Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

A model reduction method for solving the eigenvalue problem of semiclassical random Schr\"odinger operators

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A POD-based multiscale finite element method computes the expected minimal eigenvalue of a semiclassical random Schrödinger operator with root-mean-square error $O(H^6 + \rho + s^{-2/p+1} + N^{-\alpha})$, and the expected value of a…

desk verdict The method is practical and the numerics are good, but Lemma 5.4 uses a false scaling for the multiscale basis, so the central error bound is unproven. read the letter →

arxiv 2502.07574 v1 pith:UM3E7N5A submitted 2025-02-11 math.NA cs.NA

Panchi Li , Zhiwen Zhang This is my paper · ORCID
classification math.NAcs.NA MSC 35J1065N2565D3065N3081Q05
keywords eigenvalueproblemsemiclassicalrandomSchrödingeroperatorproperorthogonaldecompositionmultiscalemodelreductionquasi-MonteCarloconvergenceanalysisAndersonlocalization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper builds a reduced-order numerical method for the eigenvalue problem of a semiclassical Schrödinger operator whose random potential is written as an infinite series of independent uniform random variables. After truncating the series to $s$ terms, the method solves the resulting parametric eigenvalue problem on a coarse mesh using multiscale finite element basis functions, then compresses those basis functions with a proper orthogonal decomposition so that constructing the basis for a new random sample costs only a few POD coefficients, independent of the fine mesh. Statistics of the eigenvalue and ground state are collected with a quasi-Monte Carlo lattice rule. The central result is the combined root-mean-square error bound of Theorem 5.3: $O(H^6 + \rho + s^{-2/p+1} + N^{-\alpha})$ for the expected minimal eigenvalue and $O(H^3 + \sqrt{\rho} + s^{-2/p+1} + N^{-\alpha})$ for a linear functional of the ground state, where $H$ is the coarse mesh size, $\rho$ the POD tolerance, $s$ the truncation dimension, and $N$ the number of qMC samples. If correct, this means high-accuracy eigenvalue statistics can be obtained with online degrees of freedom equal to the coarse-mesh nodes rather than the fine-mesh resolution.

What carries the argument

The load-bearing object is the constrained multiscale basis function $\phi_i$, defined as the minimizer of the energy $a(\phi,\phi) = \frac{\epsilon^2}{2}\int_D |\nabla\phi|^2 dx + \int_D v_0 \phi^2 dx$ subject to $\int_D \phi \phi_j^H dx = \alpha \delta_{ij}$ for every coarse finite element basis function $\phi_j^H$, with $\alpha = (1,\phi_j^H)$ chosen to remove mesh-size dependence. The POD stage replaces each random snapshot $\phi_i(x,\omega_j)$ by its mean $\zeta_i^0$ plus a few fluctuation modes $\zeta_i^1,\dots,\zeta_i^{m_i}$; the orthogonality identity $(\zeta_i^k,\phi_j^H)=0$ for $k\ge 1$ means that only the single constraint for $j=i$ survives in the reduced optimal problem (3.15), so constructing $\hat\phi_i(\omega)$ online costs only $m_i+1$ POD coefficients. This reduction preserves the MsFEM orthogonality—the multiscale space is the $a$-orthogonal complement of the kernel of the coarse $L^2$ projection—which is the mechanism behind the $H^6$ eigenvalue superconvergence in Theorem 5.1.

What would settle it

Fix a coarse mesh $H$ and a smooth 1D potential, compute the constrained multiscale basis function $\phi_i$ on fine meshes $h = 1/64, 1/128, 1/256$, and measure $\|\phi_i\|_{L^2}$; the proof of Lemma 5.4 requires this norm to shrink like $h$, while the constraint $\int_D \phi_i \phi_j^H dx = \alpha$ suggests it should stay at a scale set by $H$. If the norm does not shrink with $h$, the step leading to (5.41) is false and the claimed $\sqrt{\rho}$ POD error for $\omega$ outside the snapshot set is unsupported. A complementary check is to fix $H$, $\rho$, and $m_i$, take $\omega$ outside the snapshot set, and observe whether $\|\phi_i(\omega)-\hat\phi_i(\omega)\|$ stays bounded as $h\to 0$ as the theorem predicts or grows as the alternative scaling would predict.

Watch

Extended reading notes

Core claim

The paper claims that the combined MsFEM-POD-qMC approximation achieves the superconvergent error rates stated in Theorem 5.3: under Assumption 3.1 (uniformly bounded potentials, $\ell^p$-summability of the component norms, and $W^{1,\infty}$ regularity), the root-mean-square error of the expected minimal eigenvalue is bounded by $C(H^6 + \rho + s^{-2/p+1} + N^{-\alpha})$, and the root-mean-square error of the expected value of any continuous linear functional of the ground state is bounded by $C(H^3 + \sqrt{\rho} + s^{-2/p+1} + N^{-\alpha})$, with $\alpha = \min\{1-\delta, 1/p - 1/2\}$. The proof splits the total error into four additive parts: the MsFEM discretization error, whose key deterministic estimate is $|\lambda_{\mathrm{ms}} - \lambda| \le C H^6$ (Theorem 5.1); the dimension-truncation error $s^{-2/p+1}$; the qMC integration error $N^{-\alpha}$; and a POD error contribution $\sqrt{\rho}$ (or $\rho$ for the eigenvalue) arising from the low-dimensional approximation of the random multiscale basis functions. Numerical experiments in Section 6 confirm the predicted rates and show that the method reproduces Anderson-localized eigenfunctions for white-noise-like potentials when the coarse mesh satisfies $H < \epsilon$.

Load-bearing premise

The proof that the reduced basis works for random potentials not included in the offline snapshot set assumes that each constrained multiscale basis function has $L^2$ norm of order $h^d$, where $h$ is the fine mesh size; if the constraint $\int_D \phi_i \phi_j^H dx = \alpha$ actually sets the scale through the coarse mesh $H$, then the perturbation term in the argument cannot be made uniformly smaller than the POD tolerance $\rho$, and the POD error bound for such potentials would fail.

Editorial extensions

If this is right

  • The expected minimal eigenvalue of the random Schrödinger operator can be approximated to sixth order in the coarse mesh size $H$ while the online linear system has only $N_H$ (the number of coarse nodes) degrees of freedom; fine-mesh work is confined to the offline POD snapshot generation.
  • For a fixed POD tolerance $\rho$ and truncation dimension $s$, increasing the number of qMC samples $N$ reduces the error at rate $N^{-\alpha}$, which is almost first order when the potential components decay fast enough ($p \le 2/3$), compared with $N^{-1/2}$ for Monte Carlo.
  • Because the total error is additive in $H^6$, $\rho$, $s^{-2/p+1}$, and $N^{-\alpha}$, a user can balance the four contributions independently when choosing the coarse mesh, POD rank, truncation dimension, and sample count.
  • For random potentials with decaying high-frequency components ($q>1$ in the examples), the method needs no constraint linking $H$ to the semiclassical parameter $\epsilon$; for the $q=0$ white-noise limit, the experiments indicate that $H<\epsilon$ suffices to resolve localized eigenfunctions.
  • The numerical experiments show the method reproducing Anderson-localized eigenfunctions in 1D and 2D at a fraction of the fine-FEM cost, with 29.72 seconds versus 341.17 seconds in the reported 1D comparison.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the reduction acts on the basis functions rather than on the eigenpairs, the same POD-MsFEM construction should extend to higher eigenpairs and to nonlinear eigenvalue problems such as Gross-Pitaevskii ground states, provided the uniform spectral-gap assumption used in the parametric regularity analysis holds for the eigenvalue of interest.
  • The asymmetry in the error bound (linear $\rho$ for the eigenvalue, $\sqrt{\rho}$ for the eigenfunction) suggests that a looser POD tolerance can be used when only eigenvalue statistics are wanted; one could adaptively choose $m_i$ per coarse node instead of the fixed $m_i=3$ used in the experiments.
  • The $q=0$ white-noise experiments fall outside the assumptions of the theorem (Assumption 3.1(2)-(3) fail), so the paper's claim that $H<\epsilon$ suffices there is better read as a separate conjecture; a systematic study as $s\to\infty$ with $H$ scaled by $\epsilon$ would test it.
  • If the $H^6$ rate is generic for constrained multiscale spaces, the same POD reduction could be applied to other self-adjoint elliptic eigenvalue problems with affine parametric coefficients, not only Schrödinger operators.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a multiscale finite element method combined with proper orthogonal decomposition (POD) and quasi-Monte Carlo (qMC) sampling for computing statistics of the minimal eigenvalue and ground state of semiclassical random Schrödinger operators. After truncating the stochastic dimension, an MsFEM approximation is used in the physical domain, POD is used to reduce the cost of constructing the multiscale basis, and randomized lattice rules are used for the stochastic expectation. The main theoretical results are Theorems 5.1–5.3, which claim superconvergent spatial error rates (H^6 for the eigenvalue, H^3 in H1 and H^4 in L2 for the eigenfunction) plus POD and qMC error terms. Numerical experiments validate the deterministic and stochastic convergence rates and explore Anderson localization for spatially random potentials.

Significance. If correct, the combined estimate would be attractive: it would provide a UQ method for random Schrödinger eigenvalue problems with superconvergent spatial accuracy and a POD offline stage whose dimension is independent of the fine mesh. The deterministic MsFEM superconvergence part (Theorem 5.1) is plausible and consistent with earlier LOD/MsFEM results, and the truncation and qMC estimates are carefully imported from Gilbert et al. [15]. However, the new POD error analysis is the load-bearing ingredient for the random case, and it contains a false scaling estimate and a projection-coefficient identification that are not justified. Since Theorems 5.2 and 5.3 inherit these errors, the central claim of the paper is not established. The numerical section also does not report the POD tolerance rho or systematically test online parameters away from the snapshots, so it does not fill the gap.

major comments (4)
  1. [§5.4, Lemma 5.4 and Eq. (5.41)] The uniform bound in Lemma 5.4 for omega outside the snapshot set relies on the assertion, immediately before Eq. (5.41), that ||phi_i(x,omega_j)|| <= C h^d. This scaling is inconsistent with the constraint (3.5), which requires integral phi_i phi_j^H dx = alpha = (1, phi_j^H) = O(H^d). Since phi_i has O(1) amplitude on a patch of diameter O(H), its L2 norm is O(H^{d/2}), independent of the fine-mesh size h. With the correct O(H^{d/2}) scaling, the first term in (5.41) becomes C ||omega - omega_j||_inf H^{d/2}, and because a finite snapshot set Omega_0 cannot cover the continuum in the sup norm, this term cannot be made uniformly smaller than sqrt(rho). Consequently the bound ||phi_i(x,omega) - hat phi_i(x,omega)|| <= C sqrt(rho) in Lemma 5.4, and the sqrt(rho) terms in Theorems 5.2 and 5.3, are not established.
  2. [§5.4, proof of Lemma 5.4 for omega in Omega_0] For omega in the snapshot set Omega_0, the proof sets c_j^i(omega) = (tilde phi_i, zeta_j^i), i.e., it identifies the coefficients of the solution of the reduced constrained energy problem (3.15) with the coefficients of the L2 projection onto the POD subspace. Equation (5.35) bounds only the L2 projection error; it does not bound the error of the energy minimizer of (3.15). The proof therefore does not justify the claimed equality ||phi_i(x,omega) - hat phi_i(x,omega)|| = ||tilde phi_i(x,omega) - sum_j c_j^i(omega) zeta_j^i(x)|| nor the resulting sqrt(rho) bound.
  3. [§5.4, Eqs. (5.42) and (5.47)] The passage from the L2 bound of Lemma 5.4 to the H1 error is not sound. Equation (5.42) states ||nabla phi_i - nabla hat phi_i|| <= C sqrt(rho)/h^2 with no justification; an inverse inequality would give an h^{-1} factor at best, and in any case the stated h^{-2} factor means the bound cannot be uniform as h -> 0. The subsequent line 'Therefore, there exists a constant C such that (5.47)' does not follow from a bound whose right-hand side depends explicitly on h. Since Theorem 5.2 uses the uniform H1 error ||psi_pod_ms - psi_ms||_1 <= C sqrt(rho), this gap affects the eigenvalue and eigenfunction estimates for the random problem.
  4. [§6.2, Example 6.3 and POD error discussion] The numerical experiments do not report the POD tolerance rho, and the text explicitly states 'the POD error rho is not discussed here' before Example 6.4. Figure 3 tests one online point outside the snapshot set, but there is no systematic study of the error as a function of rho or as a function of the sup-norm distance from omega to the nearest snapshot. The numerics therefore do not supply the missing uniform control needed for Lemma 5.4 and Theorems 5.2 and 5.3.
minor comments (5)
  1. [Introduction, Eq. (1.9)] Equation (1.9) states the eigenvalue error bound with a sqrt(rho) term, while Theorem 5.3, Eq. (5.53), states the same bound with a rho term; these should be made consistent.
  2. [§3.3, Eq. (3.15)] The reduced optimal problem in Eq. (3.15) uses the bilinear form a(.,.) without specifying whether this is the deterministic form from (3.3) or the random/truncated form A_s(omega;.,.); the notation should be aligned with the random setting used in Algorithm 2.
  3. [§5.4, Eq. (5.34)] The reference to 'Proposition 1 [23]' should be made precise, and the left-hand side of Eq. (5.34) should use the fluctuation tilde phi_i consistently with Eq. (5.35).
  4. [§6.3, Remark 6.1] Remark 6.1 correctly notes that for q = 0 the random potentials do not satisfy Assumption 3.1(2)–(3), so the convergence theory does not apply to Example 6.6. This is a useful caveat, but the Anderson-localization simulations in Figures 9 and 10 should be presented as numerical demonstrations outside the theorem rather than as validations of the theoretical error estimates.
  5. [§5.4, Eq. (5.42)] The notation in Eq. (5.42) mixes the constant C and the mesh size h without indicating which quantities are kept fixed; the constant should be tracked explicitly through the inverse inequality to make the h-dependence clear.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central error bound combines an internally proved MsFEM estimate with external qMC/truncation results and a POD residual bound; the only self-citations are minor and not load-bearing.

full rationale

The derivation chain for Theorem 5.3 is a sum of four independent contributions. First, the MsFEM superconvergence bounds in Theorem 5.1 are proved in Section 5.1 using the orthogonal splitting (5.18)-(5.19) and the duality argument of Lemma 5.1; Remark 5.1 cites earlier work only as an alternative route, not as the proof. Second, the dimension-truncation and qMC error bounds, Propositions 5.1 and 5.2, are imported from Gilbert et al. [15], an external source, and the parametric regularity needed for those bounds is proved in the present paper as Lemma 4.1. Third, the POD contribution is an estimate of the new basis error in terms of the user-specified POD tolerance rho via (3.10)-(3.11); this is a standard residual bound, not a fitted parameter later renamed as a prediction. Fourth, Theorem 5.2 combines the MsFEM and POD estimates by the triangle inequality. The only self-citations are to the authors' earlier work for the alpha-normalization in (3.5), attributed to [26], and for an alternative superconvergence proof, [27]; neither is load-bearing, since the alpha-constraint is explicitly defined in the text and Theorem 5.1 is proved in this paper. The proof of Lemma 5.4 does contain a questionable scaling assertion, ||phi_i(x,omega_j)|| <= C h^d, and it identifies the reduced optimal solution with the L2 projection, but these are correctness risks in the POD error analysis rather than cases where a prediction reduces by construction to an input. Therefore the paper's central claim is not circular; the score reflects only the presence of minor self-citations that do not carry the argument.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The algorithm has hand-chosen numerical parameters (POD rank, snapshot count), and the error analysis relies on boundedness and summability assumptions plus an unproved scaling assertion in the POD error lemma.

free parameters (2)
  • POD basis rank m_i per coarse node = 3 (set in Section 6)
    The number of POD modes per coarse node is fixed to 3 in all experiments; the corresponding POD tolerance rho is never reported, so the rho term in Theorem 5.3 is not validated.
  • Offline snapshot count Q = 200 (chosen after Table 4)
    Q is selected by trial from Table 4 as a trade-off; it does not enter the final error theorem, but it determines which potentials are represented in the POD basis.
assumptions (5)
  • domain assumption The random potential V is uniformly bounded: Vmax >= V(x,omega) >= Vmin >= 0, V not identically zero, and H sqrt(Vmax)/epsilon <= 1.
    Assumption 3.1(1); gives uniform coercivity and boundedness of the bilinear form and well-posedness of the MsFEM optimal problems.
  • domain assumption The coefficient functions v_j satisfy sum_j ||v_j||_{L_inf}^p < infinity for some p in (0,1) and sum_j ||v_j||_{W^{1,inf}} < infinity.
    Assumption 3.1(2)-(3); used for dimension truncation and qMC convergence (Propositions 5.1 and 5.2) and derivative bounds (Lemma 4.1).
  • domain assumption The minimal eigenvalue lambda_1(omega) is separated by a uniform spectral gap C_gap from the rest of the spectrum.
    Lemma 2.1, quoted from Gilbert et al. Lemma 3.1; needed for derivative estimates and for eigenfunction error bounds.
  • standard math Standard Sobolev regularity and duality estimates for the deterministic Schrödinger operator, including elliptic regularity used in Lemma 5.1.
    Used in the H^6 superconvergence proof in Section 5.1.
  • ad hoc to paper For omega outside the snapshot set, the perturbation term C||omega - omega_j||_inf ||phi_i|| can be made O(sqrt(rho)) by taking h small.
    This is the unproved step in Lemma 5.4; it relies on the asserted scaling ||phi_i|| <= C h^d and is the weakest premise in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A model reduction method for solving the eigenvalue problem of semiclassical random Schr\"odinger operators." pith.science (2026). https://pith.science/paper/UM3E7N5A

@misc{pith2026250207574,
  author       = {Pith},
  title        = {Pith review of: A model reduction method for solving the eigenvalue problem of semiclassical random Schr\"odinger operators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UM3E7N5A}},
  note         = {Machine review of arXiv:2502.07574}
}
read the original abstract

In this paper, we compute the eigenvalue problem (EVP) for the semiclassical random Schr\"odinger operators, where the random potentials are parameterized by an infinite series of random variables. After truncating the series, we introduce the multiscale finite element method (MsFEM) to approximate the resulting parametric EVP. We then use the quasi-Monte Carlo (qMC) method to calculate empirical statistics within a finite-dimensional random space. Furthermore, using a set of low-dimensional proper orthogonal decomposition (POD) basis functions, the referred degrees of freedoms for constructing multiscale basis are independent of the spatial mesh. Given the bounded assumption on the random potentials, we then derive and prove an error estimate for the proposed method. Finally, we conduct numerical experiments to validate the error estimate. In addition, we investigate the localization of eigenfunctions for the Schr\"odinger operator with spatially random potentials. The results show that our method provides a practical and efficient solution for simulating complex quantum systems governed by semiclassical random Schr\"odinger operators.

Figures

Figures reproduced from arXiv: 2502.07574 by the authors.

Figure 1
Figure 1. Numerical convergence rates of the FEM and MsFEM approximation for the EVP of the Schr¨odinger operator with the 1D double-well potential. Furthermore, we check the approximation error of the MsFEM method for the first five eigen￾values and the corresponding eigenfunctions. Numerical results are depicted in [PITH_FULL_IMAGE:figures/full_fig_p020_1.png] view at source ↗
Figure 2
Figure 2. The checkboard potential and the numerical convergence rates of the FEM and MsFEM methods. The reference solutions are computed by E[λk] = 1 N X N i=1 λk(ω i ), E[ψk] = 1 N X N i=1 ψk(ω i ), where (λk, ψk) are the FEM solution on a fine mesh. The empirical expectations of numerical solutions (E[λms,k], E[ψms,k]) are calculated similarly. Since the convergence rate of eigenvalues will be mainly concerned, we define t… view at source ↗
Figure 3
Figure 3. The basis functions and the error between the multiscale basis solved by (3.4)-(3.5) and (3.15). 1st column: sketches of basis functions. 2nd column: the case for ω ∈ Ω0. 3rd column: the case for ω ∈ Ω/Ω0. Next, we check the convergence of the proposed MsFEM-POD method. We fix Q = 200 and mi = 3 for all i = 1, · · · , NH in the rest of examples. Notice that the POD error ρ is not discussed here. Interested readers c… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Numerical convergence rates with respect to s. The red and blue sym￾bols denote the results corresponding to q = 4/3 and q = 3, respectively. Next, we verify the convergence of MsFEM-POD in the physical space. The reference solution is computed by the FEM with q = 4 3 …
Figure 5
Figure 5. Figure 5: Numerical convergence rates of FEM and MsFEM-POD in physic space and the comparison of the CPU time. At last, we compare the convergence rates of the qMC and MC methods. Both the FEM and MsFEM-POD are employed with the same computational setups. As shown in [PITH_FULL…
Figure 6
Figure 6. Figure 6: Numerical convergence rates of the FEM and the MsFEM-POD with respect to N. The ”MsFEM” in the figure denotes the results provided by the MsFEM-POD method. 6.3. Localization of eigenfunctions. At the end of this section, we employ the random poten￾tials over the domain…
Figure 7
Figure 7. Figure 7: The first 5 eigenvalues computed by the FEM and MsFEM-POD for different semiclassical constant ϵ. We next consider the 2D case. We set s = 32, and the mesh size h = 1 320 and H = 1 10 . The means and variances of the minimal eigenvalue as ϵ varies are recorded in [PIT…
Figure 8
Figure 8. Figure 8: The 2D ground states for different ϵ. 1st column: FEM solution; 2nd column: MsFEM-POD solution; 3rd column: error distribution. s = 256, ϵ = 1 16 , and h = 1 15000 . Numerical tests show that H should be slightly smaller than ϵ but is independent of s. We set the coars…
Figure 10
Figure 10. Figure 10: A realization of the 2D parameterized random potential and the lo￾calized eigenfunctions computed by the MsFEM-POD method. The corresponding eigenvalues are shown in [PITH_FULL_IMAGE:figures/full_fig_p027_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A quasi-Monte Carlo multiscale method for the wave propagation in random media

    math.NA 2025-07 reject novelty 5.0 of 10

    A quasi-Monte Carlo plus localized multiscale method for the random Helmholtz equation is proposed, but the proof of its central wavenumber-explicit error estimate has serious gaps for d=2.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages · cited by 1 Pith paper

  1. [15]

    A. D. Gilbert, I. G. Graham, F. Y. Kuo, R. Scheichl, and I. H. Sloan, Analysis of quasi-Monte Carlo methods for elliptic eigenvalue problems with stochastic coefficients, Numer. Math. 142 (2019), 863–915

    work page 2019

  2. [1]

    Altmann and D

    R. Altmann and D. Peterseim, Localized computation of eigenstates of random Schr¨ odingeroperators, SIAM J. Sci. Comput. 41 (2019), no. 6, B1211–B1227

    work page 2019

  3. [2]

    P. W. Anderson, Absence of diffusion in certain random lattices, Phys. Rev. 109 (1958), 1492–1505

    work page 1958

  4. [3]

    D. N. Arnold, G. David, M. Filoche, D. Jerison, and S. Mayboroda,Computing spectra without solving eigenvalue problems, SIAM J. Sci. Comput. 41 (2019), no. 1, B69–B92

    work page 2019

  5. [4]

    N Arnold, G

    D. N Arnold, G. David, D. Jerison, S. Mayboroda, and M. Filoche, Effective confining potential of quantum states in disordered media, Phys. Rev. Lett. 116 (2016), no. 5, 056602

    work page 2016

  6. [5]

    Benner, S

    P. Benner, S. Gugercin, and K. Willcox, A survey of projection-based model reduction methods for parametric dynamical systems, SIAM Rev. 57 (2015), no. 4, 483–531

    work page 2015

  7. [6]

    Bertrand, D

    F. Bertrand, D. Boffi, and A. Halim, Data-driven reduced order modeling for parametric PDE eigenvalue problems using Gaussian process regression, J. Comput. Phys. 495 (2023), 112503

    work page 2023

  8. [7]

    Bertrand, D

    F. Bertrand, D. Boffi, and A. Halim, A reduced order model for the finite element approximation of eigenvalue problems, Comput. Methods. Appl. Mech. Eng. 404 (2023), 115696

    work page 2023

Show all 39 references
  1. [8]

    A. G. Buchan, C. C. Pain, F. Fang, and I. M. Navon, A POD reduced-order model for eigenvalue problems with application to reactor physics, Int. J. Numer. Methods Eng. 95 (2013), no. 12, 1011–1032

    2013

  2. [9]

    Chab´ e, G

    J. Chab´ e, G. Lemari´ e, B. Gr´ emaud, D. Delande, P. Szriftgiser, and J. C. Garreau,Experimental observation of the Anderson metal-insulator transition with atomic matter waves, Phys. Rev. Lett. 101 (2008), 255702

    2008

  3. [10]

    P. Chen, A. Quarteroni, and G. Rozza, Reduced basis methods for uncertainty quantification, SIAM/ASA J. Uncertain. Quantif. 5 (2017), no. 1, 813–869

    2017

  4. [11]

    E. T. Chung, Y. Efendiev, W. T. Leung, and Z. Zhang, Cluster-based generalized multiscale finite element method for elliptic PDEs with random coefficients, J. Comput. Phys. 371 (2018), 606–617

    2018

  5. [12]

    J. Dick, F. Y. Kuo, and I. H. Sloan, High-dimensional integration: The quasi-Monte Carlo way, Acta Numer. 22 (2013), 133–288

    2013

  6. [13]

    Filoche and S

    M. Filoche and S. Mayboroda, Universal mechanism for Anderson and weak localization, Proc. Natl. Acad. Sci. USA 109 (2012), no. 37, 14761–14766

    2012

  7. [14]

    Fumagalli, A

    I. Fumagalli, A. Manzoni, N. Parolini, and M. Verani, Reduced basis approximation and a posteriori error estimates for parametrized elliptic eigenvalue problems, ESAIM: M2AN 50 (2016), no. 6, 1857–1885

    2016

  8. [16]

    Henning, A

    P. Henning, A. M ˚ alqvist, and D. Peterseim,Two-level discretization techniques for ground state computations of Bose-Einstein condensates, SIAM J. Numer. Anal. 52 (2014), no. 4, 1525–1550

    2014

  9. [17]

    Henning and A

    P. Henning and A. Persson, On optimal convergence rates for discrete minimizers of the Gross-Pitaevskii energy in localized orthogonal decomposition spaces, Multiscale Model. Simul. 21 (2023), no. 3, 993–1011

    2023

  10. [18]

    Holmes, J

    P. Holmes, J. L. Lumley, and G. Berkooz, Turbulence, coherent structures, dynamical systems and symmetry, Cambridge Monographs on Mechanics, Cambridge University Press, 1996

    1996

  11. [19]

    Horger, B

    T. Horger, B. Wohlmuth, and T. Dickopf, Simultaneous reduced basis approximation of parameterized elliptic eigenvalue problems, ESAIM: M2AN 51 (2017), no. 2, 443–465

    2017

  12. [20]

    T. Y. Hou, D. Ma, and Z. Zhang, A model reduction method for multiscale elliptic PDEs with random coefficients using an optimization approach, Multiscale Model. Simul. 17 (2019), no. 2, 826–853

    2019

  13. [21]

    Jiang and M

    L. Jiang and M. Presho, A resourceful splitting technique with applications to deterministic and stochastic multiscale finite element methods, Multiscale Model. Simul. 10 (2012), no. 3, 954–985

    2012

  14. [22]

    Jin and Z

    B. Jin and Z. Zhou, An analysis of Galerkin proper orthogonal decomposition for subdiffusion, ESAIM: M2AN 51 (2017), no. 1, 89–113

    2017

  15. [23]

    Kunisch and S

    K. Kunisch and S. Volkwein, Galerkin proper orthogonal decomposition methods for parabolic problems, Numer. Math. 90 (2001), no. 1, 117–148. 29

    2001

  16. [24]

    Laurent, O

    D. Laurent, O. Legrand, P. Sebbah, C. Vanneste, and F. Mortessagne, Localized modes in a finite-size open disordered microwave cavity, Phys. Rev. Lett. 99 (2007), 253902

    2007

  17. [25]

    Lemari´ e, H

    G. Lemari´ e, H. Lignier, D. Delande, P. Szriftgiser, and J. C. Garreau, Critical state of the Anderson transition: Between a metal and an insulator, Phys. Rev. Lett. 105 (2010), 090601

    2010

  18. [26]

    Li and Z

    P. Li and Z. Zhang, Efficient finite element methods for semiclassical nonlinear Schr¨ odingerequations with random potentials, preprint (2024)

    2024

  19. [27]

    Li and Z

    S. Li and Z. Zhang, Computing eigenvalues and eigenfunctions of Schr¨ odingerequations using a model reduction approach, Commun. Comput. Phys. 24 (2018), no. 4, 1073–1100

    2018

  20. [28]

    J. S. Liu, Monte Carlo strategies in scientific computing, Springer Series in Statistics, Springer New York, 2001

    2001

  21. [29]

    Ma, W.-K

    D. Ma, W.-K. Ching, and Z. Zhang, Proper orthogonal decomposition method for multiscale elliptic PDEs with random coefficients, J. Comput. Appl. Math. 370 (2020), 112635

    2020

  22. [30]

    Machiels, Y

    L. Machiels, Y. Maday, I. B. Oliveira, A. T. Patera, and D. V. Rovas, Output bounds for reduced-basis approximations of symmetric positive definite eigenvalue problems, C. R. Acad. Sci. Paris S´ er. I Math. 331 (2000), no. 2, 153–158

    2000

  23. [31]

    M ˚ alqvist and D

    A. M ˚ alqvist and D. Peterseim,Numerical homogenization by localized orthogonal decomposition, Society for Industrial and Applied Mathematics, Philadelphia, PA, 2020

    2020

  24. [32]

    G. S. H. Pau, Reduced-basis method for band structure calculations, Phys. Rev. E 76 (2007), no. 4, 046704

    2007

  25. [33]

    Peterseim, J

    D. Peterseim, J. W¨ arneg ˚ ard, and C. Zimmer,Super-localised wave function approximation of Bose-Einstein condensates, J. Comput. Phys. 510 (2024), 113097

    2024

  26. [34]

    Riboli, P

    F. Riboli, P. Barthelemy, S. Vignolini, F. Intonti, A. De Rossi, S. Combrie, and D. S. Wiersma, Anderson localization of near-visible light in two dimensions, Opt. Lett. 36 (2011), no. 2, 127–129

    2011

  27. [35]

    Sapienza, H

    L. Sapienza, H. Thyrrestrup, S. Stobbe, P. D. Garcia, S. Smolka, and P. Lodahl,Cavity quantum electrodynamics with Anderson-localized modes, Science 327 (2010), no. 5971, 1352–1355

    2010

  28. [36]

    Segev, Y

    M. Segev, Y. Silberberg, and D. N. Christodoulides, Anderson localization of light, Nat. Photonics 7 (2013), no. 3, 197–204

    2013

  29. [37]

    Sun and A

    J. Sun and A. Zhou, Finite element methods for eigenvalue problems, Chapman and Hall/CRC, 2016

    2016

  30. [38]

    Tuffin, Randomization of quasi-Monte Carlo methods for error estimation: Survey and normal approximation, Monte Carlo Methods Appl

    B. Tuffin, Randomization of quasi-Monte Carlo methods for error estimation: Survey and normal approximation, Monte Carlo Methods Appl. 10 (2004), no. 3-4, 617–628

    2004

  31. [39]

    Zhang, M

    Z. Zhang, M. Ci, and T. Y. Hou, A multiscale data-driven stochastic method for elliptic PDEs with random coefficients, Multiscale Model. Simul. 13 (2015), no. 1, 173–204. Department of Mathematics, The University of Hong Kong, Hong Kong Email address : lipch@hku.hk Department ...

    2015

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.