betaFeedback

WorksPhysics and Astronomy

Quasicrystals are Functionally Graded Materials

Abstract

Quasicrystals exhibit long-range order without translational periodicity, posing challenges to compute their electronic properties. Here we show that a quasicrystal can be represented exactly as a periodic approximant with a spatially graded phason, establishing a connection to functionally graded materials. Building on this connection, we develop a controlled quantum theory of quasicrystals. We derive a state-dependent control parameter governed by the quantum metric of local Bloch states, determining when finite approximations are reliable. We further construct a Hamiltonian that correctly predicts Bragg gap positions, quasimomentum equivalence, and quasiperiodic eigenstates to all orders in phason strain within a finite Fourier window. We verify the theory numerically and propose a diffraction-based experimental test of the gap shift induced by phason strain. Our framework opens a route to predicting the electronic structure and functional properties of quasicrystals using methods developed for functionally graded materials.

Quasicrystals and functionally graded materials (FGMs) represent two distinct classes of spatially structured materials: Quasicrystals possess long-range order without translational periodicity [1, 2], with applications ranging from wear-resistant and low-friction coatings [3] to prospective thermoelectric materials [4] and engineered photonic structures [5]. FGMs, on the other hand, exhibit spatial variations in composition, microstructure, or orientation, enabling tailored properties for thermal-barrier coatings in aerospace and turbines [6], wear-resistant cutting tools [7], and orthopedic implants [8]. Although traditionally treated as unrelated material families, both demonstrate how spatial organization can be exploited to control material properties.

Despite their technological importance, the spatial complexity also poses a fundamental challenge to theoretical predictions: the translational symmetry underlying conventional Bloch theory is absent in quasicrystals and generally broken by grading. For quasicrystals, a widely used approach replaces the aperiodic structure with a periodic approximant having a large unit cell, allowing conventional band-structure calculations [9, 10, 11]. Alternative approaches include direct real-space calculations [12], higher-dimensional superspace formulations [13], and quasiperiodic operator theory [14, 15, 16]. Recent developments have established spectral convergence bounds and certified calculations for infinite quasiperiodic systems [13, 17]. However, a fundamental challenge remains in connecting the electronic states of practical periodic approximants to those of the underlying quasicrystal: how are their band energies, wavefunctions, and quantum numbers related, and when does a finite approximant faithfully describe them?

In this Letter, we establish a connection between quasicrystals and FGMs, enabling a controlled quantum theory to study quasicrystal electronic structure. We show that a quasicrystal can be represented exactly as a peri-

odic approximant with a spatially graded phason, arising from the difference between irrational and rational cuts of a higher-dimensional crystal [18, 19, 20, 21] (Figure 1). Building on the recent quantum theory of FGMs [22], we derive a state-dependent control parameter governed by the quantum metric of local Bloch states, establishing when finite approximations are reliable and when they fail. We further construct a Hamiltonian that correctly predicts Bragg gap positions, quasimomentum equivalence, and quasiperiodic eigenstates to all orders in phason strain within a finite Fourier window. Our framework thus transforms periodic approximants into a controlled computational approach to quasicrystals, opening a route to predicting their electronic structure and functional properties beyond conventional band theory.

Figure 1. A quasicrystal is a graded approximant. (a) A potential that is periodic in the physical coordinate xx and internal coordinate ww (gray scale). The Bloch crystal, quasicrystal, and approximant are the slices with slopes w=0w = 0, irrational α\alpha, and nearby rational α~=c/2a\tilde{\alpha} = c/2a, respectively, with ξ=αa/c\xi= \alpha a/c the inverse golden mean. The separation δx\delta x between the last two, δ\delta as in Eq. (2), grows linearly in xx; that is, the quasicrystal is the approximant with the drifting phason w0+δxw_0 + \delta x. (b) The Fourier module of the 3/5 approximant in the plane of wavevector and internal index, with δ′\delta' exaggerated. Each supercell reciprocal vector Q~\tilde{Q} carries a column of quasicrystal wavevectors Q~+νδ′\tilde{Q} + \nu\delta'. The approximant keeps one point per column (filled), the one in the window IqI_q of qq consecutive internal indices, −q/2<ν≤q/2-q/2 < \nu\le q/2 (gray band), and identifies ν\nu with ν+q\nu+ q. Our theory moves each filled point to its true position, but it cannot reach the open ones. (c) The theory of functionally graded materials can be applied to the phasons of the quasicrystal and its approximant. The FGM parameter λ\lambda becomes defined in terms of δx\delta x, which varies slowly on the scale of the lattice constant, though not on the scale of the supercell (see The small parameter).

The quasicrystal as a graded approximant.— We consider a minimal representative quasicrystal with one phason, and show it can be represented exactly as a spatially varying phason modulation of a periodic approximant [23]. Denote the quasicrystal potential by VQ(r∈R3)V_Q(\mathbf{r}\in\mathbb{R}^3), defined as the restriction of a higher-dimensional periodic potential V(w,r)V(\mathbf{w},\mathbf{r}) over R4\mathbb{R}^4 to a 3D3D hyperplane, with discrete symmetries w→w+c\mathbf{w}\to\mathbf{w}+\mathbf{c} and r→r+Ri\mathbf{r}\to\mathbf{r}+\mathbf{R}_i for basis Ri\mathbf{R}_i of lattice Λ\Lambda with reciprocal lattice Λ∗\Lambda^*. Defining G\mathbf{G} as the wavevector and nn as the superspace harmonic, we have V=∑nGVnGe2πinw/ceiG⋅rV=\sum_{n\mathbf{G}}V_{n\mathbf{G}}e^{2\pi in\mathbf{w}/c}e^{i\mathbf{G}\cdot\mathbf{r}}. The quasicrystal potential is then

VQ(r)=V(w0+α⋅r,r)=∑nGVnGe2πinw0/cei(nβ+G)⋅r,(1)V_Q(\mathbf{r})=V(\mathbf{w}_0+\boldsymbol{\alpha}\cdot\mathbf{r},\mathbf{r})=\sum_{n\mathbf{G}}V_{n\mathbf{G}}e^{2\pi in\mathbf{w}_0/c}e^{i(n\boldsymbol{\beta}+\mathbf{G})\cdot\mathbf{r}}, \tag*{(1)}

with α\boldsymbol{\alpha} and w0\mathbf{w}_0 the hyperplane slope and intercept (the phason) respectively, and wavevector Q=nβ+G\mathbf{Q}=n\boldsymbol{\beta}+\mathbf{G} with β=2πα/c\boldsymbol{\beta}=2\pi\boldsymbol{\alpha}/c. These wavevectors form a Fourier module M=Λ∗+Zβ\mathcal{M}=\Lambda^*+\mathbb{Z}\boldsymbol{\beta}, which are dense in R3\mathbb{R}^3 whenever 11 and the hyperplane slopes ξ=αRi/c\boldsymbol{\xi}=\boldsymbol{\alpha}\mathbf{R}_i/c are rationally independent [24]; in one dimension, ξ=αa/c\xi=\alpha a/c reduces to a period ratio. The decomposition Q=nβ+G\mathbf{Q}=n\boldsymbol{\beta}+\mathbf{G} is then unique in nn, so nn is a function ν(Q)\nu(Q) on the module, which we define as the internal harmonic index. The corresponding operator ν^\hat{\nu} scales plane waves of form ei(k+Q)⋅re^{i(k+Q)\cdot r} with Bloch wavevector kk by ν(Q)\nu(Q), and is a generator of phason translations via the unitary shift operator e2πiν^w0/ce^{2\pi i\hat{\nu}w_0/c}, which preserves the energy spectrum [25, 26].

Supercells approximate α\alpha by a rational slope α~\tilde{\alpha}, where α~⋅Ri/c=pi/q\tilde{\alpha}\cdot R_i/c = p_i/q. The supercell then holds qq cells. The supercell’s lattice is Λ~={R∈Λ:α~⋅R∈cZ}\tilde{\Lambda} = \{R \in\Lambda: \tilde{\alpha}\cdot R \in c\mathbb{Z}\} with reciprocal lattice Λ~∗=Λ∗+Zβ~\tilde{\Lambda}^{*} = \Lambda^{*} + \mathbb{Z}\tilde{\beta} (vectors Q~\tilde{Q}), β~=2πα~/c\tilde{\beta} = 2\pi\tilde{\alpha}/c, and its Fourier module is the same but ν\nu has been reduced modulo qq.

We now cast the quasicrystal potential into an approximant. Let Vλ(r)≡V(α~⋅r+λ,r)V_\lambda(r) \equiv V(\tilde{\alpha}\cdot r + \lambda, r) for some λ\lambda. Then

VQ(r)=V(w0+δ⋅r+α~⋅r,r)=Vλ(r)(r),V_Q(r) = V(w_0 + \delta\cdot r + \tilde{\alpha}\cdot r, r) = V_{\lambda(r)}(r),
λ(r)≡w0+δ⋅r,δ≡α−α~,(2)\lambda(r) \equiv w_0 + \delta\cdot r,\qquad\delta\equiv\alpha- \tilde{\alpha}, \tag*{(2)}

where we have parameterized the set of approximants by a spatially varying phason, exactly reproducing an FGM potential Vλ(r)(r)V_{\lambda(r)}(r) [22]. Previous diffraction experiments have measured δ\delta, the phason strain, which displaces every diffraction peak of internal index ν\nu by νδ′\nu\delta', with δ′≡2πδ/c\delta' \equiv2\pi\delta/c its wavevector [27, 28].

Two key observations.— (1) Translating VλV_\lambda by a lattice vector in Λ\Lambda yields Vλ+α~⋅RV_{\lambda+\tilde{\alpha}\cdot R}, where α~⋅R\tilde{\alpha}\cdot R is a multiple of c/qc/q. Let En(k;λ)\mathcal{E}_n(k;\lambda) be the bands of VλV_\lambda and unk(r;λ)u_{nk}(r;\lambda) the periodic parts of its Bloch functions, normalized over the supercell. As band structure is translation-invariant,

En(k;λ+c/q)=En(k;λ).(3)\mathcal{E}_n(k;\lambda+ c/q) = \mathcal{E}_n(k;\lambda). \tag*{(3)}

Thus the local band structure must repeat qq times per phason period, with an effective modulation length ℓeff=ℓmod/q\ell_{\mathrm{eff}} = \ell_{\mathrm{mod}}/q for a phason modulation length ℓmod=c/∣δ∣\ell_{\mathrm{mod}} = c/|\delta|.

(2) What does shifting the phason λ\lambda do to the approximant potential? To answer this question, notice that for the superlattice, ν(Q~)\nu(\tilde{Q}) is not unique but is defined only modulo qq; we therefore restrict ν(Q~)∈Iq\nu(\tilde{Q}) \in I_q, a window of qq consecutive integers, e.g., −q/2<ν≤q/2-q/2 < \nu\le q/2. Notice that the harmonics of VV connecting Q~\tilde{Q} to Q~′\tilde{Q}' have internal index ν(Q~)−ν(Q~′)+ℓq\nu(\tilde{Q}) - \nu(\tilde{Q}') + \ell q, ℓ∈Z\ell\in\mathbb{Z}. For ℓ=0\ell= 0 terms, potentials of different λ\lambda are connected by a unitary transformation Uλ≡e2πiν^λ/cU_\lambda\equiv e^{2\pi i\hat{\nu}\lambda/c}. We account for the ℓ≠0\ell\ne0 terms in a wrap-around potential, denoted V~λ\tilde{V}_\lambda, which contains the high internal harmonics of VV and couplings across the two edges of the window, resulting in

Vλ=UλV0Uλ†+V~λ.(4)V_\lambda= U_\lambda V_0 U_\lambda^\dagger+ \tilde{V}_\lambda. \tag*{(4)}

UλU_\lambda is unitary, so all phason-modulated band structure effects originate from V~λ\tilde{V}_\lambda. As Q=Q~+νδ′Q = \tilde{Q} + \nu\delta' Figure 1(b), the approximant is the quasicrystal with two operations applied: ν∼ν+q\nu\sim\nu+ q, and (k+Q~+νδ′)2→(k+Q~)2(k + \tilde{Q} + \nu\delta')^2 \to(k + \tilde{Q})^2 in the kinetic energy. The former, unfortunately, cannot be undone; the approximant is only accurate up to the supercell window. We will return to this point when we define a window Hamiltonian.

The small parameter.— For a supercell of size qaqa, there are two naive candidate parameters. We might expect the small parameter to be either κ0=qa/ℓmod=q∣ξ−p/q∣\kappa_0 = qa/\ell_{\mathrm{mod}} = q|\xi- p/q| or κ=qa/ℓeff=q2∣ξ−p/q∣\kappa= qa/\ell_{\mathrm{eff}} = q^2|\xi- p/q|, based on the irrational hyperplane slope. We can bound the lowest possible error of such an approximant, taking ξ\xi to be the golden mean for example. Using a continued fraction expansion with convergents pn/qnp_n/q_n for a supercell of qnq_n cells, [29, 30] show that κn≡qn2∣ξ−pn/qn∣\kappa_n \equiv q_n^2|\xi- p_n/q_n| is bounded below by 1/(an+1+2)1/(a_{n+1}+2) for an+1a_{n+1} the next partial quotient. By Hurwitz’s theorem [31] we find κn\kappa_n converges to 1/51/\sqrt{5}. There is therefore a trade-off between a small gradient and a small supercell: the band structure of a golden-mean approximant repeats every 1/κn≈2.241/\kappa_n \approx2.24 supercells, irrespective of the supercell size. A similar argument holds for any quadratic irrational, and observed quasicrystals likewise have ξ\xi quadratic irrational [32].

It turns out that neither κ0\kappa_0 nor κ\kappa is the correct small parameter for all states. When the wrap-around potential is negligible, Eq. (4) reduces to u(λ)=Uλu(0)u(\lambda)=U_\lambda u(0) for the periodic part of the Bloch function, with slow derivative

∇~u=δ∂λu=iδν^u.(5)\widetilde{\nabla}u=\delta\partial_\lambda u=i\delta\hat{\nu}u. \tag*{(5)}

Each plane-wave component accumulates phase (winds) proportional to the internal index, via eiδν^⋅re^{i\delta\hat{\nu}\cdot\mathbf{r}}: components with ν=±1,±q/2\nu=\pm1,\pm q/2 wind once over ℓmod,ℓeff\ell_{\mathrm{mod}},\ell_{\mathrm{eff}} respectively. As a representative example, we let δ′∝e^z\delta'\propto\hat{\mathbf{e}}_z, for which the Hilbert-space distance between the Bloch states of neighboring points, 1−∣⟨u(z)∣u(z+dz)⟩∣2=gzzdz21-\lvert\langle u(z)|u(z+dz)\rangle\rvert^2=g_{zz}dz^2, with gzzg_{zz} the quantum metric, is therefore

gzz=∣δ′∣2Δν2,Δν2≡⟨ν^2⟩−⟨ν^⟩2,(6)g_{zz}=\lvert\delta'\rvert^2\Delta\nu^2,\qquad\Delta\nu^2\equiv\langle\hat{\nu}^2\rangle-\langle\hat{\nu}\rangle^2, \tag*{(6)}

with ⟨⋅⟩\langle\cdot\rangle the expectation value in uu. A natural small parameter can then be defined in terms of this metric: κeff=qagzz/2π=κ0Δν\kappa_{\mathrm{eff}}=qa\sqrt{g_{zz}}/2\pi=\kappa_0\Delta\nu. As shown in Ref. [22], a slow theory can be obtained by averaging the action over a supercell, and such an average discards supercell harmonics that have slow phases ei(ν−ν′)δ′⋅re^{i(\nu-\nu')\delta'\cdot\mathbf{r}}. Each such term can be integrated by parts, each time gaining a factor of κ0∣ν−ν′∣\kappa_0|\nu-\nu'|. The average is therefore exact up to terms that wind by κ0Δν\kappa_0\Delta\nu per supercell. In particular, then, an approximant is a good approximation to the quasicrystal depending on both the size of the supercell and on the state of interest. For example, a state comprising a few plane waves has Δν\Delta\nu of order one and κeff→0\kappa_{\mathrm{eff}}\to0 with qq; a state spread over the supercell window has Δν∼q\Delta\nu\sim q and κeff≃κ\kappa_{\mathrm{eff}}\simeq\kappa.

The complete slow theory.—We start from the modulated Bloch state of Ref. [22], ψ=unk(r)(r;λ(r))AeiS\psi=u_{n\mathbf{k}(\mathbf{r})}(\mathbf{r};\lambda(\mathbf{r}))\mathcal{A}e^{iS}, with A\mathcal{A} a slowly varying amplitude and k=∇S\mathbf{k}=\nabla S the slow lattice momentum. Inserting ψ\psi into the Schrödinger action and averaging over a supercell gives a Lagrangian for the slow fields with at most two slow derivatives [22],

L‾=−A2[εn(k+a)+δεn+Ug−E]−12∇A⋅M−1⋅∇A,(7)\overline{\mathcal{L}}=-\mathcal{A}^2\left[\varepsilon_n(\mathbf{k}+\mathbf{a})+\delta\varepsilon_n+U_g-E\right]-\frac{1}{2}\nabla\mathcal{A}\cdot\mathbf{M}^{-1}\cdot\nabla\mathcal{A}, \tag*{(7)}

exact to all orders in κeff\kappa_{\mathrm{eff}}. Here k+a\mathbf{k}+\mathbf{a} is the gauge-invariant momentum, with a=⟨u∣i∇~∣u⟩\mathbf{a}=\langle u|i\widetilde{\nabla}|u\rangle the Berry connection of the local Bloch states [22], a pseudo vector potential; δεn=Im⁡⟨u∣v^i(1−Pn)∣∇~iu⟩\delta\varepsilon_n=\operatorname{Im}\langle u|\hat{v}_i(1-P_n)|\widetilde{\nabla}_i u\rangle is the gradient energy of Sundaram and Niu [33], with v^=p^/m\hat{\mathbf{v}}=\hat{\mathbf{p}}/m the velocity, p^∣k+Q~⟩=(k+Q~)∣k+Q~⟩\hat{\mathbf{p}}|\mathbf{k}+\widetilde{\mathbf{Q}}\rangle=(\mathbf{k}+\widetilde{\mathbf{Q}})|\mathbf{k}+\widetilde{\mathbf{Q}}\rangle, and PnP_n the band projector; Ug=tr⁡(g/2m)U_g=\operatorname{tr}(g/2m) is the quantum-metric potential; M−1=∂2εn/∂k∂k\mathbf{M}^{-1}=\partial^2\varepsilon_n/\partial\mathbf{k}\partial\mathbf{k} is the inverse band mass. For a generic graded crystal these are unwieldy sums over bands. For a quasicrystal, though, we can use Eq. (5) as v^\hat{\mathbf{v}} and ν^\hat{\nu} are both diagonal in the plane-wave basis:

a=δ′⟨ν^⟩,δεn=δ′⋅Cov⁡(v^,ν^),Ug=δ′2Δν22m,(8)\mathbf{a}=\delta'\langle\hat{\nu}\rangle,\qquad\delta\varepsilon_n=\delta'\cdot\operatorname{Cov}(\hat{\mathbf{v}},\hat{\nu}),\qquad U_g=\frac{\delta'^2\Delta\nu^2}{2m}, \tag*{(8)}

with Cov⁡(v^,ν^)=∑Q~(ν−⟨ν^⟩)(k+Q~)∣uQ~∣2/m\operatorname{Cov}(\hat{\mathbf{v}},\hat{\nu})=\sum_{\widetilde{Q}}(\nu-\langle\hat{\nu}\rangle)(\mathbf{k}+\widetilde{\mathbf{Q}})|u_{\widetilde{Q}}|^2/m the correlation between the velocity of a plane-wave component and its internal index and uQ~u_{\widetilde{Q}} the plane-wave coefficients of uu. Expanding Eq. (7) about k\mathbf{k} (equivalently, applying second-order perturbation theory to the two-window Hamiltonian below), we find that the energy of the state (n,k)(n,\mathbf{k}) through second order is

Enk=εn+δ′⋅Cn+δ′2⟨ν^2⟩2m−∑n′≠n∣δ′⋅Cn′n∣2εn′−εn,(9)E_{n\mathbf{k}}=\varepsilon_n+\delta'\cdot\mathbf{C}_n+\frac{\delta'^2\langle\hat{\nu}^2\rangle}{2m}-\sum_{n'\ne n}\frac{\lvert\delta'\cdot\mathbf{C}_{n'n}\rvert^2}{\varepsilon_{n'}-\varepsilon_n}, \tag*{(9)}

with Cn′n≡⟨un′∣12{v^,ν^}∣un⟩C_{n'n}\equiv\langle u_{n'}|\frac{1}{2}\{\hat{\mathbf{v}},\hat{\nu}\}|u_n\rangle and Cn=Cnn=⟨ν^⟩vn+Cov⁡(v^,ν^)\mathbf{C}_n=\mathbf{C}_{nn}=\langle\hat{\nu}\rangle\mathbf{v}_n+\operatorname{Cov}(\hat{\mathbf{v}},\hat{\nu}), vn=∂kεn\mathbf{v}_n=\partial_{\mathbf{k}}\varepsilon_n the band velocity. Equation (9) holds if ∣δ′⋅Cn′n∣≪∣εn′−εn∣\lvert\delta'\cdot\mathbf{C}_{n'n}\rvert\ll\lvert\varepsilon_{n'}-\varepsilon_n\rvert; otherwise the bands must be diagonalized, which we consider in Eq. (11). The first-order term is the Hellmann-Feynman shift produced by moving every plane-wave component of the approximant’s Bloch state from k+Q~\mathbf{k}+\widetilde{\mathbf{Q}} to its true position k+Q~+νδ′\mathbf{k}+\widetilde{\mathbf{Q}}+\nu\delta'. At the band edge, the state is an equal superposition of two plane waves whose internal indices differ by n⊥n_\perp, for a module wavevector Q=n⊥β+G\mathbf{Q}=n_\perp\boldsymbol{\beta}+\mathbf{G}. Thus the band velocity, and in turn a⋅vn\mathbf{a}\cdot\mathbf{v}_n are both 0. The covariance term δ′⋅Cov⁡(v^,ν^)\delta'\cdot\operatorname{Cov}(\hat{\mathbf{v}},\hat{\nu}) is n⊥δ′⋅Q~/4mn_\perp\delta'\cdot\widetilde{\mathbf{Q}}/4m with Q~=Q−n⊥δ′\widetilde{\mathbf{Q}}=\mathbf{Q}-n_\perp\delta'. For a simple (ungrounded) approximant, the wavevector is Q~\widetilde{\mathbf{Q}}, so the Bragg crossing is predicted to be Q~/2=Q/2−n⊥δ′/2\widetilde{\mathbf{Q}}/2=\mathbf{Q}/2-n_\perp\delta'/2 instead of Q/2\mathbf{Q}/2. As a⋅vn=0\mathbf{a}\cdot\mathbf{v}_n=0, retaining the Berry connection does not improve the Bragg crossing.

Eigenstates, quantum numbers, band structure.—When the wrap-around potential is negligible, the local band structure does not depend on the phason, so the equations of motion of Eq. (7), En(k+a)+δεn+Ug=E\mathcal{E}_n(\mathbf{k}+\mathbf{a})+\delta\varepsilon_n+U_g=E and ∇⋅(A2vn)=0\nabla\cdot(\mathcal{A}^2\mathbf{v}_n)=0, have constant solutions. For a volume V\mathcal{V}, the eigenstate is then

ψnk=1V∑Q~∈Λ∗unk,Q~e2πi(Q~ω)/cei(k+Q~+ν(Q~)δ′)⋅r.(10)\psi_{n\mathbf{k}}=\frac{1}{\sqrt{\mathcal{V}}}\sum_{\widetilde{\mathbf{Q}}\in\Lambda^*}u_{n\mathbf{k},\widetilde{\mathbf{Q}}}e^{2\pi i(\widetilde{\mathbf{Q}}\omega)/c}e^{i(\mathbf{k}+\widetilde{\mathbf{Q}}+\nu(\widetilde{\mathbf{Q}})\delta')\cdot\mathbf{r}}. \tag*{(10)}

Its wavevectors k+Q~+νδ′=k+nβ+G\mathbf{k}+\widetilde{\mathbf{Q}}+\nu\delta'=\mathbf{k}+n\boldsymbol{\beta}+\mathbf{G} are exactly the module points k+M\mathbf{k}+\mathcal{M}. Equation (10) is therefore simply the plane-wave superposition solution from KAM theory for the extended states of a weak quasiperiodic potential [34, 35], with the approximant supplying coefficients. As the approximant’s state is periodic, only the phases eiνδ′⋅re^{i\nu\delta'\cdot\mathbf{r}} make Eq. (10) quasiperiodic, so the corresponding momentum observable by photoemission would be at ℏ(k+Q)\hbar(\mathbf{k}+\mathbf{Q}) rather than ℏ(k+Q~)\hbar(\mathbf{k}+\widetilde{\mathbf{Q}}). The quantum numbers are the supercell band index nn and Bloch wavevector k\mathbf{k}. We quickly show that the quasicrystal band structure is correctly periodic under module M\mathcal{M}. The approximant identifies (n,k)∼(n,k+Q~)(n,k)\sim(n,k+\widetilde{Q}) as the same state, as un,k+Q~=e−iQ~⋅runku_{n,k+\widetilde{Q}}=e^{-i\widetilde{Q}\cdot r}u_{nk}, but e−iQ~⋅re^{-i\widetilde{Q}\cdot r} also shifts the internal indices by ν(Q~)\nu(\widetilde{Q}), and by Eq. (5) induces a slow phase on uu, so kk shifts by ν(Q~)δ′\nu(\widetilde{Q})\delta'. Thus, (n,k)∼(n,k+Q~;ν→ν+ν(Q~))∼(n,k+Q)(n,k)\sim(n,k+\widetilde{Q};\nu\to\nu+\nu(\widetilde{Q}))\sim(n,k+Q) with Q∈MQ\in\mathcal{M}. We now consider band gaps from a kk-space viewpoint. In one dimension, a gap corresponding to a module wavevector n⊥β+2πm/an_\perp\beta+2\pi m/a can be identified by the internal and lattice harmonics, (n⊥,m)(n_\perp,m), and by the gap-labeling theorem [14, 15, 36], the integrated density of states converges correctly regardless of the choice of approximant ∣n⊥p+mq∣/qa→∣n⊥ξ+m∣/a|n_\perp p+mq|/qa\to|n_\perp\xi+m|/a. Equation (9) moves only the positions of the gaps, not their labels. Finally, we have assumed each gap is resolved by a single energy band. This condition requires the gap to be wider than the maximum possible phason-induced displacement, Δgap≫∣n⊥∣∣δ′∣∣v∣\Delta_{\mathrm{gap}}\gg|n_\perp||\delta'||v|, with v=Q~/2mv=\widetilde{Q}/2m the velocity at the Bragg plane. As Δgap\Delta_{\mathrm{gap}} shrinks geometrically with ∣n⊥∣|n_\perp|, any finite size approximant can resolve at most finitely many gaps, and will necessarily miss the Cantor-set-like fine structure.

All orders.—Thus far, we have treated Eq. (9) as a perturbative expansion. We now show it is an expansion of the supercell window Hamiltonian. Consider the subspace Sk\mathcal{S}_k spanned by the window module wavevectors k+Q~+ν(Q~)δ′k+\widetilde{Q}+\nu(\widetilde{Q})\delta'. The interior potential consists of the ℓ=0\ell=0 terms of Eq. (4), so

HQ∣Sk=Uw0H~kUw0∗,H~k=H0(k)+δ′⋅{p^,ν^}+δ′2ν^22m,(11)H_Q\big|_{\mathcal{S}_k}=U_{w_0}\widetilde{H}_kU_{w_0}^*,\qquad\widetilde{H}_k=H_0(k)+\frac{\delta'\cdot\{\hat{p},\hat{\nu}\}+\delta'^2\hat{\nu}^2}{2m}, \tag*{(11)}

exactly, with HQ=−∇2/2m+VQH_Q=-\nabla^2/2m+V_Q the true Hamiltonian and H0(k)H_0(k) the approximant Hamiltonian at λ=0\lambda=0. The FGM action is the perturbation theory of this window Hamiltonian HQ∣SkH_Q\big|_{\mathcal{S}_k}, with H~k\widetilde{H}_k quadratic in the strain. But H~k\widetilde{H}_k is just the approximant’s plane-wave Hamiltonian with each kinetic energy moved to (k+Q~+νδ′)2/2m(k+\widetilde{Q}+\nu\delta')^2/2m, as the wrap-around potential does not act within Sk\mathcal{S}_k. To solve for the true eigenstates, we instead diagonalize H~k\widetilde{H}_k and thus HQ∣SkH_Q\big|_{\mathcal{S}_k} to all orders of δ′\delta', with the only remaining error due to the window Sk\mathcal{S}_k. Diagonalizing H~k\widetilde{H}_k is no more difficult than diagonalizing the approximant’s Hamiltonian, and allows computation of eigenstates by inserting the eigenvectors into Eq. (10). The spectrum is also correctly phason-invariant, as ω0\omega_0 enters only through the unitarity Uw0U_{w_0}. Now consider the error due to Sk\mathcal{S}_k. Assuming an extended state, such that the coefficients decay with the internal index on a scale ζ\zeta, the error is of order e−q/ζe^{-q/\zeta}. Physically, the local band structure repeats after a phason advance of c/qc/q, Eq. (3). The local Bloch state therefore has degenerate copies of uu with internal indices shifted by jqjq, carrying slow momentum jqδ′jq\delta', by Eq. (5). These copies place the plane waves at Q~+(ν+jq)δ′\widetilde{Q}+(\nu+jq)\delta', the open points of Figure 1(b), and since ∣jqδ′∣eff|jq\delta'|_{\mathrm{eff}} they never separate from the state itself. The wrap-around potential is simply the coupling to these degenerate copies of uu: its ℓ\ellth term carries phase e2πiℓqλ/ce^{2\pi i\ell q\lambda/c}, which at λ(r)\lambda(r) becomes the slow phase eiℓqδ′⋅re^{i\ell q\delta'\cdot r}, scattering the slow momentum by ℓqδ′\ell q\delta'. For a window Hamiltonian (2J+1)q(2J+1)q wide, we can couple copies with ∣j∣≤J|j|\le J. Therefore enlarging the supercell window is equivalent to coupling degenerate local Bloch states. We now describe ways to compute extended, localized, and critical states. Critical states are neither extended nor localized, decaying as a power law rather than exponentially. For an extended state, a small finite supercell suffices, with a weight on the jjth Bloch state proportional to e−(2∣j∣−1)q/2ζe^{-(2|j|-1)q/2\zeta}. For a localized state, coefficients are spread uniformly over the internal index, so no finite window captures it [Fig. 2(b)], though in real space a supercell that includes the environment is sufficient. For a critical state, we require sufficient resolution in both ν\nu and rr, so those states are not directly computable. A possible indirect approach would treat critical states as fixed points of the process of enlarging the supercell window; we return to this point in the discussion.

Numerical tests.— We test the theory in one dimension under a bichromatic potential (Figure 2), which is an example continuum model of cold atoms in two incommensurate optical lattices [37, 38, 39]. At weak coupling, the approximant’s band-edge error is n⊥δ′Q~/4mn_{\perp}\delta' \tilde{Q}/4m, decreasing as ∣δ′∣∝1/q2|\delta'| \propto1/q^2; after first-order correction [Eq. (9)] the residual is ∝∣δ′∣2\propto|\delta'|^2. The Berry connection component a⋅v\mathbf{a}\cdot\mathbf{v} does not affect the edges (to 10−610^{-6}); and finally the window Hamiltonian, at its own extremum over kk, is accurate to better than 10−1210^{-12} for all three gaps from q=13q=13 on. Here, we use the window Hamiltonian of the 233/377233/377 approximant as the reference solution, whose only error is its own leakage. One might worry that we are merely comparing the window Hamiltonian with itself. To rule this out, we also computed the same band edges by diagonalizing HQH_Q directly on the module points k+2π(m+nξ)/ak+2\pi(m+n\xi)/a with ∣n∣≤40|n|\leq40 and kinetic energy below 400ℏ2/2ma2400\hbar^2/2ma^2, about 530 plane waves, with no approximant and no folding; the two references agree to 2×10−132\times10^{-13}, the numerical floor of Figure 2(a). Checking Eq. (3), the approximant’s energy spectra at phason phase φ=2πw0/c\varphi=2\pi w_0/c and φ+2π/q\varphi+2\pi/q agree to 10−1310^{-13}, with sub-period variation decaying as e−1.8qe^{-1.8q}. At strong coupling, the same potential can host both extended and localized states: high-level states have Δν≃0.5\Delta\nu\simeq0.5 and behave as if at weak coupling, while low-lying states are localized in a few wells of the potential, with Δν≃0.29q≃q/12\Delta\nu\simeq0.29q\simeq q/\sqrt{12}, and are not captured due to state leakage [Fig. 2(b)]. Critical states occur in this potential only at a mobility edge, which would require careful tuning [40, 41], so we do not consider them.

Figure 2. Numerical test on the one-dimensional potential 2W0cos⁡(2πx/a)+2W1cos⁡(2πξx/a+φ)2W_0\cos(2\pi x/a)+2W_1\cos(2\pi\xi x/a+\varphi), with ξ\xi the inverse golden mean and p/qp/q its Fibonacci approximants. (a) Weak coupling, W0=0.3W_0=0.3 and W1=0.5W_1=0.5 in ℏ2/2ma2\hbar^2/2ma^2 units: lower edge error of three gaps, relative to the 233/377233/377 window Hamiltonian, as computed from the approximant (filled blue), the first-order correction of Eq. (9) (open orange), and the window Hamiltonian of Eq. (11) (green). The dashed and dotted lines are proportional to ∣δ′∣|\delta'| and ∣δ′∣2|\delta'|^2. The window Hamiltonian is exact except where the state has nonnegligible weight at the window edge, as seen at q=5,8q=5,8. Inset: predicted gap (1,0)(1,0) positions by the approximant and window Hamiltonian at Q~/2\widetilde{Q}/2 and Q/2Q/2 respectively. (b) Leakage sjs_j, the weight of a state in the shell qj/2<∣ν∣≤qj+1/2q_j/2<|\nu|\le q_{j+1}/2 that the approximant qjq_j omits and the next convergent retains. Extended states (gray) reach the numerical floor within two convergents. The two low-lying states at strong coupling, W0=3W_0=3 and W1=5W_1=5 (purple), are localized in a few wells of the potential, and their leakage never decreases; the purple dotted line, 1−1/τ1-1/\tau with τ\tau the golden ratio, is the leakage of a state spread uniformly over the window.

Discussion.— We have shown that a quasicrystal can be represented exactly as a phason-graded periodic approximant [Eq. (2)], establishing a connection to functionally graded materials. This viewpoint provides a controlled framework for calculating quasicrystal electronic structure, including band energies, quantum numbers, and quasiperiodic eigenstates, while identifying which spectral features can be reliably resolved by finite approximations.

In particular, our framework yields the correct wavevector equivalence (n,k)∼(n,k+Q)(n,k)\sim(n,k+Q), reconstructs quasiperiodic eigenstates [Eq. (10)], and corrects Bragg crossing wavevectors by n⊥δ′/2n_{\perp}\delta'/2, predicting photoemission momenta at ℏ(k+Q)\hbar(k+Q) rather than ℏ(k+Q~)\hbar(k+\widetilde{Q}). The resulting window Hamiltonian gives band energies to all orders in phason strain within a finite Fourier window, while preserving spectral invariance under phason translations. Furthermore, the quantum-metric control parameter establishes when a finite approximation is reliable and identifies unresolved fine spectral structures with Δgap≫∣n⊥δ′′v∣\Delta_{\mathrm{gap}}\gg|n_{\perp}\delta''\mathbf{v}|. These advances provide a foundation for predicting not only the electronic structure of quasicrystals, but also their spectroscopic, optical, and transport properties through extensions of existing response theories.

Our framework also suggests new opportunities for understanding and engineering quasiperiodic matter. In particular, the explicit connection between phason modulation and electronic structure offers a route to exploring how spatial variations in quasiperiodic order influence electronic properties, potentially enabling the design of functional quasicrystalline structures. For localized and critical states, where the control parameter becomes large, further developments are required. We conjecture that critical states may be understood as fixed points of a renormalization procedure involving successively larger Fourier windows. Another important direction is extending the theory to metallic quasicrystals, where discontinuous atomic rearrangements under phason variation challenge the present smooth-potential assumption.

Finally, we propose a direct experimental test connecting phason strain to electronic structure. The phason strain can be measured through the displacement νδ′\nu\delta' of diffraction peaks, while our theory predicts a corresponding Bragg-crossing energy shift of n⊥δ′⋅v/2n_{\perp}\delta'\cdot\mathbf{v}/2, with v=Q~/2m\mathbf{v}=\widetilde{Q}/2m, relative to the periodic approximant. This relation could be tested by combining diffraction with electronic spectroscopy in quasicrystals [42], or through analogous measurements in detuned bichromatic optical lattices and moiré quasicrystals [43, 44, 45]. More broadly, our work extends the computational framework of functionally graded materials to quasiperiodic matter, opening possibilities for predictive modeling and ultimately the design of quasicrystals with tailored quantum properties.

Acknowledgments

The authors acknowledge support from the National Science Foundation (NSF) Convergence Accelerator Award No. 2345084 and the U.S. Department of Energy (DOE), Basic Energy Sciences (BES), Award No. DE-SC0020148. D.C.C acknowledges support from the Jack Kent Cooke Foundation and The Paul and Daisy Soros Fellowships for New Americans. M. L. acknowledges support from R. Wachnik.

References

† [email protected]

References

  1. [1]D. Shechtman, I. Blech, D. Gratias, and J. W. Cahn, Physical Review Letters 53, 1951 (1984).
  2. [2]D. Levine and P. J. Steinhardt, Physical Review Letters 53, 2477 (1984).
  3. [3]J. Dubois, Chemical Society Reviews 41, 6760 (2012).
  4. [4]Y. Takagiwa, Materials 15, 6816 (2022).
  5. [5]Z. V. Vardeny, A. Nahata, and A. Agrawal, Nature Photonics 7, 177 (2013).
  6. [6]M. Niino and S. Maeda, ISIJ International 30, 699 (1990).
  7. [7]K. Cai, J. Sun, H. Wang, R. Bao, Z. Guo, W. Zhang, L. Zhao, X. Li, R. Cao, X. Yun, and J. Zhao, Progress in Solid State Chemistry 74, 100454 (2024).
  8. [8]A. Sola, D. Bellucci, and V. Cannillo, Biotechnology Advances 34, 504 (2016).
  9. [9]V. Elser and C. L. Henley, Physical Review Letters 55, 2883 (1985).
  10. [10]A. I. Goldman and R. F. Kelton, Reviews of Modern Physics 65, 213 (1993).
  11. [11]T. Fujiwara, Physical Review B 40, 942 (1989).
  12. [12]K. J. U. Vidarte and C. Lewenkopf, Frontiers in Carbon 3, 1496179 (2024).
  13. [13]B. Davies and C. Thalhammer, Proceedings of the Royal Society A 481, 20240894 (2025).
  14. [14]R. Johnson and J. M. Moser, Communications in Mathematical Physics 84, 403 (1982).
  15. [15]J. Bellissard, in From Number Theory to Physics, edited by M. Waldschmidt, P. Moussa, J.-M. Luck, and C. Itzykson (Springer Berlin Heidelberg, Berlin, 1992) pp. 538–630.
  16. [16]A. Jagannathan, Reviews of Modern Physics 93, 045001 (2021).
  17. [17]J. Chok, M. J. Colbrook, M. Embree, and J. Fillman, Physical Review Letters 137, 176403 (2026).
  18. [18]P. M. de Wolff, Acta Crystallographica Section A 30, 777 (1974).
  19. [19]A. Janner and T. Janssen, Physical Review B 15, 643 (1977).
  20. [20]P. A. Kalugin, A. Y. Kitayev, and L. S. Levitov, Journal de Physique Lettres 46, 601 (1985).
  21. [21]M. Duneau and A. Katz, Physical Review Letters 54, 2688 (1985).
  22. [22]M. J. Landry, R. Okabe, C. Fu, and M. Li, Quantum Theory of Functionally Graded Materials (2026), arXiv:2603.03424v1, arXiv:2603.03424 [cond-mat.mtrl-sci].arxiv.org/abs/2603.03424
  23. [23]We consider the single phason model for simplicity, given
  24. [24]R. Lifshitz, Foundations of Physics 33, 1703 (2003).
  25. [25]P. Bak, Physical Review Letters 54, 1517 (1985).
  26. [26]M. Baggioli and M. Landry, SciPost Phys. 9, 062 (2020), arXiv:2008.05339 [hep-th].DOI
  27. [27]T. C. Lubensky, J. E. Socolar, P. J. Steinhardt, P. A. Bancel, and P. A. Heiney, Physical Review Letters 57, 1440 (1986).
  28. [28]Y. Ishii, Physical Review B 45, 5228 (1992).
  29. [29]G. H. Hardy and E. M. Wright, An Introduction To The Theory Of Numbers (Oxford University Press, Oxford, 2008).
  30. [30]A. Y. Khinchin, Continued Fractions (Dover Publications, New York, 1997).
  31. [31]A. Hurwitz, Mathematische Annalen 39, 279 (1891).
  32. [32]L. S. Levitov, Europhysics Letters (EPL) 6, 517 (1988).
  33. [33]G. Sundaram and Q. Niu, Physical Review B 59, 14915 (1999).
  34. [34]E. I. Dinaburg and Y. G. Sinai, Functional Analysis and Its Applications 9, 279 (1975).
  35. [35]L. H. Eliasson, Communications in Mathematical Physics 146, 447 (1992).
  36. [36]J. Avron and B. Simon, Duke Mathematical Journal 50, 369 (1983).
  37. [37]G. Roati, C. D’Errico, L. Fallani, M. Fattori, C. Fort, M. Zaccanti, G. Modugno, M. Modugno, and M. Inguscio, Nature 453, 895 (2008).
  38. [38]M. Modugno, New Journal of Physics 11, 033023 (2009).
  39. [39]H. P. Lüschen, S. Scherg, T. Kohler, M. Schreiber, P. Bordia, X. Li, S. Das Sarma, and I. Bloch, Physical Review Letters 120, 160401 (2018).
  40. [40]S. Aubry and G. André, Ann. Israel Phys. Soc. 3, 133 (1980).
  41. [41]J. Biddle and S. Das Sarma, Physical Review Letters 104, 070601 (2010).
  42. [42]K. Edagawa, P. Mandal, K. Hosono, K. Suzuki, and S. Takeuchi, Physical Review B 70, 184202 (2004).
  43. [43]S. J. Ahn, P. Moon, T.-H. Kim, H.-W. Kim, H.-C. Shin, E. H. Kim, H. W. Cha, S.-J. Kahng, P. Kim, M. Koshino, Y.-W. Son, C.-W. Yang, and J. R. Ahn, Science 361, 782 (2018).
  44. [44]W. Yao, E. Wang, C. Bao, Y. Zhang, K. Zhang, K. Bao, C. K. Chan, C. Chen, J. Avila, M. C. Asensio, J. Zhu, and S. Zhou, Proceedings of the National Academy of Sciences 115, 6928 (2018).
  45. [45]A. Uri, S. C. de la Barrera, M. T. Randeria, D. Rodan-Legrain, T. Devakul, P. J. D. Crowley, N. Paul, K. Watanabe, T. Taniguchi, R. Lifshitz, L. Fu, R. C. Ashoori, and P. Jarillo-Herrero, Nature 620, 762 (2023).

Paper details

vlnc:W_8k72ys2g
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