betaFeedback

WorksPhysics and Astronomy

Quadrupole Origin of the Orbital Hall Effect in an sp3sp^3 Centrosymmetric Lattice: Robustness and Tunability under Disorder

Abstract

We investigate how disorder affects orbital Hall transport in the centrosymmetric spinless sp3sp^3 tight-binding model on a square lattice, widely used as a minimal model for the orbital Hall effect [1]. In the pristine system, the intrinsic orbital Hall response is related to the quadrupolar orbital texture generated by ss-pp hybridization. Using the Kubo-Bastin formalism expanded in Chebyshev polynomials, we analyze the effect of vacancies, orbital-dependent Anderson on-site disorder, and selective VspσV_{spσ}-bond disorder on the orbital transport. For Anderson disorder and vacancies, the longitudinal charge conductivity decreases faster than the orbital Hall conductivity over the range studied, leading to an increase of the orbital Hall angle and, in the Anderson case, saturation at stronger disorder. Selective VspσV_{spσ}-bond disorder behaves differently: by modifying only the ss-pp hybridization channel while leaving the remaining hopping network intact, it can produce a non-monotonic orbital Hall response and enhance the orbital Hall conductivity at intermediate dilution. Overall, our results show that the orbital Hall response remains sizable under vacancy and on-site disorder, while selective modification of the ss-pp hybridization provides an additional route to tune the orbital Hall conductivity.

Introduction

Orbitronics studies non-equilibrium phenomena associated with electronic orbital angular momentum in solids, including its generation, transport, accumulation, and detection [2, 3, 4]. A widely studied transport response is the orbital Hall effect (OHE), in which an applied electric field drives a transverse flow of orbital angular momentum [1, 5, 6, 7, 8, 9]. Closely related current-induced orbital responses include the orbital Rashba–Edelstein effect [10, 11, 12, 13, 14] and orbital torque [15, 16, 17].

Unlike its spin counterpart, the OHE does not require strong spin-orbit coupling, so it can be large in light elements and in materials where spin-orbit-driven responses are weak [18, 19]. However, those same materials are rarely the clean crystals in which the intrinsic mechanism is defined: the metals in which the OHE is measured and used are alloyed, polycrystalline, or amorphous. Whether the intrinsic response survives in such systems, and how it is modified by disorder, is therefore an important question.

Recent theoretical works have addressed different disorder regimes, including the robustness of intrinsic orbital Hall contributions in weakly disordered crystalline metals, the role of vertex corrections, and the emergence of extrinsic orbital Hall mechanisms in dilute or ultra-clean limits [20, 21, 22, 23, 24, 25, 26]. At the same time, experiments are now probing orbital Hall signals in systems where the conductivity is tuned by alloying, structural disorder, or crystalline order [27, 28, 29]. These results show that orbital Hall responses can persist in disordered samples, but also that their scaling with longitudinal transport depends on the microscopic disorder and transport regime.

The intrinsic OHE has been connected to orbital textures in momentum space, understood as the momentum-dependent orbital composition of Bloch states [1]. In the pp-orbital sector, this texture can be decomposed into dipolar and quadrupolar components. For centrosymmetric and time-reversal-symmetric systems, the equilibrium orbital dipole is quenched, so the OHE cannot be understood as arising from a pre-existing texture of LzL_z. The quadrupoles provide the relevant structure: quadrupole expectation values can form non-trivial momentum-space textures without breaking either symmetry. Since the dynamics of the orbital dipole is coupled to the quadrupole degrees of freedom [30, 31], these quadrupolar textures provide the microscopic structure from which the transverse LzL_z orbital current emerges. Very recently, polarization-resolved Kerr microscopy has directly observed current-induced orbital-quadrupole accumulation in Ti and Pt, isolating the time-reversal-even quadrupolar response from the dipolar contribution and finding the quadrupolar optical response larger in both metals [32].

Here we make this connection explicit in a spinless sp3sp^{3} tight-binding model on a square lattice. The ss-pp hybridization generates a quadrupolar texture and a low-energy theory relates this texture to the intrinsic LzL_z-projected orbital Hall response. This clean limit provides the reference point for studying how different microscopic forms of disorder modify orbital Hall transport.

We consider orbital-dependent Anderson disorder, vacancies, and selective VspσV_{sp\sigma}-bond disorder. The first two provide standard forms of static disorder, while the last one acts directly on the ss-pp hybridization channel. Using the Kubo–Bastin formalism expanded in Chebyshev polynomials, we evaluate the orbital Hall conductivity, the longitudinal conductivity, and the orbital Hall angle, allowing us to distinguish changes in orbital-current generation from changes in charge transport as disorder is increased. We find that the orbital Hall conductivity is markedly less sensitive to disorder than the longitudinal conductivity, so that the orbital Hall angle increases over a wide disorder range before saturating, consistent with trends reported in recent experiments [28, 29].

Theory and Method

Model Hamiltonian

To analyze the orbital Hall response and its relation to orbital textures, we consider a tight-binding model on a simple square lattice with four orbitals s,px,py,pzs,p_x,p_y,p_z at each lattice point. We ignore the spin degree of freedom to focus solely on orbital response. The real-space Hamiltonian reads

H=∑i,αεiαciα†ciα−∑⟨i,j⟩,α,β[tijαβciα†cjβ+h.c.](1)H = \sum_{i,\alpha} \varepsilon_i^\alpha c_{i\alpha}^{\dagger} c_{i\alpha} - \sum_{\langle i,j\rangle,\alpha,\beta} \left[t_{ij}^{\alpha\beta} c_{i\alpha}^{\dagger}c_{j\beta} + \mathrm{h.c.}\right] \tag*{(1)}

where εiα\varepsilon_i^\alpha represents the on-site energy of the α\alpha orbital at site ii, tijαβt_{ij}^{\alpha\beta} is the transfer integral from orbital β\beta at site jj to orbital α\alpha at site ii. To determine the hopping amplitudes we adopt the Slater-Koster approximation [33],

tijss=Vssσ,t_{ij}^{ss} = V_{ss\sigma},
tijspx=lVspσ,t_{ij}^{sp_x} = lV_{sp\sigma},
tijpxpx=l2Vppσ+(1−l2)Vppπ,t_{ij}^{p_xp_x} = l^2V_{pp\sigma} + (1-l^2)V_{pp\pi},
tijpxpy=lm(Vppσ−Vppπ),(2)t_{ij}^{p_xp_y} = lm(V_{pp\sigma} - V_{pp\pi}), \tag*{(2)}

where ll and mm are the direction cosines of the vector connecting sites ii and jj. Numerical values of the Hamiltonian parameters are as follows: εis=3.2\varepsilon_i^s = 3.2 eV, and εipx=εipy=εipz=−0.5\varepsilon_i^{p_x} = \varepsilon_i^{p_y} = \varepsilon_i^{p_z} = -0.5 eV for on-site energies, Vssσ=−0.5V_{ss\sigma} = -0.5 eV, Vspσ=0.5V_{sp\sigma} = 0.5 eV, Vppσ=0.5V_{pp\sigma} = 0.5 eV, and Vppπ=−0.2V_{pp\pi} = -0.2 eV for nearest-neighbor hopping amplitudes.

When we write this Hamiltonian in k-space, the different parities of the ss and pp orbitals appear as complex electronic couplings. The ss orbital has even parity, while the pxp_x and pyp_y orbitals have odd parity, so the ss-pp hopping changes sign when the bond direction is reversed. For a bond along the direction μ=x,y\mu=x,y, the two opposite contributions are VspσeikμaV_{sp\sigma}e^{ik_\mu a} and −Vspσe−ikμa-V_{sp\sigma}e^{-ik_\mu a}, whose sum gives Hspμ(k)=2iVspσsin⁡(kμa)H_{sp_\mu}(\mathbf{k})=2iV_{sp\sigma}\sin(k_\mu a) with μ=x,y\mu=x,y. The ss-pp mixing is therefore purely imaginary and momentum dependent. It acts as an effective momentum-space vector potential and generates a finite orbital Berry curvature in the model [1].

Microscopic features of the quadrupole texture

While related works have discussed the OHE in terms of orbital textures [34], here we decompose these textures into multipolar components [31]. In a full pp-orbital manifold, the non-charge orbital degrees of freedom can be organized into three orbital dipoles and five orbital quadrupoles: population imbalances among pxp_x, pyp_y, and pzp_z are described by diagonal quadrupoles such as Qx2−y2Q_{x^2-y^2} and Q3z2−r2Q_{3z^2-r^2}, real coherences between pairs of orbitals by QxyQ_{xy}, QyzQ_{yz}, and QzxQ_{zx}, and chiral complex combinations by the components of L\mathbf{L}. Thus, real linear combinations of pp orbitals correspond to quadrupolar textures, while chiral complex combinations carry orbital angular momentum. These operators are equivalent to the orbital angular position (OAP) operators introduced in Ref. [30]. This multipolar decomposition is useful because QxyQ_{xy} and Qx2−y2Q_{x^2-y^2} are Hermitian orbital operators with a direct physical meaning: they describe, respectively, real inter-orbital coherence and orbital population imbalance. In this sense, the quadrupole language provides a concrete operator description of the orbital texture rather than only a characterization of the Bloch-state composition.

While the orbital dipole vanishes at equilibrium in the presence of both time-reversal and inversion symmetries [1, 30], the momentum-dependent ss-pp hybridization generates a non-uniform quadrupolar texture. The symmetry argument behind this distinction is simple. Under inversion the pp orbitals are odd, so any bilinear combination in them is even, and both L\mathbf{L} and Q\mathbf{Q} are inversion even; inversion symmetry alone does not force either response to vanish. Under time reversal, however, L\mathbf{L} is odd while Q\mathbf{Q} is even. Combining the two, inversion symmetry requires ⟨Lz⟩k=⟨Lz⟩−k\langle L_z\rangle_{\mathbf{k}}=\langle L_z\rangle_{-\mathbf{k}} while time-reversal symmetry requires ⟨Lz⟩k=−⟨Lz⟩−k\langle L_z\rangle_{\mathbf{k}}=-\langle L_z\rangle_{-\mathbf{k}}, so the orbital dipole vanishes at every k\mathbf{k}. For the quadrupole both symmetries impose the same condition, ⟨Q⟩k=⟨Q⟩−k\langle Q\rangle_{\mathbf{k}}=\langle Q\rangle_{-\mathbf{k}}, which is satisfied by any even function of k\mathbf{k} and leaves the quadrupole texture unconstrained.

We first analyze the clean system in momentum space by Fourier transforming the real-space lattice model of Eq. (1) and obtaining the Bloch Hamiltonian H(k)H(\mathbf{k}). Using the basis Ψk=(∣sk⟩,∣pxk⟩,∣pyk⟩,∣pzk⟩)T\Psi_{\mathbf{k}}=(|s\mathbf{k}\rangle, |p_x\mathbf{k}\rangle, |p_y\mathbf{k}\rangle, |p_z\mathbf{k}\rangle)^{\mathsf{T}}, the matrix elements are given by the Slater-Koster parameters specified in Eq. (2).

Because the model is two-dimensional, the pzp_z or

The pzp_z orbital remains decoupled from the s,px,pys,p_x,p_y sector. As a result, the only quadrupole operators that can develop non-trivial momentum-dependent expectation values in the px,pyp_x,p_y subspace are Qxy={Lx,Ly}=−ℏ2(∣px⟩⟨py∣+∣py⟩⟨px∣)Q_{xy}=\{L_x,L_y\}=-\hbar^2\left(|p_x\rangle\langle p_y|+|p_y\rangle\langle p_x|\right), which describes the real linear combination between pxp_x and pyp_y, and Qx2−y2=Lx2−Ly2=−ℏ2(∣px⟩⟨px∣−∣py⟩⟨py∣)Q_{x^2-y^2}=L_x^2-L_y^2=-\hbar^2\left(|p_x\rangle\langle p_x|-|p_y\rangle\langle p_y|\right), which describes the imbalance between pxp_x and pyp_y. The remaining quadrupoles either involve pzp_z or reduce to a trivial constant in this subspace, and therefore do not generate additional textures for the present model.

Figure 1 summarizes the clean-limit structure of the full lattice model. Panel 1(a) defines the high-symmetry path used throughout the paper, while Figure 1(b) shows the corresponding band structure resolved by orbital character. The pxp_x and pyp_y bands are degenerate at both Γ\Gamma

Clean-limit band structure, orbital quadrupole texture, and $L_z$-projected Berry curvature of the full lattice model

Figure 1. Clean-limit band structure, orbital quadrupole texture, and LzL_z-projected Berry curvature of the full lattice model. (a) Square-lattice Brillouin zone showing the Γ\Gamma-X-M-Γ\Gamma-M′' path. (b) Band structure along the path, marker color indicating dominant orbital character (ss: orange; pxp_x: red; pyp_y: blue star; pzp_z: gray) and marker size the corresponding orbital weight at each (k,ϵ)(k,\epsilon). (c),(d) ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle along the same path, color indicating sign (blue negative, red positive), marker size the orbital weight. (e)-(g) ⟨Qxy⟩+\langle Q_{xy}\rangle_+, ⟨Qx2−y2⟩+\langle Q_{x^2-y^2}\rangle_+, and the LzL_z-projected Berry curvature of the upper band, Ω+,yxLz(k)\Omega^{L_z}_{+,yx}(k), evaluated by exact diagonalization over the entire Brillouin zone. Ω+,yxLz(k)\Omega^{L_z}_{+,yx}(k) is evaluated via the exact Kubo sum.

and MM, where sin⁡(kxa)=sin⁡(kya)=0\sin(k_xa)=\sin(k_ya)=0 and the ss orbital decouples from the pp-orbital sector. Along the diagonal M−Γ−M′M-\Gamma-M', the condition Hxx(k)=Hyy(k)H_{xx}(k)=H_{yy}(k) follows from the x↔yx\leftrightarrow y symmetry of the px,pyp_x,p_y block. Away from these symmetry-enforced degeneracies, the splitting of the px,pyp_x,p_y bands is generated by the ss-pp hybridization.

Figures 1(c) and 1(d) show the corresponding quadrupole expectation values along the same path. The two bands carry opposite quadrupolar textures: QxyQ_{xy} is selected along the diagonal directions, while Qx2−y2Q_{x^2-y^2} distinguishes the pxp_x- and pyp_y-dominated sectors along the principal directions. The full Brillouin-zone maps in Figs. 1(e) and 1(f) make this texture more explicit: QxyQ_{xy} changes sign between neighboring quadrants, whereas Qx2−y2Q_{x^2-y^2} changes sign across the zone diagonals. Figure 1(g) shows that the LzL_z-projected Berry curvature is concentrated in the same regions where the quadrupolar texture varies most strongly. Thus, the full lattice calculation already displays the main microscopic ingredients of the intrinsic OHE: ss-pp hybridization produces a quadrupolar texture in the px,pyp_x,p_y sector, and the orbital Berry curvature follows from this texture. Appendix A describes how these textures are extracted from momentum- and orbital-resolved spectral functions using Quantum KITE, while Appendix B applies the same analysis to the disordered systems.

The connection between spsp hybridization and these quadrupole degrees of freedom becomes explicit when examining the low-energy effective Hamiltonian near the center of the Brillouin zone (Γ\Gamma point). By employing the Löwdin downfolding technique [34] to project out the higher-lying ss band as a remote mediator, the effective Hamiltonian projected onto the pp-subspace is, to second order,

Hpeff(k)=h0(k)1−Cℏ2kxkyQxy−Bℏ2(kx2−ky2)Qx2−y2,(3)H_p^{\mathrm{eff}}(k)=h_0(k)\mathbb{1}-\frac{C}{\hbar^2}k_xk_yQ_{xy}-\frac{B}{\hbar^2}(k_x^2-k_y^2)Q_{x^2-y^2}, \tag*{(3)}

with h0(k)=Ep+A(kx2+ky2)h_0(k)=E_p+\mathcal{A}(k_x^2+k_y^2), A=A+C/2\mathcal{A}=A+C/2, B=B+C/2\mathcal{B}=B+C/2, C=4a2Vspσ2/(ϵp−ϵs)C=4a^2V_{sp\sigma}^2/(\epsilon_p-\epsilon_s), A=a22(Vppσ+Vppπ)A=\frac{a^2}{2}(V_{pp\sigma}+V_{pp\pi}), B=a22(Vppσ−Vppπ)B=\frac{a^2}{2}(V_{pp\sigma}-V_{pp\pi}), and Ep=ϵp+2(Vppσ+Vppπ)E_p=\epsilon_p+2(V_{pp\sigma}+V_{pp\pi}). Thus the projected Hamiltonian has a simple Pauli-matrix structure in the {∣px⟩,∣py⟩}\{|p_x\rangle,|p_y\rangle\} subspace, where Qxy=−ℏ2σxQ_{xy}=-\hbar^2\sigma_x and Qx2−y2=−ℏ2σzQ_{x^2-y^2}=-\hbar^2\sigma_z.

Diagonalizing Eq. (3) gives the exact bands E±(k)=h0(k)±d(k)E_\pm(k)=h_0(k)\pm d(k), d(k)=dx2+dz2d(k)=\sqrt{d_x^2+d_z^2}, with dx=Ckxkyd_x=Ck_xk_y and dz(k)=B(kx2−ky2)d_z(k)=B(k_x^2-k_y^2). The eigenstates carry quadrupole expectation values

⟨Qxy⟩±=∓ℏ2dxd,⟨Qx2−y2⟩±=∓ℏ2dzd.(4)\langle Q_{xy}\rangle_\pm=\mp\hbar^2\frac{d_x}{d},\qquad\langle Q_{x^2-y^2}\rangle_\pm=\mp\hbar^2\frac{d_z}{d}. \tag*{(4)}

Figure 2 shows the resulting textures from this effective theory, restricted to the small-kk window where the quadratic expansion holds.

Band structure and band-resolved quadrupolar textures of the effective $p_x$-$p_y$ model

Figure 2. Band structure and band-resolved quadrupolar textures of the effective pxp_x-pyp_y model, valid near the Γ\Gamma point, along (a) kx=kyk_x=k_y, (b) kx=0k_x=0 and (c) ky=0k_y=0. (a) corresponds to ⟨Qxy⟩\langle Q_{xy}\rangle and (b)-(c) correspond to ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle. The two bands carry opposite textures, red positive and blue negative. (d), (e), and (f) show the second-band quadrupole and Berry-curvature textures from the effective theory, Eqs. (4) and (5) – ⟨Qxy⟩\langle Q_{xy}\rangle, ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle, and ΩyxLz(k)\Omega^{L_z}_{yx}(\mathbf{k}), respectively – restricted to the small-kk region kx,ky∈[−1,1]k_x,k_y\in[-1,1] near Γ\Gamma where this quadratic expansion is valid, rather than the full Brillouin zone shown in Figure 1.

Near Γ\Gamma, where the quadratic expansion is valid, the effective theory agrees with the exact diagonalization of the full lattice model. The band-resolved quadrupole textures in Figure 2(a)-(c) reproduce the corresponding features of Figure 1(c),(Figure 1d), while the upper-band maps in Figure 2(d)-(Figure 2f) reproduce the local structure of Figure 1(e)-(Figure 1g).

Figure 2(a)-(Figure 2c) shows the band-resolved quadrupole textures of the effective model near Γ\Gamma. Along the principal axes, the texture is dominated by Qx2−y2Q_{x^2-y^2}, while along the diagonal kx=kyk_x=k_y it is dominated by QxyQ_{xy}. Away from these high-symmetry directions, both quadrupole components are present, and the two bands carry opposite textures. The corresponding momentum-space maps in Figure 2(d) and (e) show the expected fourfold pattern: ⟨Qxy⟩\langle Q_{xy}\rangle changes sign between neighboring quadrants, whereas ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle changes sign across the diagonals. The angular profile is slightly distorted because the coefficients of the two quadrupole terms are different, reflecting the competition between direct pp-orbital hopping and the virtual pp-ss-pp process.

This quadrupole structure is the microscopic origin of the orbital Hall response. The velocity operators va=ℏ−1∂kaHeffv_a=\hbar^{-1}\partial_{k_a}H_{\mathrm{eff}} inherit the momentum derivatives of dxd_x and dzd_z, so the electric-field vertex vxv_x directly probes the quadrupole texture. The orbital-current operator jyLz=12{vy,Lz}j_y^{L_z}=\frac{1}{2}\{v_y,L_z\}, by contrast, anticommutes with both quadrupole operators and collapses onto the orbital-dipole channel alone, jyLz=(2Aky/ℏ)Lzj_y^{L_z}=(2Ak_y/\hbar)L_z: the current vertex identifies what is transported (LzL_z) but carries no memory of the texture itself. The texture re-enters through the interband matrix element ⟨u−∣vx∣u+⟩=(dz∂kxdx−dx∂kxdz)/(ℏd)\langle u_-|v_x|u_+\rangle=(d_z\partial_{k_x}d_x-d_x\partial_{k_x}d_z)/(\hbar d), finite only when the field rotates the quadrupole eigenbasis, and vanishing identically if C=0C=0 or B=0\mathcal{B}=0. Combining both vertices gives the projected Berry curvature

Ω±,yxLz(k)=∓ACBky2(kx2+ky2)d(k)3,(5)\Omega^{L_z}_{\pm,yx}(\mathbf{k})=\mp\frac{AC\mathcal{B}k_y^2(k_x^2+k_y^2)}{d(\mathbf{k})^3}, \tag*{(5)}

shown in Figure 2(f) with weight concentrated away from Γ\Gamma and the axes, where the interband coupling requires both quadrupole channels to be simultaneously active. This clean, disorder-free texture provides the microscopic reference for the disorder analysis below. Equation (5) shows that, in the low-energy clean limit, the LzL_z-projected Berry curvature depends on both the ss-pp-mediated coefficient CC and the pxp_x-pyp_y quadrupolar coefficient B\mathcal{B} and disorder can affect these ingredients in different ways.

The Kubo-Bastin formula for Orbital-Conductivity

To evaluate the orbital conductivity σxyO\sigma^\mathcal{O}_{xy} in large-scale disordered systems, we employ the Kubo-Bastin linear response formalism. The orbital current operator flowing along the α\alpha-direction with orbital operator O\mathcal{O} is defined as

JαO=12{O,vα}(6)J^\mathcal{O}_\alpha=\frac{1}{2}\{\mathcal{O},v_\alpha\} \tag*{(6)}

where vα=iℏ[H,rα]v_\alpha=\frac{i}{\hbar}[H,r_\alpha] is the velocity operator component. Within the independent-particle approximation, the dc orbital conductivity is given by Bastin formula [35],

σαβO(μ,T)=ieℏΩ∫−∞∞dϵ f(ϵ,μ,T)×Tr⁡[JαOδ(ϵ−H)vβdG+(ϵ)dϵ−JαOdG−(ϵ)dϵvβδ(ϵ−H)],(7)\begin{aligned} \sigma^\mathcal{O}_{\alpha\beta}(\mu,T)=\frac{i e\hbar}{\Omega}\int_{-\infty}^{\infty}\mathrm{d}\epsilon\,f(\epsilon,\mu,T) \\ &\times\operatorname{Tr}\left[J^\mathcal{O}_\alpha\delta(\epsilon-H)v_\beta\frac{\mathrm{d}G^+(\epsilon)}{\mathrm{d}\epsilon}-J^\mathcal{O}_\alpha\frac{\mathrm{d}G^-(\epsilon)}{\mathrm{d}\epsilon}v_\beta\delta(\epsilon-H)\right], \tag*{(7)} \end{aligned}

where f(ϵ,μ,T)f(\epsilon,\mu,T) is the Fermi-Dirac distribution function, G±(ϵ)=(ϵ−H±η)−1G^\pm(\epsilon)=(\epsilon-H\pm\eta)^{-1} represent the retarded and advanced Green's functions, respectively, with η\eta acting as an imaginary energy broadening, and Ω\Omega is the system volume.

Direct evaluation of this formula via full diagonalization scales as an order N3\mathcal{N}^3, where N\mathcal{N} is the dimension of the Hilbert space, severely limiting the maximum treatable system size. To overcome this limitation and calculate macroscopic transport properties in the presence of real-space disorder, we expand the spectral operators using the Chebyshev Polynomial Method [36, 37, 38, 39]. The Hamiltonian and energy variables are first rescaled into the dimensionless interval [−1,1][-1,1] to ensure convergence, and are represented by H~\widetilde{H} and ϵ~\widetilde{\epsilon}. The delta functions and Green’s functions are then expanded in a series of Chebyshev polynomials of the first kind, Tm(x)T_m(x). To filter out unphysical Gibbs oscillations resulting from the truncation of the series at a finite moment MM, we incorporate the kernel gmg_m. The resulting formula reads

σαβO(μ,T)≈4e2ℏπΩ4ΔE2∫−11dε~ f(ε~,μ,T~)(1−ε~2)ζαβO(ε~),(8)\sigma_{\alpha\beta}^{\mathcal{O}}(\mu,T) \approx\frac{4e^2\hbar}{\pi\Omega}\frac{4}{\Delta E^2}\int_{-1}^{1} \mathrm{d}\tilde{\varepsilon}\,\frac{f(\tilde{\varepsilon},\mu,\tilde{T})}{(1-\tilde{\varepsilon}^2)}\zeta_{\alpha\beta}^{\mathcal{O}}(\tilde{\varepsilon}), \tag*{(8)}

with

ζαβO(ε~)≡∑m,nMgmgn(1+δm0)(1+δn0)Γm,n(ε~)μmnαβ,O,(9)\zeta_{\alpha\beta}^{\mathcal{O}}(\tilde{\varepsilon}) \equiv\sum_{m,n}^{M}\frac{g_mg_n}{(1+\delta_{m0})(1+\delta_{n0})}\Gamma_{m,n}(\tilde{\varepsilon})\mu_{mn}^{\alpha\beta,\mathcal{O}}, \tag*{(9)}

where

Γm,n(ε~)≡(ε~−in1−ε~2)einarccos⁡(ε~)Tm(ε~)+(ε~+im1−ε~2)e−imarccos⁡(ε~)Tn(ε~),(10)\Gamma_{m,n}(\tilde{\varepsilon}) \equiv(\tilde{\varepsilon}-\mathrm{i}n\sqrt{1-\tilde{\varepsilon}^{2}})e^{\mathrm{i}n\arccos(\tilde{\varepsilon})}T_m(\tilde{\varepsilon})+(\tilde{\varepsilon}+\mathrm{i}m\sqrt{1-\tilde{\varepsilon}^{2}})e^{-\mathrm{i}m\arccos(\tilde{\varepsilon})}T_n(\tilde{\varepsilon}), \tag*{(10)}
μmnαβ,O≡Tr⁡[Tm(H~)JαOTn(H~)vβ],(11)\mu_{mn}^{\alpha\beta,\mathcal{O}} \equiv\operatorname{Tr}\left[T_m(\widetilde{H})J_{\alpha}^{\mathcal{O}}T_n(\widetilde{H})v_{\beta}\right], \tag*{(11)}

and ΔE≡E+−E−\Delta E \equiv E^{+}-E^{-} is the spectral bandwidth, where E−E^{-} and E+E^{+} are the minimum and maximum eigenvalues.

This formulation allows us to express the conductivity tensor as a sum over a two-dimensional matrix of Chebyshev moments μmnαβ,O\mu_{mn}^{\alpha\beta,\mathcal{O}}. The computation of these moments relies exclusively on sequential sparse matrix-vector multiplications, which scale linearly with both the system size NN and the number of expansion moments MM.

We implement this real-space approach using Quantum KITE [38], a high-performance open-source software designed for large-scale tight-binding simulations. By using KITE’s optimized parallelization and memory-efficient algorithms, we calculate the continuous energy dependence of σαβO\sigma_{\alpha\beta}^{\mathcal{O}} across the entire spectrum. This setup enables us to efficiently account for millions of atomic sites, making it an ideal tool for investigating the impacts of realistic, random structural defects without resorting to artificial supercells.

Results and Discussion

We consider three complementary disorder mechanisms acting on the full sp3sp^3 lattice model: vacancies, orbital-dependent Anderson on-site disorder, and VspσV_{sp\sigma}-bond disorder. They are introduced in this order to separate different ways of perturbing the orbital texture. Vacancies remove complete lattice sites and all orbital channels attached to them; orbital-dependent Anderson disorder shifts the on-site energies of different orbitals independently, producing random local crystal-field environments; and VspσV_{sp\sigma}-bond disorder acts directly on the ss-pp hybridization channel. A fourth case, dilute controlled on-site impurities, is treated separately in Appendix Appendix § 1.2.4. Each disordered calculation is compared with the pristine limit. The same Quantum KITE implementation is used to compute both bulk transport coefficients and momentum-/real-space orbital-resolved spectral quantities, with the pristine spectral function and quadrupole

textures validated against exact diagonalization in Appendix Appendix A. Below we focus on the transport coefficients — σOHELz\sigma_{\mathrm{OHE}}^{L_z}, σxx\sigma_{xx}, and θOH\theta_{\mathrm{OH}} — for the three disorder mechanisms. The corresponding momentum-resolved spectral functions and quadrupole textures are collected in Appendix Appendix B, where they provide a microscopic counterpart to the transport trends while keeping the main discussion focused on transport. All calculations reported below use identical simulation parameters across every disorder type studied [40].

Vacancy disorder

We begin with vacancy disorder, which provides a simple model of structural disorder. A subset of lattice sites is randomly selected with concentration nvn_v, and these sites are removed from the lattice by setting all hopping amplitudes connected to them to zero, H→Hvac=PVHPVH \rightarrow H_{\mathrm{vac}}=P_VHP_V, where PVP_V projects onto the subspace of non-vacant sites.

Figure 3 summarizes the resulting transport behavior for vacancy concentrations up to nv=5%n_v = 5\%. Panels (a)–(d) show the disorder-averaged density of states, orbital Hall conductivity σOHELz\sigma_{\mathrm{OHE}}^{L_z}, longitudinal conductivity σxx\sigma_{xx}, and orbital Hall angle θOH=σOHELz/σxx\theta_{\mathrm{OH}} = \sigma_{\mathrm{OHE}}^{L_z}/\sigma_{xx} as a function of energy. Increasing the vacancy concentration suppresses both σxx\sigma_{xx} and σOHELz\sigma_{\mathrm{OHE}}^{L_z}, but the two responses are affected at different rates. Over the energy range shown, the longitudinal conductivity is strongly reduced, while the orbital Hall conductivity remains comparatively robust.

Vacancy disorder transport and conductivity plots

Figure 3. Vacancy disorder. (a) Density of states, (b) orbital Hall conductivity σOHELz\sigma_{\mathrm{OHE}}^{L_z}, (c) longitudinal conductivity σxx\sigma_{xx}, and (d) orbital Hall angle θOH\theta_{\mathrm{OH}} as a function of energy, for vacancy concentrations nv≲5.0%n_v \lesssim5.0\%. The black line corresponds to the pristine system. (e) σOHELz\sigma_{\mathrm{OHE}}^{L_z} and (f) θOH\theta_{\mathrm{OH}} at fixed EF=−0.85E_F=-0.85 eV plotted against ρxx\rho_{xx}, parametrized by nvn_v.

To make the comparison more quantitative, we focus on EF=−0.85E_F = -0.85 eV, an energy close to the orbital Hall peak of the pristine model and in the same energy window commonly used in previous studies of the sp3sp^3 OHE model [1]. Figure 3(e) and Figure 3(f) show σOHELz\sigma_{\mathrm{OHE}}^{L_z} and θOH=σOHELz/σxx\theta_{\mathrm{OH}} = \sigma_{\mathrm{OHE}}^{L_z}/\sigma_{xx} at this fixed Fermi energy as a function of the longitudinal resistivity, ρxx=1/σxx\rho_{xx} = 1/\sigma_{xx}, with the color scale indicating the vacancy concentration. Using ρxx\rho_{xx} rather than nvn_v on the horizontal axis allows us to compare different disorder mechanisms through their effect on charge transport, and also makes contact with experiments where disorder is often discussed in terms of resistivity. As shown in Figure 3(e) and Figure 3(f), σOHELz\sigma_{\mathrm{OHE}}^{L_z} decreases approximately linearly with ρxx\rho_{xx} over the range studied, while θOH\theta_{\mathrm{OH}} increases. In this range, the increase of ρxx\rho_{xx} dominates over the moderate decrease of σOHELz\sigma_{\mathrm{OHE}}^{L_z}, leading to a nearly linear increase of the orbital Hall angle. This trend is similar to the increase of orbital Hall angle with resistivity reported in Ru1−x_{1-x}Cux_x and Ru1−x_{1-x}Tix_x alloy bilayers [28]. The corresponding momentum-resolved spectral function and quadrupole-texture response are shown in Appendix B, Figure B1, where the main quadrupolar character remains visible as the spectral function broadens.

Vacancy disorder spectral and orbital-texture properties

Figure B1. Vacancy disorder. Spectral and orbital-texture properties for nv=1.0%n_v = 1.0\% (left column) and 5.0%5.0\% (right column), along the Γ\Gamma–X–M–Γ\Gamma–M′ path of Figure A1(a). (a), (b) Spectral function. (c), (d) ⟨Qxy⟩\langle Q_{xy}\rangle and (e), (f) ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle, both plotted on a [−10−3-10^{-3}, 10−310^{-3}] color scale.

Anderson Disorder

We next consider Anderson disorder implemented with orbital-dependent on-site shifts,

H→H+∑i,αwiαciα†ciα,H \rightarrow H + \sum_{i,\alpha} w_i^\alpha c_{i\alpha}^{\dagger}c_{i\alpha},

where wiαw_i^\alpha is drawn independently for each site ii and orbital α\alpha from a uniform distribution wiα∈[−W/2,W/2]w_i^\alpha\in[-W/2, W/2]. This differs from the usual scalar Anderson disorder, in which all orbitals on the same site would be shifted by the same random potential. Here, the disorder also changes the relative on-site energies of the ss, pxp_x, pyp_y, and pzp_z orbitals. As a result, it introduces random local crystal-field variations and modifies the local orbital-hybridization environment, in addition to ordinary site-diagonal disorder. Thus, although the hopping amplitudes themselves are unchanged, the disorder changes the local energy offsets that control ss-pp mixing and the quadrupolar texture of the clean model.

Figure 4 summarizes the transport behavior for Anderson disorder implemented with orbital-dependent on-site

Anderson disorder transport results

Figure 4. Anderson disorder. (a) Density of states, (b) orbital Hall conductivity σOHELz\sigma_{\mathrm{OHE}}^{L_z}, (c) longitudinal conductivity σxx\sigma_{xx}, and (d) orbital Hall angle θOH=σOHELz/σxx\theta_{\mathrm{OH}} = \sigma_{\mathrm{OHE}}^{L_z}/\sigma_{xx} as a function of energy, for disorder strengths W≲2.5W \lesssim2.5 eV. The black line corresponds to the pristine system. (e) σOHELz\sigma_{\mathrm{OHE}}^{L_z} and (f) θOH\theta_{\mathrm{OH}} at fixed EF=−0.85E_F = -0.85 eV plotted against the longitudinal resistivity ρxx=1/σxx\rho_{xx} = 1/\sigma_{xx}, parameterized by WW up to 3.53.5 eV.

shifts. Panels (a)–(d) show the disorder-averaged density of states, σOHELz\sigma_{\mathrm{OHE}}^{L_z}, σxx\sigma_{xx}, and orbital Hall angle θOH\theta_{\mathrm{OH}} as a function of energy, for disorder strengths up to W=2.5W = 2.5 eV. Panels (e) and (f) extend the analysis to W=3.5W = 3.5 eV and show σOHELz\sigma_{\mathrm{OHE}}^{L_z} and θOH\theta_{\mathrm{OH}} at fixed Fermi energy EF=−0.85E_F = -0.85 eV as a function of the longitudinal resistivity.

Increasing WW broadens the spectral features and suppresses both σxx\sigma_{xx} and σOHELz\sigma_{\mathrm{OHE}}^{L_z}, but the effect on the longitudinal conductivity is much stronger. Compared with vacancies, Anderson disorder gives access to a broader range of ρxx\rho_{xx}, allowing us to follow the evolution from the weak-to-moderate disorder regime to a more strongly disordered regime. At low and intermediate resistivities, σOHELz\sigma_{\mathrm{OHE}}^{L_z} decreases approximately linearly with ρxx\rho_{xx}, similar to the vacancy case, σOHELz≈σ0−γWρxx\sigma_{\mathrm{OHE}}^{L_z} \approx\sigma_0 - \gamma_W\rho_{xx}, with σ0=1.102 e/2π\sigma_0 = 1.102\ e/2\pi and γW=1.512 ℏ/2πe\gamma_W = 1.512\ \hbar/2\pi e. At larger resistivities, the behavior changes: σOHELz\sigma_{\mathrm{OHE}}^{L_z} crosses over to an approximately inverse dependence on ρxx\rho_{xx},

σOHELz≈Cρxx(12)\sigma_{\mathrm{OHE}}^{L_z} \approx\frac{C}{\rho_{xx}} \tag*{(12)}

with C=0.1775 e/2πC = 0.1775\ e/2\pi. Since θOH=σOHELzρxx\theta_{\mathrm{OH}} = \sigma_{\mathrm{OHE}}^{L_z}\rho_{xx}, this second regime corresponds to the saturation of the orbital Hall angle, as seen in Figure 4(f). The crossover between the two regimes occurs around ρxx≈0.24 ℏ/e2\rho_{xx} \approx0.24\ \hbar/e^2.

The resulting behavior connects with recent experiments on orbital Hall systems with tunable disorder. At weak to moderate disorder, the increase of θOH\theta_{\mathrm{OH}} with resistivity is similar to the trend reported in Ru1−xCux\mathrm{Ru}_{1-x}\mathrm{Cu}_x and Ru1−xTix\mathrm{Ru}_{1-x}\mathrm{Ti}_x alloy bilayers [28]. At stronger disorder, θOH\theta_{\mathrm{OH}} approaches an approximately constant value, which means that σOHELz\sigma_{\mathrm{OHE}}^{L_z} scales with σxx\sigma_{xx}. This is the same type of scaling reported in Mn thin films spanning amorphous to single-crystalline order [29]. The corresponding momentum-resolved spectral function and quadrupole-texture response are shown in Appendix B, Figure B2; as for vacancies, the dominant quadrupolar character remains identifiable as the spectral function broadens.

Spectral and orbital-texture properties for Anderson disorder

Figure B2. Anderson disorder. Spectral and orbital-texture properties for W=0.4W = 0.4 eV (left column) and W=1.0W = 1.0 eV (right column), along the Γ\Gamma-X-M-Γ\Gamma-M′' path of Figure A1(a). (a),(b) Spectral function. (c),(d) ⟨Qxy⟩\langle Q_{xy}\rangle and (e),(f) ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle, both on a [−10−3,10−3][-10^{-3}, 10^{-3}] color scale rather than the [−1,1][-1, 1] scale of Figure 2: the two principal bands retain O(1)O(1) texture (saturated red/blue), while the disorder-induced leakage into the nominally forbidden regions is a weak, three-order-of-magnitude-smaller correction.

Vsp Bond Disorder

Finally, we consider a disorder mechanism that acts directly on the ss-pp hybridization responsible for the orbital Hall response in the clean model. A random subset of sites, with fraction nbn_b, is selected, and at every selected site the four VspσV_{sp\sigma} bonds touching it – along +x^+\hat{x}, −x^-\hat{x}, +y^+\hat{y}, and −y^-\hat{y} – are set to zero. All other hopping amplitudes, VssσV_{ss\sigma}, VppσV_{pp\sigma}, and VppπV_{pp\pi}, are kept unchanged. This construction locally suppresses the ss-pp mixing, and therefore the hybridization channel that generates the quadrupolar texture and the orbital Berry curvature, without removing the remaining hopping network. Charge transport can still proceed through the ss-ss and pp-pp channels, while the orbital Hall response is perturbed more selectively through the ss-pp sector.

Equation (5) shows that, in the clean low-energy model, Ω+,yxLz∝ACB\Omega_{+,yx}^{L_z} \propto\mathcal{A}\mathcal{C}\mathcal{B}, with C∝Vspσ2\mathcal{C} \propto V_{sp\sigma}^2. Thus, continuously tuning Vspσ→0V_{sp\sigma} \to0 suppresses the orbital Hall response. This limiting behavior was previously demonstrated on a cubic lattice using the same sp3sp^3 framework [1]; the clean full-lattice dependence on VspσV_{sp\sigma} for the present square lattice is analyzed in Appendix C. Here, instead of globally rescaling the hopping parameter, we ask what happens when the ss-pp hybridization is removed locally and stochastically, while the rest of the hopping network remains intact.

Figure 5 shows the transport response for VspσV_{sp\sigma}-bond dilution with nb=0n_b = 0–25%, where nb=0n_b = 0 corresponds to the pristine system. Panels (a) and (b) show σxx\sigma_{xx} and σOHELz\sigma_{\mathrm{OHE}}^{L_z} as functions of energy. Since only the ss-pp hopping is diluted, while the ss-ss and pp-pp hoppings remain intact, the longitudinal conductivity is only moderately affected over this range. The orbital Hall conductivity, by contrast, is modified more directly because the ss-pp hybridization is the channel that generates the quadrupolar texture and the orbital Berry curvature in the clean model. The inset in Figure 5(b) highlights the crossings of the σOHELz(ε)\sigma_{\mathrm{OHE}}^{L_z}(\varepsilon) curves near ε≃−1 eV\varepsilon\simeq-1\ \mathrm{eV}, which motivates the

Transport response for Vspσ-bond disorder

Figure 5. VspσV_{sp\sigma} bond disorder. (a) Longitudinal conductivity σxx\sigma_{xx} and (b) orbital Hall conductivity σOHELz\sigma_{\mathrm{OHE}}^{L_z} as a function of energy, for dilution fraction nb=0n_b = 0–25%; inset in (b) zooms into the curve crossings near ε≃−1 eV\varepsilon\simeq-1\ \mathrm{eV}. (c) σOHELz\sigma_{\mathrm{OHE}}^{L_z} at fixed EF=−1.01 eVE_F = -1.01\ \mathrm{eV} against nbn_b, showing a non-monotonic peak near nb≈18%n_b \approx18\%. (d) Same data as σOHELz\sigma_{\mathrm{OHE}}^{L_z} against σxx\sigma_{xx}, parameterized by nbn_b.

choice of EF=−1.01 eVE_F = -1.01\ \mathrm{eV} for the fixed-energy analysis. At this Fermi energy, Figure 5(c) shows a non-monotonic dependence on the dilution fraction: σOHELz\sigma_{\mathrm{OHE}}^{L_z} first increases with nbn_b, reaches a maximum near nb≃18%n_b \simeq18\%, and then decreases as the ss-pp hybridization is further diluted. Panel (d) shows the same data plotted against σxx\sigma_{xx}, with nbn_b as the tuning parameter.

To understand this behavior, we can resort to the low-energy model and ask how the clean orbital Hall response depends on VspσV_{sp\sigma}. In this model, Eq. (5) shows that the Berry curvature is controlled by the product ACB\mathcal{A}\mathcal{C}\mathcal{B}, with C=4a2Vspσ2/(εp−εs)\mathcal{C} = 4a^2V_{sp\sigma}^2/(\varepsilon_p-\varepsilon_s), A=A+C/2\mathcal{A} = A + C/2, and B=B+C/2\mathcal{B} = B + C/2. The dependence on VspσV_{sp\sigma} is therefore not monotonic in general. Increasing VspσV_{sp\sigma} increases the magnitude of CC, which is the coefficient associated with the QxyQ_{xy}-like texture generated by virtual pp-ss-pp processes. At the same time, for the parameters used here, C<0C < 0, so the same hybridization reduces both A\mathcal{A} and B\mathcal{B}, including the coefficient associated with the Qx2−y2Q_{x^2-y^2}-like texture. The orbital Hall response is then maximized not by making the ss-pp hybridization as large as possible, but by balancing the two quadrupolar channels: the QxyQ_{xy} texture must be sizable, while the Qx2−y2Q_{x^2-y^2} texture and the orbital-current prefactor remain sufficiently large.

This interpretation is supported by the clean full-lattice calculation reported in Appendix C. There, σOHELz\sigma_{\mathrm{OHE}}^{L_z} is computed for several uniformly rescaled values of VspσV_{sp\sigma}, and the response is already non-monotonic in the pristine system near EF≃−1E_F \simeq-1 eV. The precise position of the maximum depends on energy: for example, the maximum occurs near Vspσ≃0.50V_{sp\sigma} \simeq0.50 eV at EF=−1.01E_F = -1.01 eV, while it shifts toward Vspσ≃0.55V_{sp\sigma} \simeq0.55 eV at the nearby energy EF=−1.035E_F = -1.035 eV. The low-energy estimate should therefore be viewed as a guide to the underlying hybridization balance, rather than as a quantitative prediction for the full lattice. If the stochastic bond dilution is interpreted only as a crude effective reduction of the hybridization, Vspσeff≃Vspσ0(1−nb)V_{sp\sigma}^{\mathrm{eff}} \simeq V_{sp\sigma}^{0}(1-n_b), the observed maximum near nb≃18%n_b \simeq18\% gives Vspσeff≃0.336V_{sp\sigma}^{\mathrm{eff}} \simeq0.336 eV, close to the small-kk stationary point Vspσ≃0.348V_{sp\sigma} \simeq0.348 eV. This agreement should not be read as an exact mapping between random bond dilution and a clean hopping renormalization. Bond dilution also modifies the local electronic environment around the removed bonds, redistributes spectral weight, broadens spectral features, and mixes momenta, which can shift the maximum relative to the clean estimate. Rather, the comparison indicates that the non-monotonic clean dependence on the hybridization is one ingredient of the bond-dilution peak, while the numerical calculation retains the disorder-specific effects absent from the estimate. The corresponding momentum-resolved spectral function and quadrupole-texture response are shown in Appendix B, Figure B3; bond dilution redistributes the QxyQ_{xy} texture while the Qx2−y2Q_{x^2-y^2} sign structure remains recognizable at the transport maximum.

Spectral and orbital-texture properties for Vspσ bond disorder

Figure B3. VspσV_{sp\sigma} bond disorder. Spectral and orbital-texture properties for nb=18%n_b = 18\% (left column, the anomalous peak dilution fraction of Figure 5(c)) and nb=2%n_b = 2\% (right column), along the Γ\Gamma-X-M-Γ\Gamma-M′' path of Figure A1(a). (a),(b) Spectral function. (c), (d) ⟨Qxy⟩\langle Q_{xy}\rangle and (e), (f) ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle, both plotted on a [−10−3,10−3][-10^{-3}, 10^{-3}] color scale.

Additional disorder and real-space results

The main text has focused on vacancies, orbital-dependent Anderson disorder, and VspσV_{sp\sigma}-bond dilution. Two additional sets of calculations, reported in the Appendices, complement this analysis.

Appendix D considers dilute controlled on-site impurities, where a fixed fraction of sites receives the same energy shift δU\delta U. This case is useful because it separates dilute impurity physics from the fully random, orbital-dependent Anderson disorder discussed in Sec. III B. The main result is that positive and negative impurity potentials of the same magnitude do not affect the transport coefficients in the same way. This asymmetry is consistent with resonant impurity scattering: the effect depends on where the impurity-induced spectral feature lies relative to the Fermi energy, not only on the magnitude of δU\delta U. Although δU\delta U shifts all orbitals on an impurity site by the same amount, it changes their energy detuning relative to orbitals on neighboring host sites and can therefore modify the local ss-pp hybridization in addition to producing resonant scattering.

Appendix E complements the momentum-space spectral analysis with real-space maps of the LDOS and local orbital multipoles. These maps clarify an important distinction between orbital quadrupoles and orbital angular momentum in the present spinless model. The quadrupolar operators QxyQ_{xy} and Qx2−y2Q_{x^2-y^2} are real orbital operators, so local disorder can generate finite real-space quadrupolar patterns by breaking the spatial cancellation present in the pristine lattice. By contrast, LzL_z is purely imaginary in the real px,pyp_x,p_y orbital basis, and all disorder potentials considered here preserve the reality of the Hamiltonian. The eigenstates can therefore still be chosen real, forcing the local expectation value ⟨Lz⟩\langle L_z\rangle to vanish site by site, not only after spatial averaging. Thus, disorder creates local orbital anisotropy, visible in the quadrupolar textures, but it does not create a static local orbital chirality in this model.

Conclusions

We have studied how different forms of disorder affect the orbital Hall response of a centrosymmetric spinless sp3sp^3 tight-binding model on a square lattice. The clean model provides the microscopic reference for the analysis: ss-pp hybridization generates quadrupolar orbital textures in the px,pyp_x,p_y sector, and the low-energy theory relates the projected orbital Berry curvature to the product ACB\mathcal{A}\mathcal{C}\mathcal{B}. This shows that the orbital Hall response is controlled by the interplay between the QxyQ_{xy} and Qx2−y2Q_{x^2-y^2} channels, rather than by a single orbital texture or by the magnitude of VspσV_{sp\sigma} alone.

For the more generic forms of disorder studied here, vacancies and orbital-dependent Anderson disorder, the results are most clearly organized in terms of the longitudinal resistivity ρxx\rho_{xx}. Both disorder mechanisms increase ρxx\rho_{xx} and suppress the longitudinal conductivity more strongly than the orbital Hall conductivity over the range studied. As a result, the orbital Hall angle increases with ρxx\rho_{xx}. In the Anderson case, where a broader resistivity range is accessible, this increase crosses over to a saturation regime, corresponding to an approximately linear scaling between σOHELz\sigma_{\mathrm{OHE}}^{L_z} and σxx\sigma_{xx}. Thus, although vacancies and Anderson disorder act differently at the microscopic level, they support the same qualitative conclusion: the orbital Hall response remains sizable even when charge transport is strongly degraded.

The momentum-resolved spectral functions provide a microscopic counterpart to this transport trend. Vacancy and Anderson disorder broaden the bands and produce weak texture leakage away from the clean-limit symmetry directions, but the dominant quadrupolar character of the main bands remains visible. The persistence of the dominant quadrupolar texture is consistent with the transport observation that σOHELz\sigma_{\mathrm{OHE}}^{L_z} is suppressed more slowly than σxx\sigma_{xx}, allowing θOH\theta_{\mathrm{OH}} to increase as the system becomes more resistive.

Bond dilution behaves differently because it does not simply act as a generic source of resistivity. It selectively modifies the ss-pp hybridization channel that generates the QxyQ_{xy}-like texture and contributes to the orbital Berry curvature. Instead of a monotonic suppression, the orbital Hall conductivity develops a maximum near nb≃18%n_b \simeq18\%. This behavior is consistent with the clean low-energy analysis: reducing the effective ss-pp hybridization can in- lar patterns by breaking the spatial cancellation present in the pristine lattice. By contrast, LzL_z is purely imaginary in the real px,pyp_x,p_y orbital basis, and all disorder potentials considered here preserve the reality of the Hamiltonian. The eigenstates can therefore still be chosen real, forcing the local expectation value ⟨Lz⟩\langle L_z\rangle to vanish site by site, not only after spatial averaging. Thus, disorder creates local orbital anisotropy, visible in the quadrupolar textures, but it does not create a static local orbital chirality in this model.

Conclusions

We have studied how different forms of disorder affect the orbital Hall response of a centrosymmetric spinless sp3sp^3 tight-binding model on a square lattice. The clean model provides the microscopic reference for the analysis: ss-pp hybridization generates quadrupolar orbital textures in the px,pyp_x,p_y sector, and the low-energy theory relates the projected orbital Berry curvature to the product ACB\mathcal{A}\mathcal{C}\mathcal{B}. This shows that the orbital Hall response is controlled by the interplay between the QxyQ_{xy} and Qx2−y2Q_{x^2-y^2} channels, rather than by a single orbital texture or by the magnitude of VspσV_{sp\sigma} alone.

For the more generic forms of disorder studied here, vacancies and orbital-dependent Anderson disorder, the results are most clearly organized in terms of the longitudinal resistivity ρxx\rho_{xx}. Both disorder mechanisms increase ρxx\rho_{xx} and suppress the longitudinal conductivity more strongly than the orbital Hall conductivity over the range studied. As a result, the orbital Hall angle increases with ρxx\rho_{xx}. In the Anderson case, where a broader resistivity range is accessible, this increase crosses over to a saturation regime, corresponding to an approximately linear scaling between σOHELz\sigma_{\mathrm{OHE}}^{L_z} and σxx\sigma_{xx}. Thus, although vacancies and Anderson disorder act differently at the microscopic level, they support the same qualitative conclusion: the orbital Hall response remains sizable even when charge transport is strongly degraded.

The momentum-resolved spectral functions provide a microscopic counterpart to this transport trend. Vacancy and Anderson disorder broaden the bands and produce weak texture leakage away from the clean-limit symmetry directions, but the dominant quadrupolar character of the main bands remains visible. The persistence of the dominant quadrupolar texture is consistent with the transport observation that σOHELz\sigma_{\mathrm{OHE}}^{L_z} is suppressed more slowly than σxx\sigma_{xx}, allowing θOH\theta_{\mathrm{OH}} to increase as the system becomes more resistive.

Bond dilution behaves differently because it does not simply act as a generic source of resistivity. It selectively modifies the ss-pp hybridization channel that generates the QxyQ_{xy}-like texture and contributes to the orbital Berry curvature. Instead of a monotonic suppression, the orbital Hall conductivity develops a maximum near nb≃18%n_b \simeq18\%. This behavior is consistent with the clean low-energy analysis: reducing the effective ss-pp hybridization can increase the response over an intermediate range because the Berry curvature depends on the balance between AA, CC, and BB. In the disordered calculation this tuning is achieved by stochastic bond removal rather than by a uniform hopping renormalization, so the peak combines the clean hybridization balance with disorder-induced broadening and momentum mixing.

The additional calculations reported in the Appendices extend this picture beyond the three main disorder mechanisms. Dilute controlled on-site impurities reveal a sign-dependent response: attractive and repulsive impurity potentials of the same magnitude affect the transport coefficients differently, consistent with resonant scattering and with the fact that the impurity potential changes the local intersite ss-pp detuning relative to neighboring host sites. Real-space texture maps show that disorder generates local quadrupolar patterns around the perturbed regions, while the local orbital angular momentum remains zero site by site. This distinction reflects the reality of the spinless Hamiltonian: the quadrupolar operators are real and can acquire local expectation values when spatial cancellation is broken, whereas LzL_z is purely imaginary in the real px,pyp_x,p_y basis.

Overall, our results show that disorder affects orbital Hall transport in two complementary ways. When disorder mainly increases ρxx\rho_{xx}, as in the vacancy and Anderson cases, the orbital Hall conductivity is comparatively robust and the orbital Hall angle grows with resistivity, eventually reaching a saturation regime in the strongly disordered Anderson case. When disorder selectively modifies the ss-pp hybridization, as in bond dilution, it can tune the quadrupolar texture itself and enhance the orbital Hall response over an intermediate range. The orbital Hall response of the sp3sp^3 model is therefore not controlled by disorder strength alone, but by the combined effect of resistivity, spectral broadening, and the microscopic orbital channel that is being perturbed.

Acknowledgments

TGR, acknowledges the FCT - Fundação para a Ciência e Tecnologia for financial support (Grant No. 2023.11755.PEX and DOI identifier https://doi.org/10.54499/2023.11755.PEX) TGR acknowledges FCT project reference numbers UIDB/04650/2020, and 2022.07471.CEECIND/CP1718/CT0001 (with DOI identifier: 10.54499/2022.07471.CEECIND/CP1718/CT0001) and CNPq (Grant No 305013/2024-6). TGR and KJUV acknowledge support from the EIC Pathfinder OPEN grant 101129641 “OBELIX”. TPC acknowledges CNPq (Grant No. 305647/2024-5). HPV acknowledges the University of York High-Performance Computing service, Viking. Further support from FCT through the PhD Grant. No. 2024.00560.BD (HPV) is acknowledged.

Orbital-resolved spectral functions using Quantum KITE

Beyond the linear-response conductivities discussed in Sec. II, Quantum KITE [38] also gives access to momentum- and orbital-resolved spectral quantities. Two target functions are used throughout this Appendix, arpes⁡(⋅)\operatorname{arpes}(\cdot) and spectral_map⁡()\operatorname{spectral\_map}(). The first evaluates the spectral function A(k,ε)A(\mathbf{k},\varepsilon) along a user-specified path in the Brillouin zone. The second evaluates the spectral function over the full two-dimensional Brillouin zone at a single fixed energy. Both target functions accept a weight argument that assigns a coefficient to each orbital in the unit cell, allowing the spectral trace to be projected onto an arbitrary coherent combination of orbitals rather than onto a single orbital. For example, using the basis {∣s⟩,∣px⟩,∣py⟩,∣pz⟩}\{\lvert s\rangle,\lvert p_x\rangle,\lvert p_y\rangle,\lvert p_z\rangle\} from our Hamiltonian, Eq. (1), the weight parameter lets target functions isolate the eigenstates of each orbital multipole individually: weight = (0,1,±1,0)(0,1,\pm1,0) projects onto the Qxy=±1Q_{xy}=\pm1 eigenstates (∣px⟩±∣py⟩)/2\left(\lvert p_x\rangle\pm\lvert p_y\rangle\right)/\sqrt{2}, weight = (0,1,1,0)(0,1,1,0) or (0,0,1,0)(0,0,1,0) projects onto the Qx2−y2=±1Q_{x^2-y^2}=\pm1 eigenstates, which are simply the bare ∣px⟩,∣py⟩\lvert p_x\rangle,\lvert p_y\rangle orbitals; and weight = (0,1,±i,0)(0,1,\pm i,0) projects onto the Lz=±ℏL_z=\pm\hbar orbital-dipole eigenstates (∣px⟩±i∣py⟩)/2\left(\lvert p_x\rangle\pm i\lvert p_y\rangle\right)/\sqrt{2}. This is what allows KITE to resolve the orbital-quadrupole and orbital-dipole textures separately in momentum space, in addition to the ordinary spectral function.

Figure A1(b) shows the resulting spectral function A(k,ε)A(\mathbf{k},\varepsilon); panels (c) and (d) show ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle resolved along the same path, and panels (e) and (f) show both textures projected onto the full Brillouin zone at fixed energy ε=0.15\varepsilon=0.15 eV (dashed line in (b)-(d)). Note the path revisits Γ\Gamma before continuing to M′M', the same Brillouin-zone corner as MM reached via the opposite diagonal (kx=−kyk_x=-k_y instead kx=kyk_x=k_y): since ⟨Qxy⟩∝kxky\langle Q_{xy}\rangle\propto k_xk_y changes sign between the two diagonals, this choice of path makes the fourfold quadrupole texture of Figure A1(c) directly visible, rather than hidden by the point-group equivalence of MM and M′M'. To validate this approach, we compare KITE’s output against the exact diagonalization of the full lattice Hamiltonian of Sec. II A — i.e., without the quadractic Löwdin truncation used to obtain Eq. (3) — along the path shown in Figure A1(a).

Spectral properties and texture maps calculated via Quantum KITE from the full lattice Hamiltonian

Figure A1. Spectral properties and texture maps calculated via Quantum KITE from the full lattice Hamiltonian. (a) Square-lattice Brillouin zone showing the Γ\Gamma–X–M–Γ\Gamma–M′ path. (b) Spectral function A(k,ε)A(\mathbf{k},\varepsilon) along the path. (c), (d) Expectation values ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle resolved along the path. (e), (f) Corresponding textures projected onto the full Brillouin zone at fixed energies indicated by the green lines in panels (c) and (d), respectively.

Momentum-Resolved Orbital-Texture Response to Disorder

This appendix collects the momentum-resolved spectral functions and quadrupole-texture response for the three disorder mechanisms discussed in Sec. III, complementing the transport results shown there.

Vacancy disorder

Figure B1 shows the corresponding spectral function and quadrupolar textures along the Γ\Gamma–X–M–Γ\Gamma–M′ path, for nv=1.0%n_v = 1.0\% and 5.0%5.0\%. The spectral function broadens visibly as the vacancy concentration increases, reflecting the loss of translational invariance and the stronger momentum mixing. The quadrupolar textures, however, remain largely concentrated on the same main band branches as in the clean system. This is most apparent in the saturated red and blue features of ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle, which persist even at nv=5.0%n_v = 5.0\%.

Vacancy disorder also produces weak texture leakage into regions where the corresponding clean-limit quadrupole component vanishes: ⟨Qxy⟩\langle Q_{xy}\rangle appears along Γ\Gamma–X, and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle appears along M–Γ\Gamma–M′. This leakage increases with vacancy concentration, but remains small on the [−10−3-10^{-3}, 10−310^{-3}] scale used in the figure. Thus, the spectral functions show clear disorder broadening, while the underlying quadrupolar character of the main

bands is comparatively robust. This provides a microscopic counterpart to the transport result that σOHELz\sigma_{\mathrm{OHE}}^{L_z} is less affected by vacancies than σxx\sigma_{xx}.

Anderson disorder

Figure B2 shows the disorder-averaged spectral function together with the ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle textures along the same Γ\Gamma–X–M–Γ\Gamma–M′ path used in Appendix A. The two columns correspond to W=0.4W = 0.4 eV and W=1.0W = 1.0 eV. As WW increases, the spectral function broadens and spectral weight spreads into regions that are forbidden in the pristine limit, reflecting the loss of exact translational symmetry.

The quadrupolar textures show the same qualitative effect, but on a much smaller scale. Weak disorder-induced weight appears in the clean-limit forbidden regions: ⟨Qxy⟩\langle Q_{xy}\rangle along Γ\Gamma–X and ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle along M–Γ\Gamma–M′. This leakage is larger for W=1.0W = 1.0 eV than the leakage, as indicated by the [−10−3,10−3][-10^{-3}, 10^{-3}] color scale used in Figure B2(c)–(Figure B2f). Thus, Anderson disorder visibly broadens the spectral function, while the dominant quadrupolar character of the main bands remains identifiable over the disorder range shown.

directly tied to the ss-pp hybridization through C∝Vspσ2C \propto V_{sp\sigma}^{2}. Thus, bond dilution acts on the same microscopic ingredient that controls the clean QxyQ_{xy} texture, but it does so locally and stochastically, producing spectral broadening and momentum mixing rather than a simple rigid rescaling of the clean bands.

VspσV_{sp\sigma} Bond Disorder

Figure B3 shows the corresponding spectral function and quadrupolar textures for nb=18%n_b = 18\%, close to the transport maximum of Figure 5(c), and for a weakly diluted case, nb=2%n_b = 2\%, along the Γ\Gamma-X-M-Γ\Gamma-M′' path. The main effect of increasing the bond dilution is not to destroy the band structure, but to broaden and redistribute the orbital texture within the low-energy px,pyp_x,p_y manifold. This is particularly clear in the ⟨Qxy⟩\langle Q_{xy}\rangle maps: at nb=18%n_b = 18\%, the texture is more spread over the low-energy branches than at nb=2%n_b = 2\%, reflecting the fact that this channel is directly tied to the ss-pp hybridization through C∝Vspσ2C \propto V_{sp\sigma}^{2}.

The ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle texture is affected in a different way. Its coefficient B=B+C/2B = B + C/2 contains a finite VspσV_{sp\sigma}-independent contribution BB, so reducing the ss-pp hybridization does not eliminate this channel. Consistently, the ⟨Qx2−y2⟩\langle Q_{x^2-y^2}\rangle maps retain a recognizable sign structure on the low-energy branches even at nb=18%n_b = 18\%, although the texture is broadened by disorder. This contrast between the directly diluted QxyQ_{xy} channel and the partially retained Qx2−y2Q_{x^2-y^2} channel supports the interpretation of the bond-dilution peak as a combination of clean hybridization physics and disorder-induced redistribution of spectral weight.

Non-monotonic VspσV_{sp\sigma} Dependence in the Pristine Limit

The non-monotonic σOHELz(nb)\sigma_{\mathrm{OHE}}^{L_z}(n_b) found in Sec. III C suggests that bond dilution does more than simply suppress the ss-pp hybridization. To isolate the part of this behavior that is already present in the clean model, we first ask how the low-energy Berry curvature depends on a uniform change of VspσV_{sp\sigma}. Eq. (5) shows that the projected Berry curvature contains the product ACBACB, with C=4a2Vspσ2/(ϵp−ϵs)C = 4a^2V_{sp\sigma}^2/(\epsilon_p-\epsilon_s), A=A+C/2A = A + C/2, and B=B+C/2B = B + C/2. For the parameters used here, C<0C < 0, so increasing VspσV_{sp\sigma} increases the magnitude of CC while reducing both AA and BB. The response is therefore not expected to be monotonic in VspσV_{sp\sigma}.

This can be seen directly from the small-kk prefactor F(Vspσ)=ACBF(V_{sp\sigma}) = ACB. With A=0.15A = 0.15 eV, B=0.35B = 0.35 eV, and ϵp−ϵs=−3.7\epsilon_p-\epsilon_s = -3.7 eV, FF vanishes at Vspσ=0V_{sp\sigma} = 0, at Vspσ=0.527V_{sp\sigma} = 0.527 eV where A=0A = 0, and at Vspσ=0.805V_{sp\sigma} = 0.805 eV where B=0B = 0. The corresponding extrema occur at Vspσ≃0.348V_{sp\sigma} \simeq0.348 eV and Vspσ≃0.704V_{sp\sigma} \simeq0.704 eV. Thus, even in the pristine effective Hamiltonian, the orbital Hall response is controlled by a balance between the QxyQ_{xy}-like channel associated with CC, the Qx2−y2Q_{x^2-y^2}-like channel associated with BB, and the orbital-current prefactor AA.

Clean full-lattice dependence on Vspσ

Figure C1. Clean full-lattice dependence on VspσV_{sp\sigma}. (a) σOHELz(EF)\sigma_{\mathrm{OHE}}^{L_z}(E_F) for several values of VspσV_{sp\sigma} close to the pristine value Vspσ=0.50V_{sp\sigma}=0.50 eV (black). Both the height and the energy position of the main orbital Hall peak near EF≈−1E_F \approx-1 eV depend on VspσV_{sp\sigma}. (b) σOHELz(Vspσ)\sigma_{\mathrm{OHE}}^{L_z}(V_{sp\sigma}) at two fixed Fermi energies, EF=−1.01E_F=-1.01 eV and EF=−1.035E_F=-1.035 eV, illustrating that the position of the maximum is energy dependent.

Figure C1 shows that the non-monotonic dependence on VspσV_{sp\sigma} survives beyond the small-kk expansion when the full lattice Hamiltonian, Eq. (1), is used without disorder. At EF=−1.01E_F=-1.01 eV, σOHELz\sigma_{\mathrm{OHE}}^{L_z} increases and then decreases as VspσV_{sp\sigma} is varied from 0.400.40 to 0.600.60 eV, with a maximum near the pristine value Vspσ=0.50V_{sp\sigma}=0.50 eV. At the nearby energy EF=−1.035E_F=-1.035 eV, the maximum is shifted toward Vspσ≃0.55V_{sp\sigma}\simeq0.55 eV. The full-lattice maximum is therefore not expected to coincide quantitatively with the small-kk estimate: the conductivity receives contributions from a finite energy window and from regions of the Brillouin zone beyond the quadratic expansion around Γ\Gamma. The important point is qualitative: the clean system already has a non-monotonic dependence on VspσV_{sp\sigma}.

This observation is useful for interpreting the bond-dilution result, but it should not be turned into a one-to-one mapping. If one makes a crude virtual-crystal estimate, a VspσV_{sp\sigma} bond survives only when neither endpoint belongs to the diluted set, giving Vspσeff/Vspσ≃(1−nb)2V_{sp\sigma}^{\mathrm{eff}}/V_{sp\sigma}\simeq(1-n_b)^2. For nb≃18%n_b\simeq18\%, this gives Vspσeff≃0.336V_{sp\sigma}^{\mathrm{eff}}\simeq0.336 eV, close to the lower small-kk extremum at Vspσ≃0.348V_{sp\sigma}\simeq0.348 eV. This agreement is suggestive, but only at the level of the low-energy hybridization balance. Stochastic bond dilution is not equivalent to a uniform rescaling of VspσV_{sp\sigma}: it changes the local electronic environment around the removed bonds, redistributes spectral weight, broadens the bands, and mixes momenta. These disorder-specific effects can shift the maximum relative to both the small-kk estimate and the clean full-lattice calculation. The appendix therefore supports a more cautious interpretation: the bond-dilution peak appears to combine the non-monotonic clean dependence on ss-pp hybridization with additional spectral effects produced by random bond removal.

On-site Disorder

As a final disorder mechanism, we consider a dilute concentration of identical, controlled on-site impurities, complementary to the continuous, all-site Anderson disorder of Sec. III B in that here every orbital at a small, fixed fraction non_o of randomly selected sites receives the same energy shift δU\delta U, εiα→εiα+δU\varepsilon_i^\alpha\to\varepsilon_i^\alpha+\delta U, i∈Si\in\mathcal{S}, where S\mathcal{S} is a random subset of sites of concentration non_o, and δU\delta U is a controlled parameter rather than a random variable.

Figures D1(a)–(d) show the resulting spectral and transport properties. Each nonzero δU\delta U introduces sharp, resonance-like features in the density of states [panel (a)] whose energy position shifts systematically with δU\delta U, consistent with the appearance of a localized impurity level pulled away from the host bands—a characteristic signature of a resonant impurity [41, 42], here realized at finite but dilute concentration. The pristine system (δU=0\delta U=0, black) retains the sharpest and tallest features in σxx\sigma_{xx} [panel (c)], as expected in the absence of any impurity scattering; every nonzero δU\delta U broadens and suppresses σxx\sigma_{xx}, with σOHELz\sigma_{\mathrm{OHE}}^{L_z} and σOH\sigma_{\mathrm{OH}} [panels (b),(d)] responding correspondingly.

Panels (e) and (f) reveal a clear asymmetry between positive and negative δU\delta U at fixed EF=−0.85E_F=-0.85 eV. At comparable ∣δU∣|\delta U|, positive shifts (δU>0\delta U>0) suppress σxx\sigma_{xx} more strongly than negative shifts of the same magnitude, and correspondingly suppress σOHELz\sigma_{\mathrm{OHE}}^{L_z} more strongly as well. Thus, the two branches do not simply collapse onto a single curve parameterized by ∣δU∣|\delta U|. This sign-dependence is the expected hallmark of resonant impurity scattering, whose efficiency depends on where the impurity level falls relative to EFE_F, rather than on the impurity strength alone; a repulsive (δU>0\delta U>0) and an attractive (δU<0\delta U<0) impurity of equal magnitude are therefore not equivalent at this Fermi energy.

Density of states, orbital Hall conductivity, longitudinal conductivity, and orbital Hall angle as functions of energy, with conductivity plots versus resistivity

Figure D1. On-site disorder at fixed concentration no=2%n_{\mathrm{o}}=2\%. (a) Density of states, (b) orbital Hall conductivity σOHEL\sigma_{\mathrm{OHE}}^{L}, (c) longitudinal conductivity σxx\sigma_{xx}, and (d) orbital Hall angle θOH\theta_{\mathrm{OH}} as a function of energy, for impurity strengths δU=−3.0\delta U=-3.0 to 3.03.0 eV. The black line (δU=0.0\delta U=0.0) corresponds to the pristine system. (e) σOHEL\sigma_{\mathrm{OHE}}^{L} and (f) θOH\theta_{\mathrm{OH}} at fixed EF=−0.85E_{\mathrm{F}}=-0.85 eV plotted against ρxx\rho_{xx}, parameterized by δU\delta U.

Figure D2 shows the corresponding spectral function and orbital textures for δU=−2\delta U=-2 eV (left column) and δU=+2\delta U=+2 eV (right column), along the same Γ\Gamma-X-M-Γ\Gamma-M′' path used throughout this work. Beyond the disorder-induced leakage already seen for Anderson, vacancy, and bond disorder – ⟨Qxy⟩\langle Q_{xy}\rangle into Γ\Gamma-X, ⟨Qx2−y2⟩\langle Q_{x^{2}-y^{2}}\rangle into M-Γ\Gamma-M′' – panels (a),(b) reveal an additional, qualitatively distinct feature: narrow, nearly kk-independent bands appearing at fixed energies away from the host dispersion. Unlike the kk-mixing leakage common to every disorder type studied in this work, a genuinely localized or resonant impurity state disperses weakly with kk, appearing instead as a horizontal streak in the spectral function – a characteristic signature of the dilute resonant-impurity problem [41, 42]. The energy at which this feature appears shifts with the sign of δU\delta U, consistent with the impurity level’s position relative to the host bands.

Spectral function and orbital-texture properties for on-site disorder

Figure D2. On-site disorder at fixed concentration no=2%n_{\mathrm{o}}=2\%. Spectral and orbital-texture properties for δU=−2\delta U=-2 eV (left column) and δU=+2\delta U=+2 eV (right column), along the Γ\Gamma-X-M-Γ\Gamma-M′' path of Figure A1(a). (a),(b) Spectral function. (c),(d) ⟨Qxy⟩\langle Q_{xy}\rangle and (e),(f) ⟨Qx2−y2⟩\langle Q_{x^{2}-y^{2}}\rangle.

Real-Space Orbital Textures

While Appendix A resolves the quadrupole textures in momentum space, KITE’s ldos_map() target function additionally gives access to their real-space counterpart: the local density of states (LDOS) and the local expectation values of QxyQ_{xy}, Qx2−y2Q_{x^{2}-y^{2}}, and LzL_z, resolved site by site, at fixed energy ε≈−0.5\varepsilon\approx-0.5 eV. This complements the momentum-space picture by directly visualizing how each disorder mechanism perturbs the orbital texture in the region where it is actually applied.

Each calculation is performed on a large system of 384×384 nm2384\times384\ \mathrm{nm}^{2}, but Figure E1 displays only a 10×10 nm210\times10\ \mathrm{nm}^{2} window centered on the region where disorder was introduced. To allow a direct, realization-by-realization comparison across disorder types at fixed concentration, the same random seed is used for vacancy and on-site disorder at 2% concentration (and at 2% dilution fraction for bond disorder), so that all three share the identical random site selection. The LDOS shown is restricted to the px,pyp_x,p_y manifold (excluding ss and pzp_z).

LDOS and real-space orbital textures for four disorder types

Figure E1. LDOS and real-space orbital textures computed via KITE over a 10×10 nm210 \times10\ \mathrm{nm}^{2} window at fixed energy ε≈−0.5 eV\varepsilon\approx-0.5\ \mathrm{eV}, for four disorder types at matched concentration and random seed. Columns: LDOS (px,pyp_x,p_y manifold only), ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^{2}-y^{2}}\rangle.

In the pristine system (not shown), the LDOS is spatially homogeneous and the local quadrupole and dipole expectation values vanish identically at every site, exactly as established in Sec. II B: the kx→−kxk_x \to-k_x, ky→−kyk_y \to-k_y reflection symmetry forces ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^{2}-y^{2}}\rangle to cancel over the full Brillouin zone at any single energy, while the reality of the Hamiltonian forces ⟨Lz⟩=ℏ⟨σy⟩=0\langle L_z\rangle= \hbar\langle\sigma_y\rangle= 0 identically for every eigenstate. The pristine row is therefore omitted from Figure E1, and every pattern visible in the remaining rows is a genuine disorder-induced effect.

The dipole expectation value, ⟨Lz⟩\langle L_z\rangle, remains zero at every site for all four disorder types. This is expected: Anderson, vacancy, bond, and on-site disorder are all real perturbations of a real Hamiltonian, so every disordered eigenstate can still be chosen real, and ⟨Lz⟩=ℏ⟨σy⟩\langle L_z\rangle= \hbar\langle\sigma_y\rangle continues to vanish identically regardless of the specific disorder realization. By contrast, ⟨Qxy⟩\langle Q_{xy}\rangle and ⟨Qx2−y2⟩\langle Q_{x^{2}-y^{2}}\rangle develop a markedly inhomogeneous real-space distribution once disorder is introduced, with strongly populated regions of both positive (red) and negative (blue) sign concentrated at the disordered sites themselves – the real-space counterpart of the momentum-space leakage discussed in Secs. III B-III C and Appendix § 1.2.4.

The LDOS panels reveal a distinct spatial signature for each disorder mechanism. Anderson disorder produces a genuinely inhomogeneous LDOS across the entire window, consistent with every site being independently perturbed. Vacancies instead produce sharply localized enhanced regions (high LDOS) immediately surrounding each vacant site, while the vacant site itself shows exactly zero LDOS, appearing as an isolated white spot at its center. Bond disorder produces a distinctive checkerboard-like pattern in the affected region, alternating between neighboring sites – a spatial signature not seen for any of the other three disorder types. We verified that a single bond-diluted site reproduces the same alternation along both the axis directions away from the defect, confirming it is an intrinsic response to the VspσV_{sp\sigma} dilution itself rather than an interference effect between multiple nearby defects. For on-site disorder, the randomly selected impurity sites themselves show suppressed LDOS at this energy, while their immediate neighbors show enhanced LDOS, consistent with the impurity resonance discussed in Appendix § 1.2.4 being centered at a different energy and locally redistributing spectral weight around it at ε≈−0.5 eV\varepsilon\approx-0.5\ \mathrm{eV}.

References

References

  1. [1]D. Go, D. Jo, C. Kim, and H.-W. Lee, Intrinsic spin and orbital hall effects from orbital texture, Phys. Rev. Lett. 121, 086602 (2018).DOI
  2. [2]D. Go, D. Jo, H.-W. Lee, M. Kläui, and Y. Mokrousov, Orbitronics: Orbital currents in solids, Europhys. Lett. 135, 37001 (2021).DOI
  3. [3]S. Fukami, K.-J. Lee, and M. Kläui, Challenges and opportunities in orbitronics, Nat. Phys. 10.1038/s41567-025-03143-w (2025).arxiv.org/abs/2509.23954
  4. [4]T. P. Cysne, L. M. Canonico, M. Costa, R. B. Muniz, and T. G. Rappoport, Orbitronics in two-dimensional materials, npj Spintronics 3, 39 (2025).DOI
  5. [5]B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Orbitronics: The intrinsic orbital current in p-doped silicon, Phys. Rev. Lett. 95, 066601 (2005).pubmed.ncbi.nlm.nih.gov/16090968
  6. [6]H. Kontani, T. Tanaka, D. S. Hirashima, K. Yamada, and J. Inoue, Giant orbital hall effect in transition metals: Origin of large spin and anomalous hall effects, Phys. Rev. Lett. 102, 016601 (2009).DOI
  7. [7]Y.-G. Choi, D. Jo, K.-H. Ko, D. Go, K.-H. Kim, H. G. Park, C. Kim, B.-C. Min, G.-M. Choi, and H.-W. Lee, Observation of the orbital hall effect in a light metal ti, Nature 619, 52 (2023).DOI
  8. [8]I. Lyalin, S. Alikhani, M. Berritta, P. M. Oppeneer, and R. K. Kawakami, Magneto-optical detection of the orbital hall effect in chromium, Phys. Rev. Lett. 131, 156702 (2023).DOI
  9. [9]G. Sala, H. Wang, W. Legrand, and P. Gambardella, Orbital hall magnetoresistance in a 3d transition metal, Phys. Rev. Lett. 131, 156703 (2023).DOI
  10. [10]X. Chen, Y. Liu, G. Yang, H. Shi, C. Hu, M. Li, and H. Zeng, Giant antidamping orbital torque originating from the orbital rashba-edelstein effect in ferromagnetic heterostructures, Nature Communications 9, 2569 (2018).DOI
  11. [11]L. Salemi, M. Berritta, A. K. Nandy, and P. M. Oppeneer, Orbitally dominated rashba-edelstein effect in noncentrosymmetric antiferromagnets, Nature Communications 10, 10.1038/s41467-019-13367-z (2019).DOI
  12. [12]S. Ding, Z. Liang, D. Go, C. Yun, M. Xue, Z. Liu, S. Becker, W. Yang, H. Du, C. Wang, Y. Yang, G. Jakob, M. Kläui, Y. Mokrousov, and J. Yang, Observation of the orbital rashba-edelstein magnetoresistance, Phys. Rev. Lett. 128, 067201 (2022).DOI
  13. [13]A. El Hamdi et al., Observation of the orbital inverse rashba-edelstein effect, Nat. Phys. 19, 1855 (2023).DOI
  14. [14]W. Gao, L. Liao, H. Isshiki, N. Budai, J. Kim, H.-W. Lee, K.-J. Lee, D. Go, Y. Mokrousov, S. Miwa, and Y. Otani, Nonlocal electrical detection of reciprocal orbital Edelstein effect, Nature Communications 16, 6380 (2025).arxiv.org/abs/2502.11040
  15. [15]D. Go and H.-W. Lee, Orbital torque: Torque generation by orbital current injection, Phys. Rev. Res. 2, 013177 (2020).DOI
  16. [16]S. Ding, A. Ross, D. Go, L. Baldrati, Z. Ren, F. Freimuth, S. Becker, F. Kammerbauer, J. Yang, G. Jakob, Y. Mokrousov, and M. Kläui, Harnessing orbital-to-spin conversion of interfacial orbital currents for efficient spin-orbit torques, Phys. Rev. Lett. 125, 177201 (2020).DOI
  17. [17]D. Lee, D. Go, H.-J. Park, W. E. Yang, H.-W. Ko, D. Yun, D. Jo, S. Lee, G. Go, J. H. Oh, K.-J. Kim, B.-G. Park, B.-C. Min, H. C. Koo, H.-W. Lee, O. Lee, and K.-J. Lee, Orbital torque in magnetic bilayers, Nature Communications 12, 10.1038/s41467-021-26650-9 (2021).DOI
  18. [18]D. Jo, D. Go, and H.-W. Lee, Gigantic intrinsic orbital hall effects in weakly spin-orbit coupled metals, Phys. Rev. B 98, 214405 (2018).DOI
  19. [19]T. G. Rappoport, First light on orbitronics as a viable alternative to electronics, Nature 619, 38 (2023).DOI
  20. [20]P. Tang and G. E. W. Bauer, Role of disorder in the intrinsic orbital hall effect, Phys. Rev. Lett. 133, 186302 (2024).DOI
  21. [21]S. Mankovsky and H. Ebert, Spin and orbital Hall effect in nonmagnetic transition metals: Extrinsic versus intrinsic contributions, Phys. Rev. B 110, 184417 (2024).arxiv.org/abs/2408.07814
  22. [22]Y. Liu, Y. Xu, A. Fert, H.-Y. Jaffres, S. Eimer, T. Nie, X. Zhang, and W. Zhao, Efficient terahertz generation from copt-based terahertz emitters via orbital-to-charge conversion, arXiv preprint arXiv:2402.09228 (2024).arxiv.org/abs/2402.09228
  23. [23]H. Liu and D. Culcer, Dominance of extrinsic scattering mechanisms in the orbital hall effect: Graphene, transition metal dichalcogenides, and topological antiferromagnets, Phys. Rev. Lett. 132, 186302 (2024).DOI
  24. [24]A. Veneri, T. G. Rappoport, and A. Ferreira, Extrinsic orbital hall effect: Orbital skew scattering and crossover between diffusive and intrinsic orbital transport, Phys. Rev. Lett. 134, 136201 (2025).DOI
  25. [25]A. L. R. Barbosa, H.-W. Lee, and T. G. Rappoport, Extrinsic orbital Hall effect and orbital relaxation in mesoscopic devices, arXiv preprint 10.48550/arXiv.2507.01941 (2025), arXiv:2507.01941 [cond-mat.mes-hall].DOI
  26. [26]D. Joy and V. Pandey, Skew scattering induced contribution to orbital Hall response, arXiv preprint (2026), author list incomplete — verify against arXiv listing, arXiv:2607.24671 [cond-mat.mes-hall].arxiv.org/abs/2607.24671
  27. [27]S. Peng, X. Zheng, S. Li, et al., Unconventional scaling of the orbital Hall effect, Nature Materials 24, 1749 (2025), author list truncated — complete from the article page.DOI
  28. [28]Y. Yang, Y. Chen, W. Xu, and D. Wei, Effects of impurity scattering on orbital hall conductivity and orbital transport in ru-based alloys, arXiv:2607.03076 (2026), no DOI/ISSN – preprint; update once/if published.arxiv.org/abs/2607.03076
  29. [29]M.-G. Kang, F. Nasr, G. Sala, R. Schlitz, S. Ding, S. Bartsch, M. D. Rossell, S. F. Alvarado, and P. Gambardella, Orbital hall conductivity and relaxation in thin films with variable disorder, Nat. Phys. 10.1038/s41567-026-03334-z (2026).DOI
  30. [30]S. Han, H.-W. Lee, and K.-W. Kim, Orbital dynamics in centrosymmetric systems, Phys. Rev. Lett. 128, 176601 (2022).DOI
  31. [31]A. Manchon, C. Sun, X. Ning, T. Sato, T. Kato, and T. G. Rappoport, Multipolar orbital relaxation of the t2g states, Physical Review Letters 136, 268201 (2026).
  32. [32]G.-H. Lee, Y. Ji, Y. Park, C. An, S. Ko, H.-W. Ko, J. Lim, J. H. Oh, F. Mahfouzi, B.-G. Park, K.-J. Kim, M. D. Stiles, K.-W. Kim, P. M. Haney, and K.-J. Lee, Observation of current-induced orbital quadrupole accumulation (2026), arXiv:2608.22838.arxiv.org/abs/2608.22838
  33. [33]J. C. Slater and G. F. Koster, Simplified LCAO method for the periodic potential problem, Phys. Rev. 94, 1498 (1954).DOI
  34. [34]S. Han, H.-W. Lee, and K.-W. Kim, Microscopic study of orbital textures, Curr. Appl. Phys. 50, 13 (2023).DOI
  35. [35]A. Bastin, C. Lewiner, O. Betbeder-Matibet, and P. Nozières, Quantum oscillations of the Hall effect of a fermion gas with random impurity scattering, J. Phys. Chem. Solids 32, 1811 (1971).
  36. [36]A. Weiße, G. Wellein, A. Alvermann, and H. Fehske, The kernel polynomial method, Rev. Mod. Phys. 78, 275 (2006).DOI
  37. [37]J. H. García, L. Covaci, and T. G. Rappoport, Real-space calculation of the conductivity tensor for disordered topological matter, Phys. Rev. Lett. 114, 116602 (2015).pubmed.ncbi.nlm.nih.gov/25839298
  38. [38]S. M. João, M. Andelkovic, L. Covaci, T. G. Rappoport, J. a. M. V. P. Lopes, and A. Ferreira, KITE: high-performance accurate modelling of electronic structure and response functions of large molecules, disordered crystals and heterostructures, R. Soc. Open Sci. 7, 191809 (2020).
  39. [39]K. J. U. Vidarte, H. P. Veiga, J. M. V. P. Lopes, R. Cardias, A. Ferreira, T. P. Cysne, and T. G. Rappoport, Real-space spectral approach to orbital magnetization, Phys. Rev. B 113, 224438 (2026).DOI
  40. [40]System size N × N sites with N = 448, M = 1024 Chebyshev moments, R = 200 random vectors for the stochastic trace evaluation, averaged over D = 200 independent disorder realizations. Identical across all disorder types studied (Anderson, vacancy, bond, and the on-site variant of Appendix D).
  41. [41]G. F. Koster and J. C. Slater, Wave functions for impu rity levels, Phys. Rev. 95, 1167 (1954).
  42. [42]G. F. Koster and J. C. Slater. Simplified impurity calculation, Phys. Rev. 96, 1208 (1954).DOI

Paper details

vlnc:W_d8x9ztgu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents