Abstract
In classical and singleparticle settings, nontrivial band topology always gives rise to robust boundary modes. For quantum manybody systems, however, multiple topological fermions are not always able to coexist, since Pauli exclusion prevents additional fermions from occupying the limited number of available topological modes. In this work, we show, through IBM quantum computers, how one can robustly stabilize more fermions than the number of topological modes through specially designed 2fermion interactions. Our demonstration hinges on the realization of BDI and Dclass topological Hamiltonians on transmonbased quantum hardware, and relied on a tensor networkaided circuit recompilation approach. We also achieved the full reconstruction of multiplefermion topological band structures through iterative quantum phase estimation (IQPE). All in all, our work showcases how advances in quantum algorithm implementation enable noisy intermediatescale quantum (NISQ) devices to be exploited for topological stabilization beyond the context of singleparticle topological invariants.
Introduction
Manybody quantum effects like particle statistics and Hubbard interactions underscore some of the most exciting condensed matter phenomena, such as superconductivity and fractionalization^{1,2,3}. Their interplay with singleparticle properties like band topology is particularly fascinating, with rich fractionalized quasiparticles emerging out of Coulomb repulsion within dispersionless Landau levels, for instance. But unlike in purely singlebody settings, particle statistics often play a central role, determining all Fermi surface properties like electrical conductivity. In topological systems, Pauli exclusion also implies that topological robustness is only conferred upon the few electrons occupying the limited number of topological modes^{4}.
Compared to singlebody topological phenomena realizable in classical metamaterials, novel manybody phases are traditionally much harder to engineer and physically demonstrate. This is due to challenges in accessing and manipulating intrinsically fragile quantum states, unlike the macroscopic classical degrees of freedom of photonic, acoustic and electrical circuit lattices^{5,6,7,8}. Fortunately, there is considerable recent experimental progress in various quantum systems such as ultracold atomic lattices,^{9}, photonic systems^{10}, silicon^{11} and trapped ion systems^{12}, which make Richard Feynman’s vision^{13} of utilizing quantum systems to simulate quantum Hamiltonians a reality.
Of particular versatility are universal quantum simulators, also known as quantum computers, which allow arbitrary quantum systems to be simulated, thereby in principle enabling any quantum phenomenon to be physically realized. Through the appropriate design of sequences of quantum operations, known collectively as quantum algorithms, quantum computers are capable of performing a vast array of computational tasks, some at polynomial or exponential resource advantage over classical algorithms. Indeed, even in the current noisy intermediatescale quantum (NISQ) era, quantum computers have already shown great promise, with demonstrations of quantum advantage in limited settings^{14,15}, mapping topology in parameter space^{16}, the achievement of chemical accuracy in intermediatescale electronic structure calculations^{17}, and in neutron scattering and exotic magnetic phenomena simulations^{18,19}.
This work shall employ computations made with IBM quantum computers, which not only rank favourably with other cuttingedge platforms^{20,21} in hardware performance, such as gate fidelities and decoherence times, but is also fully accessible via the cloud. IBM Quantum (IBM Q) currently offers access to up to 65qubit machines based on superconducting transmon qubits, and provides an opensource software development kit called Qiskit^{22,23}. Thus far, IBM Q has been successfully utilized in simulating spin models^{24}, global quantum quenches^{25}, quantum chemistry problems^{26}, topological phenomena^{27,28,29,30}, machine learning^{31} and various other applications^{32}.
We emphasize that quantum computation in the current NISQera is still plagued with significant limitations. Main bottlenecks include low gate fidelity, decoherence, limited qubit connectivity and limited number of qubits^{33}, which together impose constraints on circuit depth and structure. Amidst qubit noise and readout error, typical depths of \({{{\mathcal{O}}}}(1{0}^{1})\) entangling gate layers are presently feasible for precision results. Of utmost current priority is hence the development of error mitigation and circuit optimization approaches that maximize computational capability within hardware bounds, thereby enabling the practical use of quantum simulators in contexts where classical simulators, for example, purposedesigned electrical circuits^{34}, are inadmissible.
In this work, we stabilize BDI and Dclass topological boundary states, as well as demonstrate a full band structure reconstruction of the fermionic extended Kitaev chain (KC) on IBM quantum computers. Compared to existing quantum computer realizations of other topological states^{27,28,29}, some of which are performed in parameter space, ours was performed on a longer (12qubit) chain with physical open boundaries that host topological modes. Furthermore, our extended KC Hamiltonian contains multiple nonlocal couplings which presented heavy demands on circuit depth and complexity, necessitating the use of tensor networkaided circuit recompilation techniques beyond traditional trotterization. Crucially, by exploiting the quantum nature of the IBM machines, we engineered effective interactions to stabilize more fermions than originally allowed by the number of topological zero modes, hence physically realizing a fewbody phenomenon that has not been possible in existing classical realizations.
Results and discussion
Physical motivation and models
Topological robustness is a highly soughtafter property exemplified by the extraordinarily long survival duration of boundary states in specially designed topological lattices^{35,36}. This robustness has inspired various potential applications like topological lasers and sensors^{37,38}, and originates from the integer quantization of topological invariants characterizing the lattice band structure^{4}. While nontrivial topology has been demonstrated in a wide variety of classical metamaterials, true quantization of the response can only be observed in quantum settings (but see refs. ^{39,40}).
Yet, a fully quantum topological fermionic system is also limited by the fact that there can only be as many topological fermions as available topological modes. Since the latter is determined by the topological invariant, which is typically an integer of order \({{{\mathcal{O}}}}(1{0}^{0})\), it will be of great scientific and practical interest to probe how quantum interactions can also enhance the number of robustly surviving fermions beyond what is dictated by the topological invariant. We shall first introduce our topological lattice models, and later discuss how specific interactions preserve the fidelity of multiple fermions by hosting manybody states that are adiabatically connected to noninteracting topological states.
In this work we simulate on a quantum computer the extended Kitaev model^{41,42}, which is a 1D open chain containing nextnearest neighbour (NNN) couplings in addition to its underlying nearestneighbour (NN) structure, with two sites per unit cell:
for N unit cells, chemical potential μ, tunnelling coefficients v_{1,2}, superconducting pairing Δ_{1,2}, relative phase ϕ, and fermionic operators c_{j}, d_{j} respectively acting on the A and Bsublattices of the jth unit cell. Our model H^{KC} is an extension of the standard Kitaev chain model that has been intensely studied for hosting Majorana zero modes^{43}, which is of particular interest to faulttolerant topological quantum computing^{44}. Its extra degrees of freedom in Eq. (1) allow for smooth tuning across the BDI and Dsymmetry protected topological (SPT) classes elaborated below, and will ultimately also be useful in the design of interactions that preserve the robustness of multiple fermions.
When ϕ = 0, H^{KC} preserves timereversal symmetry (TRS) in addition to parity and charge conjugation symmetry, and belongs to the BDI class of the tenfoldway topological classification^{45}, characterized by a winding number \(\nu \in {\mathbb{Z}}\). Our particular model possesses ν = 0, 1, 2 regimes, the first trivial and the latter two respectively exhibiting twofold and fourfold degeneracy of midgap topological modes localized at either boundary. Mathematically, 2πν corresponds to the winding of its Berry phase across one Brillouin zone period (see Supplementary Note 1).
When ϕ ≠ 0, TRS is broken and H^{KC} falls into the Dclass characterized by \({{\mathbb{Z}}}_{2}\) invariant, with its topologically trivial/nontrivial phases respectively labelled by Berry phase winding γ = 0, π which encapsulates the relative configurations of the k and −k paths^{41,42}. This Dclass phase minimally requires NNN couplings, not present in the BDI class which contains the extremely wellknown SuSchriefferHeeger (SSH) model^{46,47}, and is thus much less frequently investigated let alone realized in quantum settings^{48}. We note that the SSH model, which we shall also simulate, is a special case of the extended Kitaev model upon unitary rotation:
with v, w intracell and intercell hopping coefficients. We remark that H^{SSH} and H^{KC} are quadratic in the fermionic basis and are, in principle, efficiently simulable classically^{49}; but with the addition of quartic interactions, both models cannot be efficiently classically simulated.
Persistent boundary modes from topological protection
As the first experiments on the IBM quantum computer, we demonstrate that initial states at the boundary survive much longer when the Hamiltonian is topological. We emphasize that this is true for a rather large parameter space of initial states, not just exact topological eigenstates (see Supplementary Note 4). For a start, various perfectly localized 1fermion initial states defined in Table 1 are evolved via H^{SSH} and H^{KC}.
We utilize a tensor networkaided recompilation technique (see Methods) to construct quantum circuits for timeevolution, in order to overcome circuit depth limitations on NISQ hardware. Our approach is based on prior literature^{50,51,52}, with added sectorspecificity to enhance performance at larger fermion numbers and qubit counts. Furthermore, we employ a suite of error mitigation techniques to improve data quality, in particular readout mitigation (RO)^{25,53}, postselection (PS)^{25,54}, and averaging across machines and qubit chains (see Methods). Through computationalbasis measurements on simulation qubits, the occupancy \({{{\bf{O}}}}={[\left\langle {O}_{1}\right\rangle \left\langle {O}_{2}\right\rangle \ldots \left\langle {O}_{n}\right\rangle ]}^{\top }\) for \({O}_{j}={c}_{j}^{{\dagger} }{c}_{j}=(1{\sigma }_{j}^{z})/2\) along the chain is retrieved, and the extent of evolution away from \(\left{\psi }_{0}\right\rangle\), whose occupancy is O_{0}, can be assessed via the occupancy fidelity
Note \({{{{\mathcal{F}}}}}_{O}\) is distinct from state fidelity \({{{\mathcal{F}}}}={\left\left\langle {\psi }_{0} \psi \right\rangle \right}^{2}\)—see Supplementary Note 3 for a more detailed discussion.
We present 1particle timeevolution results for H^{SSH} in Fig. 1a, comparing raw data, data with RO and PS mitigation, and data with both. The effectiveness of the error mitigation methods is clear, with the occupancy fidelity of all initial states approaching close to exact diagonalization (ED) results with RO and PS applied. The stability, i.e. persistence, of the initial state is quantified via the occupancy fidelity \({{{{\mathcal{F}}}}}_{O}\); the slower the decay in \({{{{\mathcal{F}}}}}_{O}\), the more stable the state. As expected, the decay of the initial states \(\left{1}^{{{{\rm{L}}}}}\right\rangle\), \(\left{1}^{{{{\rm{R}}}}}\right\rangle\) localized at the left and right boundary sites are slow in the topological regime, compared to that of the middlelocalized state \(\left{1}^{{{{\rm{M}}}}}\right\rangle\), which overlaps negligibly with topological boundary modes; by contrast, in the trivial regime topological protection does not exist and all states decay quickly. We remark that edgelocalized states can be stable even with significant perturbations, which is a hallmark of topological robustness. For instance, the stability of \(\left{1}^{{{{\rm{L}}}}}\right\rangle\), in the sense of slow decay of \({{{{\mathcal{F}}}}}_{O}\), is preserved even when mixed with nonSPT states localized on the neighbouring and nextneighbouring sites (see Supplementary Note 4).
For H^{KC} with NNN couplings, even relatively delocalized initial states can exhibit slow decay. We determine ‘idealized’ maximally localized boundary modes that are adiabatically connected to topological eigenstates (Supplementary Note 1), that exhibits negligible decay, such as to facilitate the design of more general persistent states. In the BDI class, they are \(\left{1}_{1}^{{{{\rm{L}}}}}\right\rangle\) and \(\left{1}_{2}^{{{{\rm{L}}}}}\right\rangle\) as defined in Table 1, which are localized on the first and second unit cells respectively. In Fig. 1c, \(\left{1}_{1}^{{{{\rm{L}}}}}\right\rangle\) (and \(\left{1}_{1}^{{{{\rm{R}}}}}\right\rangle\)) are robust for both ν = 1 and 2 topological sectors, but \(\left{1}_{2}^{{{{\rm{L}}}}}\right\rangle\) (and \(\left{1}_{2}^{{{{\rm{R}}}}}\right\rangle\)) are robust only for the ν = 2 sector, which supports two topological boundary modes at each end. For Dclass H^{KC} systems with nonzero ϕ and NNN couplings, the four idealized boundary states are each localized on the four boundary sites with ϕdependent relative phases between site orbitals, as defined in Table 1. Indeed, almost negligible decay was observed (Fig. 1c) for these states in the topological regimes, compared to the middlelocalized \(\left{1}^{{{{\rm{M}}}}}\right\rangle\) which do not benefit from topological robustness; in the trivial regime, all states decay quickly as expected. The extent of delocalization of \(\left{1}^{{{{\rm{M}}}}}\right\rangle\) in all regimes, and of all initial states in the topologically trivial regime, are detailed in the occupancy density maps of Fig. 1c. While state initialization in the timeevolution circuits for H^{SSH} and BDIclass H^{KC} are performed through explicit circuit components (see Supplementary Note 3), that for \(\left{1}_{1,2}^{{{{\rm{L}}}},{{{\rm{R}}}}}\right\rangle\) of Dclass H^{KC} are absorbed into our dynamically optimized recompilation ansatz, so as to minimize incurred costs in circuit depth.
Band structure reconstruction
Our boundary states owe their slow decay to the gapped nature of their dominant (topological) eigenstate component, as is readily understood through a twocomponent approximate treatment. Consider \(\left\psi (t)\right\rangle =\alpha \left{\psi }_{1}\right\rangle +\beta \left{\psi }_{2}\right\rangle {e}^{i{{\Delta }}Et}\) with \({\left\alpha \right}^{2}+{\left\beta \right}^{2}=1\); then the state fidelity is \({{{\mathcal{F}}}}(t)={\left\left\langle \psi (t) \psi (0)\right\rangle \right}^{2}=14{\left\alpha \right}^{2}{\left\beta \right}^{2}{\sin }^{2}({{\Delta }}Et/2)\), which stays near unity as long as \({\left\alpha \right}^{2}{\left\beta \right}^{2}\ll 1\), that is, when a dominant eigenstate component exists. This can be checked by comparing all our initial states with exact topological eigenstates (Fig. 1b). In realistic settings where the initial boundary state generically overlaps with arbitrarily many eigenstates, the energy gap between the dominant topological eigenstate and all other states is crucial to stability. It introduces a separation of frequency scales that avoids overwhelming destructive interference. Experimentally, this energy gap can be verified from a full reconstruction of the topological band structure. Below, we first describe our approach for mapping the band structure, and then present its results for both singlefermion and twofermion systems.
To probe and reconstruct the band structure on quantum hardware, we perform iterative quantum phase estimation (IQPE)^{55,56}, which supports the estimation of eigenenergies of a simulated Hamiltonian, in principle to arbitrary precision (see Methods). Compared to quantum phase estimation (QPE)^{57,58}, IQPE circuits are shallower and require fewer qubits, and are thus suited for implementation on currentgeneration NISQ hardware. IQPE results for H^{SSH} and H^{KC} in the 1particle sector are shown in Fig. 1d. The BDI topological phase transition in H^{SSH} at w = v is apparent, with a pair of midgap states separating from the bulk and approaching degeneracy at E = 0 as w/v increases. The H^{KC} model possesses richer behaviour—in the BDI case, transitions from the ν = 1 phase into the trivial (ν = 0) and then into ν = 2 phase occur as v/μ increases, for illustrative v ≡ v_{1} = v_{2} = Δ_{1} = Δ_{2}. The ν = 1 phase exhibits twofold degeneracy of midgap states like in the SSH model, and the ν = 2 phase exhibits fourfold degeneracy at E = 0 due to having two zero modes at each boundary. In the D class, twofold midgap topological degeneracy occurs in the γ = π but not γ = 0 phase. For the latter, fourfold degeneracy is broken as the eigenenergies split into pairs of positive energy and negative energy states, which can be regarded as remnants of topological midgap states. In all cases, the reconstructed band structure from hardware execution of IQPE closely match ED results.
Topological stability for multiple fermions—noninteracting case
Compared to existing classical realizations of BDI and Dclass topological states, our IBM Q realizations possess the advantage of granting natural access to quantum manybody effects like fermionic statistics and interactions. We shall first discuss the former, specifically on the various ways whereby up to four fermions can simultaneously enjoy topological robustness.
We first consider H^{SSH} and the twofermion sector, in which twofermion boundary states can be constructed either as \(\left{2}^{{{{\rm{LR}}}}}\right\rangle =c{[{1}^{{{{\rm{L}}}}}]}^{{\dagger} }c{[{1}^{{{{\rm{R}}}}}]}^{{\dagger} }\left{{{\rm{vac}}}}\right\rangle\) with 1 fermion at each boundary, or alternatively as \(\left{2}_{{{{\rm{AA}}}}}^{{{{\rm{L}}}}}\right\rangle\) or \(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{L}}}}}\right\rangle\) with one fermion in the \(\left{1}^{{{{\rm{L}}}}}\right\rangle\) orbital and the other in the nearest A or B site. Of these, only \(\left{2}^{{{{\rm{LR}}}}}\right\rangle\) has both particles overlapping significantly with topological boundary modes, so only it would be conferred stability in the topological phase (Fig. 2b). The \(\left{2}_{{{{\rm{AA}}}}}^{{{{\rm{L}}}}}\right\rangle\) and \(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{L}}}}}\right\rangle\) states, like \(\left{1}^{{{{\rm{M}}}}}\right\rangle\), are unstable whether topological modes exist or not, despite one of the fermions being in the topologically stable orbital \(\left{1}^{{{{\rm{L}}}}}\right\rangle\). The upshot is that since H^{SSH} possesses only one topological state at each end of the chain, it is impossible to construct a 2particle topological state localized at only one boundary. Generalization to multiple fermions follows.
To realize 2fermion topological states localized at a single boundary, the fourfold midgap topological modes of H^{KC}, with two at each boundary, can be utilized. In particular, the two fermions can occupy \(\left{1}_{1}^{{{{\rm{L}}}}}\right\rangle\) and \(\left{1}_{2}^{{{{\rm{L}}}}}\right\rangle\) near the left boundary, resulting in \(\left{2}_{12}^{{{{\rm{L}}}}}\right\rangle\), or analogously \(\left{2}_{12}^{{{{\rm{R}}}}}\right\rangle\) for the right boundary. Relaxing the requirement of localization on a single edge, robust states such as \(2_{ij}^{\rm{LR}}\rangle =c{[1_i^{\rm{L}}]}^{\dagger }c{[1_j^{\rm{R}}]}^{\dagger} {\rm{vac}}\rangle\) are all possible, where i, j ∈ {1, 2}. Indeed, hardware results indicate these states are conferred stability in the topological regime; the contrast against the non topologically protected middlelocalized \(\left{2}^{{{{\rm{M}}}}}\right\rangle\) is drastic (Fig. 2c). The occupancy density maps concur that these 2fermion edge localized states persist almost perfectly in the topological phase, but diffuses in the trivial phase.
As a generalization to larger numbers of effective particles, we demonstrate that up to 4 fermions can enjoy stability in the Dclass H^{KC} system, with each occupying a different topological mode. Indeed, the 4fermion state \(\left{4}^{{{{\rm{LLRR}}}}}\right\rangle\) is robust compared to the middlelocalized \(\left{4}^{{{{\rm{M}}}}}\right\rangle\) state in the topological phase (Fig. 2c).
Stability for multiple fermions—interacting case
Quantum interactions can present new phase transitions^{59} and avenues of stability with no noninteracting analogues^{60,61}. The understanding of strongly correlated topological models is also crucial in understanding the phenomenology of real materials^{62,63}. While weak interactions may only perturb the stability of SPT boundary modes, strong interactions can drastically break the symmetry protection altogether, leading to new preferred states. Our timeevolution and IQPE methods, hinging on circuit recompilation, readily supports the study of strongly correlated models, facilitating digital quantum computers as alternative experimental platforms from existing ultracold atomic lattices^{64} or photonic crystals^{65}.
We consider twobody Hubbard interactions which have been commonly used to model strong densitydensity interactions^{64,66}. Among the simplest 1D topological model with interactions is the SSHHubbard model, which has been explored in various contexts^{67,68}. We are interested in Hubbard interactions that directly compete with the boundarylocalization of SSH topological modes. Specifically, we consider interactions between all successive sites (U_{2}), sites within the same unit cell (U_{4}), and same sublattice sites across adjacent unit cells (U_{5}). Our interacting SSH model is then obtained by adding these interaction terms to Eq. (2), \({H}_{{{{\rm{full}}}}}^{{{{\rm{SSH}}}}}={H}^{{{{\rm{SSH}}}}}+{H}_{{{{\rm{int}}}}}^{{{{\rm{SSH}}}}}\), with
where \({c}_{j}^{{\dagger} }\) (\({d}_{j}^{{\dagger} }\)) creates a fermion in the A (B) sublattice of the jth unit cell. We emphasize that these interactions are homogeneous across all unit cells, and do not induce boundary effects on their own, without the interplay with topological boundary localization. The schematic in Fig. 3a illustrates the interaction terms in Eq. (4). The U_{4} and U_{5} terms are intentionally chosen to induce asymmetry across the two sublattices—they impact states with inhomogeneous polarizations differently, such as SSH topological modes which occupy one sublattice exclusively. We dynamically evolve on IBM Q various boundarylocalized 2fermion states \(\left{2}^{{{{\rm{LR}}}}}\right\rangle\), \(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{L}}}}}\right\rangle\), \(\left{2}_{{{{\rm{AA}}}}}^{{{{\rm{L}}}}}\right\rangle\), \(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{R}}}}}\right\rangle\), \(\left{2}_{{{{\rm{BB}}}}}^{{{{\rm{R}}}}}\right\rangle\), schematically shown in Fig. 3b. As before, the superscripts indicate the edge localization (left/right) of the 2 fermions, while the subscripts indicate the occupied sublattices; we remind the reader that the chain starts with an Asite on the left.
Our chosen 2particle interactions have enabled us to achieve stable multifermion boundary states due to either topological localization or interactioninduced polarization, and often the interplay of both. As such, we can for instance produce stable states localized on a single boundary in \({H}_{{{{\rm{full}}}}}^{{{{\rm{SSH}}}}}\), reminiscent of topological states, but yet existing even in the topologically trivial regime of the noninteracting part H^{SSH}. The conferred stability generally increases with interaction strength at diminishing returns; beyond U/w ~ 10, little additional stability is gained from stronger interactions. Yet, the impact of the topological character of the noninteracting terms in our systems remains significant even in the face of arbitrarily strong interactions.
In Fig. 3c–d, we present the evolution of various boundarylocalized 2fermion initial states subject to strong interactions U_{2}, U_{4} or U_{5} of relative strength 10. While such strong interactions may seem clearly dominant compared to the noninteracting H^{SSH} part, the topological character of H^{SSH} still affects the evolution of most of these states significantly. When U_{2} is switched on (Fig. 3c), \(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{L}}}}}\right\rangle\) (and equivalently \(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{R}}}}}\right\rangle\), not shown) retain nearperfect fidelity, even when the chain is topologically trivial. For the \(\left{2}^{{{{\rm{LR}}}}}\right\rangle\) initial state localized at both boundaries, topology enables a longer survival time. To obtain stable 2fermion states localized at a single boundary, we may switch on U_{4} and U_{5}; when U_{5} > 0, both types of left boundary modes (\(\left{2}_{{{{\rm{AB}}}}}^{{{{\rm{L}}}}}\right\rangle\) and \(\left{2}_{{{{\rm{AA}}}}}^{{{{\rm{L}}}}}\right\rangle\)) are stabilized. On the other hand, to obtain stable right boundary states, we require U_{4} > 0 as well, such as to balance with U_{5}. This is reflected from the stability plot in parameter space (Fig. 3e)—the timeaveraged occupancy fidelity \(\overline{{{{{\mathcal{F}}}}}_{O}}\) for the leftlocalized 2particle boundary states is minimal along the U_{4} = U_{5} diagonal, being greatly destabilized relative to the rightlocalized modes. The preferential stabilization at either boundary by U_{4} and U_{5} is a consequence of their coupling of specific sublattice pairs on finite chains; when the interaction couples all sublattices like for U_{2}, twofermion boundary modes are stable on both edges, and the preferential stability is not observed.
Earlier, we showed that the stability of initial boundary states can be maintained even when they differ rather significantly from topological ansatz states (Supplementary Note 4). We now extend the same remark to boundary states that are not topologically protected, but which are conferred robustness under the combination of topology and interactioninduced effects. In Fig. 3f, we mix \(\left{2}_{{{{\rm{AA}}}}}^{{{{\rm{L}}}}}\right\rangle\) with \(\left{2}_{{{{\rm{BB}}}}}^{{{{\rm{L}}}}}\right\rangle\) perturbation, neither of which are SPT, and examine their evolution under varying U_{2} interactions. When U_{2} = 0, as expected, \(\overline{{{{{\mathcal{F}}}}}_{O}}\) is relatively low; but for U_{2} > 0 an increase in \(\overline{{{{{\mathcal{F}}}}}_{O}}\) is observed, reaching ≳ 90% for U_{2} ≳ 5/2. This stabilization holds across a wide range of mixing amplitudes. Most interestingly, the stabilization occurs only in the topological phase of H^{SSH}—indeed in the trivial regime, U_{2} has virtually no effect on state stability. Hence, neither topology nor interactions alone suffices to enable the observed perturbationrobust stabilization of nonSPT states.
Similarly, one can study Hubbard interactions on the extended Kitaev model. We again focus on interaction terms that induce sublattice asymmetry, and thus left/right boundary asymmetry for topological modes. The extended Kitaev chain (Eq. (1)), however, has longerrange couplings compared to the SSH model (Eq. (2)). The NNN hoppings and pairings were crucial in the noninteracting model to demonstrate a richer topology, exhibiting more than one pair of topological modes in the ν = 2 phase of the BDI class. We thus consider interactions that compete with this topology, specifically between occupied Asites of NN (U_{2}) an NNN (U_{3}) unit cells, with \({H}_{{{{\rm{full}}}}}^{{{{\rm{KC}}}}}={H}^{{{{\rm{KC}}}}}+{H}_{{{{\rm{int}}}}}^{{{{\rm{KC}}}}}\),
These interactions are schematically illustrated in Fig. 3a. In the BDI class of the noninteracting part H^{KC}, we dynamically evolve the various 2fermion states used earlier, \(\left{2}_{{{{\rm{AA}}}}}^{{{{\rm{L}}}}}\right\rangle\) and \(\left{2}_{{{{\rm{BB}}}}}^{{{{\rm{L}}}}}\right\rangle\), which now have large overlaps with the topological modes of the noninteracting Kitaev chain H^{KC} (see Fig. 3b for schematics). Recall that the BDIclass of H^{KC} can host up to two pairs of topological modes at the two boundaries. As shown in Fig. 4a, the presence of either interaction (U_{2} or U_{3}) worsens the fidelity of all 2particle boundary modes when ν = 2. These interactions hence disrupt the symmetry protection conferred in the BDI class. When ν = 1, there is one fewer pair of topological boundary modes, but the states \(\left{2}_{{{{\rm{A0A}}}}}^{{{{\rm{L}}}}}\right\rangle\) and \(\left{2}_{{{{\rm{B0B}}}}}^{{{{\rm{R}}}}}\right\rangle\)—which do not overlap significantly with any noninteracting topological mode—now enjoy enhanced robustness in the presence of interactions. Also, all edgelocalized modes can be made robust in the topologically trivial case ν = 0.
In the D class of H^{KC}, we demonstrate that U_{3} interactions can confer stability to 3 and 4 fermion states (Fig. 4b), which carry a greater number of fermions on a boundary than available topological modes, and are hence topologically unprotected. The stability of SPT \(\left{2}_{{{{\rm{12}}}}}^{{{{\rm{L}}}}}\right\rangle\) is not destroyed when U_{3} is imposed, and simultaneously the nonSPT \(\left{3}^{{{{\rm{LLL}}}}}\right\rangle\) and \(\left{4}^{{{{\rm{LLLR}}}}}\right\rangle\) are stabilized, in both the γ = 0 topological and trivial phases of H^{KC}, the latter more pronounced. Though not shown in the main text, U_{2} has a similar but weaker effect (see Supplementary Note 4). We are thus able to engineer, in both H^{SSH} and (both BDI and Dclasses of) H^{KC}, enhanced stability of edge modes that goes beyond what can be achieved from standard topology alone. These results demonstrate how NNN Hubbard interactions can lead to avenues of interesting dynamical behaviour, which is not wholly surprising, given that they are known to lead to new phases like topological Mott phases in other contexts imbued with appropriate lattice structure, for example, the honeycomb lattice^{69}.
Finally, we present the full twofermion band structure of the interacting SSH model (Eq. (4)) reconstructed on quantum hardware using IQPE (Fig. 3g). Similar results are obtained against U_{5} instead of U_{2} (Supplementary Note 4). With strong interactions, an additional band of energies whose large gap scales linearly with interaction strength appears. However, the interpretation of this additional set of bands is not straightforward. The interacting Hamiltonian does not share the same eigenstates as its noninteracting counterpart; in fact differences in eigenstate wavefunctions are drastic even for modest U ~ v, w comparable to the noninteracting H^{SSH}. One therefore cannot hope to understand the separation of the band purely using the noninteracting eigenstates. Moreover, ED reveals that the band does not necessarily contain 2fermion boundarylocalized modes, even though initial 2fermion boundarylocalized states can be stabilized in some cases. The nearperfect fidelity of these stabilized 2particle boundary modes, some representatives having been previously demonstrated (Fig. 3), is due to strong overlap with one of the many eigenstates \(\left\psi \right\rangle\) of the interacting Hamiltonian; but the location of \(\left\psi \right\rangle\) in the band structure may be in either set of bands.
Discussion and outlook
By realizing the interacting extended Kitaev Chain on IBM quantum computers, we have demonstrated how various boundary states can be robustly preserved by topological mechanisms of BDI and Dclass symmetries. Importantly, this topological protection is not limited to topological eigenstates, and interplays nontrivially with 2body interactions and Pauli exclusion statistics in multifermion settings. Most spectacularly, we discovered avenues where interactions allow more fermions to be stabilized at one boundary than suggested by the number of available topological modes alone. Our work also illustrates how tensor networkaided circuit recompilation techniques beyond traditional trotterization enable the simulation and full band structure reconstruction of complex topological Hamiltonians on quantum circuits. Our approach can be further extended to more sophisticated manybody interacting Hamiltonians, presenting new opportunities and raising the stateoftheart in the quantum simulation of strongly correlated topological systems.
Methods
Quantum simulation of state evolution
Given an initial state \(\left{\psi }_{0}\right\rangle\) and a timeindependent Hamiltonian H ∈ {H^{SSH}, H^{KC}}, the timeevolved state is \(\left\psi (t)\right\rangle =U(t)\left{\psi }_{0}\right\rangle\), with propagator U(t) = e^{−iHt}. A schematic of the quantum circuit performing timeevolution is given in Fig. 5a. Traditional implementation of U(t) on quantum circuits entails expanding the Hamiltonian H in the spin1/2 basis and employing trotterization^{70,71}; but acceptable trotterization error requires numerous time steps, yielding infeasibly deep circuits. Furthermore, each trotter step, comprising terms \({e}^{i{\beta }_{{{{\boldsymbol{\sigma }}}}}{{{\boldsymbol{\sigma }}}}{{\Delta }}t}\) for Pauli strings σ and coefficients β_{σ} (see Supplementary Note 3), requires layers of entangling gates scaling with the weight of σ, thus impairing its usefulness for models with longerrange couplings.
To transcend these limitations, we employ an implementation strategy known as circuit recompilation, built upon prior studies^{50,51,52}. A circuit ansatz (Fig. 5c) comprising n_{L} ≤ 8 repetitive layers of general U_{3} rotation and CX entangling gates is iteratively optimized, through tensor networkbased quantum simulation^{72}, to approach the intended unitary. Specifically, the U_{3} angles χ = (θ, ϕ, λ) are finetuned with LBFGSB with basinhopping (see Supplementary Note 3). We design the circuit ansatz to contain no longerrange CXs, in order to conform to qubit connectivity on hardware, and the CXs are placed in a brickwork pattern to maximize entangling power, to accommodate the intended unitary in as few layers as possible. This technique yields orderofmagnitude shallower circuits than trotterization at comparable error rates, and is critical in our acquisition of highquality experiment data on NISQ hardware. For larger fermion numbers (≥3), we improve the optimization procedure to focus on the relevant sectors, referred to as sectorspecific recompilation, so as to overcome the impairment in performance brought about by the increased entanglement in evolved states.
We employ error mitigation techniques to improve the quality of hardware results—first, readout error mitigation (RO) reverses bitflips during measurement based on prior calibration of qubits^{25,53}; second, postselection (PS) is performed on particle number^{25,54}. Indeed, since \({H}_{{{{\rm{full}}}}}^{{{{\rm{SSH}}}}}\) and \({H}_{{{{\rm{full}}}}}^{{{{\rm{KC}}}}}\) are numberconserving, timeevolved states that fall outside the particle number sector of \(\left{\psi }_{0}\right\rangle\) are unphysical and can be discarded; no additional circuit depth is incurred since the measurements for O suffice. To feasibly accommodate the number of qubits n used in our simulations (7 ≤ n ≤ 13), we adopt a tensored RO scheme that differs from the opensource implementation in Qiskit^{22}. See Supplementary Note 2 for an introductory overview of quantum computation, and Supplementary Note 3 for technical details of our abovementioned quantum simulation methods.
We end the section with a conceptual remark pertaining to the mapping between the fermionic models and the spin1/2 qubits on the quantum computers. In traditional trotterization, the JordanWigner or BravyiKitaev transforms^{70,71} map the fermionic Hamiltonian into spin chains; the constituent Pauli strings are then naturally expressible on quantum circuits. In comparison, using circuit recompilation, instead of assembling quantum circuits from a fixed set of formulae, an approximate mapping between the fermionic propagator and spin1/2 quantum gates is dynamically determined during optimization.
Iterative quantum phase estimation
Given \(U\left\psi \right\rangle ={e}^{2\pi i\phi }\left\psi \right\rangle\) for unitary U and eigenstate \(\left\psi \right\rangle\), IQPE estimates the eigenphase ϕ ∈ [0, 1), in principle to arbitrary precision. Setting U = e^{−iHt} allows the inference of eigenenergy E = −2πϕ/t of \(\left\psi \right\rangle\). As mentioned, compared to QPE^{57,58}, IQPE circuits are shallower and require fewer qubits—only a single ancilla qubit and a single controlledunitary block is required. There is no need for multiqubit inverse Fourier transforms. Truncating the binary expansion ϕ = 0. ϕ_{1}ϕ_{2}…ϕ_{m} to m bits, IQPE iterates from k = m to k = 1; in iteration k, a controlled\({U}^{{2}^{k1}}\) block and a feedback R_{z}(ω_{k}) = −2π(0.0ϕ_{k+1}ϕ_{k+2}…ϕ_{m}) rotation are applied, and the ancilla qubit is measured to determine ϕ_{k}. An IQPE circuit diagram is shown in Fig. 5b.
To minimize circuit depth, the initialization of \(\left\psi \right\rangle\) and the controlledunitary block are implemented via recompilation; RO mitigation is applied to all qubits, and PS is applied to the simulation qubits to select for specific particle number sectors. Indeed, by performing IQPE over numerous \(\left\psi \right\rangle\)—which need not be exact eigenstates, since superposition states collapse at measurement and yield expected eigenenergies nonetheless—the band structure of the Hamiltonian H can be recovered at arbitrarily high resolution. See Supplementary Note 3 for further implementation details.
Data availability
The data that support the findings of this study are available from the corresponding authors upon reasonable request.
Code availability
The code used in this study is available from the corresponding authors upon reasonable request.
References
Orenstein, J. & Millis, A. Advances in the physics of hightemperature superconductivity. Science 288, 468–474 (2000).
Bolotin, K. I., Ghahari, F., Shulman, M. D., Stormer, H. L. & Kim, P. Observation of the fractional quantum hall effect in graphene. Nature 462, 196–199 (2009).
Lee, C. H., Ho, W. W., Yang, B., Gong, J. & Papić, Z. Floquet mechanism for nonabelian fractional quantum hall states. Phys. Rev. Lett. 121, 237401 (2018).
Hasan, M. Z. & Kane, C. L. Colloquium: Topological insulators. Rev. Mod. Phys. 82, 3045–3067 (2010).
Ozawa, T. et al. Topological photonics. Rev. Mod. Phys. 91, 015006 (2019).
Yang, Z. et al. Topological acoustics. Phys. Rev. Lett. 114, 114301 (2015).
Ma, G., Xiao, M. & Chan, C. T. Topological phases in acoustic and mechanical systems. Nat. Rev. Phys. 1, 281–294 (2019).
Lee, C. H. et al. Imaging nodal knots in momentum space through topolectrical circuits. Nat. Commun. 11, 1–13 (2020).
Bernien, H. et al. Probing manybody dynamics on a 51atom quantum simulator. Nature 551, 579–584 (2017).
Slussarenko, S. & Pryde, G. J. Photonic quantum information processing: a concise review. Appl. Phys. Rev. 6, 041303 (2019).
Yang, C. H. et al. Silicon qubit fidelities approaching incoherent noise limits via pulse engineering. Nat. Electron. 2, 151–158 (2019).
Bruzewicz, C. D., Chiaverini, J., McConnell, R. & Sage, J. M. Trappedion quantum computing: Progress and challenges. Appl. Phys. Rev. 6, 021314 (2019).
Feynman, R. P. Simulating physics with computers. Int. J. Theor. Phys. 21, 467–488 (1982).
Arute, F. et al. Quantum supremacy using a programmable superconducting processor. Nature 574, 505–510 (2019).
Zhong, H.S. et al. Quantum computational advantage using photons. Science 370, 1460–1463 (2021).
Roushan, P. et al. Observation of topological transitions in interacting quantum circuits. Nature 515, 241–244 (2014).
Quantum, G. A. et al. Hartreefock on a superconducting qubit quantum computer. Science 369, 1084–1089 (2020).
Kairys, P. et al. Simulating the shastrysutherland ising model using quantum annealing. PRX Quantum 1, 020320 (2020).
Chiesa, A. et al. Quantum hardware simulating fourdimensional inelastic neutron scattering. Nat. Phys. 15, 455–459 (2019).
LaRose, R. Overview and comparison of gate level quantum software platforms. Quantum 3, 130 (2019).
Blinov, S., Wu, B. & Monroe, C. Comparison of cloudbased ion trap and superconducting quantum computer architectures. Preprint at https://arxiv.org/abs/2102.00371 (2021).
Andersson, S. et al. Learn quantum computation using qiskit. http://community.qiskit.org/textbook (2020).
Aleksandrowicz, G. et al. Qiskit: an opensource framework for quantum computing. https://doi.org/10.5281/zenodo.2562111 (2019).
Zhukov, A. A., Remizov, S. V., Pogosov, W. V. & Lozovik, Y. E. Algorithmic simulation of farfromequilibrium dynamics using quantum computer. Quantum Inf. Process. 17, 223 (2018).
Smith, A., Kim, M. S., Pollmann, F. & Knolle, J. Simulating quantum manybody dynamics on a current digital quantum computer. npj Quantum Inf. 5, 106 (2019).
McCaskey, A. J. et al. Quantum chemistry as a benchmark for nearterm quantum computers. npj Quantum Inf. 5, 99 (2019).
Choo, K., Von Keyserlingk, C. W., Regnault, N. & Neupert, T. Measurement of the entanglement spectrum of a symmetryprotected topological state using the ibm quantum computer. Phys. Rev. Lett. 121, 086808 (2018).
Smith, A., Jobst, B., Green, A. G. & Pollmann, F. Crossing a topological phase transition with a quantum computer. Preprint at https://arxiv.org/abs/1910.05351 (2019).
Azses, D. et al. Identification of symmetryprotected topological states on noisy quantum computers. Phys. Rev. Lett. 125, 120502 (2020).
Mei, F. et al. Digital simulation of topological matter on programmable quantum processors. Phys. Rev. Lett. 125, 160503 (2020).
Havlíček, V. et al. Supervised learning with quantumenhanced feature spaces. Nature 567, 209–212 (2019).
Behera, B. K., Reza, T., Gupta, A. & Panigrahi, P. K. Designing quantum router in ibm quantum computer. Quantum Inf. Process. 18, 328 (2019).
Preskill, J. Quantum computing in the nisq era and beyond. Quantum 2, 79 (2018).
Lee, C. H. et al. Topolectrical circuits. Commun. Phys. 1, 1–9 (2018).
Haldane, F. D. M. Model for a quantum hall effect without landau levels: condensedmatter realization of the “parity anomaly". Phys. Rev. Lett. 61, 2015–2018 (1988).
Ezawa, M. Higherorder topological insulators and semimetals on the breathing kagome and pyrochlore lattices. Phys. Rev. Lett. 120, 026801 (2018).
Bandres, M. A. et al. Topological insulator laser: experiments. Science 359, eaar4005 (2018).
Budich, J. C. & Bergholtz, E. J. Nonhermitian topological sensors. Phys. Rev. Lett. 125, 180403 (2020).
Li, L., Mu, S., Lee, C. H. & Gong, J. Quantized classical response from spectral winding topology. Nat. Commun. 12, 1–11 (2021).
Leykam, D. & Smirnova, D. A. Probing bulk topological invariants using leaky photonic lattices. Nat. Phys. 17, 632–638 (2021).
Li, L., Yang, C. & Chen, S. Topological invariants for phase transition points of onedimensional z 2 topological systems. Eur. Phys. J. B 89, 195 (2016).
Lee, C. H., Li, L., Thomale, R. & Gong, J. Unraveling nonhermitian pumping: emergent spectral singularities and anomalous responses. Phys. Rev. B 102, 085151 (2020).
Kitaev, A. Y. Unpaired majorana fermions in quantum wires. Phys.Uspekhi 44, 131 (2001).
Aasen, D. et al. Milestones toward majoranabased quantum computing. Phys. Rev. X 6, 031016 (2016).
Ryu, S., Schnyder, A. P., Furusaki, A. & Ludwig, A. W. Topological insulators and superconductors: tenfold way and dimensional hierarchy. N. J. Phys. 12, 065010 (2010).
Su, W. P., Schrieffer, J. R. & Heeger, A. J. Solitons in polyacetylene. Phys. Rev. Lett. 42, 1698–1701 (1979).
Meier, E. J., An, F. A. & Gadway, B. Observation of the topological soliton state in the SuSchriefferHeeger model. Nat. Commun. 7, 13986 (2016).
Bagrets, D. & Altland, A. Class d spectral peak in majorana quantum wires. Phys. Rev. Lett. 109, 227005 (2012).
Terhal, B. M. & DiVincenzo, D. P. Classical simulation of noninteractingfermion quantum circuits. Phys. Rev. A 65, 032325 (2002).
Sun, S.N. et al. Quantum computation of finitetemperature static and dynamical properties of spin systems using quantum imaginary time evolution. PRX Quantum 2, 010317 (2021).
Khatri, S. et al. Quantumassisted quantum compiling. Quantum 3, 140 (2019).
Heya, K., Suzuki, Y., Nakamura, Y. & Fujii, K. Variational quantum gate optimization. Preprint at https://arxiv.org/abs/1810.12745 (2018).
Kandala, A. et al. Error mitigation extends the computational reach of a noisy quantum processor. Nature 567, 491–495 (2019).
McArdle, S., Yuan, X. & Benjamin, S. Errormitigated digital quantum simulation. Phys. Rev. Lett. 122, 180501 (2019).
Dobšíček, M., Johansson, G., Shumeiko, V. & Wendin, G. Arbitrary accuracy iterative quantum phase estimation algorithm using a single ancillary qubit: A twoqubit benchmark. Phys. Rev. A 76, 030306 (2007).
Mohammadbagherpoor, H. et al. An improved implementation approach for quantum phase estimation on quantum computers. In 2019 IEEE International Conference on Rebooting Computing (ICRC), 1–9 (IEEE, 2019).
Whitfield, J. D., Biamonte, J. & AspuruGuzik, A. Simulation of electronic structure hamiltonians using quantum computers. Mol. Phys. 109, 735–750 (2011).
AspuruGuzik, A., Dutoi, A. D., Love, P. J. & HeadGordon, M. Simulated quantum computation of molecular energies. Science 309, 1704–1707 (2005).
Wang, C., Potter, A. C. & Senthil, T. Classification of interacting electronic topological insulators in three dimensions. Science 343, 629–631 (2014).
Rachel, S. Interacting topological insulators: a review. Rep. Prog. Phys. 81, 116501 (2018).
Tang, E. & Wen, X.G. Interacting onedimensional fermionic symmetryprotected topological phases. Phys. Rev. Lett. 109, 096403 (2012).
Shitade, A. et al. Quantum spin hall effect in a transition metal oxide na_{2}iro_{3}. Phys. Rev. Lett. 102, 256403 (2009).
Pesin, D. & Balents, L. Mott physics and band topology in materials with strong spin–orbit interaction. Nat. Phys. 6, 376–381 (2010).
Jünemann, J. et al. Exploring interacting topological insulators with ultracold atoms: the synthetic creutzhubbard model. Phys. Rev. X 7, 031057 (2017).
Verbin, M., Zilberberg, O., Kraus, Y. E., Lahini, Y. & Silberberg, Y. Observation of topological phase transitions in photonic quasicrystals. Phys. Rev. Lett. 110, 076403 (2013).
Grusdt, F., Höning, M. & Fleischhauer, M. Topological edge states in the onedimensional superlattice bosehubbard model. Phys. Rev. Lett. 110, 260405 (2013).
Kuno, Y. Phase structure of the interacting SuSchriefferHeeger model and the relationship with the grossneveu model on lattice. Phys. Rev. B 99, 064105 (2019).
Marques, A. M. & Dias, R. G. Multihole edge states in SuSchriefferHeeger chains with interactions. Phys. Rev. B 95, 115443 (2017).
Raghu, S., Qi, X.L., Honerkamp, C. & Zhang, S.C. Topological mott insulators. Phys. Rev. Lett. 100, 156401 (2008).
Georgescu, I. M., Ashhab, S. & Nori, F. Quantum simulation. Rev. Mod. Phys. 86, 153–185 (2014).
Ortiz, G., Gubernatis, J. E., Knill, E. & Laflamme, R. Quantum algorithms for fermionic simulations. Phys. Rev. A 64, 022319 (2001).
Gray, J. quimb: A python package for quantum information and manybody calculations. J. Open Source Softw. 3, 819 (2018).
Acknowledgements
J.M.K. thanks ShiNing Sun of Caltech for helpful discussions on quantum computing and algorithms. We acknowledge the use of IBM Quantum services for this work. The views expressed are those of the authors, and do not reflect the official policy or position of IBM or the IBM Quantum team.
Funding
This work is supported by MOE Tier 1 grant 210048A00010.
Author information
Authors and Affiliations
Contributions
C.H.L. initiated and supervised the project, and wrote parts of the manuscript. J.M.K. developed most of the quantum simulation codebase, ran experiments on emulators and hardware, and wrote most of the manuscript. T.T. contributed to the codebase, ran experiments, and wrote parts of the manuscript. Y.H.P. and W.E.N. provided support on emulators and hardware usage, and edited the manuscript. All authors analyzed computational and experiment results. The manuscript reflects the contributions of all authors.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Koh, J.M., Tai, T., Phee, Y.H. et al. Stabilizing multiple topological fermions on a quantum computer. npj Quantum Inf 8, 16 (2022). https://doi.org/10.1038/s41534022005271
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41534022005271
This article is cited by

Observing and braiding topological Majorana modes on programmable quantum simulators
Nature Communications (2023)

Quenched topological boundary modes can persist in a trivial system
Communications Physics (2021)