betaFeedback

WorksPhysics and Astronomy

Response theory at low temperature for weakly interacting quantum spin systems

Abstract

Recent work by Greenblatt, Lange, Marcelli, and Porta (Commun. Math. Phys. 2024) provides the first rigorous justification of adiabatic theory for interacting quantum systems at positive temperature. Their analysis treats weakly interacting lattice fermions and combines a rigorous Wick rotation with fermionic cluster expansion techniques. Here, we study a class of weakly interacting quantum spin systems and derive response theory at positive temperature, thereby covering non-Fermi systems for the first time. Our proof identifies and exploits the locality property of local indistinguishability as a key ingredient in carrying out low-temperature adiabatic expansions. The fact that these systems satisfy local indistinguishability was recently established jointly with Adhikari (arXiv:2508.15907) using cluster expansion methods together with a probabilistic swapping trick.

Introduction and main result

An important topic in the mathematical analysis of non-equilibrium physics is the response of a quantum system to adiabatic switching, i.e., to slowly turning on a weak perturbation. Of particular interest is the justification of Kubo’s formula [27] and its higher-order refinements. The rigorous response theory of this kind is closely connected to adiabatic theory, which was pioneered by Kato [25] and generalized by Nenciu [35]. An influential interpretation through the geometry of vector bundles was discovered by Berry [7, 37]. For subsequent developments of the by now classical adiabatic theory, such as space-adiabatic perturbation theory, we refer to the book [39] by Teufel.

Beyond the classical framework, two fundamental obstacles arise when studying adiabatic theory for quantum many-body systems: First, the relevant perturbations tend to close the spectral gap that was crucial to standard adiabatic theory. Second, the error terms diverge with growing system size, making them thermodynamically unstable for extended systems. In the past years, it has been realized that the key to circumvent both of these obstacles is to exploit locality properties of gapped ground states. Indeed, Hastings’ spectral flow [15, 4] has allowed to replace the standard Kato construction of the inverse Liouvillian operator by a quasi-local variant. Locality provides both dynamical robustness even when the gap closes and error estimates that are uniform in volume, hence thermodynamically stable. These ideas have been used to derive response theory for local observables in gapped ground states and to prove quantization of the Hall conductance at zero temperature in various settings [14, 11, 2, 3, 32, 40, 16, 17, 19, 20, 42, 41, 6]. We also mention that linear response theory can also be studied outside of an adiabatic framework, typically in a weak coupling regime; see e.g. [38, 22].

Recent work [12] by Greenblatt, Lange, Marcelli, and Porta then initiated rigorous response theory for positive temperature many-body systems. Now, the initial state is an equilibrium (Gibbs) state at temperature T>0T > 0. This is the physically more realistic situation, but it requires fundamentally new mathematical techniques to treat mixed states instead of gapped pure states. We also mention the recent companion works [24, 23], which studied structural aspects of adiabatic theory at positive temperature, as well as a recent physics paper concerned with adiabatic theory at positive temperature [21].

In [12], the authors consider a system of weakly interacting lattice fermions in DD dimensions subjected to a weak and adiabatically slow perturbation. They derive response theory through a convergent expansion in terms of the small perturbation size. Their strategy consists of two main steps: First, they use Lieb-Robinson bounds to justify a rigorous version of the Wick rotation which reformulates kk-point dynamical Duhamel correlation functions as kk-point Euclidean correlation functions. Second, they calculate these Euclidean correlation functions by fermionic cluster expansion techniques, which heavily exploit diagrammatic cancellations that arise from the antisymmetry. The errors they obtain are small provided that the temperature TT satisfies T≪ηD+2T \ll\eta^{D+2}, where η∈(0,1]\eta\in(0, 1] is a small parameter denoting the inverse of the long time scale on which the adiabatically slow switching occurs.

Main result

In the present paper, we give the first derivation of response theory for non-Fermi quantum many-body systems: On ZD\mathbb{Z}^{D}, equipped with the usual graph ℓ1\ell^{1}-metric d:ZD×ZD→N0d:\mathbb{Z}^{D}\times\mathbb{Z}^{D}\to\mathbb{N}_{0}, denote an arbitrary finite connected subset by Λ⋐ZD\Lambda\Subset\mathbb{Z}^{D}. With every site x∈ZDx\in\mathbb{Z}^{D}, we associate a finite-dimensional Hilbert space Hx:=Cq\mathcal{H}_{x}:=\mathbb{C}^{q} with q≥2q\ge2 and the corresponding space of linear operators Ax:=L(Hx)\mathcal{A}_{x}:=\mathcal{L}(\mathcal{H}_{x}). For every X⊂ΛX\subset\Lambda we introduce the many-particle Hilbert space on XX as HX:=⨂x∈XHx\mathcal{H}_{X}:=\bigotimes_{x\in X}\mathcal{H}_{x} and denote the associated C∗\mathrm{C}^{*}-algebra of observables by AX:=L(HX)\mathcal{A}_{X}:=\mathcal{L}(\mathcal{H}_{X}). Note that, whenever X⊂X′X\subset X', we have that AX\mathcal{A}_{X} is naturally embedded as a subalgebra of AX′\mathcal{A}_{X'}.

We consider self-adjoint Hamiltonians as the sum of an interacting and non-interacting part,

H0=∑x∈Λhx+∑x∈Λ:BR(x)⊂Λjx.(1)H_{0}=\sum_{x\in\Lambda}h_{x}+\sum_{\substack{x\in\Lambda:\\ B_{R}(x)\subset\Lambda}}j_{x}. \tag*{(1)}

Here, in the non-interacting part, every hx≥0h_{x}\ge0 acts only on the site x∈Λx\in\Lambda, hx∈A{x}h_{x}\in\mathcal{A}_{\{x\}}, and has a unique zero-energy ground state separated from the rest of the spectrum by a gap of size at least one. In the interacting part, every jxj_{x} is a small finite-range interaction supported on BR(x)B_{R}(x), a ball of radius RR around site xx, i.e. jx∈ABR(x)j_{x}\in\mathcal{A}_{B_{R}(x)}, and we denote ∥h∥∞=sup⁡x∈ZD∥hx∥∞\lVert h\rVert_{\infty}=\sup_{x\in\mathbb{Z}^{D}}\lVert h_{x}\rVert_{\infty} and ∥j∥∞=sup⁡x∈ZD∥jx∥∞\lVert j\rVert_{\infty}=\sup_{x\in\mathbb{Z}^{D}}\lVert j_{x}\rVert_{\infty}. Additionally, we assume that every jxj_{x} is relatively form bounded with respect to the sum of all hxh_{x} in BR(x)B_{R}(x) in the following sense: There exists a small a∈(0,1)a\in(0,1) such that

∣⟨ψ,jxψ⟩∣≤aκR∣⟨ψ,∑z∈BR(x)hzψ⟩∣for all states ψ and κR=sup⁡x∈Λ∣BR(x)∣.(2)\left|\langle\psi,j_{x}\psi\rangle\right|\le\frac{a}{\kappa_{R}}\left|\left\langle\psi,\sum_{z\in B_{R}(x)}h_{z}\psi\right\rangle\right| \quad\text{for all states }\psi\text{ and }\kappa_{R}=\sup_{x\in\Lambda}\left|B_{R}(x)\right|. \tag*{(2)}

with Br(x):={y∈Λ∣d(x,y)≤r}B_{r}(x):=\{y\in\Lambda\mid d(x,y)\le r\}. Examples of Hamiltonians that we cover are Heisenberg XXZXXZ Hamiltonians on any DD-dimensional lattice, subjected to a strong external field in the ZZ-direction (which then plays the role of hxh_{x}). In particular, (2) implies that the product ground state of ∑xhx\sum_{x}h_{x} remains a ground state of H0H_{0}.

Starting from the equilibrium state

ρβ=e−βH0tr⁡(e−βH0)\rho_{\beta}=\frac{e^{-\beta H_{0}}}{\operatorname{tr}(e^{-\beta H_{0}})}

as the initial state, we study the response of H0H_{0} from (1) to a small static perturbation εV\varepsilon V which is slowly (adiabatically) turned on. That is, given a small parameter ε>0\varepsilon>0 and a smooth switching function f:R→[0,1]f:\mathbb{R}\to[0,1] with f(t)=0f(t)=0 for t≤−1t\le-1 and f(t)=1f(t)=1 for t≥0t\ge0, we consider the time-dependent Hamiltonian

Hε(t)=H0+εf(t)V,V=∑x∈Λ:BR(x)⊂Λvx,(3)H_{\varepsilon}(t)=H_{0}+\varepsilon f(t)V,\qquad V=\sum_{\substack{x\in\Lambda:\\ B_{R}(x)\subset\Lambda}}v_{x}, \tag*{(3)}

where vx∈ABR(x)v_{x}\in\mathcal{A}_{B_{R}(x)} is self-adjoint.

To study adiabatic time scales, we fix another small parameter η>0\eta> 0. We write ρβε,η,f(t)\rho_{\beta}^{\varepsilon,\eta,f}(t) for the solution of the time-dependent adiabatic Schrödinger equation

iηddtρβε,η,f(t)=[Hε(t),ρβε,η,f(t)](4)\mathrm{i}\eta\frac{\mathrm{d}}{\mathrm{d}t}\rho_{\beta}^{\varepsilon,\eta,f}(t) = [H_{\varepsilon}(t),\rho_{\beta}^{\varepsilon,\eta,f}(t)] \tag*{(4)}

and initial datum ρβε,η,f(t)=ρβ\rho_{\beta}^{\varepsilon,\eta,f}(t)=\rho_{\beta} for all t≤−1t\leq-1, where ρβ\rho_{\beta} is the Gibbs state of H0H_0 at inverse temperature β∈(0,∞)\beta\in(0,\infty). For an observable A∈AXA\in\mathcal{A}_X, the response to the perturbation is now given by

Σβ,Aε,η,f(t):=tr⁡(Aρβε,η,f(t))−tr⁡(Aρβ).\Sigma_{\beta,A}^{\varepsilon,\eta,f}(t):=\operatorname{tr}(A\rho_{\beta}^{\varepsilon,\eta,f}(t))-\operatorname{tr}(A\rho_{\beta}).

Our main result, Theorem 1.1 below, establishes an expansion of Σβ,Aε,η,f(t)\Sigma_{\beta,A}^{\varepsilon,\eta,f}(t) to any order in ε\varepsilon with an additional temperature dependent error term, i.e. response theory to any order for low temperature Gibbs states, uniformly in the size of the system.

Theorem 1.1 (Response theory at low temperature). Let DD, qq, R∈NR\in\mathbb{N}, and CintC_{\mathrm{int}}, Cswitch>0C_{\mathrm{switch}}>0. Then there exist constants vLR>0v_{\mathrm{LR}}>0 and a∈(0,1)a\in(0,1) such that the following holds: For all nn, m∈Nm\in\mathbb{N} and α∈(0,1)\alpha\in(0,1) there exist constants CC, c>0c>0 such that for all Λ⋐ZD\Lambda\Subset\mathbb{Z}^D and Hamiltonians H0H_0, VV as defined in (1) with ∥h∥∞,∥j∥∞,∥v∥∞≤Cint\lVert h\rVert_\infty,\lVert j\rVert_\infty,\lVert v\rVert_\infty\leq C_{\mathrm{int}}, jxj_x and vxv_x are of range R∈NR\in\mathbb{N}, and jxj_x is relatively aa-bounded w.r.t. the hxh_x’s as in (2), the following holds:

Take any inverse temperature β∈(0,∞)\beta\in(0,\infty), let X⊂ΛX\subset\Lambda be connected and take A∈AXA\in\mathcal{A}_X. Then there exist response coefficients σβ,j\sigma_{\beta,j} in the following sense: For ε>0\varepsilon>0 and smooth switching function f:R→[0,1]f:\mathbb{R}\to[0,1] satisfying f(t)=0f(t)=0 for t≤−1t\leq-1, f(t)=1f(t)=1 for t≥0t\geq0, and ∥f∥CCn,m(R)<Cswitch\lVert f\rVert_{C^{\mathcal{C}_{n,m}}(\mathbb{R})}<C_{\mathrm{switch}} with Cn,m:=⌈m(n+2D+1)⌉\mathcal{C}_{n,m}:=\lceil m(n+2D+1)\rceil, consider the time-dependent Hamiltonian (3). Let ρβε,η,f(t)\rho_{\beta}^{\varepsilon,\eta,f}(t) be the solution of the time-dependent adiabatic Schrödinger equation (4) with adiabatic parameter η∈(0,1]\eta\in(0,1] and initial datum ρβε,η,f(t)=ρβ\rho_{\beta}^{\varepsilon,\eta,f}(t)=\rho_\beta for all t≤−1t\leq-1, where ρβ\rho_\beta is the Gibbs state of H0H_0 at inverse temperature β∈(0,∞)\beta\in(0,\infty).

Then, the response to the perturbation satisfies

∣Σβ,Aε,η,f(t)−∑j=1nεjσβ,jA∣≤C∥A∥e∣X∣2/αεn+1((1+t)2D+1+evLR(1+t)/η−(n+1)log⁡(ε)e−cβ(1−α)/D),(5)\left|\Sigma_{\beta,A}^{\varepsilon,\eta,f}(t)-\sum_{j=1}^{n}\varepsilon^j\sigma_{\beta,j}^{A}\right| \leq C\lVert A\rVert e^{|X|^{2/\alpha}}\varepsilon^{n+1}\left((1+t)^{2D+1}+e^{v_{\mathrm{LR}}(1+t)/\eta-(n+1)\log(\varepsilon)}e^{-c\beta^{(1-\alpha)/D}}\right), \tag*{(5)}

for any α∈(0,1)\alpha\in(0,1), ε∈(0,1)\varepsilon\in(0,1), all t≥0t\geq0, and uniformly for η∈[εm,ε1/m]\eta\in[\varepsilon^m,\varepsilon^{1/m}]. The first order coefficient is given by Kubo’s formula, see (15) below.

For β=∞\beta=\infty, our main result recovers the usual well known response theory for the unique ground state of H0H_0, see e.g. [40 Theorem 4.1]. Moreover, it shows that, whenever the inverse temperature β=1/T\beta=1/T satisfies

β≥(vLR(1+t)cη−n+1clog⁡(ε))D1−α(6)\beta\geq\left(\frac{v_{\mathrm{LR}}(1+t)}{c\eta}-\frac{n+1}{c}\log(\varepsilon)\right)^{\frac{D}{1-\alpha}} \tag*{(6)}

we effectively have response theory up to order nn. This should be compared to the single previous rigorous work on positive temperature response theory for extended systems [12], which relied on the condition

β≫1ηD+2\beta\gg\frac{1}{\eta^{D+2}}

for linear response theory to be valid. Taking α≪1\alpha\ll1, our bound (6) is slightly stronger, i.e. we can allow larger temperatures. Moreover, it is obtained by completely different methods and holds for quantum spin systems instead of lattice Fermions which were considered in [12]. Furthermore, our bound drastically improves if the Hamiltonian (3) has slow transport properties, as we now explain.

Remark 1.2 (Effect of slower transport). The modular nature of the proof immediately implies an improved final result if a stronger Lieb-Robinson bound [29] holds. More precisely, whenever the dynamics τt,s\tau_{t,s} generated by the Hamiltonian (3) satisfies a Lieb-Robinson bound of the form

∥[τt,s(A),B]∥≤C∥A∥∥B∥min⁡{∣X∣,∣Y∣}fLR(∣t−s∣)e−d(X,Y)\left\lVert[\tau_{t,s}(A),B]\right\rVert\leq C\lVert A\rVert\lVert B\rVert\min\{|X|,|Y|\}f_{\mathrm{LR}}(|t-s|)e^{-d(X,Y)}

for observables A∈AXA \in\mathcal{A}_X and B∈AYB \in\mathcal{A}_Y, we can make the replacement

evLR(1+t)/η→fLR((1+t)/η)e^{v_{\mathrm{LR}}(1+t)/\eta} \to f_{\mathrm{LR}}((1+t)/\eta)

in the second line of (5).

As an example, if H0H_0 is locally commuting then the LR velocity is bounded by ε\sqrt{\varepsilon} [36] and so

∥[τt,s(A),B]∥≤C∥A∥∥B∥min⁡{∣X∣,∣Y∣}eCε1/2∣t−s∣e−d(X,Y).\left\lVert[\tau_{t,s}(A),B]\right\rVert\leq C\lVert A\rVert\lVert B\rVert\min\{|X|,|Y|\}e^{C\varepsilon^{1/2}|t-s|}e^{-d(X,Y)}.

In this case, the lower bound from (6) drastically improves to

β≥(C(1+t)cε1/2η−n+1clog⁡(ε))D1−α.\beta\geq\left(\frac{C(1+t)}{c}\frac{\varepsilon^{1/2}}{\eta}-\frac{n+1}{c}\log(\varepsilon)\right)^{\frac{D}{1-\alpha}}.

In particular, for ε=η2\varepsilon=\eta^2 and neglecting time, we only need β≫∣log⁡(η)∣D1−α\beta\gg|\log(\eta)|^{\frac{D}{1-\alpha}}. Another class of systems where substantially improved Lieb-Robinson bounds can be established are strongly disordered quantum spin systems [5, 9, 31, 10] and long-range commuting interactions [28].

About the proof

Our proof technique is completely different from that of [12]; we do not use the Wick rotation and fermionic cluster expansion techniques are unavailable here. Instead, we identify and exploit two key drivers to connect low-temperature adiabatic theory to ground state expansions in a way that is uniform in the system size:

  • local indistinguishability (LI), uniformly in the temperature

  • controlled low-energy density of states (CLEDOS).

Local indistinguishability is a locality concept for Gibbs states originating in quantum information theory [26], where it was dubbed locality of temperature. It is closely related to exponential decay of correlations [8] and establishing it at low temperature is a substantial task in its own right, because it essentially amounts to excluding phase transitions at low temperature. Controlled low-energy density of states was identified as an important condition by Hastings in [13] and further discussed in [30]. Roughly speaking, CLEDOS says that the density of states of low-lying energy intervals is at most polynomial in system size. CLEDOS has to our knowledge only been verified for non-interacting systems, but is widely expected to hold much more generally for gapped systems, though, there are also gapped Hamiltonians known in which an exponential-in-system-size ground state degeneracy occurs for open boundary conditions [13, 30].

Our proof is then structured in an abstract, modular fashion as displayed in Figure 1. Roughly speaking, local indistinguishability and the controlled low-energy spectrum enable us to approximate the Gibbs state by a local Gibbs state. Then, the combination of a unique gapped ground state and CLEDOS allows us to approximate the local Gibbs state by the local ground state, the idea being that the gap quantifies the energetic cost and CLEDOS the degeneracy of excited states. For the local ground state, we can then apply zero-temperature adiabatic theory uniformly in the system size, as done e.g. in [32, 40, 16, 19]. A side aspect of the modular approach is that it is easy to track how improvements of different ingredients affect the main result. For example, stronger Lieb-Robinson bounds (e.g. for strongly disordered systems, which are covered by our analysis) directly improve the scaling of η\eta with temperature, cf. Remark 1.2.

Structure of the argument, showing local indistinguishability, Gibbs state convergence, response theory for Gibbs states with $\beta \to \infty$, controlled low-energy density of states, and unique gapped ground state

Figure 1. Structure of the argument. Taking (LI) for the Hamiltonian (1) as input, in Theorem 4.1 we verify Assumptions (CONV) and (UGGS) based on Proposition 5.2.

Importantly, the class of Hamiltonians H0H_0 introduced in (1) provides examples where the modular assumptions, LI and CLEDOS, can be verified. Indeed, the proof that LI holds uniformly in the temperature was given in our recent joint work [1] with Adhikari. The fact that CLEDOS holds for (1) is an easy consequence of the variational principle, but it does not seem to have been observed before that CLEDOS is stable under such perturbations and thus it provides a wider class of Hamiltonians where CLEDOS is known compared to [13, 30].

Mathematical setup

In this section, we briefly introduce the (standard) mathematical framework for locally interacting quantum spin systems (see e.g. [40, 17, 16, 3, 20] for similar setups), which generalizes the one used in the introduction.

Spatial structure and algebra of observables

We consider a quantum spin system on a finite connected graph Λ\Lambda equipped with the usual graph distance d(⋅,⋅)d(\cdot,\cdot). Denote the ball of radius rr centred at x∈Λx \in\Lambda by Br(x):={y∈Λ∣d(x,y)≤r}B_r(x) := \{y \in\Lambda\mid d(x,y) \le r\}. The graph Λ\Lambda is assumed to have dimension (at most) D>0D > 0, in the sense that there exists a constant Cvol>0C_{\mathrm{vol}} > 0 such that

sup⁡x∈Λ∣Br(x)∣≤1+CvolrD,(7)\sup_{x \in\Lambda} |B_r(x)| \le1 + C_{\mathrm{vol}} r^D, \tag*{(7)}

uniformly in rr, where ∣X∣|X| denotes the number of sites in X⊂ΛX \subset\Lambda. The set of all such graphs Λ\Lambda is denoted by

G(D,Cvol):={Λ finite graph | sup⁡x∈Λ∣Br(x)∣≤1+CvolrD for all r>0}.\mathcal{G}(D,C_{\mathrm{vol}}) := \left\{ \Lambda\ \text{finite graph} \,\middle|\, \sup_{x \in\Lambda} |B_r(x)| \le1 + C_{\mathrm{vol}} r^D\ \text{for all } r > 0 \right\}.

To each vertex x∈Λx \in\Lambda, we associate a single-particle Hilbert space Hx\mathcal{H}_x, which we assume to be of uniformly bounded dimension,

q:=sup⁡x∈Λdim⁡Hx<∞.(8)q := \sup_{x \in\Lambda} \dim\mathcal{H}_x < \infty. \tag*{(8)}

We then denote the many-particle Hilbert space HΛ=⨂x∈ΛHx\mathcal{H}_{\Lambda} = \bigotimes_{x \in\Lambda} \mathcal{H}_x and the C*-algebra of observables AΛ:=L(HΛ)\mathcal{A}_{\Lambda} := \mathcal{L}(\mathcal{H}_{\Lambda}). For every X⊂ΛX \subset\Lambda there is a subalgebra AX⊂AΛ\mathcal{A}_X \subset\mathcal{A}_{\Lambda} isomorphic to L(HX)\mathcal{L}(\mathcal{H}_X).

Note that we will obtain bounds uniform in the graph Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}), but unlike the more common setups, we do not consider a fixed infinite volume Γ\Gamma and subsets Λ⋐Γ\Lambda\Subset\Gamma. Instead, our setup naturally covers torus and cylinder geometries directly.

Since a very similar construction is common for fermionic lattice systems (see e.g. [33, 19]), all the results almost immediately translate to this setting.

Interactions and SLT operators

An interaction is a map

Φ:{X⊂Λ}→AΛ,X↦Φ(X)=Φ(X)∗∈AX.\Phi: \{X \subset\Lambda\} \to\mathcal{A}_{\Lambda}, \qquad X \mapsto\Phi(X) = \Phi(X)^* \in\mathcal{A}_X.

Given an interaction one associates a (family of) sum of local terms (SLT) operators or local Hamiltonians via

HΛ′:=∑X⊂Λ′Φ(X)∈AΛfor every Λ′⊂Λ.(9)H^{\Lambda'} := \sum_{X \subset\Lambda'} \Phi(X) \in\mathcal{A}_{\Lambda} \qquad\text{for every } \Lambda' \subset\Lambda. \tag*{(9)}

We will often use the shorthand notation H=HΛH = H^\Lambda and call HH and all HΛ′H^{\Lambda'} of the form (9) SLT operators. Note that, while every interaction defines a unique operator, there are multiple interactions realizing the same operator, i.e. the assignment Φ↦H(Φ)\Phi\mapsto H(\Phi) is not invertible. Note that, by definition, all interactions and SLT operators are always self-adjoint.

For every b>0b > 0 and p∈(0,1]p \in(0,1], we consider the stretched exponential function

χb,p:[0,∞)→(0,1],x↦e−bxp(10)\chi_{b,p} : [0,\infty) \to(0,1], \quad x \mapsto e^{-b x^p} \tag*{(10)}

as localization functions and define the associated SLT interaction norm

∥Φ∥b,p:=sup⁡z∈Λ∑Z⊂Λz∈Z∥Φ(Z)∥χb,p(diam⁡(Z))(11)\lVert\Phi\rVert_{b,p} := \sup_{z\in\Lambda}\sum_{\substack{Z\subset\Lambda\\ z\in Z}} \frac{\lVert\Phi(Z)\rVert}{\chi_{b,p}(\operatorname{diam}(Z))} \tag*{(11)}

measuring the quality of localization for an interaction Φ\Phi. Since we consider only finite graphs, ∥Φ∥b,p<∞\lVert\Phi\rVert_{b,p} < \infty. Crucially, our results do not depend on the interaction Φ\Phi, but only on ∥Φ∥b,p\lVert\Phi\rVert_{b,p}. Operators with this property will often be referred to as (b,p)(b,p)-localized SLT operators.

Gibbs states

For any Hilbert space H\mathcal{H} and self-adjoint operator H∈L(H)H \in\mathcal{L}(\mathcal{H}), we denote the Gibbs state at inverse temperature β∈(0,∞)\beta\in(0,\infty) by

ρβH[H]:=e−βHtr⁡e−βH,(12)\rho^\mathcal{H}_\beta[H] := \frac{e^{-\beta H}}{\operatorname{tr}e^{-\beta H}}, \tag*{(12)}

where tr⁡(⋅)\operatorname{tr}(\cdot) is the full (non-normalized) trace on H\mathcal{H}. For a system on Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}) and Λ′⊂Λ\Lambda' \subset\Lambda, we abbreviate

ρβΛ′[K]:=ρβHΛ′[K]for K∈AΛ′andρβΛ′:=ρβΛ′[HΛ′],\rho^{\Lambda'}_\beta[K] := \rho^{\mathcal{H}_{\Lambda'}}_\beta[K] \quad\text{for } K \in\mathcal{A}_{\Lambda'} \quad\text{and}\quad\rho^{\Lambda'}_\beta:= \rho^{\Lambda'}_\beta[H^{\Lambda'}],

whenever it is clear, which operator HΛ′H^{\Lambda'} is meant. Moreover, if Λ′=Λ\Lambda' = \Lambda, we will often drop the Λ\Lambda superscript entirely, i.e. write ρβ:=ρβΛ\rho_\beta:= \rho^\Lambda_\beta. For β=∞\beta= \infty, we denote with ρ∞H(H)\rho^\mathcal{H}_\infty(H) the normalized ground state projection of HH and likewise use the above abbreviations.

Response theory at low temperature

Our main result, Theorem 1.1, will be derived as a special case from a much more general result, Theorem 3.1 below, which extends zero temperature response theory to low temperature Gibbs states. It relies on two main assumptions: A uniform gapped ground state (Assumption (UGGS)) and the convergence of Gibbs states as β→∞\beta\to\infty (Assumption (CONV)). In Theorem 4.1, these assumptions will then be shown to be satisfied for the systems considered in Theorem 1.1.

We now formulate our two main assumptions.

(UGGS) Unique (uniform) gapped ground state. We say that the SLT Hamiltonian HH has a unique uniformly gapped ground state with gap size g>0g > 0 and minimal system size R0R_0 (both independent of Λ\Lambda) if and only if the following holds: For all connected Λ′⊂Λ\Lambda' \subset\Lambda, such that there exists x∈Λ′x \in\Lambda' with BR0(x)⊂Λ′B_{R_0}(x) \subset\Lambda', we have that E0Λ′:=inf⁡spec⁡(HΛ′)E_0^{\Lambda'} := \inf\operatorname{spec}(H^{\Lambda'}) is a non-degenerate eigenvalue of HΛ′H^{\Lambda'} and dist⁡(E0Λ′,spec⁡(HΛ′)∖{E0Λ′})≥g\operatorname{dist}(E_0^{\Lambda'},\operatorname{spec}(H^{\Lambda'}) \setminus\{E_0^{\Lambda'}\}) \ge g. If this holds only for Λ′=Λ\Lambda' = \Lambda, we say that HH has a unique gapped ground state with gap size g>0g > 0. ◇

Assumption (UGGS) is standard in works on response theory for the ground state, see e.g. [3, 32, 40, 18, 20].

(CONV) Convergence of Gibbs states as β→∞\beta\to\infty. We say that an SLT Hamiltonian HH with a unique ground state ρ∞\rho_\infty has a converging low-temperature Gibbs state with respect to two monotonically increasing/decreasing functions fconv,ζconv:[0,∞)→[0,∞)f_{\mathrm{conv}}, \zeta_{\mathrm{conv}} : [0,\infty) \to[0,\infty), respectively, such that lim⁡β→∞ζconv(β)=0\lim_{\beta\to\infty} \zeta_{\mathrm{conv}}(\beta) = 0 and a threshold β0∈[0,∞)\beta_0 \in[0,\infty), if and only if

∣tr⁡(ρβA)−tr⁡(ρ∞A)∣≤∥A∥fconv(∣X∣)ζconv(β)(13)\left|\operatorname{tr}(\rho_\beta A) - \operatorname{tr}(\rho_\infty A)\right| \le\lVert A\rVert f_{\mathrm{conv}}(|X|)\zeta_{\mathrm{conv}}(\beta) \tag*{(13)}

for all connected X⊂ΛX \subset\Lambda, A∈AXA \in\mathcal{A}_X and β>β0\beta> \beta_0. ◇

In a nutshell, Assumption (CONV) says that, when tested on local observables, the Gibbs state converges to the ground state uniformly in the system size. Moreover, by taking the maximum over A∈AXA \in\mathcal{A}_X with ∥A∥=1\lVert A\rVert= 1 on the left-hand side of (13), it expresses a convergence in a local trace norm. We stress that the uniformity in system size is crucial: In fact, when allowing a ∣Λ∣|\Lambda|-dependent prefactor fconvf_{\mathrm{conv}} in (13) and assuming that the ground state of HH is separated by a gap of size gg from the rest of the spectrum, such a bound is certainly always true with an exponential rate ζconv(β)=e−gβ\zeta_{\mathrm{conv}}(\beta) = e^{-g\beta}; see Lemma A.1.

Moreover, we can easily obtain bounds on the best possible forms of the two functions fconvf_{\mathrm{conv}} and ζconv\zeta_{\mathrm{conv}} by considering the following non-interacting example. The simplest non-trivial Hamiltonian having a uniformly gapped ground state with gap size g>0g > 0, namely modelling non-interacting spin-1/2 particles in an external magnetic field, is

H=g2∑x∈Λσx3H = \frac{g}{2}\sum_{x\in\Lambda}\sigma_x^3

where σx3\sigma_x^3 is the third Pauli matrix acting only on x∈Λx \in\Lambda. For this concrete Hamiltonian, it is straightforward to see that (cf. also (22) and the proof of Lemma A.1)

max⁡A∈AX∥A∥=1∣tr⁡(ρβA)−tr⁡(ρ∞A)∣=2(1−1∑j=0∣X∣(∣X∣j)e−βgj)=2(1−(1+e−βg)−∣X∣)\begin{aligned} \max_{\substack{A\in\mathcal{A}_X\\ \lVert A\rVert=1}}\left|\operatorname{tr}(\rho_\beta A)-\operatorname{tr}(\rho_\infty A)\right| =2\left(1-\frac{1}{\sum_{j=0}^{|X|}\binom{|X|}{j}e^{-\beta g j}}\right) \\ &=2\left(1-\left(1+e^{-\beta g}\right)^{-|X|}\right) \end{aligned}
∈[cmin⁡{1,∣X∣e−βg},C∣X∣e−βg]\in\left[c\min\{1,|X|e^{-\beta g}\},C|X|e^{-\beta g}\right]

for some constants C,c>0C,c>0. Hence, in (13), fconv(∣X∣)f_{\mathrm{conv}}(|X|) must necessarily grow at least linearly, and ζconv(β)\zeta_{\mathrm{conv}}(\beta) can decay at most exponentially fast with rate given by the gap size gg.

Assuming (UGGS) and (CONV), we have response theory for the low temperature Gibbs state to any order, as the following theorem states.

Theorem 3.1 (Response theory to all orders). Fix nn, mm, DD, q∈Nq \in\mathbb{N}, 0<p′<p≤10 < p' < p \le1, 0<b′<10 < b' < 1 and b>b′b > b'. Fix also CvolC_{\mathrm{vol}}, CintC_{\mathrm{int}}, gg, Cswitch>0C_{\mathrm{switch}} > 0. Then there exist constants CC, vLR>0v_{\mathrm{LR}} > 0, independent of Λ\Lambda, such that the following holds for every Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}) and all SLT operators H0H_0, VV with interactions satisfying ∥ΦH0∥b,p,∥ΦV∥b,p≤Cint\lVert\Phi_{H_0}\rVert_{b,p}, \lVert\Phi_V\rVert_{b,p} \le C_{\mathrm{int}}. Assume that H0H_0 has a unique gapped ground state with gap size g>0g > 0 and satisfies Assumption (CONV) w.r.t. the functions fconvf_{\mathrm{conv}}, ζconv\zeta_{\mathrm{conv}} and threshold β0∈[0,∞)\beta_0 \in[0,\infty). Take any inverse temperature β>β0\beta> \beta_0. Let X⊂ΛX \subset\Lambda be connected and take A∈AXA \in\mathcal{A}_X. Then there exist response coefficients σβ,jA\sigma^A_{\beta,j} in the following sense: For ε∈(0,1]\varepsilon\in(0,1] and a smooth switching function f:R→[0,1]f:\mathbb{R} \to[0,1] satisfying f(t)=0f(t)=0 for t≤−1t \le-1, f(t)=1f(t)=1 for t≥0t \ge0, and ∥f∥CCn,m(R)<Cswitch\lVert f\rVert_{C^{\mathfrak{C}_{n,m}}(\mathbb{R})}<C_{\mathrm{switch}}, where Cn,m:=⌈m(n+1+2D/p)⌉\mathfrak{C}_{n,m}:=\lceil m(n+1+2D/p)\rceil, consider the time-dependent Hamiltonian

Hε(t):=H0+εf(t)V.H_\varepsilon(t):=H_0+\varepsilon f(t)V.

Let ρβε,η,f(t)\rho^{\varepsilon,\eta,f}_\beta(t) be the solution of the time-dependent adiabatic Schrödinger equation

iηddtρβε,η,f(t)=[Hε(t),ρβε,η,f(t)]\mathrm{i}\eta\frac{\mathrm{d}}{\mathrm{d}t}\rho^{\varepsilon,\eta,f}_\beta(t)=\left[H_\varepsilon(t),\rho^{\varepsilon,\eta,f}_\beta(t)\right]

with adiabatic parameter η∈(0,1]\eta\in(0,1] and initial datum ρβε,η,f(t)=ρβ\rho^{\varepsilon,\eta,f}_\beta(t)=\rho_\beta for all t≤−1t \le-1, where ρβ\rho_\beta is the Gibbs state of H0H_0 at inverse temperature β\beta. Then, the response to the perturbation, Σβ,Aε,η,f(t):=tr⁡(Aρβε,η,f(t))−tr⁡(Aρβ)\Sigma^{\varepsilon,\eta,f}_{\beta,A}(t):=\operatorname{tr}(A\rho^{\varepsilon,\eta,f}_\beta(t))-\operatorname{tr}(A\rho_\beta), satisfies

∣Σβ,Aε,η,f(t)−∑j=1nεjσβ,jA∣≤C∥A∥∣X∣(3+n)D(1+t)2D/p+1εn+1+C∥A∥∣X∣ninf⁡r>0(evLR(1+t)/ηχb′,p′(r)+fconv(∣X∣(1+CvolrD))ζconv(β)),(14)\left|\Sigma^{\varepsilon,\eta,f}_{\beta,A}(t)-\sum_{j=1}^{n}\varepsilon^j\sigma^A_{\beta,j}\right| \le C\lVert A\rVert\lvert X\rvert^{(3+n)D}(1+t)^{2D/p+1}\varepsilon^{n+1} +C\lVert A\rVert\lvert X\rvert^{n}\inf_{r>0}\left(e^{v_{\mathrm{LR}}(1+t)/\eta}\chi_{b',p'}(r)+f_{\mathrm{conv}}\left(\lvert X\rvert(1+C_{\mathrm{vol}}r^D)\right)\zeta_{\mathrm{conv}}(\beta)\right), \tag*{(14)}

for all t≥0t \ge0, uniformly for η∈[εm,ε1/m]\eta\in[\varepsilon^m,\varepsilon^{1/m}]. The first order coefficient is given by Kubo’s formula

σβ,1A=−itr⁡([IH0,g(V),A]ρβ),(15)\sigma^A_{\beta,1}=-\mathrm{i}\operatorname{tr}\left([\mathcal{I}_{H_0,g}(V),A]\rho_\beta\right), \tag*{(15)}

where IH0,g(⋅)\mathcal{I}_{H_0,g}(\cdot) is the quasi-local inverse of the Liouvillian with suitable filter function introduced in Remark 3.2 below.

Remark 3.2 (Explicit response coefficients). The response coefficients from Theorem 3.1 can be explicitly calculated from the initial state. The key object IH0,g(⋅)\mathcal{I}_{H_0,g}(\cdot) is the inverse Liouvillian

IH,g(O):=∫Rdt wg(t)∫0tds eiHsOe−iHs,\mathcal{I}_{H,g}(O):=\int_{\mathbb{R}}\mathrm{d}t\,w_g(t)\int_0^t\mathrm{d}s\,e^{\mathrm{i}Hs}Oe^{-\mathrm{i}Hs},

defined via a filter function wg∈L1(R)w_g \in L^1(\mathbb{R}) that is positive, normalized to ∫wg=1\int w_g=1, and has Fourier transform1^1 w^g\widehat{w}_g with support

supp⁡(w^g)⊂[−g,g],\operatorname{supp}(\widehat{w}_g)\subset[-g,g],

and for all p<1p < 1 decays as

∣wg(t)∣≤Ce−∣t∣p.(16)\lvert w_g(t) \rvert\le C e^{-\lvert t\rvert^p}. \tag*{(16)}

A function satisfying these requirements can be explicitly constructed (see e.g. [15, 4]) and is discussed in [20 Appendix A.3] in more detail. The higher orders are constructed iteratively; see [20 Section 6.2 and Proposition 4.5].

The proof of Theorem 3.1 relies on the method of constructing non-equilibrium almost stationary states (NEASSs) (see e.g. [32, 40, 16, 17, 19, 20, 42, 6]) and is given in Section 6.

Proof of Theorem 1.1 via convergence of Gibbs states

The following theorem shows that the systems considered in Theorem 1.1 in fact satisfy Assumptions (UGGS) and (CONV).

Theorem 4.1 (Rapid ground state convergence). Fix DD, qq, R∈NR \in\mathbb{N} and Cint>0C_{\mathrm{int}} > 0. Then there exist a∈(0,1)a \in(0,1) and C1,C2,C3>0C_1, C_2, C_3 > 0 such that the following holds. Consider a lattice Λ⋐ZD\Lambda\Subset\mathbb{Z}^D and a Hamiltonian H0H_0 as defined in (1) with ∥h∥∞,∥j∥∞≤Cint\lVert h \rVert_\infty, \lVert j \rVert_\infty\le C_{\mathrm{int}}, jxj_x of range RR, and jxj_x relatively aa-bounded w.r.t. the hxh_x’s in the sense (2).

Then, H0H_0 has a unique gapped ground state ρ∞\rho_\infty with gap g≥1−ag \ge1-a, and we have that its Gibbs state ρβ\rho_\beta converges exponentially fast to the ground state as β→∞\beta\to\infty, i.e.

∣tr⁡(Aρβ)−tr⁡(Aρ∞)∣≤C1∥A∥exp⁡(C2∣X∣)exp⁡(−C3β)\lvert\operatorname{tr}(A\rho_\beta) - \operatorname{tr}(A\rho_\infty) \rvert\le C_1 \lVert A \rVert\exp(C_2 \lvert X\rvert)\exp(-C_3\beta)

for all connected sets X⊂ΛX \subset\Lambda and observables A∈AXA \in\mathcal{A}_X.

The proof of Theorem 4.1 is given in Section 5.

Proof of Theorem 1.1. This is an immediate consequence of Theorems 3.1 and 4.1. The minimum over rr is estimated by applying the Young inequality ∣X∣rD≤α∣X∣1/α+(1−α)rD/(1−α)\lvert X\rvert r^D \le\alpha\lvert X\rvert^{1/\alpha} + (1-\alpha)r^{D/(1-\alpha)} and choosing r∼β(1−α)/Dr \sim\beta^{(1-\alpha)/D}. Afterwards, we adjust the exponent of ∣X∣\lvert X\rvert to absorb multiplicative prefactors. □

Proof of Theorem 4.1

The goal of this section is to give the proof of Theorem 4.1. In order to do so, in Proposition 5.2 below, we will show that Assumption (CONV) is a consequence of local indistinguishability, see Assumption (LI), and a controlled low-energy density of states, see Assumption (CLEDOS).

Local indistinguishability and controlled low-energy density of states

We start by formulating the local indistinguishability assumption precisely.

(LI) Local indistinguishability. We say that an SLT Hamiltonian HH satisfies local indistinguishability (LI) at inverse temperature β∈(0,∞)\beta\in(0,\infty) on Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}) with minimal system size R0R_0 (independent of Λ\Lambda) and with respect to two monotonically increasing/decreasing functions fLI,ζLI:[0,∞)→[0,∞)f_{\mathrm{LI}},\zeta_{\mathrm{LI}}:[0,\infty)\to[0,\infty), respectively, with ζLI(r)→0\zeta_{\mathrm{LI}}(r)\to0 as r→∞r\to\infty, if and only if the following holds: For all connected Λ′⊂Λ\Lambda' \subset\Lambda such that there exists x∈Λ′x\in\Lambda' with BR0(x)⊂Λ′B_{R_0}(x)\subset\Lambda', we have that

∣tr⁡(ρβΛA)−tr⁡(ρβΛ′A)∣≤∥A∥fLI(∣X∣)ζLI(d(X,Λ∖Λ′))(17)\left|\operatorname{tr}\left(\rho_{\beta}^{\Lambda}A\right)-\operatorname{tr}\left(\rho_{\beta}^{\Lambda'}A\right)\right|\leq\lVert A\rVert f_{\mathrm{LI}}(|X|)\zeta_{\mathrm{LI}}\left(d(X,\Lambda\setminus\Lambda')\right) \tag*{(17)}

for all connected X⊂Λ′X\subset\Lambda' and A∈AXA\in\mathcal{A}_X. Moreover, if there exists β0∈[0,∞)\beta_0\in[0,\infty) such that HH satisfies LI for all β>β0\beta>\beta_0 with the same functions fLI,ζLIf_{\mathrm{LI}},\zeta_{\mathrm{LI}}, we say that HH satisfies LI uniformly in small temperatures w.r.t. fLI,ζLIf_{\mathrm{LI}},\zeta_{\mathrm{LI}} and threshold β0\beta_0.

Next, we formulate the assumption of a controlled low-energy density of states precisely: To do so, for a fixed constant δ>0\delta>0, we denote the intervals

Ik=Ik(δ):=[kδ,(k+1)δ)for k∈N0.I_k=I_k(\delta):=[k\delta,(k+1)\delta)\quad\text{for }k\in\mathbb{N}_0.

(CLEDOS) Controlled low-energy density of states. Let HH be an SLT Hamiltonian on Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}) with ground state eigenvalue zero, i.e. inf⁡spec⁡(H)=0\inf\operatorname{spec}(H)=0. We say that HH has a uniform controlled low-energy density of states (CLEDOS) with minimal system size R0R_0, w.r.t. δ=δcledos>0\delta=\delta_{\mathrm{cledos}}>0 and m=mcledos∈Nm=m_{\mathrm{cledos}}\in\mathbb{N} if and only if the following holds: For all connected Λ′⊂Λ\Lambda'\subset\Lambda such that there exists x∈Λ′x\in\Lambda' with BR0(x)⊂Λ′B_{R_0}(x)\subset\Lambda', we have that

inf⁡spec⁡(HΛ′)=0andtr⁡(χIk(δ)(HΛ′))≤1+km∣Λ′∣mkfor all k∈N0.(18)\inf\operatorname{spec}(H^{\Lambda'})=0\quad\text{and}\quad\operatorname{tr}\left(\chi_{I_k(\delta)}(H^{\Lambda'})\right)\leq1+k^m|\Lambda'|^{mk}\quad\text{for all }k\in\mathbb{N}_0. \tag*{(18)}

where χI\chi_I denotes the characteristic function of an interval I⊂RI\subset\mathbb{R}.

Note that (18) for k=0k=0 implies that, in particular, zero is a non-degenerate ground state eigenvalue of HΛ′H^{\Lambda'}, which is separated by a gap of size at least δ\delta. More precisely, (CLEDOS) implies (UGGS) with g=δg=\delta. For general kk, Assumption (CLEDOS) states that the number of eigenvalues in the interval Ik(δ)I_k(\delta) can grow at most polynomially in the system size ∣Λ′∣|\Lambda'|.

Further, we point out that, upon adjusting m→3mm\to3m, the bound (18) is equivalent to controlling

tr⁡(χ[0,kδ)(HΛ′))=∑ℓ=0k−1tr⁡(χIℓ(δ)(HΛ′)).(19)\operatorname{tr}\left(\chi_{[0,k\delta)}(H^{\Lambda'})\right)=\sum_{\ell=0}^{k-1}\operatorname{tr}\left(\chi_{I_\ell(\delta)}(H^{\Lambda'})\right). \tag*{(19)}

Remark 5.1 (On Assumption (CLEDOS)). As we exploit in the proof of Theorem 4.1, Assumption (CLEDOS) holds true for non-interacting Hamiltonians with a uniform on-site gap. Additionally, as employed in the proof of Theorem 4.1 as well, relatively form-bounded perturbations of such non-interacting on-site gapped Hamiltonians (with a form bound constant smaller than the minimal on-site gap size) preserve Assumption (CLEDOS) upon adjusting the parameters δ\delta and mm. More generally, we expect Assumption (CLEDOS) to be generically true for weak perturbations of on-site gapped Hamiltonians, i.e. not necessarily satisfying the form bound. In fact, the essence of Assumption (CLEDOS) is that HH has a unique gapped ground state and at low excited energies, the many-body density of states (mbDoS) is at most polynomially large in the system size.

We can now formulate the main result of this section, stating that, under additional spectral conditions, (LI) implies (CONV). The proof is given in Section 5.3.

Proposition 5.2 ((LI) implies (CONV)). Let HH be an SLT Hamiltonian that satisfies Assumption (LI) with minimal system size R0R_0, uniformly in small temperatures w.r.t. fLIf_{\mathrm{LI}}, ζLI\zeta_{\mathrm{LI}} and threshold β0∈[0,∞)\beta_0 \in[0,\infty). Suppose, additionally, that HH has a unique ground state. Then we have the following:

(i) [Using (CLEDOS)] Assume additionally that HH satisfies Assumption (CLEDOS) w.r.t. δ=δcledos\delta=\delta_{\mathrm{cledos}}, m=mcledosm=m_{\mathrm{cledos}} and the same minimal size R0R_0. Then HH has a converging low-temperature Gibbs state w.r.t. the threshold β0\beta_0 and the functions

fconv(∣X∣)=2(fLI(∣X∣)+Cm(1+∣X∣m))f_{\mathrm{conv}}(|X|)=2\left(f_{\mathrm{LI}}(|X|)+C_m\left(1+|X|^m\right)\right)

and

ζconv(β)=inf⁡r∈(R0,∞)(ζLI(r)+Cm(1+(1+CvolrD)m)e−δβ),\zeta_{\mathrm{conv}}(\beta)=\inf_{r\in(R_0,\infty)}\left(\zeta_{\mathrm{LI}}(r)+C_m\left(1+\left(1+C_{\mathrm{vol}}r^D\right)^m\right)e^{-\delta\beta}\right),

where CmC_m is a constant depending only on mm.

(ii) [Using (UGGS)] Assume additionally that HH satisfies Assumption (UGGS) with parameters gg and R0R_0. Then HH has a converging low-temperature Gibbs state w.r.t. the threshold β0\beta_0 and the functions

fconv(∣X∣)=2(fLI(∣X∣)+elog⁡(q)α∣X∣1/α)f_{\mathrm{conv}}(|X|)=2\left(f_{\mathrm{LI}}(|X|)+e^{\log(q)\alpha|X|^{1/\alpha}}\right)

and

ζconv(β)=inf⁡r∈(R0,∞)(ζLI(r)+elog⁡(q)(1−α)(1+CvolrD)11−αe−gβ)\zeta_{\mathrm{conv}}(\beta)=\inf_{r\in(R_0,\infty)}\left(\zeta_{\mathrm{LI}}(r)+e^{\log(q)(1-\alpha)(1+C_{\mathrm{vol}}r^D)^{\frac{1}{1-\alpha}}}e^{-g\beta}\right)

for any α∈(0,1)\alpha\in(0,1).

Rapid ground state convergence via LI: Proof of Theorem 4.1

We will show Theorem 4.1 based on Proposition 5.2. In order to do so, we first show that H0H_0 from (1) satisfies Assumption (CLEDOS). Let Λ′⊂Λ\Lambda' \subset\Lambda and denote the non-interacting part of (1) by

HnonΛ′=∑x∈Λ′hx.H_{\mathrm{non}}^{\Lambda'}=\sum_{x\in\Lambda'}h_x.

Then, using the fact that HnonΛ′H_{\mathrm{non}}^{\Lambda'} is non-interacting and each hx≥0h_x \ge0 has a gap of size at least one above its ground state eigenvalue zero, it holds that

tr⁡(χ[0,k)(HnonΛ′))≤∑ℓ=0k−1(∣Λ′∣ℓ)(q−1)ℓ≤∑ℓ=0k−11ℓ!∣Λ′∣ℓ(q−1)ℓ≤1+1k≥2eq−1∣Λ′∣k.\operatorname{tr}\left(\chi_{[0,k)}\left(H_{\mathrm{non}}^{\Lambda'}\right)\right) \le\sum_{\ell=0}^{k-1} \binom{|\Lambda'|}{\ell}(q-1)^\ell \le\sum_{\ell=0}^{k-1} \frac{1}{\ell!}|\Lambda'|^\ell(q-1)^\ell\le1+\mathbb{1}_{k\ge2}e^{q-1}|\Lambda'|^k.

Taking m=⌈(q−1)/(log⁡2)⌉m=\lceil(q-1)/(\log2)\rceil, we then have that 1k≥2eq−1≤km\mathbb{1}_{k\ge2}e^{q-1}\le k^m.

Next, by the form bound (2), we have (1−a)HnonΛ′≤H0Λ′≤(1+a)HnonΛ′(1-a)H_{\mathrm{non}}^{\Lambda'}\le H_0^{\Lambda'}\le(1+a)H_{\mathrm{non}}^{\Lambda'}. Hence, by the min-max principle, we have that the jthj^{\mathrm{th}} eigenvalue Ej0,Λ′E_j^{0,\Lambda'} of H0Λ′H_0^{\Lambda'} remains “of the same order” as the jthj^{\mathrm{th}} eigenvalue Ejnon,Λ′E_j^{\mathrm{non},\Lambda'} of HnonΛ′H_{\mathrm{non}}^{\Lambda'}, i.e. Ej0,Λ′∈[(1−a)Ejnon,Λ′,(1+a)Ejnon,Λ′]E_j^{0,\Lambda'}\in[(1-a)E_j^{\mathrm{non},\Lambda'},(1+a)E_j^{\mathrm{non},\Lambda'}]; cf. also Remark 5.1. Hence,

tr⁡(χ[0,(1−a)k)(H0Λ′))≤∑ℓ=0k−1(∣Λ′∣ℓ)(q−1)ℓ≤1+1k≥2eq−1∣Λ′∣k.\operatorname{tr}\left(\chi_{[0,(1-a)k)}\left(H_0^{\Lambda'}\right)\right)\le\sum_{\ell=0}^{k-1}\binom{|\Lambda'|}{\ell}(q-1)^\ell\le1+\mathbb{1}_{k\ge2}e^{q-1}|\Lambda'|^k.

and thus, taking m=⌈(q−1)/log⁡2⌉m=\lceil(q-1)/\log2\rceil, we have that 1k≥2eq−1≤km\mathbb{1}_{k\ge2}e^{q-1}\le k^m.

In particular, involving the argument around (19), H0H_0 satisfies Assumption (CLEDOS) w.r.t. the parameters δ=1−a2\delta=\frac{1-a}{2} and m=3⌈(q−1)/log⁡2⌉m=3\lceil(q-1)/\log2\rceil.

The claim now follows from Proposition 5.2 (i) by involving the fact that H0H_0 from (1) satisfies Assumption (LI) at every inverse temperature β∈(0,∞)\beta\in(0,\infty) with

fLI(∣X∣)=C1exp⁡(C2∣X∣)andζLI(d(X,Λ∖Λ′))=exp⁡(−d(X,Λ∖Λ′)/ξLI)f_{\mathrm{LI}}(|X|)=C_1\exp(C_2|X|)\quad\text{and}\quad\zeta_{\mathrm{LI}}(d(X,\Lambda\setminus\Lambda'))=\exp(-d(X,\Lambda\setminus\Lambda')/\xi_{\mathrm{LI}})

for some constants C1,C2,ξLI>0C_1,C_2,\xi_{\mathrm{LI}}>0, as proven in [1 Theorem 2.3]. The exponential decay of ζLI\zeta_{\mathrm{LI}} allows to choose r∼βr\sim\beta in the infimum of ζconv\zeta_{\mathrm{conv}} provided in Proposition 5.2 and obtain the exponential decay e−C3βe^{-C_3\beta} in the statement.

From LI to CONV: Proof of Proposition 5.2

Let us first notice that (LI) extends to the ground state under the assumptions of the proposition. Therefore, notice that (CLEDOS) at k=0k=0 and (UGGS) each imply that the ground states ρ∞Λ′\rho_{\infty}^{\Lambda'} of HΛ′H^{\Lambda'} are non-degenerate for the relevant Λ′\Lambda' allowed in the assumptions. Hence, ρβΛ′→ρ∞Λ′\rho_{\beta}^{\Lambda'}\to\rho_{\infty}^{\Lambda'} in trace norm. And since the LI bound is uniform in β>β0\beta>\beta_0, it extends to the limit.

Now, let X⊂ΛX\subset\Lambda be connected, A∈AXA\in\mathcal{A}_X and β>β0\beta>\beta_0, so that (17) is available with the functions fLI,ζLIf_{\mathrm{LI}},\zeta_{\mathrm{LI}}. We begin with the first statement in Proposition 5.2 and estimate

∣tr⁡(ρβΛ′A)−tr⁡(ρβΛA)∣≤∣tr⁡(ρβΛA)−tr⁡(ρβΛ′A)∣+∣tr⁡(ρβΛ′A)−tr⁡(ρ∞Λ′A)∣+∣tr⁡(ρ∞Λ′A)−tr⁡(ρ∞ΛA)∣(20)\left|\operatorname{tr}(\rho_\beta^{\Lambda'}A)-\operatorname{tr}(\rho_\beta^\Lambda A)\right| \le\left|\operatorname{tr}(\rho_\beta^\Lambda A)-\operatorname{tr}(\rho_\beta^{\Lambda'}A)\right|+\left|\operatorname{tr}(\rho_\beta^{\Lambda'}A)-\operatorname{tr}(\rho_\infty^{\Lambda'}A)\right|+\left|\operatorname{tr}(\rho_\infty^{\Lambda'}A)-\operatorname{tr}(\rho_\infty^\Lambda A)\right| \tag*{(20)}

for some connected Λ′\Lambda^{\prime} satisfying X⊂Λ′⊂ΛX \subset\Lambda^{\prime} \subset\Lambda to be chosen below. Thus the first and third terms in (20) satisfy

∣tr⁡(ρβΛA)−tr⁡(ρβΛ′A)∣+∣tr⁡(ρ∞Λ′A)−tr⁡(ρ∞ΛA)∣≤2∥A∥fLI(∣X∣)ζLI(d(X,Λ∖Λ′)).(21)\left|\operatorname{tr}(\rho_{\beta}^{\Lambda} A)-\operatorname{tr}(\rho_{\beta}^{\Lambda^{\prime}} A)\right|+\left|\operatorname{tr}(\rho_{\infty}^{\Lambda^{\prime}} A)-\operatorname{tr}(\rho_{\infty}^{\Lambda} A)\right| \le2\lVert A\rVert_{f_{\mathrm{LI}}(|X|)}\zeta_{\mathrm{LI}}(d(X,\Lambda\setminus\Lambda^{\prime})). \tag*{(21)}

For the second term on the right-hand side of (20), we employ Assumption (CLEDOS). In order to do this, let

ZβΛ′≔tr⁡e−βHΛ′Z_{\beta}^{\Lambda^{\prime}} \coloneqq\operatorname{tr}e^{-\beta H^{\Lambda^{\prime}}}

be the partition function of HΛ′H^{\Lambda^{\prime}} at inverse temperature β\beta. Since zero is a non-degenerate ground state eigenvalue of HΛ′H^{\Lambda^{\prime}}, we find

∣tr⁡(ρβΛ′A)−tr⁡(ρ∞Λ′A)∣≤∥A∥tr⁡∣ρβΛ′−ρ∞Λ′∣=2∥A∥ZβΛ′−1ZβΛ′.(22)\left|\operatorname{tr}(\rho_{\beta}^{\Lambda^{\prime}} A)-\operatorname{tr}(\rho_{\infty}^{\Lambda^{\prime}} A)\right| \le\lVert A\rVert\operatorname{tr}\lvert\rho_{\beta}^{\Lambda^{\prime}}-\rho_{\infty}^{\Lambda^{\prime}}\rvert=2\lVert A\rVert\frac{Z_{\beta}^{\Lambda^{\prime}}-1}{Z_{\beta}^{\Lambda^{\prime}}}. \tag*{(22)}

where we used that

tr⁡∣ρβΛ′−ρ∞Λ′∣=(1−1ZβΛ′)+ZβΛ′−1ZβΛ′\operatorname{tr}\lvert\rho_{\beta}^{\Lambda^{\prime}}-\rho_{\infty}^{\Lambda^{\prime}}\rvert=\left(1-\frac{1}{Z_{\beta}^{\Lambda^{\prime}}}\right)+\frac{Z_{\beta}^{\Lambda^{\prime}}-1}{Z_{\beta}^{\Lambda^{\prime}}}

with the two contributions coming from the ground state eigenvalue and the remaining excited eigenvalues, respectively, since inf⁡spec⁡(HΛ′)=0\inf\operatorname{spec}(H^{\Lambda^{\prime}})=0. We are hence left with proving an upper bound on the partition function.

In order to do so, we employ Assumption (CLEDOS) and find the partition function to be bounded as

ZβΛ′≤1+∑k=1∞(1+km∣Λ′∣mk)e−βδk≤1+Cm(1+∣Λ′∣m)e−δβ(23)Z_{\beta}^{\Lambda^{\prime}}\le1+\sum_{k=1}^{\infty}(1+k^{m}\lvert\Lambda^{\prime}\rvert^{mk})e^{-\beta\delta k}\le1+C_{m}(1+\lvert\Lambda^{\prime}\rvert^{m})e^{-\delta\beta} \tag*{(23)}

for some constant CmC_{m} depending only on mm, if β\beta is large enough such that (1+∣Λ′∣m)e−δβ<1/2(1+\lvert\Lambda^{\prime}\rvert^{m})e^{-\delta\beta}<1/2 (otherwise the following bound (24) is anyway trivial). Hence, combining (20) with (21) and (22)–(23), we obtain

∣tr⁡(ρβΛA)−tr⁡(ρ∞ΛA)∣≤2∥A∥(fLI(∣X∣)ζLI(d(X,Λ∖Λ′))+Cm(1+∣Λ′∣m)e−δβ).(24)\left|\operatorname{tr}(\rho_{\beta}^{\Lambda} A)-\operatorname{tr}(\rho_{\infty}^{\Lambda} A)\right|\le2\lVert A\rVert\left(f_{\mathrm{LI}}(|X|)\zeta_{\mathrm{LI}}(d(X,\Lambda\setminus\Lambda^{\prime}))+C_{m}(1+\lvert\Lambda^{\prime}\rvert^{m})e^{-\delta\beta}\right). \tag*{(24)}

We now pick Λ′=Xr={x∈Λ∣d(x,X)≤r}\Lambda^{\prime}=X_{r}=\{x\in\Lambda\mid d(x,X)\le r\}, the rr-fattening of XX in Λ\Lambda with r>R0r>R_{0}, and find that we can choose

fconv(∣X∣)=2(fLI(∣X∣)+Cm(1+∣X∣m))f_{\mathrm{conv}}(|X|)=2\left(f_{\mathrm{LI}}(|X|)+C_{m}(1+|X|^{m})\right)

and

ζconv(β)=inf⁡r∈(R0,∞)(ζLI(r)+Cm(1+(1+CvolrD)m)e−δβ).\zeta_{\mathrm{conv}}(\beta)=\inf_{r\in(R_{0},\infty)}\left(\zeta_{\mathrm{LI}}(r)+C_{m}(1+(1+C_{\mathrm{vol}}r^{D})^{m})e^{-\delta\beta}\right).

Note that fconvf_{\mathrm{conv}} is monotonically increasing and ζconv\zeta_{\mathrm{conv}} is monotonically decreasing.

We now turn to proving the second statement in Proposition 5.2. Similarly to the first statement, we apply (20) and (21). However, now for the middle term of (20) we employ Lemma A.1 instead of (22)–(23): In fact, we have that

∣tr⁡(ρβΛ′A)−tr⁡(ρ∞Λ′A)∣≤∥A∥∥ρβΛ′−ρ∞Λ′∥tr⁡≤2∥A∥elog⁡(q)∣Λ′∣e−gβ.\left|\operatorname{tr}\left(\rho_{\beta}^{\Lambda'} A\right)-\operatorname{tr}\left(\rho_{\infty}^{\Lambda'} A\right)\right| \leq\|A\| \|\rho_{\beta}^{\Lambda'}-\rho_{\infty}^{\Lambda'}\|_{\operatorname{tr}} \leq2\|A\|e^{\log(q)|\Lambda'|}e^{-g\beta}.

where qq is the dimension of the local Hilbert spaces Hx\mathcal{H}_x. Consequently,

∣tr⁡(ρβΛA)−tr⁡(ρ∞ΛA)∣≤2∥A∥(fLI(∣X∣)ζLI(d(X,Λ∖Λ′))+elog⁡(q)∣Λ′∣e−gβ).\left|\operatorname{tr}\left(\rho_{\beta}^{\Lambda} A\right)-\operatorname{tr}\left(\rho_{\infty}^{\Lambda} A\right)\right| \leq2\|A\|\left(f_{\mathrm{LI}}(|X|)\zeta_{\mathrm{LI}}\left(d(X,\Lambda\setminus\Lambda'\right))+e^{\log(q)|\Lambda'|}e^{-g\beta}\right).

Picking again Λ′=Xr={x∈Λ∣d(x,X)≤r}\Lambda'=X_r=\{x\in\Lambda\mid d(x,X)\leq r\} for r>R0r>R_0 and optimizing in rr, we find that we can choose, for any α∈(0,1)\alpha\in(0,1) by a Young inequality,

fconv(∣X∣)=2(fLI(∣X∣)+elog⁡(q)α∣X∣1/α)f_{\mathrm{conv}}(|X|)=2\left(f_{\mathrm{LI}}(|X|)+e^{\log(q)\alpha|X|^{1/\alpha}}\right)

and

ζconv(β)=inf⁡r∈(R0,∞)(ζLI(r)+elog⁡(q)(1−α)(1+CvolrD)11−αe−gβ).\zeta_{\mathrm{conv}}(\beta)=\inf_{r\in(R_0,\infty)}\left(\zeta_{\mathrm{LI}}(r)+e^{\log(q)(1-\alpha)(1+C_{\mathrm{vol}}r^D)^{\frac{1}{1-\alpha}}}e^{-g\beta}\right).

This concludes the proof.

NEASS construction: Proof of Theorem 3.1

As mentioned before, the proof of Theorem 3.1 builds on response theory for the ground state. Contrary to the positive temperature setting, response theory is well understood for gapped ground states. More specifically, one can construct so-called non-equilibrium almost stationary states (NEASSs), which approximate the state into which the system is driven very well and allow to compute expectation values of observables explicitly [3, 40, 19, 20]. We now state the ground-state version of Theorem 3.1.

For X⊂ΛX\subset\Lambda, let ΠXr:AΛ→AXr\Pi_{X_r}:\mathcal{A}_{\Lambda}\to\mathcal{A}_{X_r} be the normalized-partial-trace conditional expectation, see [34 Lemma 4.1] or [20 Lemma A.6] for details and crucial properties. It is used to restrict the support of spread out observables in the following statement.

Theorem 6.1 (Response theory for gapped ground states (adaptation of [20 Theorem 4.1]).) Fix n,m,D,q∈Nn,m,D,q\in\mathbb{N}, 0<p′<p≤10<p'<p\leq1, 0<b′<10<b'<1 and b>b′b>b'. Fix also Cvol,Cint,g,Cswitch>0C_{\mathrm{vol}},C_{\mathrm{int}},g,C_{\mathrm{switch}}>0. Then there exist constants C,Cj>0C,C_j>0 such that the following holds for every Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}) and all SLT operators H0,VH_0,V with interactions satisfying ∥ΦH0∥b,p,∥ΦV∥b,p≤Cint\|\Phi_{H_0}\|_{b,p},\|\Phi_V\|_{b,p}\leq C_{\mathrm{int}}. Assume H0H_0 has a unique gapped ground state ρ∞\rho_\infty with gap size g>0g>0. Let X⊂ΛX\subset\Lambda be connected and take A∈AXA\in\mathcal{A}_X. Then there exist response coefficients σ∞,jA\sigma_{\infty,j}^{A} in the following sense: For ε∈(0,1]\varepsilon\in(0,1] and a smooth switching function f:R→[0,1]f:\mathbb{R}\to[0,1] satisfying f(t)=0f(t)=0 for t≤−1t\leq-1, f(t)=1f(t)=1 for t≥0t\geq0, and ∥f∥Cn,m(R)<Cswitch\|f\|_{\mathcal{C}^{n,m}(\mathbb{R})}<C_{\mathrm{switch}}, where Cn,m:=⌈m(n+1+2D/p)⌉\mathcal{C}_{n,m}:=\lceil m(n+1+2D/p)\rceil, consider the time-dependent Hamiltonian

Hε(t):=H0+εf(t)V.(25)H_{\varepsilon}(t):=H_0+\varepsilon f(t)V. \tag*{(25)}

:::`

Let ρ∞ε,η,f(t)\rho_{\infty}^{\varepsilon,\eta,f}(t) be the solution of the time-dependent adiabatic Schrödinger equation

iηddtρ∞ε,η,f(t)=[Hε(t),ρ∞ε,η,f(t)](26)\mathrm{i}\eta\frac{\mathrm{d}}{\mathrm{d}t}\rho_{\infty}^{\varepsilon,\eta,f}(t)=\left[H_{\varepsilon}(t),\rho_{\infty}^{\varepsilon,\eta,f}(t)\right] \tag*{(26)}

with adiabatic parameter η∈(0,1]\eta\in(0,1] and initial datum ρ∞ε,η,f(t)=ρ∞\rho_{\infty}^{\varepsilon,\eta,f}(t)=\rho_{\infty} for all t≤−1t\leq-1. Then, the response to the perturbation, Σ∞,Aε,η,f(t):=tr⁡(Aρ∞ε,η,f(t))−tr⁡(Aρ∞)\Sigma_{\infty,A}^{\varepsilon,\eta,f}(t):=\operatorname{tr}(A\rho_{\infty}^{\varepsilon,\eta,f}(t))-\operatorname{tr}(A\rho_{\infty}), satisfies

sup⁡η∈[εm,ε1/m]∣Σ∞,Aε,η,f(t)−∑j=1nεjσ∞,jA∣≤Cεn+1∥A∥∣X∣(3+n)D(1+t)2D/p+1,(27)\sup_{\eta\in[\varepsilon^{m},\varepsilon^{1/m}]}\left|\Sigma_{\infty,A}^{\varepsilon,\eta,f}(t)-\sum_{j=1}^{n}\varepsilon^{j}\sigma_{\infty,j}^{A}\right|\leq C\varepsilon^{n+1}\lVert A\rVert\lvert X\rvert^{(3+n)D}(1+t)^{2D/p+1}, \tag*{(27)}

for all t≥0t\geq0. Moreover, there exist linear maps Kj:AΛ→AΛ\mathfrak{K}_{j}:\mathcal{A}_{\Lambda}\to\mathcal{A}_{\Lambda} such that σ∞,jA=tr⁡(Kj(A)ρ∞)\sigma_{\infty,j}^{A}=\operatorname{tr}(\mathfrak{K}_{j}(A)\rho_{\infty}). In particular K0(A)=A\mathfrak{K}_{0}(A)=A, K1(A)=−i[JH0,g(V),A]\mathfrak{K}_{1}(A)=-\mathrm{i}\left[\mathcal{J}_{H_{0,g}}(V),A\right], and, for 0≤j≤n0\leq j\leq n,

∥Kj(Λ)∥≤Cj∥A∥∣X∣j,∥(1−ΠXr)Kj(Λ)∥≤Cj∥A∥∣X∣jχb′,p′(r).(28)\lVert\mathfrak{K}_{j}(\Lambda)\rVert\leq C_{j}\lVert A\rVert\lvert X\rvert^{j},\qquad\lVert(1-\Pi_{X_{r}})\mathfrak{K}_{j}(\Lambda)\rVert\leq C_{j}\lVert A\rVert\lvert X\rvert^{j}\chi_{b',p'}(r). \tag*{(28)}

Proof. In [20], this statement is formulated for systems which are gapped only within Λgap⊂Λ\Lambda^{\mathrm{gap}}\subset\Lambda. The first part of the statement is obtained by setting Λpert=Λgap=Λ\Lambda^{\mathrm{pert}}=\Lambda^{\mathrm{gap}}=\Lambda from [20 Theorem 4.1]. Then, note that [20 Proposition 4.3] provides a sequence of (1,p′)(1,p')-localized SLT operators AμA_{\mu} (these have nothing to do with the AA used in this work). By the construction, Kj\mathfrak{K}_{j} collects the εj\varepsilon^{j} order terms of e−i∑μ=1nεμ[Aμ,⋅]\mathrm{e}^{-\mathrm{i}\sum_{\mu=1}^{n}\varepsilon^{\mu}[A_{\mu},\cdot]} after expanding the exponential. Hence, each Kj\mathfrak{K}_{j} is a sum of multicommutators with AμA_{\mu} of depth at most jj. The first bound in (28) therefore follows from [20 (93) of Lemma A.4]. The second bound is trivial for j=0j=0. For j≥1j\geq1, it follows from well-known properties of the conditional expectation [20 Lemma A.6(e)] together with a generalization of [20 (92) of Lemma A.4] to multicommutators, which we provide in Lemma B.1. This concludes the proof. □

Now, let τt,sε,η,f\tau_{t,s}^{\varepsilon,\eta,f} denote the Heisenberg time evolution generated by Hε(⋅)/ηH_{\varepsilon}(\cdot)/\eta, such that

tr⁡(Aρβε,η,f(t))=tr⁡(τt,−1ε,η,f(A)ρβ)\operatorname{tr}\left(A\rho_{\beta}^{\varepsilon,\eta,f}(t)\right)=\operatorname{tr}\left(\tau_{t,-1}^{\varepsilon,\eta,f}(A)\rho_{\beta}\right)

for all A∈AΛA\in\mathcal{A}_{\Lambda}.

To proceed with the proof of Theorem 3.1, we combine the ground state expansion with Assumption (CONV). Therefore, define the response coefficients by

σβ,jA:=tr⁡(Kj(A)ρβ)\sigma_{\beta,j}^{A}:=\operatorname{tr}\left(\mathfrak{K}_{j}(A)\rho_{\beta}\right)

for all j∈{1,…,n}j\in\{1,\ldots,n\} with Kj\mathfrak{K}_{j} provided in Theorem 6.1. With K1\mathfrak{K}_{1} provided in Theorem 6.1, this in particular gives (15). Then,

Σβ,Aε,η,f(t)−∑j=1nεjσβ,jA=tr⁡((τt,−1ε,η,f(A)−A−∑j=1nεjKj(A))ρβ)=Σ∞,Aε,η,f(t)−∑j=1nεjσ∞,jA+tr⁡((τt,−1ε,η,f(A)−A−∑j=1nεjKj(A))(ρβ−ρ∞))(29)\begin{aligned} \Sigma_{\beta,A}^{\varepsilon,\eta,f}(t)-\sum_{j=1}^{n}\varepsilon^{j}\sigma_{\beta,j}^{A} =\operatorname{tr}\left(\left(\tau_{t,-1}^{\varepsilon,\eta,f}(A)-A-\sum_{j=1}^{n}\varepsilon^{j}\mathfrak{K}_{j}(A)\right)\rho_{\beta}\right) \\ &=\Sigma_{\infty,A}^{\varepsilon,\eta,f}(t)-\sum_{j=1}^{n}\varepsilon^{j}\sigma_{\infty,j}^{A} +\operatorname{tr}\left(\left(\tau_{t,-1}^{\varepsilon,\eta,f}(A)-A-\sum_{j=1}^{n}\varepsilon^{j}\mathfrak{K}_{j}(A)\right)(\rho_{\beta}-\rho_{\infty})\right) \tag*{(29)} \end{aligned}

for all β∈(β0,∞]\beta\in(\beta_0,\infty]. The first term is bounded by the estimate (27) for the response at zero temperature. To bound the second term, denote the operator in the parenthesis FF and notice that ∥F∥≤Cn∥A∥∣X∣n\lVert F\rVert\le C_n\lVert A\rVert|X|^n by (28) for some CnC_n. We then use the conditional expectation to decompose

F=ΠXrF+(1−ΠXr)F.F=\Pi_{X_r}F+(1-\Pi_{X_r})F.

The first term ΠXrF∈AXr\Pi_{X_r}F\in\mathcal{A}_{X_r} is local and applying Assumption (CONV), together with ∣Xr∣≤∣X∣(1+CvolrD)|X_r|\le|X|(1+C_{\mathrm{vol}}r^D), gives

∣tr⁡(ΠXrF(ρβ−ρ∞))∣≤Cn∥A∥∣X∣nfconv(∣X∣(1+CvolrD))ζconv(β).(30)\left|\operatorname{tr}\left(\Pi_{X_r}F(\rho_\beta-\rho_\infty)\right)\right|\le C_n\lVert A\rVert|X|^n f_{\mathrm{conv}}\left(|X|(1+C_{\mathrm{vol}}r^D)\right)\zeta_{\mathrm{conv}}(\beta). \tag*{(30)}

The second term is bounded by

∥(1−ΠXr)F∥≤∥(1−ΠXr)τt,−1ε,η,f(A)∥+∑j=1n∥(1−ΠXr)Kj(A)∥.\left\lVert(1-\Pi_{X_r})F\right\rVert\le\left\lVert(1-\Pi_{X_r})\tau_{t,-1}^{\varepsilon,\eta,f}(A)\right\rVert+\sum_{j=1}^{n}\left\lVert(1-\Pi_{X_r})\mathcal{K}_j(A)\right\rVert.

By Lieb-Robinson bounds and the properties of the conditional expectation, there exists C,vLR>0C,v_{\mathrm{LR}}>0 such that

∥(1−ΠXr)τt,−1ε,η,f(A)∥≤C∥A∥∣X∣evLR(1+t)/ηχb′,p′(r),\left\lVert(1-\Pi_{X_r})\tau_{t,-1}^{\varepsilon,\eta,f}(A)\right\rVert\le C\lVert A\rVert|X|e^{v_{\mathrm{LR}}(1+t)/\eta}\chi_{b',p'}(r),

for all r>0r>0, t≥0t\ge0, and ε,η∈(0,1]\varepsilon,\eta\in(0,1]. For more details see [20 Lemma A.6(e) and Lemma A.3]. Together with (28), we therefore have

∥(1−ΠXr)F∥≤Cn∥A∥∣X∣nevLR(1+t)/ηχb′,p′(r).(31)\left\lVert(1-\Pi_{X_r})F\right\rVert\le C_n\lVert A\rVert|X|^n e^{v_{\mathrm{LR}}(1+t)/\eta}\chi_{b',p'}(r). \tag*{(31)}

Adding (27), (30) and (31) to bound (29) and then minimizing over r>0r>0 proves (14).

Acknowledgments

J.H. acknowledges partial financial support by the ERC Advanced Grant “RMTBeyond” No. 101020331 and the ERC Consolidator Grant “ProbQuant” (jointly with the Swiss State Secretariat for Education, Research and Innovation). The research of M.L. is supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through grants TRR 352 – Project-ID 470903074 and FOR 5413 - Project-ID 465199066 and by the European Union (ERC StG MathQuantProp, Grant Agreement 101163620).2 This project started when T. W. was still at Universität Tübingen and supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 465199066; 470903074. The research of T.W. was also partially supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy – EXC-2111 – 390814868.

AI declaration

The proof of the result and the original draft were produced in 2025 without the use of Large Language Models (LLMs). LLMs were used to scan the draft for errors, which led to various minor improvements and corrections. The authors take full responsibility for the manuscript in its entirety.

An auxiliary lemma

Lemma A.1 (General β\beta-convergence). Let H\mathcal{H} be a Hilbert space of finite dimension n∈Nn \in\mathbb{N} and H∈L(H)H \in\mathcal{L}(\mathcal{H}) a self-adjoint operator. Assume that HH has non-degenerate ground state eigenvalue separated by a gap of size gg from the second-smallest eigenvalue. Recall the notation ρβH[H]\rho_{\beta}^{\mathcal{H}}[H] for its Gibbs state at inverse temperature β∈(0,∞)\beta\in(0,\infty). Then, denoting the trace norm of operators on H\mathcal{H} by ∥ρ∥tr:=sup⁡A∈L(H):∥A∥=1∣tr⁡(ρA)∣\|\rho\|_{\mathrm{tr}} := \sup_{A \in\mathcal{L}(\mathcal{H}):\|A\|=1}|\operatorname{tr}(\rho A)| we have

∥ρβH[H]−ρ∞H[H]∥tr≤2(n−1)e−gβ.(32)\left\|\rho_{\beta}^{\mathcal{H}}[H]-\rho_{\infty}^{\mathcal{H}}[H]\right\|_{\mathrm{tr}} \leq2(n-1)e^{-g\beta}. \tag*{(32)}

where ρ∞H[H]\rho_{\infty}^{\mathcal{H}}[H] denotes the projection onto the ground state of HH.

Proof. By spectral decomposition of HH as H=∑j=1nEjPjH=\sum_{j=1}^{n}E_jP_j, where Ej≤Ej+1E_j\leq E_{j+1} are the possibly degenerate eigenvalues (we may assume w.l.o.g. that E1=0E_1=0) and PjP_j orthogonal rank-1 spectral projectors, we can write

ρβH[H]=∑j=1ne−βEjPj∑j=1ne−βEj.\rho_{\beta}^{\mathcal{H}}[H]=\frac{\sum_{j=1}^{n}e^{-\beta E_j}P_j}{\sum_{j=1}^{n}e^{-\beta E_j}}.

We use that E2≥gE_2\geq g and find that

∥ρβH[H]−ρ∞H[H]∥tr=∑k=1n∣e−βEk∑j=1ne−βEj−δk,1∣=2∑j=2ne−βEj∑j=1ne−βEj≤2(n−1)e−gβ1+(n−1)e−gβ≤2(n−1)e−gβ.\begin{aligned} \left\|\rho_{\beta}^{\mathcal{H}}[H]-\rho_{\infty}^{\mathcal{H}}[H]\right\|_{\mathrm{tr}}=\sum_{k=1}^{n}\left|\frac{e^{-\beta E_k}}{\sum_{j=1}^{n}e^{-\beta E_j}}-\delta_{k,1}\right| \\ &=2\frac{\sum_{j=2}^{n}e^{-\beta E_j}}{\sum_{j=1}^{n}e^{-\beta E_j}}\leq2\frac{(n-1)e^{-g\beta}}{1+(n-1)e^{-g\beta}}\leq2(n-1)e^{-g\beta}. \end{aligned}

Locality of multicommutators

Lemma B.1. Let D∈ND \in\mathbb{N}, Cvol>0C_{\mathrm{vol}}>0, b>b′>0b>b'>0, p∈(0,1]p\in(0,1] and k∈Nk\in\mathbb{N}. There exists a constant Ck>0C_k>0 such that for all lattices Λ∈G(D,Cvol)\Lambda\in\mathcal{G}(D,C_{\mathrm{vol}}), disjoint subsets X,Y⊂ΛX,Y\subset\Lambda, SLT operators A1,…,AkA_1,\ldots,A_k and observables O∈AXO\in\mathcal{A}_X and O~∈AY\widetilde{O}\in\mathcal{A}_Y it holds that

∥[[Ak,⋯[A1,O]⋯ ],O~]∥≤Ck∥O∥∥O~∥∣X∣kχb′,p(d(X,Y))∏j=1k∥ΦAj∥b,p.(33)\left\|\left[\left[A_k,\cdots\left[A_1,O\right]\cdots\right],\widetilde{O}\right]\right\|\leq C_k\|O\|\|\widetilde{O}\||X|^k\chi_{b',p}(d(X,Y))\prod_{j=1}^{k}\|\Phi_{A_j}\|_{b,p}. \tag*{(33)}

Proof. We write out each AlA_l as a sum over supports ZlZ_l and denote Z0=XZ_0 = X. Notice that each commutator vanishes if Zl∩⋃j<lZj=∅Z_l \cap\bigcup_{j<l} Z_j = \varnothing. Hence,

∥[[Ak,⋯[A1,O]⋯ ],O~]∥≤∑Z1⊂Λ:Z1∩Z0≠∅⋯∑Zk⊂Λ:Zk∩(Zk−1∪⋯∪Z0)≠∅1(Z1∪⋯∪Zk)∩Y≠∅2k+1∥O~∥∥ΦAk(Zk)∥⋯∥ΦA1(Z1)∥∥O∥.\left\|\left[\left[A_k,\cdots\left[A_1,O\right]\cdots\right],\widetilde{O}\right]\right\| \leq\sum_{\substack{Z_1\subset\Lambda:\\ Z_1\cap Z_0\ne\varnothing}} \cdots\sum_{\substack{Z_k\subset\Lambda:\\ Z_k\cap(Z_{k-1}\cup\cdots\cup Z_0)\ne\varnothing}} \mathbf{1}_{(Z_1\cup\cdots\cup Z_k)\cap Y\ne\varnothing} 2^{k+1}\|\widetilde{O}\| \|\Phi_{A_k}(Z_k)\|\cdots\|\Phi_{A_1}(Z_1)\|\|O\|.

Due to the intersecting condition, only terms satisfying d(X,Y)≤∑j=1kdiam⁡(Zj)d(X,Y) \leq\sum_{j=1}^{k} \operatorname{diam}(Z_j) contribute to this sum and we multiply each summand with

χb′,p(d(X,Y))∏j=1kχb′,p(diam⁡(Zj))≥1.\frac{\chi_{b',p}(d(X,Y))}{\prod_{j=1}^{k}\chi_{b',p}(\operatorname{diam}(Z_j))} \geq1.

Starting from the right, we can now upper bound each sum as

∑Zl⊂Λ:Zl∩(Zl−1∪⋯∪Z0)≠∅∥ΦAl(Zl)∥χb′,p(diam⁡(Zl))∣Zl∣k−l≤∑zl∈Zl−1∪⋯∪Z0∑Zl⊂Λ:zl∈Zl∥ΦAl(Zl)∥χb′,p(diam⁡(Zl))∣Zl∣k−l≤C∣Zl−1∪⋯∪Z0∣∥ΦAl∥b,p,\begin{aligned} \sum_{\substack{Z_l\subset\Lambda:\\ Z_l\cap(Z_{l-1}\cup\cdots\cup Z_0)\ne\varnothing}} \frac{\|\Phi_{A_l}(Z_l)\|}{\chi_{b',p}(\operatorname{diam}(Z_l))}|Z_l|^{k-l} \leq\sum_{z_l\in Z_{l-1}\cup\cdots\cup Z_0}\sum_{\substack{Z_l\subset\Lambda:\\ z_l\in Z_l}} \frac{\|\Phi_{A_l}(Z_l)\|}{\chi_{b',p}(\operatorname{diam}(Z_l))}|Z_l|^{k-l} \\ &\leq C|Z_{l-1}\cup\cdots\cup Z_0|\|\Phi_{A_l}\|_{b,p}, \end{aligned}

where the accumulating factors ∣Zl∣k−l|Z_l|^{k-l} coming from ∣Zl−1∪⋯∪Z0∣≤∣Zl−1∣⋯∣Z0∣|Z_{l-1}\cup\cdots\cup Z_0|\leq|Z_{l-1}|\cdots|Z_0| are absorbed in the interaction norms and only ∣Z0∣k|Z_0|^k remains at the end. □

References

References

  1. [1]A. Adhikari, J. Henheik, M. Lemm, and T. Wessel, “Uniform-in-temperature locality estimates for weakly interacting quantum systems”, 2025, arXiv:2508.15907, pre-published.arxiv.org/abs/2508.15907
  2. [2]S. Bachmann, W. De Roeck, and M. Fraas, “Adiabatic theorem for quantum spin systems”, Physical Review Letters 119, 060201 (2017).DOI
  3. [3]S. Bachmann, W. De Roeck, and M. Fraas, “The adiabatic theorem and linear response theory for extended quantum systems”, Communications in Mathematical Physics 361, 997–1027 (2018), arXiv:1705.02838.DOI
  4. [4]S. Bachmann, S. Michalakis, B. Nachtergaele, and R. Sims, “Automorphic equivalence within gapped phases of quantum lattice systems”, Communications in Mathematical Physics 309, 835–871 (2012), arXiv:1102.0842.DOI
  5. [5]C. L. Baldwin, A. Ehrenberg, A. Y. Guo, and A. V. Gorshkov, “Disordered Lieb-Robinson bounds in one dimension”, PRX Quantum 4, 020349 (2023), arXiv:2208.05509.DOI
  6. [6]L. Becker, S. Teufel, and M. Wesle, “The generalized adiabatic theorem for extended lattice systems”, 2025, arXiv:2510.20914.arxiv.org/abs/2510.20914
  7. [7]M. V. Berry, “Quantal phase factors accompanying adiabatic changes”, Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences 392, 45–57 (1984).
  8. [8]Á. Capel, M. Moscolari, S. Teufel, and T. Wessel, “From decay of correlations to locality and stability of the Gibbs state”, Communications in Mathematical Physics 406, 43 (2025), arXiv:2310.09182.DOI
  9. [9]W. De Roeck, L. Giacomin, F. Huveneers, and O. Prosniak, “Absence of normal heat conduction in strongly disordered interacting quantum chains”, 2024, arXiv:2408.04338.arxiv.org/abs/2408.04338
  10. [10]A. Elgart and A. Klein, “Many-body localization for the random XXZ spin chain in fixed energy intervals”, 2026, arXiv:2602.01441.arxiv.org/abs/2602.01441
  11. [11]A. Giuliani, V. Mastropietro, and M. Porta, “Universality of the Hall conductivity in interacting electron systems”, Communications in Mathematical Physics 349, 1107–1161 (2017), arXiv:1511.04047.DOI
  12. [12]R. L. Greenblatt, M. Lange, G. Marcelli, and M. Porta, “Adiabatic evolution of low-temperature many-body systems”, Communications in Mathematical Physics 405, 75 (2024), arXiv:2211.16836.DOI
  13. [13]M. B. Hastings, “Entropy and entanglement in quantum ground states”, Physical Review B 76, 035114 (2007), arXiv:cond-mat/0701055.DOI
  14. [14]M. B. Hastings and S. Michalakis, “Quantization of Hall conductance for interacting electrons on a torus”, Communications in Mathematical Physics 334, 433–471 (2015), arXiv:1306.1258.DOI
  15. [15]M. B. Hastings and X.-G. Wen, “Quasiadiabatic continuation of quantum states: the stability of topological ground-state degeneracy and emergent gauge invariance”, Physical Review B 72, 045141 (2005), arXiv:cond-mat/0503554.DOI
  16. [16]J. Henheik and S. Teufel, “Adiabatic theorem in the thermodynamic limit: systems with a gap in the bulk”, Forum of Mathematics, Sigma 10 (2022), arXiv:2012.15239.arxiv.org/abs/2012.15239
  17. [17]J. Henheik and S. Teufel, “Adiabatic theorem in the thermodynamic limit: Systems with a uniform gap”, Journal of Mathematical Physics 63, 011901 (2022), arXiv:2012.15238.DOI
  18. [18]J. Henheik and S. Teufel, “Justifying Kubo’s formula for gapped systems at zero temperature: a brief review and some new results”, Reviews in Mathematical Physics 33, 2060004 (2021).DOI
  19. [19]J. Henheik and T. Wessel, “On adiabatic theory for extended fermionic lattice systems”, Journal of Mathematical Physics 63, 121101 (2022), arXiv:2208.12220.DOI
  20. [20]J. Henheik and T. Wessel, “Response theory for locally gapped systems”, Journal of Statistical Physics 193, 151 (2026), arXiv:2410.10809.arxiv.org/abs/2410.10809
  21. [21]N. Il’in, A. Aristova, and O. Lychkovskiy, “Adiabatic theorem for closed quantum systems initialized at finite temperature”, Physical Review A 104 (2021), arXiv:2002.02947.DOI
  22. [22]V. Jakšić, Y. Ogata, and C.-A. Pillet, “The Green-Kubo formula and the Onsager reciprocity relations in quantum statistical mechanics”, Communications in Mathematical Physics 265, 721–738 (2006).DOI
  23. [23]V. Jakšić, C.-A. Pillet, and C. Tauber, “A note on adiabatic time evolution and quasi-static processes in translation-invariant quantum systems”, Annales Henri Poincaré 25, 751–771 (2024), arXiv:2204.02177.arxiv.org/abs/2204.02177
  24. [24]V. Jakšić, C.-A. Pillet, and C. Tauber, “Approach to equilibrium in translation-invariant quantum systems: some structural results”, Annales Henri Poincaré 25, 715–749 (2024), arXiv:2204.00440.arxiv.org/abs/2204.00440
  25. [25]T. Kato, “On the adiabatic theorem of quantum mechanics”, Journal of the Physical Society of Japan 5, 435–439 (1950).DOI
  26. [26]M. Kliesch, C. Gogolin, M. J. Kastoryano, A. Riera, and J. Eisert, “Locality of temperature”, Physical Review X 4, 031019 (2014), arXiv:1309.0816.DOI
  27. [27]R. Kubo, “Statistical-mechanical theory of irreversible processes. I. General theory and simple applications to magnetic and conduction problems”, Journal of the Physical Society of Japan 12, 570–586 (1957).DOI
  28. [28]M. Lemm and T. Wessel, “Enhanced Lieb-Robinson bounds for commuting long-range interactions”, Journal of Mathematical Physics 66, 091901 (2025), arXiv:2411.19241.DOI
  29. [29]E. H. Lieb and D. W. Robinson, “The finite group velocity of quantum spin systems”, Communications in Mathematical Physics 28, 251–257 (1972).
  30. [30]L. Masanes, “Area law for the entropy of low-energy states”, Physical Review A 80, 052104 (2009), arXiv:0907.4672.DOI
  31. [31]B. T. McDonough, M. Lemm, and A. Lucas, “Non-perturbatively slow spread of quantum correlations in non-resonant systems”, 2025, arXiv:2510.11831.arxiv.org/abs/2510.11831
  32. [32]D. Monaco and S. Teufel, “Adiabatic currents for interacting fermions on a lattice”, Reviews in Mathematical Physics 31, 1950009 (2019), arXiv:1707.01852.DOI
  33. [33]B. Nachtergaele, R. Sims, and A. Young, “Lieb-Robinson bounds, the spectral flow, and stability of the spectral gap for lattice fermion systems”, in Mathematical results in quantum physics, edited by F. Bonetto, D. Borthwick, E. Harrell, and M. Loss, Contemporary Mathematics 717 (2018), pp. 93–115, arXiv:1705.08553.DOI
  34. [34]B. Nachtergaele, R. Sims, and A. Young, “Quasi-locality bounds for quantum lattice systems. I. Lieb-Robinson bounds, quasi-local maps, and spectral flow automorphisms”, Journal of Mathematical Physics 60, 061101 (2019), arXiv:1810.02428.DOI
  35. [35]G. Nenciu, “On the adiabatic theorem of quantum mechanics”, Journal of Physics A: Mathematical and General 13, L15–L18 (1980).DOI
  36. [36]I. Prémont-Schwarz, A. Hamma, I. Klich, and F. Markopoulou-Kalamara, “Lieb-Robinson bounds for commutator-bounded operators”, Physical Review A 81, 040102 (2010), arXiv:0912.4544.
  37. [37]B. Simon, “Holonomy, the quantum adiabatic theorem, and Berry’s phase”, Physical Review Letters 51, 2167–2170 (1983).DOI
  38. [38]H. Spohn and J. L. Lebowitz, “Irreversible thermodynamics for quantum systems weakly coupled to thermal reservoirs”, in Advances in chemical physics, edited by S. A. Rice, 38 (John Wiley & Sons, Ltd, 1978), pp. 109–142.
  39. [39]S. Teufel, Adiabatic perturbation theory in quantum dynamics, Lecture Notes in Mathematics (Springer, Berlin, Heidelberg, 2003).DOI
  40. [40]S. Teufel, “Non-equilibrium almost-stationary states and linear response for gapped quantum systems”, Communications in Mathematical Physics 373, 621–653 (2020), arXiv:1708.03581.arxiv.org/abs/1708.03581
  41. [41]S. Teufel and M. Wesle, “A note on Hall conductance and Hall conductivity in interacting fermion systems”, 2025, arXiv:2506.13581.arxiv.org/abs/2506.13581
  42. [42]M. Wesle, G. Marcelli, T. Miyao, D. Monaco, and S. Teufel, “Near linearity of the macroscopic Hall current response in infinitely extended gapped fermion systems”, Communications in Mathematical Physics 406, 199 (2025), arXiv:2411.06967.arxiv.org/abs/2411.06967

Paper details

vlnc:W_v7hz3kbr
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