Non-Hermitian multipole skin effects challenge localization
Source abstract
Abstract
We study the effect of quenched disorder on the non-Hermitian skin effect in systems that conserve a U(1) charge and its associated multipole moments. In particular, we generalize the Hatano-Nelson argument for a localization transition in disordered, non-reciprocal systems to the interacting case. When only U(1) charge is conserved, we show that there is a transition between a skin effect phase, in which charges cluster at a boundary, and a many-body localized phase, in which charges localize at random positions. In the dynamics of entanglement, this coincides with an area to volume law transition. For systems without boundaries, the skin effect becomes a delocalized phase with a unidirectional current. If dipoles or higher multipoles are conserved, we show that the non-Hermitian skin effect remains stable to arbitrary disorder. Counterintuitively, the system is therefore always delocalized under periodic boundary conditions, regardless of disorder strength.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: Quantum Mechanics and Non-Hermitian Physics · Orbital Angular Momentum in Optics · Quantum and Classical Electrodynamics
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
Institute for Condensed Matter Theory
Max Planck Institute for the Physics of Complex Systems
Universidade de Lisboa
Funding
Office of Naval Research · Multidisciplinary University Research Initiative · Gordon and Betty Moore Foundation
References (80)
Show source-provided references
1. 10.1103/PhysRev.109.149210.1103/PhysRev.109.1492 ↗
2. 10.1016/j.aop.2005.11.01410.1016/j.aop.2005.11.014 ↗
3. 10.1103/PhysRevLett.95.20660310.1103/PhysRevLett.95.206603 ↗
4. 10.1146/annurev-conmatphys-031214-01472610.1146/annurev-conmatphys-031214-014726 ↗
5. 10.1103/RevModPhys.91.02100110.1103/RevModPhys.91.021001 ↗
6. 10.1088/1361-6633/ad975610.1088/1361-6633/ad9756 ↗
7. 10.1080/23746149.2022.210943110.1080/23746149.2022.2109431 ↗
8. 10.1080/00018732.2021.187699110.1080/00018732.2021.1876991 ↗
9. 10.1103/PhysRevLett.102.06570310.1103/PhysRevLett.102.065703 ↗
10. 10.1103/PhysRevB.84.15310110.1103/PhysRevB.84.153101 ↗
11. 10.1103/PhysRevLett.124.05680210.1103/PhysRevLett.124.056802 ↗
12. 10.1103/PhysRevLett.123.20640410.1103/PhysRevLett.123.206404 ↗
13. 10.1103/PhysRevLett.124.08680110.1103/PhysRevLett.124.086801 ↗
14. 10.1038/s42005-023-01130-210.1038/s42005-023-01130-2 ↗
15. 10.1103/PhysRevLett.121.08680310.1103/PhysRevLett.121.086803 ↗
16. 10.1103/PhysRevResearch.2.03301710.1103/PhysRevResearch.2.033017 ↗
17. 10.1103/RevModPhys.93.01500510.1103/RevModPhys.93.015005 ↗
18. 10.1103/PhysRevB.103.17430310.1103/PhysRevB.103.174303 ↗
19. 10.1103/PhysRevB.98.20513610.1103/PhysRevB.98.205136 ↗
20. 10.1103/PhysRevB.100.13430610.1103/PhysRevB.100.134306 ↗
21. 10.1103/PhysRevX.9.03100910.1103/PhysRevX.9.031009 ↗
22. 10.1103/PhysRevLett.128.05060210.1103/PhysRevLett.128.050602 ↗
23. 10.1103/PhysRevB.99.22430710.1103/PhysRevB.99.224307 ↗
24. 10.1103/PhysRevB.103.22421010.1103/PhysRevB.103.224210 ↗
25. 10.1103/PhysRevLett.77.57010.1103/PhysRevLett.77.570 ↗
26. 10.1103/PhysRevB.56.865110.1103/PhysRevB.56.8651 ↗
27. 10.1103/PhysRevB.58.838410.1103/PhysRevB.58.8384 ↗
28. 10.1103/PhysRevB.56.R433310.1103/PhysRevB.56.R4333 ↗
29. 10.1103/PhysRevE.59.643310.1103/PhysRevE.59.6433 ↗
30. 10.1103/PhysRevLett.133.13650310.1103/PhysRevLett.133.136503 ↗
31. 10.1103/PhysRevLett.123.09060310.1103/PhysRevLett.123.090603 ↗
32. 10.1103/PhysRevB.102.06420610.1103/PhysRevB.102.064206 ↗
33. 10.1103/PhysRevB.106.06420810.1103/PhysRevB.106.064208 ↗
34. 10.1103/PhysRevB.106.13420210.1103/PhysRevB.106.134202 ↗
35. 10.1103/PhysRevB.109.17420510.1103/PhysRevB.109.174205 ↗
36. 10.1038/s42005-024-01576-y10.1038/s42005-024-01576-y ↗
37. 10.1103/PhysRevB.109.L14020110.1103/PhysRevB.109.L140201 ↗
38. 10.1103/PhysRevLett.42.67310.1103/PhysRevLett.42.673 ↗
39. In the non-Hermitian case, ηα†(g) is only a creation operator for right eigenstates, such that |ψg(α)〉=ηα†(g)|0〉, and left eigenstates are created by a different operator, 〈ψg(α)|=〈0|ζα(g).
40. 10.1051/jphys:01984004508012830010.1051/jphys:019840045080128300 ↗
41. 10.1103/PhysRevB.101.12110910.1103/PhysRevB.101.121109 ↗
42. 10.1103/PhysRevResearch.4.03312210.1103/PhysRevResearch.4.033122 ↗
43. 10.1103/PhysRevB.106.L12110210.1103/PhysRevB.106.L121102 ↗
44. 10.1103/PhysRevB.105.16513710.1103/PhysRevB.105.165137 ↗
45. 10.1103/PhysRevB.75.15511110.1103/PhysRevB.75.155111 ↗
46. 10.1103/PhysRevB.82.17441110.1103/PhysRevB.82.174411 ↗
47. 10.1103/PhysRevB.91.08110310.1103/PhysRevB.91.081103 ↗
48. 10.1103/PhysRevLett.115.04660310.1103/PhysRevLett.115.046603 ↗
49. 10.1103/PhysRevLett.115.18720110.1103/PhysRevLett.115.187201 ↗
50. 10.1103/PhysRevLett.125.15660110.1103/PhysRevLett.125.156601 ↗
51. 10.1103/PhysRevLett.119.01680210.1103/PhysRevLett.119.016802 ↗
52. 10.1103/PhysRevLett.118.01680410.1103/PhysRevLett.118.016804 ↗
53. 10.1103/PhysRevB.98.17420210.1103/PhysRevB.98.174202 ↗
54. 10.1007/s10955-016-1508-x10.1007/s10955-016-1508-x ↗
55. 10.1016/j.nuclphysb.2014.12.01410.1016/j.nuclphysb.2014.12.014 ↗
56. 10.1103/PhysRevB.90.17420210.1103/PhysRevB.90.174202 ↗
57. 10.1103/PhysRevB.91.08542510.1103/PhysRevB.91.085425 ↗
58. Since both the LIOMs and the charge commute with the Hamiltonian, if [Oα,N]≠0, then there would be a degenerate subspace of states, which is almost impossible due to the onsite disorder. See also Ref. [77].
59. Other works that numerically explore spectral and dynamical signatures of the non-Hermitian MBL transition include Refs. [29, 30, 31, 32, 34], though none discuss the physical mechanism for localization via non-Hermitian LIOMs.
60. 10.1103/PhysRevB.103.06420110.1103/PhysRevB.103.064201 ↗
61. 10.1103/PhysRevB.108.18420710.1103/PhysRevB.108.184207 ↗
62. We quantify entanglement via the half-partition von Neumann entanglement entropy, SL/2(t)=−Tr[ρL/2(t)logρL/2(t)], with ρL/2(t) the bipartite reduced density matrix of the evolved state |ψ(t)〉, starting from the Néel state.
63. 10.1103/PhysRevX.13.02100710.1103/PhysRevX.13.021007 ↗
64. For the chosen disorder strength W=5.5, we verify that the dynamics is frozen [see the inset of Fig. 4] and that the level-spacing distribution is Poissonian. See Supplemental Material for additional plots.
65. 10.1103/PhysRevB.103.13420710.1103/PhysRevB.103.134207 ↗
66. Diagonal terms like ∑iφα(i)ni are also included in the sum, as they conserve dipole moment.
67. Right eigenstates define the non-Hermitian LIOMs, so creation and annihilation operators are scaled by the same factor as in Eq. (7).
68. See Supplemental Material at http://link.aps.org/supplemental/10.1103/c1jr-zqxg for a case in which dipoles are local, zero-dimensional objects, leading to a transition.
69. 10.1103/PhysRevLett.129.17060110.1103/PhysRevLett.129.170601 ↗
70. 10.1103/PhysRevX.10.01104710.1103/PhysRevX.10.011047 ↗
71. 10.1103/PhysRevLett.125.24530310.1103/PhysRevLett.125.245303 ↗
72. S. Moudgalya · 2020
73. 10.1103/PhysRevX.10.01104210.1103/PhysRevX.10.011042 ↗
74. 10.1038/s41467-021-24726-010.1038/s41467-021-24726-0 ↗
75. 10.1103/PhysRevLett.130.01020110.1103/PhysRevLett.130.010201 ↗
76. 10.1103/PhysRevX.8.03107910.1103/PhysRevX.8.031079 ↗
77. 10.1103/wztw-l8wg10.1103/wztw-l8wg ↗
78. 10.15625/0868-3166/28/1/1172210.15625/0868-3166/28/1/11722 ↗
79. 10.1121/10.002716610.1121/10.0027166 ↗
80. 10.1103/PhysRevB.94.22420610.1103/PhysRevB.94.224206 ↗