Quantum complexity of topological phases. Or lack thereof Alberto Giuseppe Catalano, Sven Benjamin Kožić, Gianpaolo Torre, Carola Ciaramelletti, Simone Paganelli, Fabio Franchini, Salvatore Marco Giampaolo Abstract Topological phases are expected to carry enhanced computational power, stemming from the long-range entanglement that characterizes them. Here we investigate whether this advantage is reflected in quantum magic, quantified through stabilizer Rényi entropies (SREs). We analyze one-dimensional models with an exact duality between a symmetry-protected topological phase and a trivial one. Surprisingly, a finite (parameters dependent) asymmetry appears only when boundary conditions break the duality, and emerges as well in the transverse-field Ising chain, which lacks topological order. These results complement recent analysis in the literature and show that the total magic of a state is unable to detect topological phases, indicating either that a different property is required to capture their unique quantum contribution, or that not all forms of long-range entanglement constitute a genuine computational resource. One of the central challenges in quantum information science is to identify and characterize the fundamental properties that a quantum state must possess in order to enable operations that outperform classical computation — in other words, to achieve quantum computational advantage or quantum supremacy Harrow and Montanaro 2017 . For a long time, the characterization of these states focused predominantly on their entanglement properties Jozsa and Linden 2003 ; Horodecki et al. 2009 . However, over time, it became increasingly evident that entanglement alone does not provide a complete picture. Notably, certain highly entangled states can still be efficiently simulated on classical computers Terhal and DiVincenzo 2002 , while some computational models exhibit a quantum advantage even in the absence of significant entanglement Knill and Laflamme 1998 . This realization has led to the identification and study of several additional resources over the years Gour 2025 . Among them, the so-called quantum nonstabilizerness, also known as quantum magic Seddon and T. 2019 ; Howard et al. 2014 ; Veitch et al. 2014 ; Anders and Browne 2009 , has acquired a relevant role in the past years. Quantum magic quantifies the extent to which a state departs from the stabilizer manifold, i.e., the manifold of the states that can be obtained from a reference one using circuits containing only Clifford gates Aaronson and Gottesman 2004 , a subset of quantum operations that can be efficiently simulated on a classical computer. In this sense, magic captures a form of non-classicality necessary for universal quantum computation. This issue is intimately connected to the study of quantum many-body systems, whose ground states often serve as natural candidates for implementing a wide range of quantum information processing tasks. As a result, quantum magic is emerging as an operational criterion to distinguish phases of matter that are classically simulable from those that support a genuine quantum advantage Oliviero et al. 2022 ; Odavić et al. 2023 ; Catalano et al. 2025 , much like entanglement has already done Amico et al. 2008 ; Latorre and Riera 2009 ; Eisert et al. 2010 ; De Chiara et al. 2012 ; Laflorencie 2016 . Among others, topological phases of matter are widely expected to possess significant computational power due to their robustness against local perturbations and their ability to encode quantum information non-locally Nayak et al. 2008 . These phases exhibit exotic correlations not captured by local order parameters and can support computational models such as topological quantum computation, where logical gates are implemented through the braiding of anyonic excitations Kitaev 2003 ; Freedman et al. 2003 . In one-dimensional systems, the topological order is more fragile, but a weaker analog exists in the form of symmetry-protected topological phases (SPTPs) Chen et al. 2011 ; Pollmann et al. 2012 . Examples include the one-dimensional cluster-Ising model Son et al. 2011 ; Smacchia et al. 2011 ; Giampaolo and Hiesmayr 2014 ; Marić et al. 2021 and the dimerized XX chain Campos Venuti et al. 2007 ; Bahovadinov et al. 2019 , which can be obtained from the Su–Schrieffer–Heeger (SSH) model Su et al. 1979 ; Batra and Sheet 2020 ; Ciaramelletti et al. 2025 via a Jordan–Wigner transformation. These systems display non-trivial entanglement structures and symmetry-protected edge modes; as long as symmetry-breaking perturbations are excluded, they retain their topological character and can outperform trivial phases. In particular, they support non-trivial forms of quantum computation, particularly in measurement-based schemes Raussendorf et al. 2006 ; Stephen et al. 2017 . Most importantly, as argued in Chen et al. 2010 ; Chen et al. 2011 and elaborated in Cui et al. 2012 ; Franchini et al. 2014 , the long-range entanglement of the SPTPs is a precious resource that cannot be generated from trivial ones without implementing circuits whose depth diverges with the system’s size. Thus, only non-trivial topological phases can serve as universal quantum simulators. If computational complexity classes are equivalent, this implies that SPTPs should carry a higher intrinsic computational complexity, measurable as an enhanced amount of quantum magic Howard et al. 2014 ; Seddon and T. 2019 . It should be remarked, however, that recent works have realized that by allowing non-unitary processes, such as measurements, in a circuit, local entanglement can be promoted to global entanglement Piroli et al. 2021 ; Piroli et al. 2024 . In this way, (abelian) topological order can be generated using much shallower circuits, thus providing a potential loophole to the presumed primacy of topological order Lu et al. 2022 ; Tantivasadakarn et al. 2023 ; Smith et al. 2023 ; Smith et al. 2024 ; Sahay and Verresen 2025 . In this letter, we test the relation between complexity and topological order by calculating the quantum magic of one-dimensional SPTPs and comparing it with that of their topologically trivial counterparts. We focus on models that possess a duality transformation connecting a SPTP to a trivial one, and find that their magic coincides, except when boundary conditions force a violation of the duality symmetry. In such cases, we observe a finite, parameter-dependent difference in quantum magic that originates from boundary effects that tend to be independent of the system size in the thermodynamic limit. The inability of the total magic to detect the long range order of a state highlights that a single number is inefficient in characterizing the various resources hidden in a state complexity. Indeed, recent studies have also focused on long-range nonstabilizerness (LRM), that is, the magic that cannot be removed by local, shallow circuits Ellison et al. 2021 ; Korbany et al. 2025 ; Wei and Liu 2025 and found that topological order comes in different fashions, which may or may not possess LRM. However, it should be pointed out that such classifications arise within a circuit-based approach, which connects the circuit simulability with an intrinsic property of a state. Such a computational point of view is indeed of growing interest in the community, but in this work we are more interested in directly assessing a state’s complexity and what it can tell us. Figure 1: Behavior of \mathcal{M}_{2} for the SSH/dimerized XX model in Eq. ( 2 ) as a function of the dimerization parameter \delta for different system sizes under OBCs. The maximum lies in the SPTP and shifts toward the critical point \delta=0 as the chain length increases. Before proceeding with our analysis, we note that, throughout this letter, magic is quantified using the stabilizer Rényi entropies (SREs) Leone et al. 2022 . For a pure state \ket{\psi} of L qubits, the SRE of order \alpha is defined as \mathcal{M}_{\alpha}(\ket{\psi})=\frac{1}{1-\alpha}\log_{2}{\left(\frac{1}{2^{L}}\sum_{\mathcal{P}}|\bra{\psi}\mathcal{P}\ket{\psi}|^{2\alpha}\right)}. (1) Here, \mathcal{P}=\bigotimes_{i=1}^{L}P_{i} runs over all possible Pauli strings, P_{i}\in\{\sigma^{0}_{i},\sigma_{i}^{x},\sigma_{i}^{y},\sigma_{i}^{z}\} with \sigma_{i}^{0} representing the identity operator, and the order \alpha is assumed to be equal or greater than 1 Haug and Piroli 2023a . Moreover, since we focus on one-dimensional models away from criticality, our numerical analysis relies on tensor-network methods based on matrix product states (MPS). Within this framework, ground states were obtained using the density-matrix renormalization group (DMRG) algorithm Schollwöck 2011 ; Orús 2014 , and their SRE was then evaluated using two complementary approaches Tarabunga et al. 2024 ; Kožić and Torre 2025 . For ground-state determination, a modest bond dimension ( \chi\approx 50 ) was sufficient, while the MPS in the Pauli basis used for the SRE evaluation was truncated at \chi\approx 200 - 450 using standard variational compression techniques Paeckel et al. 2019 . We begin our analysis with the dimerized XX model, which can be obtained from the SSH model via a Jordan-Wigner transformation. Both models serve as paradigmatic examples of one-dimensional systems exhibiting SPTPs Su et al. 1979 ; Batra and Sheet 2020 . Under open boundary conditions (OBCs), the Hamiltonian of the dimerized XX model takes the form H_{DXX}=\sum_{k=1}^{L-1}\frac{1+(-1)^{k}\delta}{2}\left(\sigma^{x}_{k}\sigma^{x}_{k+1}+\sigma^{y}_{k}\sigma^{y}_{k+1}\right), (2) where L denotes the even length of the spin chain and \delta\in[-1,1] controls the dimerization. For \delta\in[-1,0) , the system lies in a topologically trivial phase, while for \delta\in(0,1] , it enters a topologically non-trivial phase Asbóth et al. 2016 , characterized by the presence of localized edge states and long-range entanglement between the chain endpoints Campos Venuti et al. 2007 . Motivated by the considerations made in the introduction, we expect the ground states in the topologically ordered phase to exhibit greater quantum magic than those in the trivial phase. In particular, for a fixed value \bar{\delta}>0 , we expect the magic \mathcal{M}_{2} of the ground state at \delta=\bar{\delta} to exceed that of the corresponding ground state at \delta=-\bar{\delta} . To test this hypothesis, we computed \mathcal{M}_{2} as a function of \delta for various system sizes L , with the results displayed in Fig. 1 . From the figure, it is immediately evident that quantum magic exhibits a pronounced peak within the topologically ordered phase, which grows linearly in size, shifting toward the expected critical point \delta=0 as L increases. Similar behaviors are well known in one-dimensional quantum systems. For instance, a closely analogous trend has been reported in the derivative of the concurrence for the transverse-field Ising model Osterloh et al. 2002 . Figure 2: Behavior of \Delta\mathcal{M}_{2}(\delta)=\mathcal{M}_{2}(\delta)-\mathcal{M}_{2}(-\delta) , as function of \delta>0 for the model in Eq. ( 2 ). For small L , \Delta\mathcal{M}_{2} shows a size dependence, but as L increases the curves start to converge to a non-constant limiting function of \delta . This finite, parameter-dependent asymmetry is inconsistent with the behavior of a genuine topological invariant. The self-dual point \delta=0 is excluded, since \Delta\mathcal{M}_{2}=0 by definition. To isolate the potential topological contribution to SRE, for each L we define the difference in magic between the ground states at symmetric points with respect to the critical point \Delta\mathcal{M}_{2}(\delta)=\mathcal{M}_{2}(\delta)-\mathcal{M}_{2}(-\delta). These differences are shown in Fig. 2 . For \delta>0 \Delta\mathcal{M}_{2}(\delta) remains consistently positive and rapidly converges to a finite value as the system size increases. For L\geq 30 , the curves start to collapse onto a single, well-defined limiting function except close to \delta=0 , as the correlation length becomes comparable with the system sizes. However, this limiting function retains a non-trivial dependence on \delta , suggesting that the asymmetry in quantum magic does not have a topological origin. In fact, topological quantities are expected to be quantized and constant within a given macroscopic phase Micallo et al. 2020 ; Hamma et al. 2005 ; Halász and Hamma 2012 ; Torre et al. 2024 ; Odavić et al. 2023 , whereas our results feature a smooth dependence on the microscopical parameters. This suggests that the observed difference in quantum complexity is not due to topological order, but rather originates from the violation of duality symmetry induced by OBCs. To check this hypothesis, we repeated our evaluation assuming periodic boundary conditions (PBCs), which can be implemented by adding the term \tfrac{1+\delta}{2},(\sigma^{x}_{1}\sigma^{x}_{L}+\sigma^{y}_{1}\sigma^{y}_{L}) to the Hamiltonian in Eq. ( 2 ). Under these conditions, the dimerized XX model retains an exact duality symmetry around \delta=0 Asbóth et al. 2016 , since the cases \delta>0 and \delta<0 are related by a one-site translation. Consequently, Pauli strings are biunivocally mapped into Pauli strings, and it can be proven (and we numerically checked), that under PBCs the SRE remains symmetric between the two phases, i.e., \Delta\mathcal{M}_{2}(\delta)=0 for all \delta . The non-constant asymmetry observed with OBCs, and its absence with PBCs, therefore demonstrates that the increased complexity arises from a violation of duality symmetry rather than from topological order. Figure 3: Behavior of the magic \mathcal{M}_{2} for the Cluster Ising model in Eq. ( 3 ) as a function of the parameter \lambda . To assess whether the observed phenomenon is specific to the dimerized XX model or indicative of a broader trend among one-dimensional spin systems hosting SPTPs, we extended our analysis to the so-called cluster-Ising model. The cluster-Ising model Smacchia et al. 2011 ; Son et al. 2011 ; Doherty and Bartlett 2009 is a one-dimensional spin- 1/2 system whose Hamiltonian, assuming OBCs, is given by H_{\text{CI}}=-\sum_{i=1}^{L-2}\sigma_{i}^{x}\sigma_{i+1}^{z}\sigma_{i+2}^{x}+\lambda\sum_{i=1}^{L-1}\sigma_{i}^{y}\sigma_{i+1}^{y}. (3) In the thermodynamic limit, the model hosts an SPTP – known as the cluster phase – when the three-body cluster interaction dominates over the Ising coupling. In this regime, edge states appear at the boundaries. Conversely, for \lambda>1 , where the Ising term prevails, the system enters a topologically trivial antiferromagnetic phase. Similarly to the SSH model, also the Cluster Ising chain is known to possess a duality symmetry, realized between the points \lambda and 1/\lambda Smacchia et al. 2011 . As shown in Fig. 3 , this system exhibits a qualitatively similar trend to that observed in the dimerized XX/SSH model. By defining the quantity \Delta\mathcal{M}_{2}(\lambda)=\mathcal{M}_{2}(\lambda)-\mathcal{M}_{2}(1/\lambda) – which again measures the difference in quantum magic between ground states at dual points – we observe in Fig. 4 that for all \lambda\in[0,1) , this difference rapidly converges to a limiting curve as the system size increases. However, this limiting function retains a smooth, non-universal dependence on \lambda , precluding an interpretation in terms of a topological invariant. Furthermore, consistent with the findings for the dimerized XX model, when the boundary conditions are changed from open to periodic – closing the loop by adding to the Hamiltonian in Eq. ( 3 ) the terms -\sigma_{L-1}^{x}\sigma_{L}^{z}\sigma_{1}^{x}-\sigma_{L}^{x}\sigma_{1}^{z}\sigma_{2}^{x}+\lambda\sigma_{1}^{y}\sigma_{L}^{y} – the quantity \Delta\mathcal{M}_{2}(\lambda) vanishes identically. PBCs restore the exact duality \lambda\leftrightarrow 1/\lambda and remove the possibility of observing edge states, but they do not destroy the topological nature of the \lambda<1 phase, which remains signaled by a finite string-order parameter Smacchia et al. 2011 . Thus, exactly dual points have the same SRE (since, once again, the duality trasformation is of the Clifford type), demonstrating the inability of quantum magic to detect the presence of topological order. Once again, the difference in SRE arises solely from the violation of duality symmetry at the boundaries. We remark that this is in line with the results of Wei and Liu 2025 , where it was proven that in 1D cluster states have high computational complexity from the point of view of a Montecarlo simulation (due to their sign problem structure), but not from a Clifford resource point of view. Figure 4: Behavior of the magic difference between dual points \lambda\leftrightarrow 1/\lambda for the cluster-Ising model in Eq. ( 3 ) as a function of \lambda , with OBCs. For large system sizes L , the curve converges to a finite, \lambda -dependent value, indicating a non-topological origin. As a final check, let us now consider a model with an exact duality between two phases but without topological order: the transverse-field Ising chain. Its Hamiltonian with OBCs is H_{TI}=-\sum_{i=1}^{L-1}\sigma_{i}^{z}\sigma_{i+1}^{z}-h\sum_{i=1}^{L}\sigma_{i}^{x}. (4) This model undergoes a quantum phase transition at h=1 , separating a ferromagnetic phase ( h<1 ) from a paramagnetic phase ( h>1 ) Barouch and McCoy 1971 ; Franchini 2017 . Crucially, although edge modes can be isolated in the ferromagnetic phase, the asymptotic degeneracy between the two lowest-energy states makes their long-range correlations fragile, and therefore the Ising chain does not host an SPTP Pfeuty 1970 ; Kitaev 2001 . This model is also well known for its self-duality h\leftrightarrow 1/h around the critical point h=1 , via the Kramers–Wannier transformation Kramers and Wannier 1941 . However, the duality is exact only assuming PBCs, obtained by adding the term -\sigma_{1}^{z}\sigma_{L}^{z} to Eq. ( 4 ). Figure 5: Behavior of \mathcal{M}_{2} for the Ising model in eq. ( 4 ), as a function of the transverse magnetic field h , with OBCs. Although this model does not support topological order, quantum magic behaves surprisingly similarly to the topological model considered before. As can be appreciated from Fig. 5 , the behavior of quantum magic \mathcal{M}_{2} for the ground state of the transverse field Ising model is qualitatively the same as that of the previous models: with OBCs, \mathcal{M}_{2} has a peak inside the ferromagnetic phase, which grows and moves towards h=1 as the system size is increased. This behavior is sensibly different from what was obtained in Oliviero et al. 2022 , where the same model was analyzed assuming PBCs, i.e. with exact dual symmetry, and agrees with what we get under the same conditions: the maximum is always centered at h=1 . Moreover, the quantity \Delta\mathcal{M}_{2}(h)=\mathcal{M}_{2}(h)-\mathcal{M}_{2}(1/h) ( h\in[0,1) ) vanishes identically when the duality symmetry holds, independently of h and L because the Kramers-Wannier duality is magic-preserving. Instead, as also observed for the other models, breaking the duality induces an asymmetry in the SRE. As shown in Fig. 6 , with OBCs \Delta\mathcal{M}_{2}(h) vanishes only in the limit of h\rightarrow 0 . Otherwise, its value depends on L but the different curves tend to collapse as soon as the dimension of the system becomes greater than the correlation length. Obviously, since no topological phase is present in this model, the existence of such a difference cannot be justified by the presence of topological order. Figure 6: Difference \Delta\mathcal{M}_{2} between SRE calculated at mirror point h\longleftrightarrow 1/h for the Ising chain in eq. ( 4 ) as a function of the external magnetic field h , with OBCs. As observed for the topological models, \Delta\mathcal{M}_{2} is finite and non-constant, growing approaching the critical point h=1 , although this model does not support topological order. Model Ham. Boundary 1 SPTP Exact SRE Eq. Conditions phase Duality Asymmetry Dimerized XX/ ( 2 ) OBCs ✓ ✗ ✓ SSH PBCs ✓ ✓ ✗ Cluster Ising ( 3 ) OBCs ✓ ✗ ✓ PBCs ✓ ✓ ✗ Ising Chain ( 4 ) OBCs ✗ ✗ ✓ PBCs ✗ ✓ ✗ Table 1: Recap of our results: for every model we considered we report whether, with Open or Periodic Boundary Conditions (OBCs and PBCs respectively), it hosts a SPTP or not, if the duality symmetry is exact, and if the SRE is the same between dual points or asymmetric. Before concluding, we emphasize that all evaluations of quantum magic in this Letter were carried out in the computational basis. Since magic is inherently basis-dependent, one may wonder whether our results are tied to this choice or hold more generally. A possible answer to this question is that attempted for instance in Haug and Piroli 2023b ; Ellison et al. 2021 ; Korbany et al. 2025 ; Wei and Liu 2025 : minimizing magic through local unitaries in a circuit-based approach. Indeed in this way one aims at extracting the “most interesting” part of non-stabilizerness. As we are instead looking at the whole complexity of a state, we just tested local unitaries composed of independent single-spin rotations. For small systems ( L\leq 10 ), we found that our framework remains robust: with PBCs, the magic of dual points is identical in any basis, while opening the chain, breaking the duality and inducing the formation of edge states breaks this symmetry. Hence, the physical conclusions presented here are not restricted to a specific basis but acquire a general validity. Finally, from our results we recover that, the computational basis minimizes the magic for all models and parameter choices, and thus coincides with the non-local L -partite magic Cao ; Cao et al. 2024 ; Qian and Wang 2025 . In conclusion, we calculated and analyzed the rank 2 Stabilizer Renyi Entropy \mathcal{M}_{2} to look for the contribution that topological order adds to quantum magic. To isolate this element, we focused on models that have a duality symmetry between a SPTP and a trivial one, such as the dimerized XX model, which is directly mappable onto the SSH model, and the Cluster-Ising chain. The long-range entanglement that characterizes SPTPs is known to be a precious resource that is expected to provide a quantum advantage Chen et al. 2010 ; Chen et al. 2011 , and accordingly we expected an increase in SRE between dual points. Surprisingly, however, we found that they exhibit exactly the same amount of magic. A sub-extensive asymmetry appears only under open boundary conditions, which break the exact duality and introduce small corrections to the SRE. This finite difference is not a constant throughout the phase and as a final proof of its non-topological origin we checked that the quantum Ising chain presents the same phenomenology. These results are summarized in Table 1 . Our findings are rather unexpected and should be contrasted with the case of topological frustration, where a delocalized topological excitation is injected in the ground state by an appropriate choice of boundary conditions Marić et al. 2020 ; Marić et al. 2022 ; Torre et al. 2024 ; Kožić et al. 2025 . In that case, the excitation contributes an additional amount of quantum magic that grows logarithmically with system size compared to the non-frustrated counterpart Odavić et al. 2023 ; Catalano et al. 2025 . The long-range entanglement generated by the topological excitation provides an added complexity, while apparently that of a conventional SPTP does not. Indeed, \mathcal{M}_{2} measures the amount of non-Clifford resources needed to generate a given state Aaronson and Gottesman 2004 . Our results seem to imply that is possible to efficiently generate topological order using the same amount of resources needed for much simpler orders. Such a conclusion seemingly contradicts the expected complexity equivalence between different quantum computation schemes and needs to be understood better. Indeed, to the best of our knowledge, there is no resource theory that takes specific advantage of topological order, leveraging over adiabatic continuation as a free operation and measuring the contribution that crossing a phase transition or changing topology brings. To the contrary, recent results showing how Local Operations and Classical Communications (LOCC) can be utilized to convert local into global entanglement Piroli et al. 2021 ; Piroli et al. 2024 ; Lu et al. 2022 ; Tantivasadakarn et al. 2023 ; Smith et al. 2023 ; Smith et al. 2024 ; Sahay and Verresen 2025 ., indicate that in non-unitary computational schemes, topological order is not a necessary resource. Nonetheless, the classification of these latter proposals within a Clifford resource theory is not straightforward, and thus it is not clear if our findings stem from such approaches or are indicative of something else. It should also be clarified if our results extend to other topological models, even in higher dimensions: in the absence of exact dualities, one could check whether the growth of \mathcal{M}_{2} with the system size deviates from the one expected considering purely local order Leone et al. 2022 . Acknowledgements. The research leading to these results has received funding from the following organizations: European Union via ICSC - Italian Research Center on HPC, Big Data and Quantum Computing (NextGenerationEU Project No. CN00000013), project EuRyQa (Horizon 2020), project PASQuanS2 (Quantum Technology Flagship); Italian Ministry of University and Research (MUR) via: Quantum Frontiers (the Departments of Excellence 2023-2027); the World Class Research Infrastructure - Quantum Computing and Simulation Center (QCSC) of Padova University; Istituto Nazionale di Fisica Nucleare (INFN): iniziativa specifica IS-QUANTUM. S. P. acknowledges financial support from the Ministero dell’Universit‘a e della Ricerca (MUR) and the Project PRIN 2022 number 2022W9W423 funded by the European Union Next Generation EU. SBK acknowledges support from the Croatian Science Foundation (HrZZ) Projects DOK-2020-01-9938.