Charge pumps, pivot Hamiltonians, and symmetry-protected topological phases
Source abstract
Abstract
Generalized charge pumps are topological obstructions to trivializing loops in the space of symmetric gapped Hamiltonians. We show that given mild conditions on such pumps, the associated loop has high-symmetry points that must be in distinct symmetry-protected topological (SPT) phases. To further elucidate the connection between pumps and SPTs, we focus on closed paths, “pivot loops”, defined by two Hamiltonians, where the first is unitarily evolved by the second “pivot” Hamiltonian. While such pivot loops have been studied as entanglers for SPTs, here we explore their connection to pumps. We construct families of pivot loops that pump charge for various symmetry groups, often leading to SPT phases—including dipole SPTs. Intriguingly, we find examples where nontrivial pumps do not lead to genuine SPTs but still entangle representation-SPTs (RSPTs). We use the anomaly associated with the nontrivial pump to explain the “unnecessary” criticality between these RSPTs. We also find that particularly nice pivot families form circles in Hamiltonian space, which we show are equivalent to the Hamiltonians satisfying the Dolan-Grady relation—known from the study of integrable models. This additional structure allows us to derive more powerful constraints on the phase diagram. Natural examples of such circular loops arise from pivoting with the Onsager-integrable chiral clock models, containing the aforementioned RSPT example. In fact, we show that these Onsager pivots underlie general group cohomology-based pumps in one spatial dimension. Finally, we recast the above in the language of equivariant families of Hamiltonians and relate the invariants of the pump to the candidate SPTs. We also highlight how certain SPTs arise in cases where the equivariant family is labeled by spaces that are not manifolds.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Quantum many-body systems · Quantum, superfluid, helium dynamics · Quantum chaos and dynamical systems
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
University of Oxford
University of California, Los Angeles
University of Chicago
Harish-Chandra Research Institute
References (98)
Show source-provided references
1. 10.1103/PhysRevB.27.608310.1103/PhysRevB.27.6083 ↗
2. 10.1103/PhysRevB.102.24511310.1103/PhysRevB.102.245113 ↗
3. 10.21468/SciPostPhys.8.1.00110.21468/SciPostPhys.8.1.001 ↗
4. M. Hermele, Families of gapped systems and quantum pumps, Talk at Harvard CMSA (2021), https://www.youtube.com/watch?v=wtaC0tqXZMU.
5. 10.1103/PhysRevB.106.12510810.1103/PhysRevB.106.125108 ↗
6. 10.1103/PhysRevB.108.12514710.1103/PhysRevB.108.125147 ↗
7. A. Kitaev, On the classification of Short-Range Entangled states, Talk at Simons Center for Geometry and Physics (2013), scgp.stonybrook.edu/archives/7874.
8. 10.1007/JHEP05(2019)00710.1007/JHEP05(2019)007 ↗
9. 10.1007/JHEP07(2025)00710.1007/JHEP07(2025)007 ↗
10. 10.21468/SciPostPhys.14.2.01210.21468/SciPostPhys.14.2.012 ↗
11. 10.21468/SciPostPhys.14.2.01310.21468/SciPostPhys.14.2.013 ↗
12. 10.21468/SciPostPhys.18.3.09410.21468/SciPostPhys.18.3.094 ↗
13. 10.1103/PhysRevB.80.15513110.1103/PhysRevB.80.155131 ↗
14. 10.1103/PhysRevB.81.06443910.1103/PhysRevB.81.064439 ↗
15. 10.1103/PhysRevB.81.13450910.1103/PhysRevB.81.134509 ↗
16. 10.1103/PhysRevB.83.07510210.1103/PhysRevB.83.075102 ↗
17. 10.1103/PhysRevB.84.16513910.1103/PhysRevB.84.165139 ↗
18. 10.1103/PhysRevB.83.03510710.1103/PhysRevB.83.035107 ↗
19. 10.1103/PhysRevB.86.12544110.1103/PhysRevB.86.125441 ↗
20. 10.1146/annurev-conmatphys-031214-01474010.1146/annurev-conmatphys-031214-014740 ↗
21. 10.1103/PhysRevB.96.16512410.1103/PhysRevB.96.165124 ↗
22. 10.1038/ncomms450710.1038/ncomms4507 ↗
23. 10.1103/PhysRevX.11.04105910.1103/PhysRevX.11.041059 ↗
24. 10.1103/PhysRevB.107.23510410.1103/PhysRevB.107.235104 ↗
25. 10.1103/PhysRevLett.86.91010.1103/PhysRevLett.86.910 ↗
26. 10.1209/0295-5075/95/5000110.1209/0295-5075/95/50001 ↗
27. 10.1103/PhysRevB.86.11510910.1103/PhysRevB.86.115109 ↗
28. 10.1103/PhysRevB.100.16513210.1103/PhysRevB.100.165132 ↗
29. 10.1103/PhysRev.65.11710.1103/PhysRev.65.117 ↗
30. Indeed, it reduces to for N=2. Note that this is not the ZN×ZN cluster model that we discuss in Sec. 7 below [82, 83].
31. 10.1103/PhysRevB.101.23510810.1103/PhysRevB.101.235108 ↗
32. R. Verresen, P. Fendley, and N. Tantivasadakarn (unpublished).
33. A. Kitaev, Toward a topological classification of many- body quantum states with short-range entanglement, Talk at Simons Center for Geometry and Physics (2011), https://scgp.stonybrook.edu/archives/1087.
34. Although we will not require a concrete definition, in the broadest sense of the term these are also called invertible states, for each such state there is a corresponding state that tensors to give the trivial phase [93]. Their appearance is natural as we expect a pump to be reversible.
35. Such a loop is a family of Hamiltonians parameterized by a circle. One can further generalise to families parameterized by other spaces [6], see also Sec. 10.
36. Strictly speaking, this result is for compact symmetry groups G̃—we will focus on finite symmetry groups in this work.
37. 10.1007/s00220-024-05010-w10.1007/s00220-024-05010-w ↗
38. 10.1002/cpa.316007040410.1002/cpa.3160070404 ↗
39. This is a different notion of generalized symmetries to that found in Ref. [102].
40. 10.1016/j.physleta.2009.07.06010.1016/j.physleta.2009.07.060 ↗
41. 10.1088/1742-6596/343/1/01208410.1088/1742-6596/343/1/012084 ↗
42. This is not necessarily the case (we could have an R symmetry), but all examples we are aware of give rise to a U(1) symmetry on rescaling.
43. 10.1088/0305-4470/23/12/01010.1088/0305-4470/23/12/010 ↗
44. 10.1103/PhysRevB.90.23513710.1103/PhysRevB.90.235137 ↗
45. 10.1007/s00220-025-05422-210.1007/s00220-025-05422-2 ↗
46. 10.1103/PhysRevB.87.15511410.1103/PhysRevB.87.155114 ↗
47. 10.1016/j.aop.2014.09.00710.1016/j.aop.2014.09.007 ↗
48. 10.1103/PhysRevB.95.19512810.1103/PhysRevB.95.195128 ↗
49. The key general notions are that the invariant is additive if we encircle two diabolical points, and that applying g0 that commutes with G̃ does not change the value of the invariant.
50. B. Zeng · 201910.1007/978-1-4939-9084-9 ↗
51. 10.1103/RevModPhys.93.04500310.1103/RevModPhys.93.045003 ↗
52. 10.1103/PhysRevResearch.2.04202410.1103/PhysRevResearch.2.042024 ↗
53. Note that for any path of gapped local Hamiltonians, one can use the quasi-adiabatic continuation to find a local unitary evolution acting on the ground state; and for a G̃-symmetric path, the corresponding generating Hamiltonian can always be chosen G̃-symmetric [38, 51].
54. This means that we can write this operator as a tensor network with finite bond dimension. All of our pivot Hamiltonian examples are of this form, since they have commuting local terms.
55. 10.1103/PhysRevB.98.24512210.1103/PhysRevB.98.245122 ↗
56. 10.1088/1742-5468/2007/08/p0802410.1088/1742-5468/2007/08/p08024 ↗
57. 10.1103/PhysRevB.109.11515210.1103/PhysRevB.109.115152 ↗
58. 10.21468/SciPostPhys.18.5.16810.21468/SciPostPhys.18.5.168 ↗
59. A technical condition that we can read as excluding symmetry breaking. See [52] for more details.
60. 10.1103/PhysRevB.93.24514510.1103/PhysRevB.93.245145 ↗
61. 10.1103/PhysRevB.93.20110310.1103/PhysRevB.93.201103 ↗
62. 10.1103/PhysRevB.95.15512610.1103/PhysRevB.95.155126 ↗
63. 10.1103/PhysRevD.25.158710.1103/PhysRevD.25.1587 ↗
64. 10.1090/pspum/049.1/101314010.1090/pspum/049.1/1013140 ↗
65. 10.1063/1.52903610.1063/1.529036 ↗
66. 10.1088/1751-8113/49/15/15300110.1088/1751-8113/49/15/153001 ↗
67. 10.1007/BF0102086010.1007/BF01020860 ↗
68. 10.1142/S021773239100405X10.1142/S021773239100405X ↗
69. 10.1088/1742-5468/ab11c010.1088/1742-5468/ab11c0 ↗
70. 10.1103/PhysRevLett.17.113310.1103/PhysRevLett.17.1133 ↗
71. 10.1103/PhysRev.158.38310.1103/PhysRev.158.383 ↗
72. 10.1007/BF0164648710.1007/BF01646487 ↗
73. 10.1143/PTP.46.133710.1143/PTP.46.1337 ↗
74. 10.1093/ptep/ptx14810.1093/ptep/ptx148 ↗
75. 10.1007/JHEP05(2017)09110.1007/JHEP05(2017)091 ↗
76. 10.21468/SciPostPhys.8.1.00210.21468/SciPostPhys.8.1.002 ↗
77. 10.1103/PhysRevLett.134.22660310.1103/PhysRevLett.134.226603 ↗
78. Since χ(ab)=χ(a)χ(b)=χ(b)χ(a)=χ(ba) we have ν(ab)=ν(ba)mod2π. Since we chose the range 0≤ν(g)<2π, we have ν(ab)=ν(ba).
79. 10.1016/0550-3213(85)90350-510.1016/0550-3213(85)90350-5 ↗
80. 10.1103/PhysRevB.91.15515010.1103/PhysRevB.91.155150 ↗
81. 10.1103/PhysRevB.109.12512110.1103/PhysRevB.109.125121 ↗
82. 10.1103/PhysRevA.84.02230410.1103/PhysRevA.84.022304 ↗
83. 10.1016/j.nuclphysb.2017.03.02110.1016/j.nuclphysb.2017.03.021 ↗
84. 10.1103/PhysRevLett.129.09050110.1103/PhysRevLett.129.090501 ↗
85. This in fact holds for generalized cohomology theories, so these arguments apply to fermions/spin cobordism, K theory, etc.).
86. 10.21468/SciPostPhys.8.4.06210.21468/SciPostPhys.8.4.062 ↗
87. It was shown in Ref. [9] that the ambiguity in α, given ω, is given by tensoring with an anomaly free G-equivariant family. Our equivalence relation on families does not allow this, so α is an invariant.
88. Since H denotes the anomaly free subgroup, we denote Hamiltonians by Ĥ in this section.
89. 10.1063/5.005599610.1063/5.0055996 ↗
90. 10.1090/btran/5110.1090/btran/51 ↗
91. 10.1142/S0129055X1350017710.1142/S0129055X13500177 ↗
92. 10.1016/0001-8708(74)90072-310.1016/0001-8708(74)90072-3 ↗
93. 10.1007/s10955-018-2177-810.1007/s10955-018-2177-8 ↗
94. 10.1103/PhysRevLett.130.24660110.1103/PhysRevLett.130.246601 ↗
95. 10.1016/j.aop.2023.16945710.1016/j.aop.2023.169457 ↗
96. 10.1103/PhysRevB.106.16511510.1103/PhysRevB.106.165115 ↗
97. This can be further generalized, e.g., to include anti-unitary and spatial symmetries as well. We will not consider these for simplicity.
98. 10.1007/JHEP02(2015)17210.1007/JHEP02(2015)172 ↗