(\mathbf{k})\rangle=\frac{1}{\hbar}\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle,\quad\text{for }m\ne n. \tag*{(495)} ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k )⟩ = ℏ 1 ( ε n ( k ) − ε m ( k ) ) ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ , for m = n . (495)
Indeed, when n n n is the band index, it is obvious that ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ = 0 \langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle=0 ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ = 0 for m ≠ n m\ne n m = n . Furthermore, it is always possible to normalized as ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ = 1 \langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle=1 ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ = 1 at each k \mathbf{k} k . By applying ∣ ψ m ( k ) ⟩ |\psi_m(\mathbf{k})\rangle ∣ ψ m ( k )⟩ to Eq. (494) and taking a sum of m m m , we have
∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k ) ⟩ = 1 ℏ ∂ k μ ε n ( k ) ∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ψ n ( k ) ⟩ + 1 ℏ ( ε n ( k ) − ε m ( k ) ) ∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ = 1 ℏ ∂ k μ ε n ( k ) ∑ m ∣ ψ m ( k ) ⟩ δ m n + 1 ℏ ( ε n ( k ) − ε m ( k ) ) ∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ . (496) \begin{aligned}
\sum_m|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|v_\mu|\psi_n(\mathbf{k})\rangle
={}&\frac{1}{\hbar}\partial_{k_\mu}\varepsilon_n(\mathbf{k})\sum_m|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\psi_n(\mathbf{k})\rangle\\
&+\frac{1}{\hbar}\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)\sum_m|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle\\
={}&\frac{1}{\hbar}\partial_{k_\mu}\varepsilon_n(\mathbf{k})\sum_m|\psi_m(\mathbf{k})\rangle\delta_{mn}\\
&+\frac{1}{\hbar}\left(\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})\right)\sum_m|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|. \tag*{(496)}
\end{aligned} m ∑ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ v μ ∣ ψ n ( k )⟩ = = ℏ 1 ∂ k μ ε n ( k ) m ∑ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ ψ n ( k )⟩ + ℏ 1 ( ε n ( k ) − ε m ( k ) ) m ∑ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ ℏ 1 ∂ k μ ε n ( k ) m ∑ ∣ ψ m ( k )⟩ δ mn + ℏ 1 ( ε n ( k ) − ε m ( k ) ) m ∑ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣. (496)
We have
∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k ) ⟩ = ∑ m ≠ n ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ℏ v μ ∣ ψ n ( k ) ⟩ ε n ( k ) − ε m ( k ) . (497) \sum_m|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\partial_{k_\mu}|\psi_n(\mathbf{k})\rangle
=\sum_{m\ne n}|\psi_m(\mathbf{k})\rangle\frac{\langle\psi_m(\mathbf{k})|\hbar v_\mu|\psi_n(\mathbf{k})\rangle}{\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})}. \tag*{(497)} m ∑ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ∣ ψ n ( k )⟩ = m = n ∑ ∣ ψ m ( k )⟩ ε n ( k ) − ε m ( k ) ⟨ ψ m ( k ) ∣ℏ v μ ∣ ψ n ( k )⟩ . (497)
for m ≠ n m \ne n m = n . By using the complete condition ∑ m ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ = 1 \sum_m |\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|=1 ∑ m ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ = 1 , we have
∂ k μ ∣ ψ n ( k ) ⟩ = ∑ m ≠ n ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ℏ v μ ∣ ψ n ( k ) ⟩ ε n ( k ) − ε m ( k ) . (498) \partial_{k_\mu}|\psi_n(\mathbf{k})\rangle=\sum_{m\ne n}|\psi_m(\mathbf{k})\rangle\frac{\langle\psi_m(\mathbf{k})|\hbar v_\mu|\psi_n(\mathbf{k})\rangle}{\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})}. \tag*{(498)} ∂ k μ ∣ ψ n ( k )⟩ = m = n ∑ ∣ ψ m ( k )⟩ ε n ( k ) − ε m ( k ) ⟨ ψ m ( k ) ∣ℏ v μ ∣ ψ n ( k )⟩ . (498)
We use Eq. (498) to analyze the diagonal component of the quantum metric
F n n μ ν ( k ) = ⟨ ∂ k μ ψ n ( k ) ∣ Q ( k ) ∣ ∂ k ν ψ n ( k ) ⟩ = ∑ m ′ ≠ n ⟨ ψ n ( k ) ∣ ℏ v μ ∣ ψ m ′ ( k ) ⟩ ε n ( k ) − ε m ′ ( k ) ⟨ ψ m ′ ( k ) ∣ ( 1 − ∑ m ′ ′ = 1 N ∣ ψ m ′ ′ ( k ) ⟩ ⟨ ψ m ′ ′ ( k ) ∣ ) × ∑ n ′ ≠ n ∣ ψ n ′ ( k ) ⟩ ⟨ ψ n ′ ( k ) ∣ ℏ v ν ∣ ψ n ( k ) ⟩ ε n ( k ) − ε n ′ ( k ) . (499) \begin{aligned}
\mathcal{F}^{\mu\nu}_{nn}(\mathbf{k})=\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\mathcal{Q}(\mathbf{k})|\partial_{k_\nu}\psi_n(\mathbf{k})\rangle\\
&=\sum_{m'\ne n}\frac{\langle\psi_n(\mathbf{k})|\hbar v_\mu|\psi_{m'}(\mathbf{k})\rangle}{\varepsilon_n(\mathbf{k})-\varepsilon_{m'}(\mathbf{k})}\langle\psi_{m'}(\mathbf{k})|\left(1-\sum_{m''=1}^{N}|\psi_{m''}(\mathbf{k})\rangle\langle\psi_{m''}(\mathbf{k})|\right) \\
&\times\sum_{n'\ne n}|\psi_{n'}(\mathbf{k})\rangle\frac{\langle\psi_{n'}(\mathbf{k})|\hbar v_\nu|\psi_n(\mathbf{k})\rangle}{\varepsilon_n(\mathbf{k})-\varepsilon_{n'}(\mathbf{k})}. \tag*{(499)}
\end{aligned} F nn μν ( k ) = ⟨ ∂ k μ ψ n ( k ) ∣ Q ( k ) ∣ ∂ k ν ψ n ( k )⟩ = m ′ = n ∑ ε n ( k ) − ε m ′ ( k ) ⟨ ψ n ( k ) ∣ℏ v μ ∣ ψ m ′ ( k )⟩ ⟨ ψ m ′ ( k ) ∣ ( 1 − m ′′ = 1 ∑ N ∣ ψ m ′′ ( k )⟩ ⟨ ψ m ′′ ( k ) ∣ ) × n ′ = n ∑ ∣ ψ n ′ ( k )⟩ ε n ( k ) − ε n ′ ( k ) ⟨ ψ n ′ ( k ) ∣ℏ v ν ∣ ψ n ( k )⟩ . (499)
We have
⟨ ψ m ′ ( k ) ∣ ( 1 − ∑ m ′ ′ = 1 N ∣ ψ m ′ ′ ( k ) ⟩ ⟨ ψ m ′ ′ ( k ) ∣ ) ∣ ψ n ′ ( k ) ⟩ = δ m ′ n ′ − ∑ m ′ ′ = 1 N δ m ′ m ′ ′ δ m ′ ′ n ′ = 0 , \langle\psi_{m'}(\mathbf{k})|\left(1-\sum_{m''=1}^{N}|\psi_{m''}(\mathbf{k})\rangle\langle\psi_{m''}(\mathbf{k})|\right)|\psi_{n'}(\mathbf{k})\rangle
=\delta_{m'n'}-\sum_{m''=1}^{N}\delta_{m'm''}\delta_{m''n'}=0, ⟨ ψ m ′ ( k ) ∣ ( 1 − m ′′ = 1 ∑ N ∣ ψ m ′′ ( k )⟩ ⟨ ψ m ′′ ( k ) ∣ ) ∣ ψ n ′ ( k )⟩ = δ m ′ n ′ − m ′′ = 1 ∑ N δ m ′ m ′′ δ m ′′ n ′ = 0 ,
when m ′ ′ m'' m ′′ is the occupied band, while
⟨ ψ m ′ ( k ) ∣ ( 1 − ∑ m ′ ′ = 1 N ∣ ψ m ′ ′ ( k ) ⟩ ⟨ ψ m ′ ′ ( k ) ∣ ) ∣ ψ n ′ ( k ) ⟩ = δ m ′ n ′ , (500) \langle\psi_{m'}(\mathbf{k})|\left(1-\sum_{m''=1}^{N}|\psi_{m''}(\mathbf{k})\rangle\langle\psi_{m''}(\mathbf{k})|\right)|\psi_{n'}(\mathbf{k})\rangle=\delta_{m'n'}, \tag*{(500)} ⟨ ψ m ′ ( k ) ∣ ( 1 − m ′′ = 1 ∑ N ∣ ψ m ′′ ( k )⟩ ⟨ ψ m ′′ ( k ) ∣ ) ∣ ψ n ′ ( k )⟩ = δ m ′ n ′ , (500)
when m ′ ′ m'' m ′′ is the unoccupied band.
F n n μ ν ( k ) = ∑ m ∈ U n o c c u p i e d ⟨ ψ n ( k ) ∣ ℏ v μ ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ℏ v ν ∣ ψ n ( k ) ⟩ [ ε n ( k ) − ε m ( k ) ] 2 . (501) \mathcal{F}^{\mu\nu}_{nn}(\mathbf{k})=\sum_{m\in\mathrm{Unoccupied}}\frac{\langle\psi_n(\mathbf{k})|\hbar v_\mu|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\hbar v_\nu|\psi_n(\mathbf{k})\rangle}{[\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})]^2}. \tag*{(501)} F nn μν ( k ) = m ∈ Unoccupied ∑ [ ε n ( k ) − ε m ( k ) ] 2 ⟨ ψ n ( k ) ∣ℏ v μ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ℏ v ν ∣ ψ n ( k )⟩ . (501)
The quantum metric reads
g n μ ν = Re ∑ m ∈ U n o c c u p i e d ⟨ ψ n ( k ) ∣ ℏ v μ ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ℏ v ν ∣ ψ n ( k ) ⟩ [ ε n ( k ) − ε m ( k ) ] 2 , (502) g^{\mu\nu}_n=\operatorname{Re}\sum_{m\in\mathrm{Unoccupied}}\frac{\langle\psi_n(\mathbf{k})|\hbar v_\mu|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\hbar v_\nu|\psi_n(\mathbf{k})\rangle}{[\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})]^2}, \tag*{(502)} g n μν = Re m ∈ Unoccupied ∑ [ ε n ( k ) − ε m ( k ) ] 2 ⟨ ψ n ( k ) ∣ℏ v μ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ℏ v ν ∣ ψ n ( k )⟩ , (502)
while the Berry curvature reads
Ω n μ ν = 2 Im ∑ m ∈ U n o c c u p i e d ⟨ ψ n ( k ) ∣ ℏ v μ ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ℏ v ν ∣ ψ n ( k ) ⟩ [ ε n ( k ) − ε m ( k ) ] 2 . (503) \Omega^{\mu\nu}_n=2\operatorname{Im}\sum_{m\in\mathrm{Unoccupied}}\frac{\langle\psi_n(\mathbf{k})|\hbar v_\mu|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\hbar v_\nu|\psi_n(\mathbf{k})\rangle}{[\varepsilon_n(\mathbf{k})-\varepsilon_m(\mathbf{k})]^2}. \tag*{(503)} Ω n μν = 2 Im m ∈ Unoccupied ∑ [ ε n ( k ) − ε m ( k ) ] 2 ⟨ ψ n ( k ) ∣ℏ v μ ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ℏ v ν ∣ ψ n ( k )⟩ . (503)
Bulk photovoltaic effects
We apply alternating and monochromatic electric fields,
E x ( t ) = E x ( ω ) e − i ω t + E x ( − ω ) e i ω t (504) E_x(t)=E_x(\omega)e^{-i\omega t}+E_x(-\omega)e^{i\omega t} \tag*{(504)} E x ( t ) = E x ( ω ) e − iω t + E x ( − ω ) e iω t (504)
with ω > 0 \omega>0 ω > 0 . We assume that the electric field is real,
E x ( t ) = E x ∗ ( t ) , (505) E_x(t)=E_x^*(t), \tag*{(505)} E x ( t ) = E x ∗ ( t ) , (505)
or
E x ( ω ) = E x ∗ ( − ω ) . (506) E_x(\omega)=E_x^*(-\omega). \tag*{(506)} E x ( ω ) = E x ∗ ( − ω ) . (506)
Excited electrons from the valence band and the conduction band contribute to the bulk photovoltaic current.
Injection currents
We derive the injection current. We consider an acceleration due to the static electric field and an optical excitation from the valence band to the conduction band due to the oscillatory electric field, separately[218 , 219 , 220 , 221 ] . The equation of motion of electrons is
ℏ d k d t = − e ∂ A 0 ∂ t , (507) \hbar\frac{d\mathbf{k}}{dt}=-e\frac{\partial\mathbf{A}_0}{\partial t}, \tag*{(507)} ℏ d t d k = − e ∂ t ∂ A 0 , (507)
where A 0 = ( A 0 , 0 , 0 ) \mathbf{A}_0=(A_0,0,0) A 0 = ( A 0 , 0 , 0 ) is the vector potential. The static electric field is obtained from the vector potential as
E x ( 0 ) ≡ − ∂ A 0 ∂ t , (508) E_x(0)\equiv-\frac{\partial A_0}{\partial t}, \tag*{(508)} E x ( 0 ) ≡ − ∂ t ∂ A 0 , (508)
or A 0 = − E x ( 0 ) t + constant A_0=-E_x(0)t+\text{constant} A 0 = − E x ( 0 ) t + constant . The Bloch velocity under electric field is given by the minimal substitution,
v n ( k x ) → v n ( k x − e A 0 ℏ ) . (509) v_n(k_x)\to v_n\left(k_x-\frac{eA_0}{\hbar}\right). \tag*{(509)} v n ( k x ) → v n ( k x − ℏ e A 0 ) . (509)
A comment is in order. We have only considered the contribution from the static electric field in the vector potential. Even if we take into account the contribution from the oscillatory electric field in the vector potential, the time evolution of the mean momentum does not change. It is understood as follows. The equation of motion under the oscillatory electric field reads
ℏ d k d t = e E x ( 0 ) + e E x ( ω ) cos ω t , (510) \hbar\frac{d\mathbf{k}}{dt}=eE_x(0)+eE_x(\omega)\cos\omega t, \tag*{(510)} ℏ d t d k = e E x ( 0 ) + e E x ( ω ) cos ω t , (510)
whose solution is given by
k = e ℏ ( t E x ( 0 ) + E x ( ω ) ω sin ω t ) + k 0 . (511) k=\frac{e}{\hbar}\left(tE_x(0)+\frac{E_x(\omega)}{\omega}\sin\omega t\right)+k_0. \tag*{(511)} k = ℏ e ( t E x ( 0 ) + ω E x ( ω ) sin ω t ) + k 0 . (511)
Its mean for one period with respect to the time t t t is zero. Hence, it is enough only to consider the static electric field in the vector potential. We expand Eq. (509) in powers of E x ( 0 ) E_x(0) E x ( 0 ) , and obtain
d 2 v n x d t 2 = ( e ℏ ) 2 ∂ 3 w n ∂ k x 3 [ E x ( 0 ) ] 2 , (512) \frac{d^2v_n^x}{dt^2}=\left(\frac{e}{\hbar}\right)^2\frac{\partial^3w_n}{\partial k_x^3}[E_x(0)]^2, \tag*{(512)} d t 2 d 2 v n x = ( ℏ e ) 2 ∂ k x 3 ∂ 3 w n [ E x ( 0 ) ] 2 , (512)
where we have used
d v n x d t = d k x d t d v n x d k x = − ( e ℏ ) ∂ A 0 ∂ t d 2 w n d k x 2 = e ℏ d 2 w n d k x 2 E x ( 0 ) . (513) \frac{dv_n^x}{dt}=\frac{dk_x}{dt}\frac{dv_n^x}{dk_x}=-\left(\frac{e}{\hbar}\right)\frac{\partial A_0}{\partial t}\frac{d^2w_n}{dk_x^2}=\frac{e}{\hbar}\frac{d^2w_n}{dk_x^2}E_x(0). \tag*{(513)} d t d v n x = d t d k x d k x d v n x = − ( ℏ e ) ∂ t ∂ A 0 d k x 2 d 2 w n = ℏ e d k x 2 d 2 w n E x ( 0 ) . (513)
The current is given by
J = e V ∑ n , k f n v n x (514) J=\frac{e}{V}\sum_{n,\mathbf{k}}f_n v_n^x \tag*{(514)} J = V e n , k ∑ f n v n x (514)
with the velocity along the x x x direction
v n x = ∂ ω n ∂ k x . (515) v_n^x=\frac{\partial\omega_n}{\partial k_x}. \tag*{(515)} v n x = ∂ k x ∂ ω n . (515)
The injection current originates in d J x ; x 2 / d t dJ^{x;x^2}/dt d J x ; x 2 / d t . The injection current is obtained as
d J i n j e c t x ; x 2 d t = e V ∑ n ∑ s = 0 1 ( ℓ − 1 s ) d s f n d t s d 1 − s v n x d t 1 − s . (516) \frac{dJ_{\mathrm{inject}}^{x;x^2}}{dt}=\frac{e}{V}\sum_n\sum_{s=0}^{1}\binom{\ell-1}{s}\frac{d^s f_n}{dt^s}\frac{d^{1-s}v_n^x}{dt^{1-s}}. \tag*{(516)} d t d J inject x ; x 2 = V e n ∑ s = 0 ∑ 1 ( s ℓ − 1 ) d t s d s f n d t 1 − s d 1 − s v n x . (516)
The term proportional to E x ( ω ) E x ( − ω ) E_x(\omega)E_x(-\omega) E x ( ω ) E x ( − ω ) is given by taking the terms with s = 1 s=1 s = 1 from Eq. (516) as
d J i n j e c t x ; x 2 d t = e V ∑ n d f n d t v n x . (517) \frac{dJ_{\mathrm{inject}}^{x;x^2}}{dt}=\frac{e}{V}\sum_n\frac{df_n}{dt}v_n^x. \tag*{(517)} d t d J inject x ; x 2 = V e n ∑ d t d f n v n x . (517)
The Fermi golden rule reads
d f ± d t = ± 2 π e 2 ℏ 2 ∣ E ( ω ) ⋅ a + − ∣ 2 δ ( ω + − − ω ) , (518) \frac{df_\pm}{dt}=\pm\frac{2\pi e^2}{\hbar^2}\left|\mathbf{E}(\omega)\cdot\mathbf{a}_{+-}\right|^2\delta(\omega_{+-}-\omega), \tag*{(518)} d t d f ± = ± ℏ 2 2 π e 2 ∣ E ( ω ) ⋅ a +− ∣ 2 δ ( ω +− − ω ) , (518)
where
E ( ω ) ⋅ r + − 2 = ∣ E x ( ω ) a + − x ∣ 2 = E x ( ω ) a + − x E x ∗ ( ω ) a + − x ∗ = − E x ( ω ) E x ( − ω ) a + − x a − + x , (519) \mathbf{E}(\omega)\cdot\mathbf{r}_{+-}^2=\left|E_x(\omega)a_{+-}^x\right|^2=E_x(\omega)a_{+-}^xE_x^*(\omega)a_{+-}^{x*}
=-E_x(\omega)E_x(-\omega)a_{+-}^xa_{-+}^x, \tag*{(519)} E ( ω ) ⋅ r +− 2 = E x ( ω ) a +− x 2 = E x ( ω ) a +− x E x ∗ ( ω ) a +− x ∗ = − E x ( ω ) E x ( − ω ) a +− x a −+ x , (519)
with the use of the Hermitian condition Eq. (17) . Inserting it into Eq. (517) , we have
J i n j e c t x ; x 2 = 2 π e 2 ℏ 2 e V a − + x a + − x δ ( ω + − − ω ) × ∂ 3 ω + − ∂ k x 3 E x ( ω ) E x ( − ω ) . (520) J_{\mathrm{inject}}^{x;x^2}=\frac{2\pi e^2}{\hbar^2}\frac{e}{V}a_{-+}^xa_{+-}^x\delta(\omega_{+-}-\omega)\times\frac{\partial^3\omega_{+-}}{\partial k_x^3}E_x(\omega)E_x(-\omega). \tag*{(520)} J inject x ; x 2 = ℏ 2 2 π e 2 V e a −+ x a +− x δ ( ω +− − ω ) × ∂ k x 3 ∂ 3 ω +− E x ( ω ) E x ( − ω ) . (520)
By assuming a monochromic oscillation J x ; x ℓ ∝ e − i ω 0 t J^{x;x^\ell}\propto e^{-i\omega_0t} J x ; x ℓ ∝ e − i ω 0 t , we obtain the ℓ \ell ℓ -th order current J i n j e c t x ; x ℓ J_{\mathrm{inject}}^{x;x^\ell} J inject x ; x ℓ ,
J i n j e c t x ; x ℓ = 1 i ω 0 + 1 / τ 2 π e 3 ℏ 2 V a − + x a + − x δ ( ω + − − ω ) × ∂ ω + − ∂ k x E x ( ω ) E x ( − ω ) [ E x ( 0 ) ] ℓ − 2 , (521) J_{\mathrm{inject}}^{x;x^\ell}=\frac{1}{i\omega_0+1/\tau}\frac{2\pi e^3}{\hbar^2V}a_{-+}^xa_{+-}^x\delta(\omega_{+-}-\omega)\times\frac{\partial\omega_{+-}}{\partial k_x}E_x(\omega)E_x(-\omega)[E_x(0)]^{\ell-2}, \tag*{(521)} J inject x ; x ℓ = i ω 0 + 1/ τ 1 ℏ 2 V 2 π e 3 a −+ x a +− x δ ( ω +− − ω ) × ∂ k x ∂ ω +− E x ( ω ) E x ( − ω ) [ E x ( 0 ) ] ℓ − 2 , (521)
where we have introduced a cutoff by the relaxation time τ \tau τ . When we concentrate on the direct current component ω 0 = 0 \omega_0=0 ω 0 = 0 , it is proportional to τ \tau τ . The longitudinal component is
σ i n j e c t x ; x 2 σ i n j e c t = ∑ k f − + ∂ ω + − ∂ k x a − + x a + − x δ ( ω + − − ω ) , (522) \frac{\sigma_{\mathrm{inject}}^{x;x^2}}{\sigma_{\mathrm{inject}}}=\sum_{\mathbf{k}}f_{-+}\frac{\partial\omega_{+-}}{\partial k_x}a_{-+}^xa_{+-}^x\delta(\omega_{+-}-\omega), \tag*{(522)} σ inject σ inject x ; x 2 = k ∑ f −+ ∂ k x ∂ ω +− a −+ x a +− x δ ( ω +− − ω ) , (522)
with
σ i n j e c t ≡ 2 π V e 3 ℏ 2 τ . (523) \sigma_{\mathrm{inject}}\equiv2\pi V\frac{e^3}{\hbar^2}\tau. \tag*{(523)} σ inject ≡ 2 π V ℏ 2 e 3 τ . (523)
This is Eq. (120) in the main text.
Shift currents
The shift current is induced by the difference of the position between the valence band and the conduction band when electrons are excited from the valence band to the conduction band. It induces the electric dipole E x ( t ) a + − ; x x E_x(t)a_{+-;x}^x E x ( t ) a +− ; x x . Hence, the shift current originates in a quantum interference[221 ] between the oscillation of ρ m n \rho_{mn} ρ mn and the oscillation of the dipole velocity E x a n m ; x x E_xa_{nm;x}^x E x a nm ; x x ,
J s h i f t x ; x 2 = − e 2 ℏ V ∑ k E x a + − ; x x ρ + − ( 1 ) , (524) J_{\mathrm{shift}}^{x;x^2}=-\frac{e^2}{\hbar V}\sum_{\mathbf{k}}E_xa_{+-;x}^x\rho_{+-}^{(1)}, \tag*{(524)} J shift x ; x 2 = − ℏ V e 2 k ∑ E x a +− ; x x ρ +− ( 1 ) , (524)
which is defined by the interband contributions a + − ; x x a_{+-;x}^{x} a +− ; x x and ρ + − ( 1 ) ( t ) \rho_{+-}^{(1)}(t) ρ +− ( 1 ) ( t ) . The density matrix is obtained by solving the von-Neumann equation[89 , 90 , 221 ] ,
∂ ρ m n ∂ t + i ω m n ρ m n = e i ℏ ∑ s E x ( ρ m l a s n x − a m s x ρ s n ) − e ℏ E x ρ m n ; x . (525) \frac{\partial\rho_{mn}}{\partial t}+i\omega_{mn}\rho_{mn}=\frac{e}{i\hbar}\sum_sE_x\left(\rho_{ml}a_{sn}^x-a_{ms}^x\rho_{sn}\right)-\frac{e}{\hbar}E_x\rho_{mn;x}. \tag*{(525)} ∂ t ∂ ρ mn + i ω mn ρ mn = i ℏ e s ∑ E x ( ρ m l a s n x − a m s x ρ s n ) − ℏ e E x ρ mn ; x . (525)
For the two-band system, it is rewritten as
i ( ω + − − ω 0 ) ρ + − = e i ℏ E x a + − x ( ρ + + − ρ − − ) − e ℏ E x ρ + − ; x , (526) i(\omega_{+-}-\omega_0)\rho_{+-}=\frac{e}{i\hbar}E_xa_{+-}^x(\rho_{++}-\rho_{--})-\frac{e}{\hbar}E_x\rho_{+-;x}, \tag*{(526)} i ( ω +− − ω 0 ) ρ +− = i ℏ e E x a +− x ( ρ ++ − ρ −− ) − ℏ e E x ρ +− ; x , (526)
where we have used a n n x = 0 a_{nn}^x=0 a nn x = 0 and assumed a monochromatic solution ρ m n ∝ e − i ω 0 t \rho_{mn}\propto e^{-i\omega_0t} ρ mn ∝ e − i ω 0 t . This definition recovers properly the definition of the shift current for ℓ = 2 \ell=2 ℓ = 2 .
The zeroth order solution is given by
ρ ± ( 0 ) = f ± , (527) \rho_\pm^{(0)}=f_\pm, \tag*{(527)} ρ ± ( 0 ) = f ± , (527)
because electric field is not applied. The first order solution is given by
ρ + − ( 1 ) = i e ℏ ( ω + − − ω 0 ) E x a + − x ( ρ + + ( 0 ) − ρ − − ( 0 ) ) e − i ω 0 t = i e ℏ ( ω + − − ω 0 ) E x a + − x f − + e − i ω 0 t . (528) \rho_{+-}^{(1)}=\frac{ie}{\hbar(\omega_{+-}-\omega_0)}E_xa_{+-}^x\left(\rho_{++}^{(0)}-\rho_{--}^{(0)}\right)e^{-i\omega_0t}
=\frac{ie}{\hbar(\omega_{+-}-\omega_0)}E_xa_{+-}^xf_{-+}e^{-i\omega_0t}. \tag*{(528)} ρ +− ( 1 ) = ℏ ( ω +− − ω 0 ) i e E x a +− x ( ρ ++ ( 0 ) − ρ −− ( 0 ) ) e − i ω 0 t = ℏ ( ω +− − ω 0 ) i e E x a +− x f −+ e − i ω 0 t . (528)
Especially, we have ρ + + ( 1 ) = 0 \rho_{++}^{(1)}=0 ρ ++ ( 1 ) = 0 because f + + ≡ f + − f + = 0 f_{++}\equiv f_+-f_+=0 f ++ ≡ f + − f + = 0 . Hence, the shift current is obtained as
J s h i f t x ; x 2 = − e 2 ℏ V ∑ k E x a + − ; x x ρ + − ( 1 ) = − e 2 ℏ V e − i ω 0 t ∑ k a + − ; x x f − + a + − x E x l . (529) J_{\mathrm{shift}}^{x;x^2}=-\frac{e^2}{\hbar V}\sum_{\mathbf{k}}E_xa_{+-;x}^x\rho_{+-}^{(1)}
=-\frac{e^2}{\hbar V}e^{-i\omega_0t}\sum_{\mathbf{k}}a_{+-;x}^xf_{-+}a_{+-}^xE_x^l. \tag*{(529)} J shift x ; x 2 = − ℏ V e 2 k ∑ E x a +− ; x x ρ +− ( 1 ) = − ℏ V e 2 e − i ω 0 t k ∑ a +− ; x x f −+ a +− x E x l . (529)
We use the monochromatic condition (D1) and study the direct current component,
J s h i f t x ; x 2 = σ s h i f t x ; x 2 ( 0 ; ω , − ω , 0 , 0 , ⋯ , 0 ) × E x ( ω ) E x ( − ω ) . (530) J_{\mathrm{shift}}^{x;x^2}=\sigma_{\mathrm{shift}}^{x;x^2}(0;\omega,-\omega,0,0,\cdots,0)\times E_x(\omega)E_x(-\omega). \tag*{(530)} J shift x ; x 2 = σ shift x ; x 2 ( 0 ; ω , − ω , 0 , 0 , ⋯ , 0 ) × E x ( ω ) E x ( − ω ) . (530)
Then, the shift current is obtained as
J s h i f t x ; x z = − e 2 ℏ V ∑ k a + − ; x x a + − x f − + E x ( ω ) E x ( − ω ) δ ( ω + − − ω ) (531) J_{\mathrm{shift}}^{x;xz}=-\frac{e^2}{\hbar V}\sum_{\mathbf{k}}a_{+-;x}^{x}a_{+-}^{x}f_{-+}E_x(\omega)E_x(-\omega)\delta(\omega_{+-}-\omega) \tag*{(531)} J shift x ; x z = − ℏ V e 2 k ∑ a +− ; x x a +− x f −+ E x ( ω ) E x ( − ω ) δ ( ω +− − ω ) (531)
It is independent of τ \tau τ . The shift conductivity is given by
σ s h i f t x ; x z σ s h i f t = 1 V ∑ k f − + a + − ; x x a + − x δ ( ω + − − ω ) (532) \frac{\sigma_{\mathrm{shift}}^{x;xz}}{\sigma_{\mathrm{shift}}}=\frac{1}{V}\sum_{\mathbf{k}}f_{-+}a_{+-;x}^{x}a_{+-}^{x}\delta(\omega_{+-}-\omega) \tag*{(532)} σ shift σ shift x ; x z = V 1 k ∑ f −+ a +− ; x x a +− x δ ( ω +− − ω ) (532)
with
σ s h i f t ≡ − e 3 ℏ 2 (533) \sigma_{\mathrm{shift}}\equiv-\frac{e^3}{\hbar^2} \tag*{(533)} σ shift ≡ − ℏ 2 e 3 (533)
This is Eq. (123) in the main text.
Quantum distance and Zeeman quantum geometric tensor
The Zeeman quantum geometric tensor g n m μ ν g_{nm}^{\mu\nu} g nm μν is defined by the quantum distance d S H S d_{\mathrm{SHS}} d SHS for the infinitesimal translation d k d\mathbf{k} d k of the momentum as [110 , 111 , 112 , 113 ]
d S H S ( k ) ≡ 1 − ∣ ⟨ ψ n ( k ) ∣ U d θ U d k ∣ ψ n ( k ) ⟩ ∣ 2 (534) d_{\mathrm{SHS}}(\mathbf{k})\equiv\sqrt{1-\left|\langle\psi_n(\mathbf{k})|U_{d\theta}U_{d\mathbf{k}}|\psi_n(\mathbf{k})\rangle\right|^2} \tag*{(534)} d SHS ( k ) ≡ 1 − ∣ ⟨ ψ n ( k ) ∣ U d θ U d k ∣ ψ n ( k )⟩ ∣ 2 (534)
where
U d θ ≡ e − i 2 d θ ⋅ σ , U d k ≡ e − i d k ⋅ r (535) U_{d\theta}\equiv e^{-\frac{i}{2}d\theta\cdot\sigma},\qquad U_{d\mathbf{k}}\equiv e^{-i d\mathbf{k}\cdot\mathbf{r}} \tag*{(535)} U d θ ≡ e − 2 i d θ ⋅ σ , U d k ≡ e − i d k ⋅ r (535)
are the generators of the spin angular momentum d θ d\theta d θ and the momentum d k d\mathbf{k} d k . By using the relation
U d θ U d k ∣ ψ n ⟩ = ( 1 − i σ μ 2 d θ ν − σ μ σ ν 8 d θ μ d θ ν ⋯ ) × ( 1 + ∂ k μ d k μ + 1 2 ∣ ∂ k μ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν ⋯ ) ∣ ψ n ⟩ = 1 + 1 2 ∣ ∂ k μ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν − i σ μ 2 ∣ ψ n ⟩ d θ μ − 1 8 ∣ ψ n ⟩ d θ μ d θ μ − i σ ν 2 ∂ k μ ∣ ψ n ⟩ d k μ d θ ν + ⋯ (536) \begin{aligned}
U_{d\theta}U_{d\mathbf{k}}|\psi_n\rangle
&=\left(1-\frac{i\sigma_\mu}{2}d\theta_\nu-\frac{\sigma_\mu\sigma_\nu}{8}d\theta_\mu d\theta_\nu\cdots\right)\\
&\quad\times\left(1+\partial_{k_\mu}d k_\mu+\frac{1}{2}\left|\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})\right\rangle d k_\mu d k_\nu\cdots\right)|\psi_n\rangle\\
&=1+\frac{1}{2}\left|\partial_{k_\mu}\partial_{k_\nu}\psi_n(\mathbf{k})\right\rangle d k_\mu d k_\nu\\
&\quad-\frac{i\sigma_\mu}{2}|\psi_n\rangle d\theta^\mu-\frac{1}{8}|\psi_n\rangle d\theta^\mu d\theta^\mu-\frac{i\sigma_\nu}{2}\partial_{k_\mu}|\psi_n\rangle d k_\mu d\theta^\nu+\cdots \tag*{(536)}
\end{aligned} U d θ U d k ∣ ψ n ⟩ = ( 1 − 2 i σ μ d θ ν − 8 σ μ σ ν d θ μ d θ ν ⋯ ) × ( 1 + ∂ k μ d k μ + 2 1 ∂ k μ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν ⋯ ) ∣ ψ n ⟩ = 1 + 2 1 ∂ k μ ∂ k ν ψ n ( k ) ⟩ d k μ d k ν − 2 i σ μ ∣ ψ n ⟩ d θ μ − 8 1 ∣ ψ n ⟩ d θ μ d θ μ − 2 i σ ν ∂ k μ ∣ ψ n ⟩ d k μ d θ ν + ⋯ (536)
the quantum distance is given by
( d S H S ) 2 = ∑ n ≠ m g n m μ ν d k μ d k ν + ∑ n ≠ m s n m μ ν d θ μ d θ ν 4 + ∑ n ≠ m ( z n m μ ν + z m n μ ν ) d k μ d θ ν 2 (537) \begin{aligned}
(d_{\mathrm{SHS}})^2
&=\sum_{n\ne m}g_{nm}^{\mu\nu}d k_\mu d k_\nu+\sum_{n\ne m}s_{nm}^{\mu\nu}\frac{d\theta_\mu d\theta_\nu}{4}\\
&\quad+\sum_{n\ne m}\left(z_{nm}^{\mu\nu}+z_{mn}^{\mu\nu}\right)d k_\mu\frac{d\theta_\nu}{2} \tag*{(537)}
\end{aligned} ( d SHS ) 2 = n = m ∑ g nm μν d k μ d k ν + n = m ∑ s nm μν 4 d θ μ d θ ν + n = m ∑ ( z nm μν + z mn μν ) d k μ 2 d θ ν (537)
where g n m μ ν g_{nm}^{\mu\nu} g nm μν is the quantum metric, s n m μ ν s_{nm}^{\mu\nu} s nm μν is the spin quantum geometric tensor and z n m μ ν z_{nm}^{\mu\nu} z nm μν is the Zeeman quantum geometric tensor.
We determine z n m μ ν z_{nm}^{\mu\nu} z nm μν by calculating the coefficient of d k μ d θ ν 2 d k_\mu\frac{d\theta_\nu}{2} d k μ 2 d θ ν . The fidelity is calculated as
∣ ⟨ ψ n ( k ) ∣ ψ n ( k + d k ) ⟩ ∣ 2 = ⟨ ψ n ( k + d k ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ψ n ( k + d k ) ⟩ = ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ + ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k ) ⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ + ( − i 2 ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ + i 2 ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ + i 2 ⟨ ∂ k μ ψ n ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ − i 2 ⟨ ψ n ( k ) ∣ σ ν ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ d θ ν = 1 + i 2 ( ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k ) ) σ ν ∣ ψ n ( k ) ⟩ − ⟨ ψ n ( k ) ∣ ( 1 − P ( k ) ) σ ν ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ d θ ν 2 = 1 + i 2 ( ∑ m ≠ n ⟨ ∂ k μ ψ n ( k ) ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ − ⟨ ψ n ( k ) ∣ σ ν ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ d θ ν 2 = 1 − 1 2 ∑ m ≠ n ( ⟨ ψ n ( k ) ∣ ∂ k μ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ + ⟨ ψ n ( k ) ∣ σ ν ∣ ψ m ( k ) ⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k ) ⟩ ) d k μ d θ ν 2 = 1 − 1 2 ∑ m ≠ n ( a n m μ ν s n m μ ν + s n m μ ν a n m μ ν ) d k μ d θ ν 2 (538) \begin{aligned}
\left|\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}+d\mathbf{k})\rangle\right|^2
&=\langle\psi_n(\mathbf{k}+d\mathbf{k})|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}+d\mathbf{k})\rangle\\
&=\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\left(\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)d k_\mu\\
&\quad+\left(-\frac{i}{2}\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle\right.\\
&\qquad+\frac{i}{2}\langle\psi_n(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\\
&\qquad\left.+\frac{i}{2}\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle-\frac{i}{2}\langle\psi_n(\mathbf{k})|\sigma_\nu|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)d k_\mu d\theta_\nu\\
&=1+\frac{i}{2}\left(\langle\partial_{k_\mu}\psi_n(\mathbf{k})|(1-P(\mathbf{k}))\sigma_\nu|\psi_n(\mathbf{k})\rangle\right.\\
&\qquad\left.-\langle\psi_n(\mathbf{k})|(1-P(\mathbf{k}))\sigma_\nu|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)d k_\mu\frac{d\theta_\nu}{2}\\
&=1+\frac{i}{2}\left(\sum_{m\ne n}\langle\partial_{k_\mu}\psi_n(\mathbf{k})|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle\right.\\
&\qquad\left.-\langle\psi_n(\mathbf{k})|\sigma_\nu|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)d k_\mu\frac{d\theta_\nu}{2}\\
&=1-\frac{1}{2}\sum_{m\ne n}\left(\langle\psi_n(\mathbf{k})|\partial_{k_\mu}\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle\right.\\
&\qquad\left.+\langle\psi_n(\mathbf{k})|\sigma_\nu|\psi_m(\mathbf{k})\rangle\langle\psi_m(\mathbf{k})|\partial_{k_\mu}\psi_n(\mathbf{k})\rangle\right)d k_\mu\frac{d\theta_\nu}{2}\\
&=1-\frac{1}{2}\sum_{m\ne n}\left(a_{nm}^{\mu\nu}s_{nm}^{\mu\nu}+s_{nm}^{\mu\nu}a_{nm}^{\mu\nu}\right)d k_\mu\frac{d\theta_\nu}{2} \tag*{(538)}
\end{aligned} ∣ ⟨ ψ n ( k ) ∣ ψ n ( k + d k )⟩ ∣ 2 = ⟨ ψ n ( k + d k ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ψ n ( k + d k )⟩ = ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ + ( ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k )⟩ + ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) d k μ + ( − 2 i ⟨ ∂ k μ ψ n ( k ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k )⟩ + 2 i ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ∂ k μ ψ n ( k )⟩ + 2 i ⟨ ∂ k μ ψ n ( k ) ∣ σ ν ∣ ψ n ( k )⟩ − 2 i ⟨ ψ n ( k ) ∣ σ ν ∣ ∂ k μ ψ n ( k )⟩ ) d k μ d θ ν = 1 + 2 i ( ⟨ ∂ k μ ψ n ( k ) ∣ ( 1 − P ( k )) σ ν ∣ ψ n ( k )⟩ − ⟨ ψ n ( k ) ∣ ( 1 − P ( k )) σ ν ∣ ∂ k μ ψ n ( k )⟩ ) d k μ 2 d θ ν = 1 + 2 i m = n ∑ ⟨ ∂ k μ ψ n ( k ) ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ σ ν ∣ ψ n ( k )⟩ − ⟨ ψ n ( k ) ∣ σ ν ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) d k μ 2 d θ ν = 1 − 2 1 m = n ∑ ( ⟨ ψ n ( k ) ∣ ∂ k μ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ σ ν ∣ ψ n ( k )⟩ + ⟨ ψ n ( k ) ∣ σ ν ∣ ψ m ( k )⟩ ⟨ ψ m ( k ) ∣ ∂ k μ ψ n ( k )⟩ ) d k μ 2 d θ ν = 1 − 2 1 m = n ∑ ( a nm μν s nm μν + s nm μν a nm μν ) d k μ 2 d θ ν (538)
where we have used Eq. (15) . By comparing it to Eq. (537) , we obtain
z n m μ ν = a n m μ ν s n m μ ν (539) z_{nm}^{\mu\nu}=a_{nm}^{\mu\nu}s_{nm}^{\mu\nu} \tag*{(539)} z nm μν = a nm μν s nm μν (539)
which is the Zeeman geometric tensor. This is Eq. (143) in the main text.
We calculate s n m μ ν s_{nm}^{\mu\nu} s nm μν as
∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ ∣ 2 = ⟨ ψ n ( k ′ ) ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ ψ n ( k ′ ) ⟩ = ( ⟨ ψ n ( k ) ∣ + ⟨ ψ n ( k ) ∣ i σ μ 2 d θ μ + ⋯ ) ∣ ψ n ( k ) ⟩ × ⟨ ψ n ( k ) ∣ ( ∣ ψ n ( k ) ⟩ + i σ μ 2 ∣ ψ n ( k ) ⟩ d θ μ ν + ⋯ ) = ⟨ ψ n ( k ) ∣ ψ n ( k ) ⟩ + ( i 2 ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k ) ⟩ − i σ μ 2 ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k ) ⟩ ) d θ μ + ( − 1 4 ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ ) d θ μ d θ ν = 1 + 1 4 ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k ) ⟩ ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k ) ⟩ d θ μ d θ ν (540) \begin{aligned}
\left|\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle\right|^2
&=\langle\psi_n(\mathbf{k}')|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k}')\rangle\\
&=\left(\langle\psi_n(\mathbf{k})|+\langle\psi_n(\mathbf{k})|\frac{i\sigma_\mu}{2}d\theta^\mu+\cdots\right)|\psi_n(\mathbf{k})\rangle\\
&\quad\times\langle\psi_n(\mathbf{k})|\left(|\psi_n(\mathbf{k})\rangle+\frac{i\sigma_\mu}{2}|\psi_n(\mathbf{k})\rangle d\theta^{\mu\nu}+\cdots\right)\\
&=\langle\psi_n(\mathbf{k})|\psi_n(\mathbf{k})\rangle+\left(\frac{i}{2}\langle\psi_n(\mathbf{k})|\sigma_\mu|\psi_n(\mathbf{k})\rangle-\frac{i\sigma_\mu}{2}\langle\psi_n(\mathbf{k})|\sigma_\mu|\psi_n(\mathbf{k})\rangle\right)d\theta^\mu\\
&\quad+\left(-\frac{1}{4}\langle\psi_n(\mathbf{k})|\sigma_\mu|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle\right)d\theta^\mu d\theta^\nu\\
&=1+\frac{1}{4}\langle\psi_n(\mathbf{k})|\sigma_\mu|\psi_n(\mathbf{k})\rangle\langle\psi_n(\mathbf{k})|\sigma_\nu|\psi_n(\mathbf{k})\rangle d\theta^\mu d\theta^\nu \tag*{(540)}
\end{aligned} ∣ ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ ∣ 2 = ⟨ ψ n ( k ′ ) ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ ψ n ( k ′ )⟩ = ( ⟨ ψ n ( k ) ∣ + ⟨ ψ n ( k ) ∣ 2 i σ μ d θ μ + ⋯ ) ∣ ψ n ( k )⟩ × ⟨ ψ n ( k ) ∣ ( ∣ ψ n ( k )⟩ + 2 i σ μ ∣ ψ n ( k )⟩ d θ μν + ⋯ ) = ⟨ ψ n ( k ) ∣ ψ n ( k )⟩ + ( 2 i ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k )⟩ − 2 i σ μ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k )⟩ ) d θ μ + ( − 4 1 ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k )⟩ ) d θ μ d θ ν = 1 + 4 1 ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k )⟩ ⟨ ψ n ( k ) ∣ σ ν ∣ ψ n ( k )⟩ d θ μ d θ ν (540)
Then, we have
s n m μ ν ≡ s n m μ s m n ν (541) s_{nm}^{\mu\nu} \equiv s_{nm}^{\mu}s_{mn}^{\nu} \tag*{(541)} s nm μν ≡ s nm μ s mn ν (541)
with
s n m μ ≡ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k ) ⟩ . (542) s_{nm}^{\mu} \equiv\langle\psi_n(\mathbf{k})|\sigma_\mu|\psi_n(\mathbf{k})\rangle. \tag*{(542)} s nm μ ≡ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ n ( k )⟩ . (542)
is the spin geometric tensor. This is Eq. (142) in the main text.
Zeeman quantum geometry induced intrinsic cross responses
Hall current and AC Hall current
Inserting the density matrix (153) to Eq. (154) and assuming ω ≪ ϵ m n \omega\ll\epsilon_{mn} ω ≪ ϵ mn , the current is calculated as
J μ ; ν = 1 2 ∑ n m ∑ ω 1 = ± ω ∫ d k v n m μ f n m a m n ν E ν e − i ω 1 t ℏ ω 1 − ϵ m n + i η = 1 2 ∑ n m ∑ ω 1 = ± ω ∫ d k f n m i ϵ n m a n m μ a m n ν E ν e − i ω 1 t ℏ ω 1 − ϵ m n + i η . (543) \begin{aligned}
J^{\mu;\nu}={}&\frac{1}{2}\sum_{nm}\sum_{\omega_1=\pm\omega}\int d\mathbf{k}\,v_{nm}^{\mu}f_{nm}a_{mn}^{\nu}\frac{E^{\nu}e^{-i\omega_1t}}{\hbar\omega_1-\epsilon_{mn}+i\eta}\\
={}&\frac{1}{2}\sum_{nm}\sum_{\omega_1=\pm\omega}\int d\mathbf{k}\,f_{nm}i\epsilon_{nm}a_{nm}^{\mu}a_{mn}^{\nu}\frac{E^{\nu}e^{-i\omega_1t}}{\hbar\omega_1-\epsilon_{mn}+i\eta}. \tag*{(543)}
\end{aligned} J μ ; ν = = 2 1 nm ∑ ω 1 = ± ω ∑ ∫ d k v nm μ f nm a mn ν ℏ ω 1 − ϵ mn + i η E ν e − i ω 1 t 2 1 nm ∑ ω 1 = ± ω ∑ ∫ d k f nm i ϵ nm a nm μ a mn ν ℏ ω 1 − ϵ mn + i η E ν e − i ω 1 t . (543)
We assume the off-resonant condition
ℏ ω 1 − ϵ m n ≠ 0. (544) \hbar\omega_1-\epsilon_{mn}\ne0. \tag*{(544)} ℏ ω 1 − ϵ mn = 0. (544)
Then, we can expand the denominator in Eq. (543)
1 ℏ ω 1 − ϵ m n + i η = ∑ j = 1 ∞ ω j − 1 ϵ m n j . (545) \frac{1}{\hbar\omega_1-\epsilon_{mn}+i\eta}=\sum_{j=1}^{\infty}\frac{\omega^{j-1}}{\epsilon_{mn}^{j}}. \tag*{(545)} ℏ ω 1 − ϵ mn + i η 1 = j = 1 ∑ ∞ ϵ mn j ω j − 1 . (545)
By using it, we obtain
J μ ; ν = − 1 2 ∑ n m ∑ ω 1 = ± ω ∫ d k f n m i ϵ n m a n m μ a m n ν E ν e − i ω 1 t ∑ j = 1 ∞ ω 1 j − 1 ϵ m n j = 1 2 ∑ n m ∑ ω 1 = ± ω ∫ d k f n m i a n m μ a m n ν E ν e − i ω 1 t ∑ j = 0 ∞ ω 1 j ϵ m n j = 1 2 ∫ d k ∑ n > m ∑ ω 1 = ± ω ∑ j = 0 ∞ f n m { i [ a n m μ a m n ν − a m n μ a n m ν ] ω 1 2 j ϵ m n 2 j E ν e − i ω 1 t + i [ a n m μ a m n ν + a m n μ a n m ν ] ω 1 2 j + 1 ϵ m n 2 j + 1 E ν e − i ω 1 t } . (546) \begin{aligned}
J^{\mu;\nu}={}&-\frac{1}{2}\sum_{nm}\sum_{\omega_1=\pm\omega}\int d\mathbf{k}f_{nm}i\epsilon_{nm}a_{nm}^{\mu}a_{mn}^{\nu}E^{\nu}e^{-i\omega_1t}\sum_{j=1}^{\infty}\frac{\omega_1^{j-1}}{\epsilon_{mn}^{j}}\\
={}&\frac{1}{2}\sum_{nm}\sum_{\omega_1=\pm\omega}\int d\mathbf{k}f_{nm}ia_{nm}^{\mu}a_{mn}^{\nu}E^{\nu}e^{-i\omega_1t}\sum_{j=0}^{\infty}\frac{\omega_1^{j}}{\epsilon_{mn}^{j}}\\
={}&\frac{1}{2}\int d\mathbf{k}\sum_{n>m}\sum_{\omega_1=\pm\omega}\sum_{j=0}^{\infty}f_{nm}\left\{i\left[a_{nm}^{\mu}a_{mn}^{\nu}-a_{mn}^{\mu}a_{nm}^{\nu}\right]\frac{\omega_1^{2j}}{\epsilon_{mn}^{2j}}E^{\nu}e^{-i\omega_1t}\right.\\
&\left.\qquad+i\left[a_{nm}^{\mu}a_{mn}^{\nu}+a_{mn}^{\mu}a_{nm}^{\nu}\right]\frac{\omega_1^{2j+1}}{\epsilon_{mn}^{2j+1}}E^{\nu}e^{-i\omega_1t}\right\}. \tag*{(546)}
\end{aligned} J μ ; ν = = = − 2 1 nm ∑ ω 1 = ± ω ∑ ∫ d k f nm i ϵ nm a nm μ a mn ν E ν e − i ω 1 t j = 1 ∑ ∞ ϵ mn j ω 1 j − 1 2 1 nm ∑ ω 1 = ± ω ∑ ∫ d k f nm i a nm μ a mn ν E ν e − i ω 1 t j = 0 ∑ ∞ ϵ mn j ω 1 j 2 1 ∫ d k n > m ∑ ω 1 = ± ω ∑ j = 0 ∑ ∞ f nm { i [ a nm μ a mn ν − a mn μ a nm ν ] ϵ mn 2 j ω 1 2 j E ν e − i ω 1 t + i [ a nm μ a mn ν + a mn μ a nm ν ] ϵ mn 2 j + 1 ω 1 2 j + 1 E ν e − i ω 1 t } . (546)
where we have used the relations
f n m ≡ f n − f m = − f m n , (547) f_{nm}\equiv f_n-f_m=-f_{mn}, \tag*{(547)} f nm ≡ f n − f m = − f mn , (547)
and
ϵ n m ≡ ϵ n − ϵ m = − ϵ m n . (548) \epsilon_{nm}\equiv\epsilon_n-\epsilon_m=-\epsilon_{mn}. \tag*{(548)} ϵ nm ≡ ϵ n − ϵ m = − ϵ mn . (548)
By taking the sum of ω 1 = ± ω \omega_1=\pm\omega ω 1 = ± ω , we obtain
J μ ; ν = ∫ d k ∑ n > m f n m [ Ω n m μ ν ω 2 j ϵ m n 2 j E ν cos ω t + 2 g n m μ ν ω 2 j + 1 ϵ m n 2 j + 1 E ν sin ω t ] = ∫ d k ∑ n > m f n m [ Ω n m μ ν 1 1 − ( ω / ϵ m n ) 2 E ν cos ω t + 2 g n m μ ν ω / ϵ m n 1 − ( ω / ϵ m n ) 2 E ν sin ω t ] . (549) \begin{aligned}
J^{\mu;\nu}={}&\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\Omega_{nm}^{\mu\nu}\frac{\omega^{2j}}{\epsilon_{mn}^{2j}}E^{\nu}\cos\omega t\right.\\
&\left.\qquad+2g_{nm}^{\mu\nu}\frac{\omega^{2j+1}}{\epsilon_{mn}^{2j+1}}E^{\nu}\sin\omega t\right]\\
={}&\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\Omega_{nm}^{\mu\nu}\frac{1}{1-(\omega/\epsilon_{mn})^2}E^{\nu}\cos\omega t\right.\\
&\left.\qquad+2g_{nm}^{\mu\nu}\frac{\omega/\epsilon_{mn}}{1-(\omega/\epsilon_{mn})^2}E^{\nu}\sin\omega t\right]. \tag*{(549)}
\end{aligned} J μ ; ν = = ∫ d k n > m ∑ f nm [ Ω nm μν ϵ mn 2 j ω 2 j E ν cos ω t + 2 g nm μν ϵ mn 2 j + 1 ω 2 j + 1 E ν sin ω t ] ∫ d k n > m ∑ f nm [ Ω nm μν 1 − ( ω / ϵ mn ) 2 1 E ν cos ω t + 2 g nm μν 1 − ( ω / ϵ mn ) 2 ω / ϵ mn E ν sin ω t ] . (549)
Only the order of Ω 0 \Omega^0 Ω 0 and Ω \Omega Ω are valid in the linear response theory. Then, we obtain the current
J μ ; ν E ν = ∫ d k ∑ n > m f n m [ Ω n m μ ν + 2 g n m μ ν ω ϵ m n sin ω t ] , (550) \frac{J^{\mu;\nu}}{E^{\nu}}=\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\Omega_{nm}^{\mu\nu}+2g_{nm}^{\mu\nu}\frac{\omega}{\epsilon_{mn}}\sin\omega t\right], \tag*{(550)} E ν J μ ; ν = ∫ d k n > m ∑ f nm [ Ω nm μν + 2 g nm μν ϵ mn ω sin ω t ] , (550)
which is Eq. (157) in the main text.
In the static limit Ω = 0 \Omega=0 Ω = 0 , it recovers the TKNN formula
J μ ν = ∫ d k ∑ n > m f n m Ω n m μ ν E ν . (551) J^{\mu\nu}=\int d\mathbf{k}\sum_{n>m}f_{nm}\Omega_{nm}^{\mu\nu}E^{\nu}. \tag*{(551)} J μν = ∫ d k n > m ∑ f nm Ω nm μν E ν . (551)
Intrinsic gyrotropic magnetic current
The current under external magnetic field is calculated as
J μ ; ν = − g μ B 2 ∑ n m ∑ ω 1 = ± ω ∫ d k f n m s m n ν B ν e − i ω 1 t ℏ ω 1 − ϵ m n + i η = g μ B ∫ d k ∑ n > m f n m [ Z n m μ ν 1 1 − ( ω / ϵ m n ) 2 + 2 Q n m μ ν ω / ϵ m n 1 − ( ω / ϵ m n ) 2 sin ω t ] B ν . (552) \begin{aligned}
J^{\mu;\nu}={}&-\frac{g\mu_{\mathrm{B}}}{2}\sum_{nm}\sum_{\omega_1=\pm\omega}\int d\mathbf{k}f_{nm}s_{mn}^{\nu}B^{\nu}\frac{e^{-i\omega_1t}}{\hbar\omega_1-\epsilon_{mn}+i\eta}\\
={}&g\mu_{\mathrm{B}}\int d\mathbf{k}\sum_{n>m}f_{nm}\left[Z_{nm}^{\mu\nu}\frac{1}{1-(\omega/\epsilon_{mn})^2}\right.\\
&\left.\qquad+2Q_{nm}^{\mu\nu}\frac{\omega/\epsilon_{mn}}{1-(\omega/\epsilon_{mn})^2}\sin\omega t\right]B^{\nu}. \tag*{(552)}
\end{aligned} J μ ; ν = = − 2 g μ B nm ∑ ω 1 = ± ω ∑ ∫ d k f nm s mn ν B ν ℏ ω 1 − ϵ mn + i η e − i ω 1 t g μ B ∫ d k n > m ∑ f nm [ Z nm μν 1 − ( ω / ϵ mn ) 2 1 + 2 Q nm μν 1 − ( ω / ϵ mn ) 2 ω / ϵ mn sin ω t ] B ν . (552)
It is
J μ ; ν = g μ B ∫ d k ∑ n > m f n m [ Z n m μ ν cos ω t + 2 Q n m μ ν ω ϵ m n sin ω t ] B ν . (553) J^{\mu;\nu}=g\mu_{\mathrm{B}}\int d\mathbf{k}\sum_{n>m}f_{nm}\left[Z_{nm}^{\mu\nu}\cos\omega t+2Q_{nm}^{\mu\nu}\frac{\omega}{\epsilon_{mn}}\sin\omega t\right]B^{\nu}. \tag*{(553)} J μ ; ν = g μ B ∫ d k n > m ∑ f nm [ Z nm μν cos ω t + 2 Q nm μν ϵ mn ω sin ω t ] B ν . (553)
up to the linear order in ω \omega ω , which is Eq. (158) in the main text.
Intrinsic electric field induced spin density
The spin polarization under external electric field is calculated as
S μ = 1 2 ∑ n m ∑ ω 1 = ± ω ∫ d k s n m μ f n m r m n ν E ν e − i ω 1 t ℏ ω 1 − ϵ m n + i η = − ∫ d k ∑ n > m f n m [ 2 Q n m ν μ ϵ m n 1 1 − ( ω / ϵ m n ) 2 + Z n m ν μ ω / ϵ m n 2 1 − ( ω / ϵ m n ) 2 sin ω t ] E ν . (554) \begin{aligned}
S^{\mu}={}&\frac{1}{2}\sum_{nm}\sum_{\omega_1=\pm\omega}\int d\mathbf{k}s_{nm}^{\mu}f_{nm}r_{mn}^{\nu}\frac{E^{\nu}e^{-i\omega_1t}}{\hbar\omega_1-\epsilon_{mn}+i\eta}\\
={}&-\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\frac{2Q_{nm}^{\nu\mu}}{\epsilon_{mn}}\frac{1}{1-(\omega/\epsilon_{mn})^2}\right.\\
&\left.\qquad+\frac{Z_{nm}^{\nu\mu}\omega/\epsilon_{mn}^{2}}{1-(\omega/\epsilon_{mn})^2}\sin\omega t\right]E^{\nu}. \tag*{(554)}
\end{aligned} S μ = = 2 1 nm ∑ ω 1 = ± ω ∑ ∫ d k s nm μ f nm r mn ν ℏ ω 1 − ϵ mn + i η E ν e − i ω 1 t − ∫ d k n > m ∑ f nm [ ϵ mn 2 Q nm ν μ 1 − ( ω / ϵ mn ) 2 1 + 1 − ( ω / ϵ mn ) 2 Z nm ν μ ω / ϵ mn 2 sin ω t ] E ν . (554)
It is
S μ = − ∫ d k ∑ n > m f n m [ 2 Q n m μ ν ε m n cos ω t + Z n m μ ν ω ε m n 2 sin ω t ] E ν . (555) S^{\mu}=-\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\frac{2Q_{nm}^{\mu\nu}}{\varepsilon_{mn}}\cos\omega t+Z_{nm}^{\mu\nu}\frac{\omega}{\varepsilon_{mn}^{2}}\sin\omega t\right]E^{\nu}. \tag*{(555)} S μ = − ∫ d k n > m ∑ f nm [ ε mn 2 Q nm μν cos ω t + Z nm μν ε mn 2 ω sin ω t ] E ν . (555)
up to the linear order in ω \omega ω , which is Eq. (159) in the main text.
Intrinsic magnetic field induced spin density
The spin polarization under external magnetic field is calculated as
S μ ; ν = − g μ B 2 ∑ ω 1 = ± ω ∫ d k s n m μ f n m s m n ν B ν e − i ω 1 t ℏ ω 1 − ε m n + i η = g μ B ∫ d k ∑ n > m f n m [ S n m μ ν ε m n 1 1 − ( ω / ε m n ) 2 B ν + 2 A n m μ ν ω / ε m n 2 1 − ( ω / ε m n ) 2 B ν sin ω t ] . (556) \begin{aligned}
S^{\mu;\nu}=-\frac{g\mu_{\mathrm{B}}}{2}\sum_{\omega_{1}=\pm\omega}\int d\mathbf{k}\,s_{nm}^{\mu} \frac{f_{nm}s_{mn}^{\nu}B^{\nu}e^{-i\omega_{1}t}}{\hbar\omega_{1}-\varepsilon_{mn}+i\eta} \\
&=g\mu_{\mathrm{B}}\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\frac{S_{nm}^{\mu\nu}}{\varepsilon_{mn}}\frac{1}{1-(\omega/\varepsilon_{mn})^{2}}B^{\nu}+2A_{nm}^{\mu\nu}\frac{\omega/\varepsilon_{mn}^{2}}{1-(\omega/\varepsilon_{mn})^{2}}B^{\nu}\sin\omega t\right]. \tag*{(556)}
\end{aligned} S μ ; ν = − 2 g μ B ω 1 = ± ω ∑ ∫ d k s nm μ ℏ ω 1 − ε mn + i η f nm s mn ν B ν e − i ω 1 t = g μ B ∫ d k n > m ∑ f nm [ ε mn S nm μν 1 − ( ω / ε mn ) 2 1 B ν + 2 A nm μν 1 − ( ω / ε mn ) 2 ω / ε mn 2 B ν sin ω t ] . (556)
It is
S μ ; ν = g μ B ∫ d k ∑ n > m f n m [ S n m μ ν ε m n cos ω t + 2 A n m μ ν ω ε m n 2 sin ω t ] B ν (557) S^{\mu;\nu}=g\mu_{\mathrm{B}}\int d\mathbf{k}\sum_{n>m}f_{nm}\left[\frac{S_{nm}^{\mu\nu}}{\varepsilon_{mn}}\cos\omega t+2A_{nm}^{\mu\nu}\frac{\omega}{\varepsilon_{mn}^{2}}\sin\omega t\right]B^{\nu} \tag*{(557)} S μ ; ν = g μ B ∫ d k n > m ∑ f nm [ ε mn S nm μν cos ω t + 2 A nm μν ε mn 2 ω sin ω t ] B ν (557)
up to the linear order in ω \omega ω , which is Eq. (160) in the main text.
Zeeman quantum geometry for two-band systems
Zeeman Berry curvature
By inserting n = + n=+ n = + and m = − m=- m = − in Eq. (146) , we obtain
Z + − x x = i ( r + − x s − + x − r − + x s + − x ) = ∂ θ ∂ k x cos θ cos ϕ − ∂ ϕ ∂ k x sin θ sin ϕ = ∂ n x ∂ k x , (558) Z_{+-}^{xx}=i\left(r_{+-}^{x}s_{-+}^{x}-r_{-+}^{x}s_{+-}^{x}\right)
=\frac{\partial\theta}{\partial k_{x}}\cos\theta\cos\phi-\frac{\partial\phi}{\partial k_{x}}\sin\theta\sin\phi=\frac{\partial n_{x}}{\partial k_{x}}, \tag*{(558)} Z +− xx = i ( r +− x s −+ x − r −+ x s +− x ) = ∂ k x ∂ θ cos θ cos ϕ − ∂ k x ∂ ϕ sin θ sin ϕ = ∂ k x ∂ n x , (558)
Z + − y y = i ( r + − y s − + y − r − + y s + − y ) = ∂ θ ∂ k y cos θ sin ϕ + ∂ ϕ ∂ k y cos θ sin ϕ = ∂ n y ∂ k y , (559) Z_{+-}^{yy}=i\left(r_{+-}^{y}s_{-+}^{y}-r_{-+}^{y}s_{+-}^{y}\right)
=\frac{\partial\theta}{\partial k_{y}}\cos\theta\sin\phi+\frac{\partial\phi}{\partial k_{y}}\cos\theta\sin\phi=\frac{\partial n_{y}}{\partial k_{y}}, \tag*{(559)} Z +− y y = i ( r +− y s −+ y − r −+ y s +− y ) = ∂ k y ∂ θ cos θ sin ϕ + ∂ k y ∂ ϕ cos θ sin ϕ = ∂ k y ∂ n y , (559)
Z + − x y = i ( r + − x s − + y − r − + x s + − y ) = ∂ θ ∂ k x cos θ sin ϕ + ∂ ϕ ∂ k x sin θ cos ϕ = ∂ n y ∂ k x , (560) Z_{+-}^{xy}=i\left(r_{+-}^{x}s_{-+}^{y}-r_{-+}^{x}s_{+-}^{y}\right)
=\frac{\partial\theta}{\partial k_{x}}\cos\theta\sin\phi+\frac{\partial\phi}{\partial k_{x}}\sin\theta\cos\phi=\frac{\partial n_{y}}{\partial k_{x}}, \tag*{(560)} Z +− x y = i ( r +− x s −+ y − r −+ x s +− y ) = ∂ k x ∂ θ cos θ sin ϕ + ∂ k x ∂ ϕ sin θ cos ϕ = ∂ k x ∂ n y , (560)
Z + − y x = i ( r + − y s − + x − r − + y s + − x ) = ∂ θ ∂ k y cos θ cos ϕ − ∂ ϕ ∂ k y sin θ sin ϕ = ∂ n x ∂ k y , (561) Z_{+-}^{yx}=i\left(r_{+-}^{y}s_{-+}^{x}-r_{-+}^{y}s_{+-}^{x}\right)
=\frac{\partial\theta}{\partial k_{y}}\cos\theta\cos\phi-\frac{\partial\phi}{\partial k_{y}}\sin\theta\sin\phi=\frac{\partial n_{x}}{\partial k_{y}}, \tag*{(561)} Z +− y x = i ( r +− y s −+ x − r −+ y s +− x ) = ∂ k y ∂ θ cos θ cos ϕ − ∂ k y ∂ ϕ sin θ sin ϕ = ∂ k y ∂ n x , (561)
Z + − x z = i ( r + − x s − + z − r − + x s + − z ) = − ∂ θ ∂ k x sin θ = ∂ n z ∂ k x , (562) Z_{+-}^{xz}=i\left(r_{+-}^{x}s_{-+}^{z}-r_{-+}^{x}s_{+-}^{z}\right)
=-\frac{\partial\theta}{\partial k_{x}}\sin\theta=\frac{\partial n_{z}}{\partial k_{x}}, \tag*{(562)} Z +− x z = i ( r +− x s −+ z − r −+ x s +− z ) = − ∂ k x ∂ θ sin θ = ∂ k x ∂ n z , (562)
Z + − y z = i ( r + − y s − + z − r − + y s + − z ) = − ∂ θ ∂ k y sin θ = ∂ n z ∂ k y . (563) Z_{+-}^{yz}=i\left(r_{+-}^{y}s_{-+}^{z}-r_{-+}^{y}s_{+-}^{z}\right)
=-\frac{\partial\theta}{\partial k_{y}}\sin\theta=\frac{\partial n_{z}}{\partial k_{y}}. \tag*{(563)} Z +− y z = i ( r +− y s −+ z − r −+ y s +− z ) = − ∂ k y ∂ θ sin θ = ∂ k y ∂ n z . (563)
They are summarized as
Z + − μ ν = − Z − + μ ν = ∂ n ν ∂ k μ . (564) Z_{+-}^{\mu\nu}=-Z_{-+}^{\mu\nu}=\frac{\partial n_{\nu}}{\partial k_{\mu}}. \tag*{(564)} Z +− μν = − Z −+ μν = ∂ k μ ∂ n ν . (564)
This is Eq. (161) in the main text.
Zeeman quantum metric
The expectation value of the spin
s n m μ ≡ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ m ( k ) ⟩ (565) s_{nm}^{\mu}\equiv\langle\psi_{n}(\mathbf{k})|\sigma_{\mu}|\psi_{m}(\mathbf{k})\rangle \tag*{(565)} s nm μ ≡ ⟨ ψ n ( k ) ∣ σ μ ∣ ψ m ( k )⟩ (565)
is calculated as
s − + x = − i sin ϕ + cos θ cos ϕ , (566) s_{-+}^{x}=-i\sin\phi+\cos\theta\cos\phi, \tag*{(566)} s −+ x = − i sin ϕ + cos θ cos ϕ , (566)
s − + y = i cos ϕ + cos θ sin ϕ , (567) s_{-+}^{y}=i\cos\phi+\cos\theta\sin\phi, \tag*{(567)} s −+ y = i cos ϕ + cos θ sin ϕ , (567)
s − + z = − sin θ . (568) s_{-+}^{z}=-\sin\theta. \tag*{(568)} s −+ z = − sin θ . (568)
By inserting n = + n=+ n = + and m = − m=- m = − in Eq. (145) , we obtain
Q + − x x = r + − x s − + x + r − + x s + − x 2 = − sin ϕ 2 ∂ θ ∂ k x − cos ϕ 2 sin θ cos θ ∂ ϕ ∂ k x = 1 2 ( n y ∂ n z ∂ k x − n z ∂ n y ∂ k x ) , (569) Q^{xx}_{+-}=\frac{r^{x}_{+-}s^{x}_{-+}+r^{x}_{-+}s^{x}_{+-}}{2}
=-\frac{\sin\phi}{2}\frac{\partial\theta}{\partial k_x}-\frac{\cos\phi}{2}\sin\theta\cos\theta\frac{\partial\phi}{\partial k_x}
=\frac{1}{2}\left(n_y\frac{\partial n_z}{\partial k_x}-n_z\frac{\partial n_y}{\partial k_x}\right), \tag*{(569)} Q +− xx = 2 r +− x s −+ x + r −+ x s +− x = − 2 sin ϕ ∂ k x ∂ θ − 2 cos ϕ sin θ cos θ ∂ k x ∂ ϕ = 2 1 ( n y ∂ k x ∂ n z − n z ∂ k x ∂ n y ) , (569)
Q + − y y = r + − y s − + y + r − + y s + − y 2 = cos ϕ 2 ∂ θ ∂ k y − sin ϕ 2 sin θ cos θ ∂ ϕ ∂ k y = 1 2 ( n z ∂ n x ∂ k y − n x ∂ n z ∂ k y ) , (570) Q^{yy}_{+-}=\frac{r^{y}_{+-}s^{y}_{-+}+r^{y}_{-+}s^{y}_{+-}}{2}
=\frac{\cos\phi}{2}\frac{\partial\theta}{\partial k_y}-\frac{\sin\phi}{2}\sin\theta\cos\theta\frac{\partial\phi}{\partial k_y}
=\frac{1}{2}\left(n_z\frac{\partial n_x}{\partial k_y}-n_x\frac{\partial n_z}{\partial k_y}\right), \tag*{(570)} Q +− y y = 2 r +− y s −+ y + r −+ y s +− y = 2 cos ϕ ∂ k y ∂ θ − 2 sin ϕ sin θ cos θ ∂ k y ∂ ϕ = 2 1 ( n z ∂ k y ∂ n x − n x ∂ k y ∂ n z ) , (570)
Q + − x y = r + − x s − + y + r − + x s + − y 2 = cos ϕ 2 ∂ θ ∂ k x − sin ϕ 2 sin θ cos θ ∂ ϕ ∂ k x = 1 2 ( n z ∂ n x ∂ k x − n x ∂ n z ∂ k x ) , (571) Q^{xy}_{+-}=\frac{r^{x}_{+-}s^{y}_{-+}+r^{x}_{-+}s^{y}_{+-}}{2}
=\frac{\cos\phi}{2}\frac{\partial\theta}{\partial k_x}-\frac{\sin\phi}{2}\sin\theta\cos\theta\frac{\partial\phi}{\partial k_x}
=\frac{1}{2}\left(n_z\frac{\partial n_x}{\partial k_x}-n_x\frac{\partial n_z}{\partial k_x}\right), \tag*{(571)} Q +− x y = 2 r +− x s −+ y + r −+ x s +− y = 2 cos ϕ ∂ k x ∂ θ − 2 sin ϕ sin θ cos θ ∂ k x ∂ ϕ = 2 1 ( n z ∂ k x ∂ n x − n x ∂ k x ∂ n z ) , (571)
Q + − y x = r + − y s − + x + r − + y s + − x 2 = − cos ϕ 2 ∂ θ ∂ k y − sin ϕ 2 sin θ cos θ ∂ ϕ ∂ k y = 1 2 ( n y ∂ n z ∂ k y − n z ∂ n y ∂ k y ) , (572) Q^{yx}_{+-}=\frac{r^{y}_{+-}s^{x}_{-+}+r^{y}_{-+}s^{x}_{+-}}{2}
=-\frac{\cos\phi}{2}\frac{\partial\theta}{\partial k_y}-\frac{\sin\phi}{2}\sin\theta\cos\theta\frac{\partial\phi}{\partial k_y}
=\frac{1}{2}\left(n_y\frac{\partial n_z}{\partial k_y}-n_z\frac{\partial n_y}{\partial k_y}\right), \tag*{(572)} Q +− y x = 2 r +− y s −+ x + r −+ y s +− x = − 2 cos ϕ ∂ k y ∂ θ − 2 sin ϕ sin θ cos θ ∂ k y ∂ ϕ = 2 1 ( n y ∂ k y ∂ n z − n z ∂ k y ∂ n y ) , (572)
Q + − x z = sin 2 θ 2 ∂ ϕ ∂ k x = 1 2 ( n x ∂ n y ∂ k x − n y ∂ n x ∂ k x ) , (573) Q^{xz}_{+-}=\frac{\sin^2\theta}{2}\frac{\partial\phi}{\partial k_x}
=\frac{1}{2}\left(n_x\frac{\partial n_y}{\partial k_x}-n_y\frac{\partial n_x}{\partial k_x}\right), \tag*{(573)} Q +− x z = 2 sin 2 θ ∂ k x ∂ ϕ = 2 1 ( n x ∂ k x ∂ n y − n y ∂ k x ∂ n x ) , (573)
Q + − y z = sin 2 θ 2 ∂ ϕ ∂ k y = 1 2 ( n x ∂ n y ∂ k y − n y ∂ n x ∂ k y ) . (574) Q^{yz}_{+-}=\frac{\sin^2\theta}{2}\frac{\partial\phi}{\partial k_y}
=\frac{1}{2}\left(n_x\frac{\partial n_y}{\partial k_y}-n_y\frac{\partial n_x}{\partial k_y}\right). \tag*{(574)} Q +− y z = 2 sin 2 θ ∂ k y ∂ ϕ = 2 1 ( n x ∂ k y ∂ n y − n y ∂ k y ∂ n x ) . (574)
where we have used the relations
∂ n x ∂ k μ = cos θ cos ϕ ∂ θ ∂ k μ − sin θ sin ϕ ∂ ϕ ∂ k μ , (575) \frac{\partial n_x}{\partial k_\mu}=\cos\theta\cos\phi\frac{\partial\theta}{\partial k_\mu}-\sin\theta\sin\phi\frac{\partial\phi}{\partial k_\mu}, \tag*{(575)} ∂ k μ ∂ n x = cos θ cos ϕ ∂ k μ ∂ θ − sin θ sin ϕ ∂ k μ ∂ ϕ , (575)
∂ n y ∂ k μ = cos θ sin ϕ ∂ θ ∂ k μ + sin θ cos ϕ ∂ ϕ ∂ k μ , (576) \frac{\partial n_y}{\partial k_\mu}=\cos\theta\sin\phi\frac{\partial\theta}{\partial k_\mu}+\sin\theta\cos\phi\frac{\partial\phi}{\partial k_\mu}, \tag*{(576)} ∂ k μ ∂ n y = cos θ sin ϕ ∂ k μ ∂ θ + sin θ cos ϕ ∂ k μ ∂ ϕ , (576)
∂ n z ∂ k μ = − sin θ ∂ θ ∂ k μ . (577) \frac{\partial n_z}{\partial k_\mu}=-\sin\theta\frac{\partial\theta}{\partial k_\mu}. \tag*{(577)} ∂ k μ ∂ n z = − sin θ ∂ k μ ∂ θ . (577)
They are summarized as
Q + − μ ν = 1 2 ε ν ρ σ n ρ ∂ n σ ∂ k μ . (578) Q^{\mu\nu}_{+-}=\frac{1}{2}\varepsilon_{\nu\rho\sigma}n_\rho\frac{\partial n_\sigma}{\partial k_\mu}. \tag*{(578)} Q +− μν = 2 1 ε ν ρ σ n ρ ∂ k μ ∂ n σ . (578)
This is Eq. (162) in the main text.
Spin quantum geometry
By inserting n = + n=+ n = + and m = − m=- m = − in Eq. (142) , we obtain the diagonal components,
s + − x x = s + − x s − + x = 1 4 ( 3 + cos 2 θ − 2 cos 2 ϕ sin 2 θ ) = n y 2 + n z 2 = 1 − n x 2 , (579) s^{xx}_{+-}=s^x_{+-}s^x_{-+}=\frac{1}{4}\left(3+\cos2\theta-2\cos2\phi\sin^2\theta\right)
=n_y^2+n_z^2=1-n_x^2, \tag*{(579)} s +− xx = s +− x s −+ x = 4 1 ( 3 + cos 2 θ − 2 cos 2 ϕ sin 2 θ ) = n y 2 + n z 2 = 1 − n x 2 , (579)
s + − y y = s + − y s − + y = 1 4 ( 3 + cos 2 θ + 2 cos 2 ϕ sin 2 θ ) = n x 2 + n z 2 = 1 − n y 2 , (580) s^{yy}_{+-}=s^y_{+-}s^y_{-+}=\frac{1}{4}\left(3+\cos2\theta+2\cos2\phi\sin^2\theta\right)
=n_x^2+n_z^2=1-n_y^2, \tag*{(580)} s +− y y = s +− y s −+ y = 4 1 ( 3 + cos 2 θ + 2 cos 2 ϕ sin 2 θ ) = n x 2 + n z 2 = 1 − n y 2 , (580)
s + − z z = s + − z s − + z = sin 2 θ = 1 − n z 2 . (581) s^{zz}_{+-}=s^z_{+-}s^z_{-+}=\sin^2\theta=1-n_z^2. \tag*{(581)} s +− z z = s +− z s −+ z = sin 2 θ = 1 − n z 2 . (581)
They are summarized as
s + − μ μ = 1 − n μ 2 . (582) s^{\mu\mu}_{+-}=1-n_\mu^2. \tag*{(582)} s +− μμ = 1 − n μ 2 . (582)
We also obtain the off-diagonal components,
s + − x y = i cos θ − sin 2 θ sin ϕ cos ϕ = i n z − n x n y , (583) s^{xy}_{+-}=i\cos\theta-\sin^2\theta\sin\phi\cos\phi
=in_z-n_xn_y, \tag*{(583)} s +− x y = i cos θ − sin 2 θ sin ϕ cos ϕ = i n z − n x n y , (583)
s + − y x = − i cos θ − sin 2 θ sin ϕ cos ϕ = − i n z − n x n y = − ( s + − x y ) ∗ , (584) s^{yx}_{+-}=-i\cos\theta-\sin^2\theta\sin\phi\cos\phi
=-in_z-n_xn_y=-\left(s^{xy}_{+-}\right)^*, \tag*{(584)} s +− y x = − i cos θ − sin 2 θ sin ϕ cos ϕ = − i n z − n x n y = − ( s +− x y ) ∗ , (584)
s + − x z = − sin θ ( i sin ϕ cos θ + cos ϕ ) = − i n y − n z n x , (585) s^{xz}_{+-}=-\sin\theta\left(i\sin\phi\cos\theta+\cos\phi\right)
=-in_y-n_zn_x, \tag*{(585)} s +− x z = − sin θ ( i sin ϕ cos θ + cos ϕ ) = − i n y − n z n x , (585)
s + − y z = sin θ ( + i cos ϕ − cos θ sin ϕ ) = i n x − n z n y . (586) s^{yz}_{+-}=\sin\theta\left(+i\cos\phi-\cos\theta\sin\phi\right)
=in_x-n_zn_y. \tag*{(586)} s +− y z = sin θ ( + i cos ϕ − cos θ sin ϕ ) = i n x − n z n y . (586)
They are summarized as
s + − μ ν = i ε μ ν ρ n ρ − n μ n ν (587) s^{\mu\nu}_{+-}=i\varepsilon_{\mu\nu\rho}n_\rho-n_\mu n_\nu \tag*{(587)} s +− μν = i ε μν ρ n ρ − n μ n ν (587)
for μ ≠ ν \mu\ne\nu μ = ν . Eqs. (582) and (G30) are summarized as
s + − μ ν = δ μ ν + i ε μ ν ρ n ρ − n μ n ν . (588) s^{\mu\nu}_{+-}=\delta_{\mu\nu}+i\varepsilon_{\mu\nu\rho}n_\rho-n_\mu n_\nu. \tag*{(588)} s +− μν = δ μν + i ε μν ρ n ρ − n μ n ν . (588)
The spin quantum metric (147) is given by
S + − μ ν = δ μ ν − n μ n ν . (589) \mathcal{S}^{\mu\nu}_{+-}=\delta_{\mu\nu}-n_\mu n_\nu. \tag*{(589)} S +− μν = δ μν − n μ n ν . (589)
This is Eq. (163) in the main text.
The spin Berry curvature (148) is given by
A + − μ ν = − 2 ε μ ν ρ n ρ . (590) \mathcal{A}^{\mu\nu}_{+-}=-2\varepsilon_{\mu\nu\rho}n_\rho. \tag*{(590)} A +− μν = − 2 ε μν ρ n ρ . (590)
This is Eq. (164) in the main text.
Zeeman geometry induced cross reponses in X-wave magnets
We evaluate whether there is a response by integrating the Zeeman quantum geometric quantities Z + − μ ν \mathcal{Z}^{\mu\nu}_{+-} Z +− μν and Q + − μ ν Q^{\mu\nu}_{+-} Q +− μν . It is determined by the integration over ϕ \phi ϕ .
J x ; x / E x J^{x;x}/E_x J x ; x / E x
Z + − x x Z^{xx}_{+-} Z +− xx is expanded as
Z + − x x = − λ 3 k x k y ( λ 2 k 2 + B 2 ) 3 / 2 + J λ k y ( − 3 B λ 2 k x f X ( λ 2 k 2 + B 2 ) 5 / 2 + B ∂ k x f X ( λ 2 k 2 + B 2 ) 3 / 2 ) (591) Z^{xx}_{+-}=-\frac{\lambda^{3}k_xk_y}{\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}+J\lambda k_y\left(-\frac{3B\lambda^{2}k_xf_X}{\left(\lambda^{2}k^{2}+B^{2}\right)^{5/2}}+\frac{B\partial_{k_x}f_X}{\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}\right) \tag*{(591)} Z +− xx = − ( λ 2 k 2 + B 2 ) 3/2 λ 3 k x k y + J λ k y ( − ( λ 2 k 2 + B 2 ) 5/2 3 B λ 2 k x f X + ( λ 2 k 2 + B 2 ) 3/2 B ∂ k x f X ) (591)
up to the first order in J J J . The integration is
∫ Z + − x x d ϕ = 0 (592) \int Z^{xx}_{+-}\,d\phi=0 \tag*{(592)} ∫ Z +− xx d ϕ = 0 (592)
for f X = k N X cos N X ϕ f_X=k^{N_X}\cos N_X\phi f X = k N X cos N X ϕ , and f X = k N X sin N X ϕ f_X=k^{N_X}\sin N_X\phi f X = k N X sin N X ϕ with all N X N_X N X . Hence there is no response J x ; x / E x J^{x;x}/E^x J x ; x / E x .
J y ; y / E y J^{y;y}/E^y J y ; y / E y
Z + − y y Z^{yy}_{+-} Z +− y y is expanded as
Z + − y y = − λ 3 k x k y ( λ 2 k 2 + B 2 ) 3 / 2 + J λ k x ( 3 B λ 2 k y f X ( λ 2 k 2 + B 2 ) 5 / 2 − B ∂ k y f X ( λ 2 k 2 + B 2 ) 3 / 2 ) (593) Z^{yy}_{+-}=-\frac{\lambda^{3}k_xk_y}{\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}+J\lambda k_x\left(\frac{3B\lambda^{2}k_yf_X}{\left(\lambda^{2}k^{2}+B^{2}\right)^{5/2}}-\frac{B\partial_{k_y}f_X}{\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}\right) \tag*{(593)} Z +− y y = − ( λ 2 k 2 + B 2 ) 3/2 λ 3 k x k y + J λ k x ( ( λ 2 k 2 + B 2 ) 5/2 3 B λ 2 k y f X − ( λ 2 k 2 + B 2 ) 3/2 B ∂ k y f X ) (593)
up to the first order in J J J . The integration is
∫ Z + − y y d ϕ = 0 (594) \int Z^{yy}_{+-}\,d\phi=0 \tag*{(594)} ∫ Z +− y y d ϕ = 0 (594)
for f X = k N X cos N X ϕ f_X=k^{N_X}\cos N_X\phi f X = k N X cos N X ϕ , and f X = k N X sin N X ϕ f_X=k^{N_X}\sin N_X\phi f X = k N X sin N X ϕ with all N X N_X N X . Hence there is no response J y ; y / E y J^{y;y}/E^y J y ; y / E y .
J x ; y / E y J^{x;y}/E^y J x ; y / E y
Z + − x y Z^{xy}_{+-} Z +− x y is expanded as
Z + − x y = λ ( λ 2 k y 2 + B 2 ) ( λ 2 k 2 + ( B + J f X ) 2 ) 3 / 2 + J λ ( − 3 B ( λ 2 k y 2 + B 2 ) f X ( λ 2 k 2 + B 2 ) 5 / 2 + B 2 f X − k x ∂ k y f X ( λ 2 k 2 + B 2 ) 3 / 2 ) (595) Z^{xy}_{+-}=\frac{\lambda\left(\lambda^{2}k_y^{2}+B^{2}\right)}{\left(\lambda^{2}k^{2}+\left(B+Jf_X\right)^{2}\right)^{3/2}}+J\lambda\left(-\frac{3B\left(\lambda^{2}k_y^{2}+B^{2}\right)f_X}{\left(\lambda^{2}k^{2}+B^{2}\right)^{5/2}}+\frac{B^{2}f_X-k_x\partial_{k_y}f_X}{\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}\right) \tag*{(595)} Z +− x y = ( λ 2 k 2 + ( B + J f X ) 2 ) 3/2 λ ( λ 2 k y 2 + B 2 ) + J λ ( − ( λ 2 k 2 + B 2 ) 5/2 3 B ( λ 2 k y 2 + B 2 ) f X + ( λ 2 k 2 + B 2 ) 3/2 B 2 f X − k x ∂ k y f X ) (595)
up to the first order in J J J . It is nonzero for all N X N_X N X due to the first term.
S x ; x / E x S^{x;x}/E^x S x ; x / E x
We show that S x ; x / E x S^{x;x}/E^x S x ; x / E x is nonzero only for the d x 2 − y 2 d_{x^{2}-y^{2}} d x 2 − y 2 -wave altermagnet. We have S x ; x / E x ≠ 0 S^{x;x}/E^x\ne0 S x ; x / E x = 0 only for d ′ = d x 2 − y 2 d'=d_{x^{2}-y^{2}} d ′ = d x 2 − y 2 as in shown as follows. Q + − x x / ε m n Q^{xx}_{+-}/\varepsilon_{mn} Q +− xx / ε mn is expanded as
Q + − x x ε m n = − λ B 2 ( λ 2 k 2 + B 2 ) 3 / 2 − J λ 2 k 2 ( f X − k x ∂ k x f X ) 2 ( λ 2 k 2 + B 2 ) 5 / 2 + J B 2 ( 2 f X + k x ∂ k x f X ) 2 ( λ 2 k 2 + B 2 ) 5 / 2 (596) \frac{Q^{xx}_{+-}}{\varepsilon_{mn}}=-\frac{\lambda B}{2\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}-\frac{J\lambda^{2}k^{2}\left(f_X-k_x\partial_{k_x}f_X\right)}{2\left(\lambda^{2}k^{2}+B^{2}\right)^{5/2}}+\frac{JB^{2}\left(2f_X+k_x\partial_{k_x}f_X\right)}{2\left(\lambda^{2}k^{2}+B^{2}\right)^{5/2}} \tag*{(596)} ε mn Q +− xx = − 2 ( λ 2 k 2 + B 2 ) 3/2 λ B − 2 ( λ 2 k 2 + B 2 ) 5/2 J λ 2 k 2 ( f X − k x ∂ k x f X ) + 2 ( λ 2 k 2 + B 2 ) 5/2 J B 2 ( 2 f X + k x ∂ k x f X ) (596)
up to the first order in J J J .
First, we consider the case with J = 0 J=0 J = 0 , where
Q + − x x ε m n = − λ B 2 ( λ 2 k 2 + B 2 ) 3 / 2 . (597) \frac{Q^{xx}_{+-}}{\varepsilon_{mn}}=-\frac{\lambda B}{2\left(\lambda^{2}k^{2}+B^{2}\right)^{3/2}}. \tag*{(597)} ε mn Q +− xx = − 2 ( λ 2 k 2 + B 2 ) 3/2 λ B . (597)
We have
∫ 0 ∞ k d k ∫ Q + − x x ε m n d ϕ = − B 2 ∣ B ∣ λ . (598) \int_{0}^{\infty}k\,dk\int\frac{Q^{xx}_{+-}}{\varepsilon_{mn}}\,d\phi=-\frac{B}{2|B|\lambda}. \tag*{(598)} ∫ 0 ∞ k d k ∫ ε mn Q +− xx d ϕ = − 2∣ B ∣ λ B . (598)
Hence, there is nonzero response in the presence of nonzero B B B .
Next, we consider the case with B = 0 B=0 B = 0 , where
Q + − x x ε m n = − J λ 2 k 2 ( f X − k x ∂ k x f X ) 2 ( λ 2 k 2 + B 2 ) 5 / 2 . (599) \frac{Q^{xx}_{+-}}{\varepsilon_{mn}}=-\frac{J\lambda^{2}k^{2}\left(f_X-k_x\partial_{k_x}f_X\right)}{2\left(\lambda^{2}k^{2}+B^{2}\right)^{5/2}}. \tag*{(599)} ε mn Q +− xx = − 2 ( λ 2 k 2 + B 2 ) 5/2 J λ 2 k 2 ( f X − k x ∂ k x f X ) . (599)
By using
∂ k x = cos ϕ ∂ k − sin ϕ k ∂ ϕ , (600) \partial_{k_x}=\cos\phi\partial_k-\frac{\sin\phi}{k}\partial_\phi, \tag*{(600)} ∂ k x = cos ϕ ∂ k − k sin ϕ ∂ ϕ , (600)
∂ k y = sin ϕ ∂ k + cos ϕ k ∂ ϕ , (601) \partial_{k_y}=\sin\phi\partial_k+\frac{\cos\phi}{k}\partial_\phi, \tag*{(601)} ∂ k y = sin ϕ ∂ k + k cos ϕ ∂ ϕ , (601)
we have
f X − k x ∂ k x f X = f X − k cos ϕ ( cos ϕ ∂ k f X − sin ϕ k ∂ ϕ f X ) . (602) f_X-k_x\partial_{k_x}f_X=f_X-k\cos\phi\left(\cos\phi\partial_kf_X-\frac{\sin\phi}{k}\partial_\phi f_X\right). \tag*{(602)} f X − k x ∂ k x f X = f X − k cos ϕ ( cos ϕ ∂ k f X − k sin ϕ ∂ ϕ f X ) . (602)
For f X = k N X cos N X ϕ f_X=k^{N_X}\cos N_X\phi f X = k N X cos N X ϕ , we have
f X − k x ∂ k x f X = − k N X 2 ( N X cos ( N X − 2 ) ϕ + ( N X − 2 ) cos N X ϕ ) . (603) f_X-k_x\partial_{k_x}f_X=-\frac{k^{N_X}}{2}\left(N_X\cos\left(N_X-2\right)\phi+\left(N_X-2\right)\cos N_X\phi\right). \tag*{(603)} f X − k x ∂ k x f X = − 2 k N X ( N X cos ( N X − 2 ) ϕ + ( N X − 2 ) cos N X ϕ ) . (603)
It is nonzero only for N X = 2 N_X=2 N X = 2 . It is nonzero contribution only for the d x 2 − y 2 d_{x^{2}-y^{2}} d x 2 − y 2 -wave altermagnet. For f X = k N X sin N X ϕ f_X=k^{N_X}\sin N_X\phi f X = k N X sin N X ϕ , we also have
f X − k x ∂ k x f X = − k N X 2 ( N X sin ( N X − 2 ) ϕ + ( N X − 2 ) sin N X ϕ ) , (604) f_X-k_x\partial_{k_x}f_X=-\frac{k^{N_X}}{2}\left(N_X\sin\left(N_X-2\right)\phi+\left(N_X-2\right)\sin N_X\phi\right), \tag*{(604)} f X − k x ∂ k x f X = − 2 k N X ( N X sin ( N X − 2 ) ϕ + ( N X − 2 ) sin N X ϕ ) , (604)
which is zero for all N X N_X N X . It leads to
S x E x = − ∫ d 2 k f + − 2 Q + − x x ε + − = − 2 π J m m ( m λ 2 + 2 μ ) sgn ( λ ) − π B 2 λ 3 ( B J + λ 2 ) + B ln B 2 + 2 m λ 2 { m λ 2 + μ − M } B 2 + 2 m λ 2 { m λ 2 + μ + M } (605) \frac{S^x}{E^x}=-\int d^2k\,f_{+-}\frac{2Q^{xx}_{+-}}{\varepsilon_{+-}}
=-2\pi Jm\sqrt{m\left(m\lambda^{2}+2\mu\right)}\operatorname{sgn}(\lambda)
-\frac{\pi B}{2\lambda^{3}}\left(BJ+\lambda^{2}\right)
+B\ln\frac{B^{2}+2m\lambda^{2}\left\{m\lambda^{2}+\mu-M\right\}}{B^{2}+2m\lambda^{2}\left\{m\lambda^{2}+\mu+M\right\}} \tag*{(605)} E x S x = − ∫ d 2 k f +− ε +− 2 Q +− xx = − 2 π J m m ( m λ 2 + 2 μ ) sgn ( λ ) − 2 λ 3 π B ( B J + λ 2 ) + B ln B 2 + 2 m λ 2 { m λ 2 + μ + M } B 2 + 2 m λ 2 { m λ 2 + μ − M } (605)
with
M = ( m λ 2 ) 2 + 2 μ m λ 2 + B 2 (606) M=\sqrt{\left(m\lambda^{2}\right)^{2}+2\mu m\lambda^{2}+B^{2}} \tag*{(606)} M = ( m λ 2 ) 2 + 2 μ m λ 2 + B 2 (606)
up to the first order in J J J . For B = 0 B=0 B = 0 , we obtain
S x E x = − 2 m J π λ . \frac{S^x}{E^x}=-\frac{2mJ\pi}{\lambda}. E x S x = − λ 2 m J π .
It is nonzero even for B = 0 B=0 B = 0 .
S y ; y / E y S^{y;y}/E^y S y ; y / E y
We show that S y ; y / E y S^{y;y}/E^y S y ; y / E y is nonzero only for the d x y d_{xy} d x y -wave altermagnet. We have S y ; y / E y ≠ 0 S^{y;y}/E^y \ne0 S y ; y / E y = 0 only for d ′ = d x y d'=d_{xy} d ′ = d x y as shown as follows. Q + − y y / ε m n Q_{+-}^{yy}/\varepsilon_{mn} Q +− y y / ε mn is expanded as
Q + − y y ε m n = − λ B 2 ( λ 2 k 2 + B 2 ) 3 / 2 + J λ 2 k 2 ( f X − k y ∂ k y f X ) 2 ( λ 2 k 2 + B 2 ) 5 / 2 + J B 2 ( 2 f X + k y ∂ k y f X ) 2 ( λ 2 k 2 + B 2 ) 5 / 2 . (607) \frac{Q_{+-}^{yy}}{\varepsilon_{mn}}=-\frac{\lambda B}{2\left(\lambda^2 k^2+B^2\right)^{3/2}}+\frac{J\lambda^2 k^2\left(f_X-k_y\partial_{k_y}f_X\right)}{2\left(\lambda^2 k^2+B^2\right)^{5/2}}+\frac{JB^2\left(2f_X+k_y\partial_{k_y}f_X\right)}{2\left(\lambda^2 k^2+B^2\right)^{5/2}}. \tag*{(607)} ε mn Q +− y y = − 2 ( λ 2 k 2 + B 2 ) 3/2 λ B + 2 ( λ 2 k 2 + B 2 ) 5/2 J λ 2 k 2 ( f X − k y ∂ k y f X ) + 2 ( λ 2 k 2 + B 2 ) 5/2 J B 2 ( 2 f X + k y ∂ k y f X ) . (607)
up to the first order in J J J . As in the case of Q + − x x Q_{+-}^{xx} Q +− xx , there is nonzero response for B ≠ 0 B\ne0 B = 0 .
We consider the case with B = 0 B=0 B = 0 , where
Q + − y y ε m n = J λ 2 k 2 ( f X − k y ∂ k y f X ) 2 ( λ 2 k 2 + B 2 ) 5 / 2 . (608) \frac{Q_{+-}^{yy}}{\varepsilon_{mn}}=\frac{J\lambda^2 k^2\left(f_X-k_y\partial_{k_y}f_X\right)}{2\left(\lambda^2 k^2+B^2\right)^{5/2}}. \tag*{(608)} ε mn Q +− y y = 2 ( λ 2 k 2 + B 2 ) 5/2 J λ 2 k 2 ( f X − k y ∂ k y f X ) . (608)
For f X = k N X cos N X ϕ f_X=k^{N_X}\cos N_X\phi f X = k N X cos N X ϕ , we have
f X − k x ∂ k x f X = k N X 2 ( N X sin ( N X − 2 ) ϕ + N X sin N X ϕ + 2 cos N X ϕ ) . (609) f_X-k_x\partial_{k_x}f_X=\frac{k^{N_X}}{2}\left(N_X\sin\left(N_X-2\right)\phi+N_X\sin N_X\phi+2\cos N_X\phi\right). \tag*{(609)} f X − k x ∂ k x f X = 2 k N X ( N X sin ( N X − 2 ) ϕ + N X sin N X ϕ + 2 cos N X ϕ ) . (609)
On the other hand, for f X = k N X sin N X ϕ f_X=k^{N_X}\sin N_X\phi f X = k N X sin N X ϕ , we have
f X − k x ∂ k x f X = − k N X 2 ( N X cos ( N X − 2 ) ϕ + N X cos N X ϕ − 2 sin N X ϕ ) , (610) f_X-k_x\partial_{k_x}f_X=-\frac{k^{N_X}}{2}\left(N_X\cos\left(N_X-2\right)\phi+N_X\cos N_X\phi-2\sin N_X\phi\right), \tag*{(610)} f X − k x ∂ k x f X = − 2 k N X ( N X cos ( N X − 2 ) ϕ + N X cos N X ϕ − 2 sin N X ϕ ) , (610)
which is nonzero only for N X = 2 N_X=2 N X = 2 because
∫ ( f X − k x ∂ k x f X ) d ϕ = − 2 k 2 π δ N X , 2 . (611) \int\left(f_X-k_x\partial_{k_x}f_X\right)d\phi=-2k^2\pi\delta_{N_X,2}. \tag*{(611)} ∫ ( f X − k x ∂ k x f X ) d ϕ = − 2 k 2 π δ N X , 2 . (611)
It leads to
S y ; y E x = − ∫ d 2 k k f + − 2 Q + − y y ε + − ε − = − 2 π J m m ( m λ 2 + 2 μ ) sgn ( λ ) − π B 2 λ 3 ( B J + λ 2 ) + B ln B 2 + 2 m λ 2 { m λ 2 + μ − M } B 2 + 2 m λ 2 { m λ 2 + μ + M } . (612) \frac{S^{y;y}}{E^x}=-\int d^2k\,k f_{+-}\frac{2Q_{+-}^{yy}}{\varepsilon_+ - \varepsilon_-}
=-2\pi Jm\sqrt{m\left(m\lambda^2+2\mu\right)}\operatorname{sgn}(\lambda)
-\frac{\pi B}{2\lambda^3}\left(BJ+\lambda^2\right)
+B\ln\frac{B^2+2m\lambda^2\left\{m\lambda^2+\mu-M\right\}}{B^2+2m\lambda^2\left\{m\lambda^2+\mu+M\right\}}. \tag*{(612)} E x S y ; y = − ∫ d 2 k k f +− ε + − ε − 2 Q +− y y = − 2 π J m m ( m λ 2 + 2 μ ) sgn ( λ ) − 2 λ 3 π B ( B J + λ 2 ) + B ln B 2 + 2 m λ 2 { m λ 2 + μ + M } B 2 + 2 m λ 2 { m λ 2 + μ − M } . (612)
Hence, we have nonzero contribution only for the d x y d_{xy} d x y -wave altermagnet.
S x ; y / E y S^{x;y}/E^y S x ; y / E y
We show that S x ; y / E y S^{x;y}/E^y S x ; y / E y is nonzero only for the d x y d_{xy} d x y -wave altermagnet. Q + − x y Q_{+-}^{xy} Q +− x y is expanded as
Q + − x y = λ k y J ∂ k x f X 2 ( λ 2 k 2 + B 2 ) 3 / 2 (613) Q_{+-}^{xy}=\frac{\lambda k_yJ\partial_{k_x}f_X}{2\left(\lambda^2k^2+B^2\right)^{3/2}} \tag*{(613)} Q +− x y = 2 ( λ 2 k 2 + B 2 ) 3/2 λ k y J ∂ k x f X (613)
up to the first order in J J J . It is evaluated as follows. The numerator of Q + − x y Q_{+-}^{xy} Q +− x y reads
k y ∂ k x f X = N k N X cos ( N X − 1 ) ϕ sin ϕ (614) k_y\partial_{k_x}f_X=Nk^{N_X}\cos\left(N_X-1\right)\phi\sin\phi \tag*{(614)} k y ∂ k x f X = N k N X cos ( N X − 1 ) ϕ sin ϕ (614)
for f X = k N X cos N X ϕ f_X=k^{N_X}\cos N_X\phi f X = k N X cos N X ϕ , whose integration over ϕ \phi ϕ is zero.
∫ k y ∂ k x f X d ϕ = 0. (615) \int k_y\partial_{k_x}f_Xd\phi=0. \tag*{(615)} ∫ k y ∂ k x f X d ϕ = 0. (615)
We also have
k y ∂ k x f X = N k N X sin ( N X − 1 ) ϕ sin ϕ , (616) k_y\partial_{k_x}f_X=Nk^{N_X}\sin\left(N_X-1\right)\phi\sin\phi, \tag*{(616)} k y ∂ k x f X = N k N X sin ( N X − 1 ) ϕ sin ϕ , (616)
for f X = k N X sin N X ϕ f_X=k^{N_X}\sin N_X\phi f X = k N X sin N X ϕ , whose integration over ϕ \phi ϕ is obtained as
∫ k y ∂ k x f X d ϕ = − 2 k 2 π δ N X , 2 . (617) \int k_y\partial_{k_x}f_Xd\phi=-2k^2\pi\delta_{N_X,2}. \tag*{(617)} ∫ k y ∂ k x f X d ϕ = − 2 k 2 π δ N X , 2 . (617)
Then, we have the static electric-field induced spin polarization
S x ; y E y = − ∫ d 2 k k f + − 2 Q + − x y ε + − ε − = − J π λ 3 ∑ η = ± η 2 ( B 2 + m λ 2 ( m λ 2 + μ + η M ) ) B 2 + 2 m λ 2 ( m λ 2 + μ + η M ) . (618) \frac{S^{x;y}}{E^y}=-\int d^2k\,k f_{+-}\frac{2Q_{+-}^{xy}}{\varepsilon_+-\varepsilon_-}
=-\frac{J\pi}{\lambda^3}\sum_{\eta=\pm}\eta\frac{2\left(B^2+m\lambda^2\left(m\lambda^2+\mu+\eta M\right)\right)}{\sqrt{B^2+2m\lambda^2}\left(m\lambda^2+\mu+\eta M\right)}. \tag*{(618)} E y S x ; y = − ∫ d 2 k k f +− ε + − ε − 2 Q +− x y = − λ 3 J π η = ± ∑ η B 2 + 2 m λ 2 ( m λ 2 + μ + η M ) 2 ( B 2 + m λ 2 ( m λ 2 + μ + η M ) ) . (618)
up to the first order in J J J . For B = 0 B=0 B = 0 , it is simplified as
S x E y = − 2 m J π λ (619) \frac{S^x}{E^y}=-\frac{2mJ\pi}{\lambda} \tag*{(619)} E y S x = − λ 2 m J π (619)
for μ > 0 \mu>0 μ > 0 . Hence, S x ; y / E y S^{x;y}/E^y S x ; y / E y is nonzero only for the d x y d_{xy} d x y -wave altermagnet.
S y ; x / E x S^{y;x}/E^x S y ; x / E x
We show that S y ; x / E x S^{y;x}/E^x S y ; x / E x is nonzero only for the d x y d_{xy} d x y -wave altermagnet. Q + − y x Q_{+-}^{yx} Q +− y x is expanded as
Q + − y x = λ k x J ∂ k y f X 2 ( λ 2 k 2 + B 2 ) 3 / 2 (620) Q_{+-}^{yx}=\frac{\lambda k_xJ\partial_{k_y}f_X}{2\left(\lambda^2k^2+B^2\right)^{3/2}} \tag*{(620)} Q +− y x = 2 ( λ 2 k 2 + B 2 ) 3/2 λ k x J ∂ k y f X (620)
up to the first order in J J J . It is evaluated as follows. We study Q + − y x Q_{+-}^{yx} Q +− y x , whose numerator is k x ∂ k y f X k_x\partial_{k_y}f_X k x ∂ k y f X .
k x ∂ k y f X = N k N X sin ( N X − 1 ) ϕ cos ϕ (621) k_x\partial_{k_y}f_X=Nk^{N_X}\sin\left(N_X-1\right)\phi\cos\phi \tag*{(621)} k x ∂ k y f X = N k N X sin ( N X − 1 ) ϕ cos ϕ (621)
for f X = k N X cos N X ϕ f_X=k^{N_X}\cos N_X\phi f X = k N X cos N X ϕ , whose integration over ϕ \phi ϕ is
∫ k y ∂ k x f X d ϕ = 0. (622) \int k_y\partial_{k_x}f_Xd\phi=0. \tag*{(622)} ∫ k y ∂ k x f X d ϕ = 0. (622)
We also have
k x ∂ k y f X = N k N X cos ( N X − 1 ) ϕ cos ϕ (623) k_x\partial_{k_y}f_X=Nk^{N_X}\cos\left(N_X-1\right)\phi\cos\phi \tag*{(623)} k x ∂ k y f X = N k N X cos ( N X − 1 ) ϕ cos ϕ (623)
for f X = k N X sin N X ϕ f_X=k^{N_X}\sin N_X\phi f X = k N X sin N X ϕ , whose integration is obtained as
∫ k x ∂ k y f X d ϕ = 2 k 2 π δ N X , 2 . (624) \int k_x\partial_{k_y}f_Xd\phi=2k^2\pi\delta_{N_X,2}. \tag*{(624)} ∫ k x ∂ k y f X d ϕ = 2 k 2 π δ N X , 2 . (624)
Then, we have the static electric-field induced spin polarization
S y ; x E y = − ∫ d 2 k k f + − 2 Q + − y x ε + − ε − = J π λ 3 ∑ η = ± η 2 ( B 2 + m λ 2 ( m λ 2 + μ + η M ) ) B 2 + 2 m λ 2 ( m λ 2 + μ + η M ) . (625) \frac{S^{y;x}}{E^y}=-\int d^2k\,k f_{+-}\frac{2Q_{+-}^{yx}}{\varepsilon_+-\varepsilon_-}
=\frac{J\pi}{\lambda^3}\sum_{\eta=\pm}\eta\frac{2\left(B^2+m\lambda^2\left(m\lambda^2+\mu+\eta M\right)\right)}{\sqrt{B^2+2m\lambda^2}\left(m\lambda^2+\mu+\eta M\right)}. \tag*{(625)} E y S y ; x = − ∫ d 2 k k f +− ε + − ε − 2 Q +− y x = λ 3 J π η = ± ∑ η B 2 + 2 m λ 2 ( m λ 2 + μ + η M ) 2 ( B 2 + m λ 2 ( m λ 2 + μ + η M ) ) . (625)
up to the first order in J J J . For B = 0 B=0 B = 0 , it is simplified as
S x E y = 2 m J π λ (626) \frac{S^{x}}{E^{y}}=\frac{2mJ\pi}{\lambda} \tag*{(626)} E y S x = λ 2 m J π (626)
for μ > 0 \mu>0 μ > 0 . Hence, S y ; x / E x S^{y;x}/E^{x} S y ; x / E x is nonzero only for the d x y d_{xy} d x y -wave altermagnet.
Uhlmann geometry
We prove Eq. (275) in the main text. We first calculate
⟨ ψ n ( k ) ∣ d ρ ∣ ψ m ( k ) ⟩ = ∑ n ′ ∂ k μ p n ′ δ n n ′ δ n ′ m − i p n ′ a n n ′ μ ( k ) δ n ′ m − i p n ′ δ n n ′ a m n ′ μ ∗ ( k ) = δ n m ∂ k μ p n − i p n a n m μ ( k ) + i p m a n m μ ( k ) , (627) \begin{aligned}
\langle\psi_n(\mathbf{k})|d\rho|\psi_m(\mathbf{k})\rangle
&=\sum_{n'}\partial_{k_\mu}p_{n'}\delta_{nn'}\delta_{n'm}
-ip_{n'}a^\mu_{nn'}(\mathbf{k})\delta_{n'm}
-ip_{n'}\delta_{nn'}a^{\mu*}_{mn'}(\mathbf{k})\\
&=\delta_{nm}\partial_{k_\mu}p_n-ip_na^\mu_{nm}(\mathbf{k})
+ip_ma^\mu_{nm}(\mathbf{k}), \tag*{(627)}
\end{aligned} ⟨ ψ n ( k ) ∣ d ρ ∣ ψ m ( k )⟩ = n ′ ∑ ∂ k μ p n ′ δ n n ′ δ n ′ m − i p n ′ a n n ′ μ ( k ) δ n ′ m − i p n ′ δ n n ′ a m n ′ μ ∗ ( k ) = δ nm ∂ k μ p n − i p n a nm μ ( k ) + i p m a nm μ ( k ) , (627)
where we have used
d ρ = ∑ n ′ ∂ k μ p n ′ ∣ ψ n ′ ( k ) ⟩ ⟨ ψ n ′ ( k ) ∣ + ∑ n ′ p n ′ ( ∂ k μ ∣ ψ n ′ ( k ) ⟩ ) ⟨ ψ n ′ ( k ) ∣ + ∑ n ′ p n ′ ∣ ψ n ′ ( k ) ⟩ ( ∂ k μ ⟨ ψ n ′ ( k ) ∣ ) . (628) \begin{aligned}
d\rho={}&\sum_{n'}\partial_{k_\mu}p_{n'}|\psi_{n'}(\mathbf{k})\rangle\langle\psi_{n'}(\mathbf{k})|\\
&+\sum_{n'}p_{n'}\left(\partial_{k_\mu}|\psi_{n'}(\mathbf{k})\rangle\right)\langle\psi_{n'}(\mathbf{k})|\\
&+\sum_{n'}p_{n'}|\psi_{n'}(\mathbf{k})\rangle\left(\partial_{k_\mu}\langle\psi_{n'}(\mathbf{k})|\right). \tag*{(628)}
\end{aligned} d ρ = n ′ ∑ ∂ k μ p n ′ ∣ ψ n ′ ( k )⟩ ⟨ ψ n ′ ( k ) ∣ + n ′ ∑ p n ′ ( ∂ k μ ∣ ψ n ′ ( k )⟩ ) ⟨ ψ n ′ ( k ) ∣ + n ′ ∑ p n ′ ∣ ψ n ′ ( k )⟩ ( ∂ k μ ⟨ ψ n ′ ( k ) ∣ ) . (628)
On the other hand, we have
⟨ ψ n ( k ) ∣ L ρ + ρ L 2 ∣ ψ m ( k ) ⟩ = ⟨ ψ n ( k ) ∣ ( ∑ n ′ 1 2 ∂ k μ p n ′ p n ′ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ ∑ m ′ p m ′ ∣ ψ m ′ ( k ) ⟩ ⟨ ψ m ′ ( k ) ∣ + i ∑ n ′ ≠ m ′ p m ′ − p n ′ p m ′ + p n ′ a n ′ m ′ μ ∣ ψ n ′ ⟩ ⟨ ψ m ′ ∣ ∑ m ′ ′ p m ′ ′ ∣ ψ m ′ ′ ( k ) ⟩ ⟨ ψ m ′ ′ ( k ) ∣ + ∑ m ′ ′ p m ′ ′ ∣ ψ m ′ ′ ( k ) ⟩ ⟨ ψ m ′ ′ ( k ) ∣ ∑ n ′ 1 2 ∂ k μ p n ′ p n ′ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ + i ∑ n ′ ≠ m ′ p m ′ − p n ′ p m ′ + p n ′ a n ′ m ′ μ ∣ ψ m ′ ′ ⟩ ⟨ ψ m ′ ∣ ) ∣ ψ m ( k ) ⟩ . (629) \begin{aligned}
\langle\psi_n(\mathbf{k})|\frac{L\rho+\rho L}{2}|\psi_m(\mathbf{k})\rangle
={}&\langle\psi_n(\mathbf{k})|\bigg(
\sum_{n'}\frac{1}{2}\frac{\partial_{k_\mu}p_{n'}}{p_{n'}}|\psi_{n'}\rangle\langle\psi_{n'}|
\sum_{m'}p_{m'}|\psi_{m'}(\mathbf{k})\rangle\langle\psi_{m'}(\mathbf{k})|\\
&+i\sum_{n'\ne m'}\frac{p_{m'}-p_{n'}}{p_{m'}+p_{n'}}a^\mu_{n'm'}|\psi_{n'}\rangle\langle\psi_{m'}|
\sum_{m''}p_{m''}|\psi_{m''}(\mathbf{k})\rangle\langle\psi_{m''}(\mathbf{k})|\\
&+\sum_{m''}p_{m''}|\psi_{m''}(\mathbf{k})\rangle\langle\psi_{m''}(\mathbf{k})|
\sum_{n'}\frac{1}{2}\frac{\partial_{k_\mu}p_{n'}}{p_{n'}}|\psi_{n'}\rangle\langle\psi_{n'}|\\
&+i\sum_{n'\ne m'}\frac{p_{m'}-p_{n'}}{p_{m'}+p_{n'}}a^\mu_{n'm'}|\psi_{m''}\rangle\langle\psi_{m'}|\bigg)|\psi_m(\mathbf{k})\rangle. \tag*{(629)}
\end{aligned} ⟨ ψ n ( k ) ∣ 2 L ρ + ρ L ∣ ψ m ( k )⟩ = ⟨ ψ n ( k ) ∣ ( n ′ ∑ 2 1 p n ′ ∂ k μ p n ′ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ m ′ ∑ p m ′ ∣ ψ m ′ ( k )⟩ ⟨ ψ m ′ ( k ) ∣ + i n ′ = m ′ ∑ p m ′ + p n ′ p m ′ − p n ′ a n ′ m ′ μ ∣ ψ n ′ ⟩ ⟨ ψ m ′ ∣ m ′′ ∑ p m ′′ ∣ ψ m ′′ ( k )⟩ ⟨ ψ m ′′ ( k ) ∣ + m ′′ ∑ p m ′′ ∣ ψ m ′′ ( k )⟩ ⟨ ψ m ′′ ( k ) ∣ n ′ ∑ 2 1 p n ′ ∂ k μ p n ′ ∣ ψ n ′ ⟩ ⟨ ψ n ′ ∣ + i n ′ = m ′ ∑ p m ′ + p n ′ p m ′ − p n ′ a n ′ m ′ μ ∣ ψ m ′′ ⟩ ⟨ ψ m ′ ∣ ) ∣ ψ m ( k )⟩ . (629)
It is further calculated as
∑ n ′ δ n n ′ 1 2 ∂ k μ p n ′ p n ′ δ n ′ m ′ p m ′ δ m ′ m + i ∑ n ′ ≠ m ′ p m ′ − p n ′ p m ′ + p n ′ a n ′ m ′ μ δ n n ′ δ m ′ m ′ ′ δ m ′ ′ m + ∑ m ′ p m ′ δ n m ′ ∑ n ′ 1 2 ∂ k μ p n ′ p n ′ δ m ′ n ′ δ n ′ m + i ∑ n ′ ≠ m ′ δ n m ′ ′ p m ′ ′ p m ′ − p n ′ p m ′ + p n ′ a n ′ m ′ μ δ m ′ ′ n ′ δ m ′ m = δ n m 1 2 ∂ k μ p n p n p n + i p m − p n p m + p n a n m μ p m + δ n m p n 1 2 ∂ k μ p n p n + i p n p m − p n p m + p n a n m μ = δ n m ∂ k μ p n + i ( p m − p n ) a n m μ . (630) \begin{aligned}
&\sum_{n'}\delta_{nn'}\frac{1}{2}\frac{\partial_{k_\mu}p_{n'}}{p_{n'}}\delta_{n'm'}p_{m'}\delta_{m'm}
+i\sum_{n'\ne m'}\frac{p_{m'}-p_{n'}}{p_{m'}+p_{n'}}a^\mu_{n'm'}\delta_{nn'}\delta_{m'm''}\delta_{m''m}\\
&+\sum_{m'}p_{m'}\delta_{nm'}\sum_{n'}\frac{1}{2}\frac{\partial_{k_\mu}p_{n'}}{p_{n'}}\delta_{m'n'}\delta_{n'm}
+i\sum_{n'\ne m'}\delta_{nm''}p_{m''}\frac{p_{m'}-p_{n'}}{p_{m'}+p_{n'}}a^\mu_{n'm'}\delta_{m''n'}\delta_{m'm}\\
={}&\delta_{nm}\frac{1}{2}\frac{\partial_{k_\mu}p_n}{p_n}p_n
+i\frac{p_m-p_n}{p_m+p_n}a^\mu_{nm}p_m
+\delta_{nm}p_n\frac{1}{2}\frac{\partial_{k_\mu}p_n}{p_n}
+ip_n\frac{p_m-p_n}{p_m+p_n}a^\mu_{nm}\\
={}&\delta_{nm}\partial_{k_\mu}p_n+i(p_m-p_n)a^\mu_{nm}. \tag*{(630)}
\end{aligned} = = n ′ ∑ δ n n ′ 2 1 p n ′ ∂ k μ p n ′ δ n ′ m ′ p m ′ δ m ′ m + i n ′ = m ′ ∑ p m ′ + p n ′ p m ′ − p n ′ a n ′ m ′ μ δ n n ′ δ m ′ m ′′ δ m ′′ m + m ′ ∑ p m ′ δ n m ′ n ′ ∑ 2 1 p n ′ ∂ k μ p n ′ δ m ′ n ′ δ n ′ m + i n ′ = m ′ ∑ δ n m ′′ p m ′′ p m ′ + p n ′ p m ′ − p n ′ a n ′ m ′ μ δ m ′′ n ′ δ m ′ m δ nm 2 1 p n ∂ k μ p n p n + i p m + p n p m − p n a nm μ p m + δ nm p n 2 1 p n ∂ k μ p n + i p n p m + p n p m − p n a nm μ δ nm ∂ k μ p n + i ( p m − p n ) a nm μ . (630)
which is identical to Eq. (627) .
Positive Operator-Valued Measure
Positive Operator-Valued Measure (POVM) Π n \Pi_n Π n satisfies[222 ]
p n = Tr ρ Π n , (631) p_n=\operatorname{Tr}\rho\Pi_n, \tag*{(631)} p n = Tr ρ Π n , (631)
and
∑ n = 1 N Π n = 1. (632) \sum_{n=1}^{N}\Pi_n=1. \tag*{(632)} n = 1 ∑ N Π n = 1. (632)
It is a generalization of the projective measurement. These two conditions lead to the conservation of the provability,
∑ n = 1 N p n = ∑ n = 1 N Tr ρ Π n = Tr ρ ∑ n = 1 N Π n = Tr ρ = 1. (633) \sum_{n=1}^{N}p_n=\sum_{n=1}^{N}\operatorname{Tr}\rho\Pi_n=\operatorname{Tr}\rho\sum_{n=1}^{N}\Pi_n=\operatorname{Tr}\rho=1. \tag*{(633)} n = 1 ∑ N p n = n = 1 ∑ N Tr ρ Π n = Tr ρ n = 1 ∑ N Π n = Tr ρ = 1. (633)
By using it, the classical Fisher information (280) is rewritten as
F C F i s h e r μ ν = ∑ n = 1 N Tr ∂ ρ ∂ k μ Π n Tr ρ Π n Tr ∂ ρ ∂ k ν Π n Tr ρ Π n = ∑ n = 1 N 1 Tr ρ Π n Tr ( ∂ ρ ∂ k μ Π n ) Tr ( ∂ ρ ∂ k ν Π n ) . (634) \begin{aligned}
\mathcal{F}^{\mu\nu}_{\mathrm{CFisher}}
&=\sum_{n=1}^{N}\frac{\operatorname{Tr}\frac{\partial\rho}{\partial k_\mu}\Pi_n}{\operatorname{Tr}\rho\Pi_n}
\frac{\operatorname{Tr}\frac{\partial\rho}{\partial k_\nu}\Pi_n}{\operatorname{Tr}\rho\Pi_n}\\
&=\sum_{n=1}^{N}\frac{1}{\operatorname{Tr}\rho\Pi_n}
\operatorname{Tr}\left(\frac{\partial\rho}{\partial k_\mu}\Pi_n\right)
\operatorname{Tr}\left(\frac{\partial\rho}{\partial k_\nu}\Pi_n\right). \tag*{(634)}
\end{aligned} F CFisher μν = n = 1 ∑ N Tr ρ Π n Tr ∂ k μ ∂ ρ Π n Tr ρ Π n Tr ∂ k ν ∂ ρ Π n = n = 1 ∑ N Tr ρ Π n 1 Tr ( ∂ k μ ∂ ρ Π n ) Tr ( ∂ k ν ∂ ρ Π n ) . (634)
Quantum Cramér-Rao inequality
We prove the quantum Cramér-Rao inequality (282) . It is enough to show[140 ]
∑ μ ν a μ F C F i s h e r μ ν a ν < ∑ μ ν a μ F Q F i s h e r μ ν a ν (635) \sum_{\mu\nu}a_\mu\mathcal{F}^{\mu\nu}_{\mathrm{CFisher}}a_\nu
<
\sum_{\mu\nu}a_\mu\mathcal{F}^{\mu\nu}_{\mathrm{QFisher}}a_\nu \tag*{(635)} μν ∑ a μ F CFisher μν a ν < μν ∑ a μ F QFisher μν a ν (635)
for arbitrary set of a μ a_\mu a μ . By introducing
L ~ = ∑ μ a μ L μ , (636) \widetilde{\mathcal{L}}=\sum_\mu a_\mu\mathcal{L}^\mu, \tag*{(636)} L = μ ∑ a μ L μ , (636)
we define F F i s h e r \mathcal{F}_{\mathrm{Fisher}} F Fisher as
F F i s h e r ≡ ∑ n = 1 N 1 Tr ρ Π n ( Tr ( 1 2 ( L ~ ρ + ρ L ~ ) Π n ) ) 2 = ∑ n = 1 N 1 Tr ρ Π n ∑ μ ν Tr ( a μ 2 ( L μ ρ + ρ L μ ) Π n ) × Tr ( a ν 2 ( L ν ρ + ρ L ν ) Π n ) = ∑ μ ν a μ F F i s h e r μ ν a ν . (637) \begin{aligned}
\mathcal{F}_{\mathrm{Fisher}}\equiv\sum_{n=1}^{N}\frac{1}{\operatorname{Tr}\rho\Pi_n}\left(\operatorname{Tr}\left(\frac{1}{2}\left(\widetilde{\mathcal{L}}\rho+\rho\widetilde{\mathcal{L}}\right)\Pi_n\right)\right)^2 \\
&=\sum_{n=1}^{N}\frac{1}{\operatorname{Tr}\rho\Pi_n}\sum_{\mu\nu}\operatorname{Tr}\left(\frac{a_\mu}{2}\left(\mathcal{L}^\mu\rho+\rho\mathcal{L}^\mu\right)\Pi_n\right)
\times\operatorname{Tr}\left(\frac{a_\nu}{2}\left(\mathcal{L}^\nu\rho+\rho\mathcal{L}^\nu\right)\Pi_n\right) \\
&=\sum_{\mu\nu}a_\mu\mathcal{F}_{\mathrm{Fisher}}^{\mu\nu}a_\nu. \tag*{(637)}
\end{aligned} F Fisher ≡ n = 1 ∑ N Tr ρ Π n 1 ( Tr ( 2 1 ( L ρ + ρ L ) Π n ) ) 2 = n = 1 ∑ N Tr ρ Π n 1 μν ∑ Tr ( 2 a μ ( L μ ρ + ρ L μ ) Π n ) × Tr ( 2 a ν ( L ν ρ + ρ L ν ) Π n ) = μν ∑ a μ F Fisher μν a ν . (637)
First, we show the inequality
∑ μ ν a μ F F i s h e r μ ν a ν = ∑ n = 1 N 1 Tr ρ Π n ( Tr ( 1 2 ( L ~ ρ + ρ L ~ ) Π n ) ) 2 = ∑ n = 1 N 1 Tr ρ Π n ( Tr ( Re L ~ ρ Π n ) ) 2 ≤ ∑ n = 1 N 1 Tr ρ Π n ∣ Tr ( L ~ ρ Π n ) ∣ 2 . (638) \begin{aligned}
\sum_{\mu\nu}a_\mu\mathcal{F}_{\mathrm{Fisher}}^{\mu\nu}a_\nu\\
&=\sum_{n=1}^{N}\frac{1}{\operatorname{Tr}\rho\Pi_n}\left(\operatorname{Tr}\left(\frac{1}{2}\left(\widetilde{\mathcal{L}}\rho+\rho\widetilde{\mathcal{L}}\right)\Pi_n\right)\right)^2 \\
&=\sum_{n=1}^{N}\frac{1}{\operatorname{Tr}\rho\Pi_n}\left(\operatorname{Tr}\left(\operatorname{Re}\widetilde{\mathcal{L}}\rho\Pi_n\right)\right)^2 \\
&\leq\sum_{n=1}^{N}\frac{1}{\operatorname{Tr}\rho\Pi_n}\left|\operatorname{Tr}\left(\widetilde{\mathcal{L}}\rho\Pi_n\right)\right|^2. \tag*{(638)}
\end{aligned} μν ∑ a μ F Fisher μν a ν = n = 1 ∑ N Tr ρ Π n 1 ( Tr ( 2 1 ( L ρ + ρ L ) Π n ) ) 2 = n = 1 ∑ N Tr ρ Π n 1 ( Tr ( Re L ρ Π n ) ) 2 ≤ n = 1 ∑ N Tr ρ Π n 1 Tr ( L ρ Π n ) 2 . (638)
Next, setting
A † = ρ Π n Tr ρ Π n , B = Π n L ~ ρ , (639) A^\dagger=\frac{\sqrt{\rho}\sqrt{\Pi_n}}{\sqrt{\operatorname{Tr}\rho\Pi_n}},\qquad B=\sqrt{\Pi_n}\widetilde{\mathcal{L}}\sqrt{\rho}, \tag*{(639)} A † = Tr ρ Π n ρ Π n , B = Π n L ρ , (639)
we use the Schwartz inequality
∣ Tr A † B ∣ 2 ≤ Tr A † A Tr B † B . (640) \left|\operatorname{Tr}A^\dagger B\right|^2\leq\operatorname{Tr}A^\dagger A\operatorname{Tr}B^\dagger B. \tag*{(640)} Tr A † B 2 ≤ Tr A † A Tr B † B . (640)
The left-hand side is
∣ Tr A † B ∣ 2 = ∣ Tr ρ Π n Tr ρ Π n Π n L ~ ρ ∣ 2 = ∣ Tr ρ Π n L ~ ρ Tr ρ Π n ∣ 2 = ∣ Tr L ~ ρ Π n Tr ρ Π n ∣ 2 , (641) \begin{aligned}
\left|\operatorname{Tr}A^\dagger B\right|^2
=\left|\operatorname{Tr}\frac{\sqrt{\rho}\sqrt{\Pi_n}}{\sqrt{\operatorname{Tr}\rho\Pi_n}}\sqrt{\Pi_n}\widetilde{\mathcal{L}}\sqrt{\rho}\right|^2 \\
&=\left|\operatorname{Tr}\frac{\sqrt{\rho}\Pi_n\widetilde{\mathcal{L}}\sqrt{\rho}}{\sqrt{\operatorname{Tr}\rho\Pi_n}}\right|^2 \\
&=\left|\operatorname{Tr}\frac{\widetilde{\mathcal{L}}\rho\Pi_n}{\sqrt{\operatorname{Tr}\rho\Pi_n}}\right|^2, \tag*{(641)}
\end{aligned} Tr A † B 2 = Tr Tr ρ Π n ρ Π n Π n L ρ 2 = Tr Tr ρ Π n ρ Π n L ρ 2 = Tr Tr ρ Π n L ρ Π n 2 , (641)
while the right-hand side is
∑ n = 1 N Tr A † A Tr B † B = ∑ n = 1 N Tr ρ Π n Tr ρ Π n Π n ρ Tr Π n L ~ ρ ρ L ~ Π n = ∑ n = 1 N Tr ρ Π n ρ Tr ρ Π n Tr Π n L ~ ρ L ~ Π n = ∑ n = 1 N Tr ρ Π n Tr ρ Π n Tr Π n L ~ ρ L ~ = ∑ n = 1 N Tr Π n L ~ ρ L ~ = Tr L ~ ρ L ~ = Tr ρ L ~ 2 = 1 2 F F i s h e r , (642) \begin{aligned}
\sum_{n=1}^{N}\operatorname{Tr}A^\dagger A\operatorname{Tr}B^\dagger B \\
&=\sum_{n=1}^{N}\operatorname{Tr}\frac{\sqrt{\rho}\sqrt{\Pi_n}}{\sqrt{\operatorname{Tr}\rho\Pi_n}}\sqrt{\Pi_n}\sqrt{\rho}\operatorname{Tr}\sqrt{\Pi_n}\widetilde{\mathcal{L}}\sqrt{\rho}\sqrt{\rho}\widetilde{\mathcal{L}}\sqrt{\Pi_n} \\
&=\sum_{n=1}^{N}\operatorname{Tr}\frac{\sqrt{\rho}\Pi_n\sqrt{\rho}}{\operatorname{Tr}\rho\Pi_n}\operatorname{Tr}\sqrt{\Pi_n}\widetilde{\mathcal{L}}\rho\widetilde{\mathcal{L}}\sqrt{\Pi_n} \\
&=\sum_{n=1}^{N}\operatorname{Tr}\frac{\rho\Pi_n}{\operatorname{Tr}\rho\Pi_n}\operatorname{Tr}\Pi_n\widetilde{\mathcal{L}}\rho\widetilde{\mathcal{L}}
=\sum_{n=1}^{N}\operatorname{Tr}\Pi_n\widetilde{\mathcal{L}}\rho\widetilde{\mathcal{L}} \\
&=\operatorname{Tr}\widetilde{\mathcal{L}}\rho\widetilde{\mathcal{L}}
=\operatorname{Tr}\rho\widetilde{\mathcal{L}}^2
=\frac{1}{2}\mathcal{F}_{\mathrm{Fisher}}, \tag*{(642)}
\end{aligned} n = 1 ∑ N Tr A † A Tr B † B = n = 1 ∑ N Tr Tr ρ Π n ρ Π n Π n ρ Tr Π n L ρ ρ L Π n = n = 1 ∑ N Tr Tr ρ Π n ρ Π n ρ Tr Π n L ρ L Π n = n = 1 ∑ N Tr Tr ρ Π n ρ Π n Tr Π n L ρ L = n = 1 ∑ N Tr Π n L ρ L = Tr L ρ L = Tr ρ L 2 = 2 1 F Fisher , (642)
where we have defined
F F i s h e r μ ν = 1 2 Tr L ~ 2 = 1 2 Tr ( ∑ μ a μ L μ ) 2 = 1 2 ∑ μ ν a μ Tr [ ρ { L μ , L ν } ] a ν = ∑ μ ν a μ F F i s h e r μ ν a ν . (643) \begin{aligned}
\mathcal{F}_{\mathrm{Fisher}}^{\mu\nu}
=\frac{1}{2}\operatorname{Tr}\widetilde{\mathcal{L}}^2
=\frac{1}{2}\operatorname{Tr}\left(\sum_\mu a_\mu\mathcal{L}^\mu\right)^2 \\
&=\frac{1}{2}\sum_{\mu\nu}a_\mu\operatorname{Tr}\left[\rho\left\{\mathcal{L}^\mu,\mathcal{L}^\nu\right\}\right]a_\nu\\
&=\sum_{\mu\nu}a_\mu\mathcal{F}_{\mathrm{Fisher}}^{\mu\nu}a_\nu. \tag*{(643)}
\end{aligned} F Fisher μν = 2 1 Tr L 2 = 2 1 Tr ( μ ∑ a μ L μ ) 2 = 2 1 μν ∑ a μ Tr [ ρ { L μ , L ν } ] a ν = μν ∑ a μ F Fisher μν a ν . (643)
Hence, the quantum Cramér-Rao inequality (282) is proved.
By inserting
ρ = ∣ ψ n ⟩ ⟨ ψ n ∣ , (644) \rho=\lvert\psi_n\rangle\langle\psi_n\rvert, \tag*{(644)} ρ = ∣ ψ n ⟩ ⟨ ψ n ∣ , (644)
to Eq. (290) , we have
F F i s h e r μ ν = 2 Tr ( ∣ ψ n ⟩ ⟨ ψ n ∣ ( ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k μ ) ( ∂ ∣ ψ n ⟩ ∂ k ν ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k ν ) ) = 2 Tr ( ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ + ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν + ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν ) + 2 Tr ( ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k ν + ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k μ ) + 2 Tr ( ∣ ψ n ⟩ ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ∂ ⟨ ψ n ∣ ∂ k ν + ∣ ψ n ⟩ ∂ ∣ ψ n ⟩ ∂ k ν ⟨ ψ n ∣ ∂ ⟨ ψ n ∣ ∂ k μ + ∣ ψ n ⟩ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ) + 2 Tr ( ∣ ψ n ⟩ ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ∣ ψ n ⟩ ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ) . (645) \begin{aligned}
\mathcal{F}_{\mathrm{Fisher}}^{\mu\nu}
=2\operatorname{Tr}\left(\lvert\psi_n\rangle\langle\psi_n\rvert\left(\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\langle\psi_n\vert+\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\right)
\left(\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\langle\psi_n\vert+\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\right)\right) \\
&=2\operatorname{Tr}\left(\langle\psi_n\vert\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\langle\psi_n\vert\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}
+\langle\psi_n\vert\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\lvert\psi_n\rangle
+\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}
+\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\lvert\psi_n\rangle\langle\psi_n\vert\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\right) \\
&+2\operatorname{Tr}\left(\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}
+\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\langle\psi_n\vert
+\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\langle\psi_n\vert
+\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\right) \\
&+2\operatorname{Tr}\left(\lvert\psi_n\rangle\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\langle\psi_n\vert\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}
+\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\lvert\psi_n\rangle\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}
+\lvert\psi_n\rangle\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\langle\psi_n\vert\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}
+\lvert\psi_n\rangle\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n\rvert}{\partial k_\mu}\lvert\psi_n\rangle\right) \\
&+2\operatorname{Tr}\left(\lvert\psi_n\rangle\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\langle\psi_n\vert\frac{\partial\lvert\psi_n\rangle}{\partial k_\nu}
+\lvert\psi_n\rangle\frac{\partial\lvert\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n\rvert}{\partial k_\nu}\lvert\psi_n\rangle\right). \tag*{(645)}
\end{aligned} F Fisher μν = 2 Tr ( ∣ ψ n ⟩ ⟨ ψ n ∣ ( ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ) ( ∂ k ν ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ) ) = 2 Tr ( ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ + ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ) + 2 Tr ( ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ) + 2 Tr ( ∣ ψ n ⟩ ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k ν ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ⟨ ψ n ∣ + ∣ ψ n ⟩ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ) + 2 Tr ( ∣ ψ n ⟩ ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ∣ ψ n ⟩ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ) . (645)
By using the relations
⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ = ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ = 0 , (646) \begin{aligned}
\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}
+\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n|}{\partial k_\nu}|\psi_n\rangle
&=\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n|}{\partial k_\nu}|\psi_n\rangle\\
&=0, \tag*{(646)}
\end{aligned} ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ = ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ = 0 , (646)
⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ + ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ = ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ = 0 , (647) \begin{aligned}
\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}
+\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n|}{\partial k_\mu}|\psi_n\rangle
&=\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n|}{\partial k_\mu}|\psi_n\rangle\\
&=0, \tag*{(647)}
\end{aligned} ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ + ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ = ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ = 0 , (647)
we have
F Q F i s h e r μ ν = 2 Tr ( ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν − ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ − ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ) = 2 ( F μ ν + F ν μ ) = 4 g μ ν . (648) \begin{aligned}
\mathcal{F}^{\mu\nu}_{\mathrm{QFisher}}
&=2\operatorname{Tr}\left(\frac{\partial\langle\psi_n|}{\partial k_\mu}\frac{\partial|\psi_n\rangle}{\partial k_\nu}
-\frac{\partial\langle\psi_n|}{\partial k_\mu}|\psi_n\rangle\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\right.\\
&\qquad\left.+\frac{\partial\langle\psi_n|}{\partial k_\nu}\frac{\partial|\psi_n\rangle}{\partial k_\mu}
-\frac{\partial\langle\psi_n|}{\partial k_\nu}|\psi_n\rangle\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\right)\\
&=2\left(\mathcal{F}^{\mu\nu}+\mathcal{F}^{\nu\mu}\right)=4g^{\mu\nu}. \tag*{(648)}
\end{aligned} F QFisher μν = 2 Tr ( ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ − ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ − ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ) = 2 ( F μν + F ν μ ) = 4 g μν . (648)
This is Eq. (292) in the main text.
U μ ν = − 2 i Tr [ ∣ ψ n ⟩ ⟨ ψ n ∣ ( ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν − ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ) ] = − 2 i Tr ( ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν + ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ + ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν + ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν − ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ − ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ − ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ − ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ) . (649) \begin{aligned}
\mathcal{U}^{\mu\nu}
&=-2i\operatorname{Tr}\left[|\psi_n\rangle\langle\psi_n|\left(\frac{\partial|\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n|}{\partial k_\nu}
-\frac{\partial|\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n|}{\partial k_\mu}\right)\right]\\
&=-2i\operatorname{Tr}\left(\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}
+\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\frac{\partial\langle\psi_n|}{\partial k_\nu}|\psi_n\rangle\right.\\
&\qquad+\left.\frac{\partial\langle\psi_n|}{\partial k_\mu}\frac{\partial|\psi_n\rangle}{\partial k_\nu}
+\frac{\partial\langle\psi_n|}{\partial k_\mu}|\psi_n\rangle\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\right.\\
&\qquad-\left.\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}
-\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\frac{\partial\langle\psi_n|}{\partial k_\mu}|\psi_n\rangle\right.\\
&\qquad-\left.\frac{\partial\langle\psi_n|}{\partial k_\nu}\frac{\partial|\psi_n\rangle}{\partial k_\mu}
-\frac{\partial\langle\psi_n|}{\partial k_\nu}|\psi_n\rangle\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\right). \tag*{(649)}
\end{aligned} U μν = − 2 i Tr [ ∣ ψ n ⟩ ⟨ ψ n ∣ ( ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ − ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ) ] = − 2 i Tr ( ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ + ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ + ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ − ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ − ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ − ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ − ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ) . (649)
Using the relations Eqs. (646) and (I21), we have
U μ ν = − 2 i Tr ( ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ∂ k ν − ∂ ⟨ ψ n ∣ ∂ k μ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k ν − ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ ∂ k μ + ∂ ⟨ ψ n ∣ ∂ k ν ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ ∣ ψ n ⟩ ∂ k μ ) = − 2 i ( F μ ν − F ν μ ) = − 4 Ω μ ν . (650) \begin{aligned}
\mathcal{U}^{\mu\nu}
&=-2i\operatorname{Tr}\left(\frac{\partial\langle\psi_n|}{\partial k_\mu}\frac{\partial|\psi_n\rangle}{\partial k_\nu}
-\frac{\partial\langle\psi_n|}{\partial k_\mu}|\psi_n\rangle\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\nu}\right.\\
&\qquad\left.-\frac{\partial\langle\psi_n|}{\partial k_\nu}\frac{\partial|\psi_n\rangle}{\partial k_\mu}
+\frac{\partial\langle\psi_n|}{\partial k_\nu}|\psi_n\rangle\langle\psi_n|\frac{\partial|\psi_n\rangle}{\partial k_\mu}\right)\\
&=-2i\left(\mathcal{F}^{\mu\nu}-\mathcal{F}^{\nu\mu}\right)=-4\Omega^{\mu\nu}. \tag*{(650)}
\end{aligned} U μν = − 2 i Tr ( ∂ k μ ∂ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ − ∂ k μ ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k ν ∂ ∣ ψ n ⟩ − ∂ k ν ∂ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ + ∂ k ν ∂ ⟨ ψ n ∣ ∣ ψ n ⟩ ⟨ ψ n ∣ ∂ k μ ∂ ∣ ψ n ⟩ ) = − 2 i ( F μν − F ν μ ) = − 4 Ω μν . (650)
This is Eq. (293) in the main text.