Arxiv:2009.07590V2 [Quant-Ph] 9 Dec 2020

Emulating quantum teleportation of a Majorana zero mode qubit

He-Liang Huang,1, 2, 3 Marek Naro˙zniak,4, 5 Futian Liang,1, 2, 3 Youwei Zhao,1, 2, 3 Anthony D. Castellano,1, 2, 3 Ming Gong,1, 2, 3 Yulin Wu,1, 2, 3 Shiyu Wang,1, 2, 3 Jin Lin,1, 2, 3 Yu Xu,1, 2, 3 Hui Deng,1, 2, 3 Hao Rong,1, 2, 3 6, 7, 8 1, 2, 3 4, 8, 9, 5 1, 2, 3, 1, 2, 3 Jonathan P. Dowling ∗, Cheng-Zhi Peng, Tim Byrnes, Xiaobo Zhu, † and Jian-Wei Pan 1Hefei National Laboratory for Physical Sciences at the Microscale and Department of Modern Physics, University of Science and Technology of China, Hefei 230026, China 2Shanghai Branch, CAS Center for Excellence in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China 3Shanghai Research Center for Quantum Sciences, Shanghai 201315, China 4New York University Shanghai, 1555 Century Ave, Pudong, Shanghai 200122, China 5Department of Physics, New York University, New York, NY 10003, USA 6Hearne Institute for Theoretical Physics, Department of Physics and Astronomy, Louisiana State University, Baton Rouge, Louisiana 70803, USA 7Hefei National Laboratory for Physical Sciences at Microscale and Department of Modern Physics, University of Science and Technology of China, Hefei, Anhui 230026, China 8NYU-ECNU Institute of Physics at NYU Shanghai, 3663 Zhongshan Road North, Shanghai 200062, China 9State Key Laboratory of Precision Spectroscopy, School of Physical and Material Sciences, East China Normal University, Shanghai 200062, China (Dated: December 10, 2020) Topological quantum computation based on anyons is a promising approach to achieve fault- tolerant quantum computing. The Majorana zero modes in the Kitaev chain are an example of non-Abelian anyons where braiding operations can be used to perform quantum gates. Here we perform a quantum simulation of topological quantum computing, by teleporting a qubit encoded in the Majorana zero modes of a Kitaev chain. The quantum simulation is performed by mapping the Kitaev chain to its equivalent spin version, and realizing the ground states in a superconducting quantum processor. The teleportation transfers the quantum state encoded in the spin-mapped version of the Majorana zero mode states between two Kitaev chains. The teleportation circuit is realized using only braiding operations, and can be achieved despite being restricted to Clifford gates for the Ising anyons. The Majorana encoding is a quantum error detecting code for phase flip errors, which is used to improve the average fidelity of the teleportation for six distinct states from 70.76 0.35% to 84.60 0.11%, well beyond the classical bound in either case. ± ±

One of the most attractive ways of performing fault- 27]. To date, many experiments have been conducted to tolerant quantum computing [1–8] is topological quan- find the evidence for the existence of Majorana fermions, tum computing [9–15]. In topological quantum com- however, the key feature of topological protection has puting, the quantum information is stored in the states not yet been demonstrated [28]. An elementary model of anyons, which have a non-trivial effect on the total that possesses MZMs is the Kitaev chain consisting of N state when they are interchanged. The logical states of fermions with Hamiltonian [16] the anyons form a subspace distinguishing the error-free N 1 space to those with errors, and errors are suppressed via − H = ∆(c c + c† c† ) t(c† c + c† c ) µc† c , the topological gap. For non-Abelian anyons, their braid- n n+1 n+1 n − n+1 n n n+1 − n n n=1 ing can be used to construct elementary quantum gates X (1) for quantum computing. The resulting quantum gate is arXiv:2009.07590v2 [quant-ph] 9 Dec 2020 only dependent upon the topology of the braiding path, where cn is a fermionic annihilation operator on site n, thus small imperfections in the braiding can be tolerated and t is the hopping energy, ∆ is the superconducting as long as the operation is topologically equivalent. gap, and µ is a chemical potential. For finite N and One example of a non-Abelian anyon is the Majorana working in the limit ∆ = t and µ = 0, the model has a zero mode (MZM) [14, 16–20]. MZMs are zero energy ex- degenerate ground state, corresponding to the presence citations that occur typically in low-dimensional topolog- or absence of a pair of MZMs, and can be used to encode ical superconductors. Two physical systems where MZMs the state of a qubit. By braiding one of the MZMs with have been intensely investigated are fractional quantum another, quantum gates on the encoded quantum infor- Hall systems [21–24] and semiconductor nanowires [25– mation may be performed, thereby forming the basis for topological quantum computing. While the direct realization of topological quantum computing based on MZMs is still out of reach of present

∗In memory of Jonathan P. Dowling day technology, the study of the Majorana physics is now 2 possible by way of quantum simulation in other control- (a) lable systems such as superconducting and ion-trap systems [29]. Mapping the Kitaev chain to a spin model via the Jordan–Wigner transformation, the Hamiltonian

(1) takes the form of a one-dimensional transverse ﬁeld Transmission Line

s s R Ising model, which has made it attractive to numerous R R R6 R7 8 R 5 R1 R2 3 4

Q Q Q Q Q Q5 Q6 7 8 Q1 2 3 4 XY Z proposals for simulating its equivalent dynamics. Xu, KC KC KC1 KC2 3 4 Pachos, and Guo implemented the spin version of MZMs states in a Kitaev chain, and braiding of the effective Alice Bob anyons was demonstrated to realize one qubit gates with imaginary time evolution [30, 31]. Several works have also demonstrated the path-independent nature of braiding anyonic excitations in the toric code [32–37]. Simu- (b) lating the physics of Majorana fermions by artificially Representation Hamiltonian Transformation constructed lattice models can provide additional insight into the nature of the quantum states, such as allowing Spin one to tap into the existing pool of ideas on Majorana- based quantum computation [38, 39]. To date, such stud- Jordan-Wigner ies have been restricted to examining the basic properties Fermion of anyons, and we are not aware of any quantum simu- Majorana lation of topological quantum computing involving more Majorana than one encoded qubit. 2, r In this paper, we investigate the feasibility of topo- 1,r logical quantum computing by simulating the quantum teleportation [40] of a MZM state of the Kitaev chain on FIG. 1: Experimental configuration and the mapping of Ki- superconducting qubits. We realize four spin-mapped Ki- taev chain to its spin and Majorana representations. (a) The taev chains using eight superconducting qubits (see Fig. superconducting quantum processor. We choose eight adja- 1). Each chain, consisting of two physical qubits, encodes cent qubits labelled with Q1 to Q8 from the 12-qubit processor a single logical qubit, corresponding to the spin-mapped to perform the experiment. Qubits Q1 to Q4 and Q5 to Q8 are held by Alice and Bob, respectively. Pairs of qubits form MZM states. In the teleportation, Alice is in posses- a spin-mapped Kitaev chain (KC), each which encode a sin- sion of two of the Kitaev chains, and Bob holds the two gle logical qubit. Each qubit couples to a resonator for state other chains. The teleportation then transfers a single readout, marked by R1 to R8. After decoding, the resonators logical qubit, encoded as the spin-mapped MZM states, marked by “syn” are syndrome measurements to detect phase from Alice to Bob. One of the well-known issues of quan- flip errors in the qubits. An encoded qubit is teleported from tum computing based on MZMs is that braiding opera- KC1 to KC3. (b) Mapping between spin, fermions, and Ma- tions only allow for a discrete number of Clifford gates, jorana modes. The pairing of Majorana modes in the topologically non-trivial regime are indicated by the dotted ovals. which is insufficient for universal quantum computation In the topologically non-trivial phase, the MZMs are present [11, 41, 42]. Fortunately in the teleportation protocol, at the ends of the chain. only Clifford gates are required, such that it can be com- pleted entirely with braiding operations. In addition to demonstrating the feasibility of anyonic quantum com- label the two types of Majoranas, which correspond to puting, we also show the error protection capabilities of the real and imaginary part of the fermion operator, de- MZMs. The protection of quantum information via er- noted by the left and right boxes in Fig. 1(b) respectively. ror correcting codes has been demonstrated in many past Let us denote 0L a ground state of the Hamiltonian (1), | i works [43–52], including those based on topological states taken as the state with no Majorana modes throughout [44, 52]. We show that the spin-mapped MZM states are the chain. The nature of the Kitaev Hamiltonian is such capable of detecting phase errors, allowing us to improve that applying the fermion creation operator the teleportation fidelity significantly. 1 We first give a brief review of anyonic quantum com- f † = (γ1,` iγN,r), (3) puting with MZMs. Each fermion is written in terms of 2 − two Majoranas according to the definition consisting of two Majorana edge modes at the ends of the lattice, produces another orthogonal degenerate state. γ = c + c† n,` n n These two states 0 and 1 f † 0 are the MZM | Li | Li ≡ | i γn,r = icn + icn† (2) states and are used as the logical qubit states. A minimal − implementation of the Kitaev chain Hamiltonian (1) con- where n is an integer labeling the fermions, and the `, r sists of two fermions N = 2. Under the Jordan-Wigner 3

x x mapping, the Hamiltonian takes a form H = tσ1 σ2 for and performing a Jordan-Wigner transformation, we con- ∆ = t, µ = 0, and the two MZM states are − vert the MZM teleportation circuit (Fig. 2(b)) into the equivalent 8 qubit version as shown in Fig. 2(c). 1 0L = ( ++ + ) In addition to the braiding operations, we require en- | i √2 | i |−−i coding and decoding operations to prepare the logical 1 spin-mapped MZM qubit states of (4). The encoder takes 1L = ( ++ ). (4) | i √2 | i − |−−i an arbitrary qubit state and an auxiliary qubit in the state 0 and produces its associated logical spin-mapped To encode M logical qubits, one then prepares M Ki- MZM| qubiti state taev chains, each with the Hamiltonian (1). Let us label the MZMs from the mth chain as U 0 (α 0 + β 1 ) = α 0 + β 1 , (7) enc| i | i | i | Li | Li γ(m) γ(m) which can be performed using elementary gates and the ` ≡ 1,` (m) definitions (4). Here α, β are arbitrary complex coeffi- γ(m) γ , (5) 2 2 r ≡ N,r cients such that α + β = 1. The gate decompositions for the braiding| gates,| | encoder,| and decoder U = U such that we only label the left-most and right-most Ma- dec enc† are shown in Fig. 2(d). jorana mode in the chain, which are the MZMs. An We choose eight adjacent qubits from a 12-qubit su- MZM, on the mth chain that is in the left- or right-most perconducting quantum processor [56, 57] to implement position σ `, r , can be braided with another labeled the quantum circuit of Fig. 2(c). The average fidelities of by (m , σ ).∈ { The} effect of this is to apply the unitary 0 0 single-qubit gates and the controlled-Z gate are approx- braid operator [53, 54], defined as imately 0.9994 and 0.985, respectively. The six input

(m) (m ) states of 0 , 1 , + , , + i , i , corresponding to πγ γ 0 /4 1 (m) (m0) σ σ | i | i | i |−i | i | −z ix y B(m,σ)(m0,σ0) = e 0 = (1 + γσ γσ ). (6) pairs of eigenstates of the Pauli σ , σ , σ operators are √2 0 prepared on qubit 2 as the input state for teleportation. For two logical qubits, there are four MZMs, and there- To perform the classical correction steps, we run four 4 fore there are 2 = 6 possible braiding operations (see versions of the circuit with and without each of the X Fig. S4 in Supplemental Material). Due to the non- and Z classical correction gates. Then given a particu- Abelian nature of MZMs, these produce gate operations lar measurement outcome on qubits 2 and 4, the correct on MZM states. In our circuit, we utilize the fact that circuit for that outcome is selected. To perform the to- braiding MZMs on the same chain produce a logical √Z mography measurement of the teleported state on qubit operation, and braiding adjacent MZMs on two different 6, we repeat the circuit by making measurements in the chains produce a logical √X1X2. X,Y,Z basis such that the state can be tomographically The standard quantum teleportation circuit consists of reconstructed. Each of the circuit variants were run a a sequence of Hadamard and CNOT gates [55], which are total of 40000 times for statistics. not directly available by braiding operations. To match Figure 3 shows the teleportation fidelities for each of the gates that are available with braiding operations as the six input states. First we average over all measure- closely as possible, we design a modified teleportation ment outcomes on qubits 1, 3, 5, 7, 8, which corresponds scheme (see Fig. 2(a) and Supplemental Material). The to ignoring all error syndrome measurements and any protocol proceeds in a similar way to the standard tele- changes in the ancilla MZM qubit. We find the average portation circuit, except that the classically transmitted fidelity of the six states is 70.76 0.35%, which is above quantum correction (“classical correction” for short) is the classical limit of 2/3 [58] by± 11 standard deviations. done according to the modified rules shown in the classi- We have performed an explicit simulation of the circuit cal circuit of Fig. 2(a). Using this modified teleportation shown in Fig. 2(c) including dephasing and gate errors, circuit, the equivalent version with MZMs can be con- and obtain good agreement between the experimentally structed entirely using the six available braiding gates. obtained errors (see Supplemental Material). We note The one gate that is not directly available as a braiding that the experiment further suffers from readout errors, gate in the circuit of Fig. 2(a) is the X-gate for classi- which are expected to further degrade the theoretical fi- cal correction. No combination of the six braiding op- delities. From the operations on qubit 7 and 8 it is ap- erations in Fig. S4 can produce a single qubit X-gate. parent that the final state should be in the state 00 , However, by adding an extra ancilla MZM qubit (m = 4) which is consistent with the fact that the role of these| i prepared in the eigenstate with X4 = +1, and applying qubits are only to be in the X = 1 eigenstate. We exper- the braiding operation for the logical √X X twice, we imentally obtain the probability of getting the 00 state 3 4 | i can perform an X3 gate. In this way all gates appearing is 97.98%, consistent with this expectation. in the teleportation circuit can be performed natively us- One of the benefits of encoding quantum information ing only braiding operations (Fig. 2(b)). Using a minimal with MZMs is that it allows for a natural way of pro- implementation of the Kitaev chain with N = 2 fermions, tecting against errors. As stated in Kiteav’s original pa- 4

(a) (b) (1)

(1) XX r 0 (2) XX 0 0 X Z (2) r

(3) (c) syn 0 0 1 (3) enc dec r Z correction 2 XX (4) syn + X correction + 0 (4) 3 r enc dec 0 4 R y R z R y (d) 2 2 2 XX XX syn R y R z R y 0 2 2 2 5 enc Z Z dec tomo R x R z R x 0 2 2 2 6 Z XX XX Z correction R x R z R x syn 2 2 2 0 H H 7 H H X correction enc dec 0 H H 8 H H H H

FIG. 2: Quantum circuits for simulating the teleportation of a MZM encoded qubit. (a) The modiﬁed quantum teleportation scheme. (b) The braiding sequence for MZMs that performs the quantum circuit in (a). (c) The corresponding spin-mapped qubit circuit of the MZM braiding sequence shown in (b). All measurements are performed in the 0 , 1 , which the exception of the measurement on qubit 6, where tomography (“tomo”) is performed. The measurements marked| i with| i “syn” are syndrome measurements, where single qubit phase errors are detected for a measurement outcome of 1 . (d) The gate decompositions for the braiding, encoding, and decoding gates in (c). In all ﬁgures, ψ is the state to be teleported.| i Black lines connecting the quantum gates denote qubits, dark blue lines denote MZMs, and| orangei lines denote classical information transfer.

per [16] introducing the model (1), the MZM encoding is resilient against phase-flip errors because of they correspond to non-local fermion interactions, and bit-flip errors because they corresponds to a parity non-conserving process, which are both unlikely to occur naturally. Un- der the spin-mapping, the protection against bit-flip errors is lost, but protection against phase-flip errors is still present (see Supplemental Material). This can be easily seen by examining the logical states after a phase-flip

z z 1 0˜L = σ 0L = σ 0L = ( + + + ) | i 1 | i 2 | i √2 |− i | −i z z 1 1˜L = σ 1L = σ 1L = ( + + ). (8) | i 1 | i − 2 | i √2 |− i − | −i

The states 0˜L , 1˜L span an orthogonal subspace to that spanned by| thei logical| i states, and are produced when any FIG. 3: Teleportation fidelities with and without error syn- single qubit phase error occurs. Using the relation drome detection. The fidelity is calcuated according to F = ψ ρ ψ , where ψ = 0 , 1 , + , , + i , i are the Udec(α 0˜L + β 1˜L ) = 1 (α 0 + β 1 ), (9) idealh | | statesi to be| i teleported.{| i | i | Thei |−i dashed| i line| − denotesi} the | i | i | i | i | i F = 2/3 threshold. The error bars denote one standard devi- we can see that the measurements outcomes of 1 on the ation. auxiliary qubits allows one to deduce that a phase| i flip error has occurred on any of the qubits. This constitutes an error detecting code [5, 59–61], which can be used to 5

(a) (b) tation have been performed to date in qubit [62–69] and higher dimensional systems [70–74], but never in combination with quantum error correction. Purely from the perspective of the spin formulation, our work can also be viewed as a demonstration of an encoded qubit teleportation. (c) (d) The benefit of the MZM encoding in the original fermion model is protection against both bit- and phase- flip errors [16]. By performing the spin-mapping, the protection against bit-flip errors is lost, as evident by ex-

(e) (f) amining the states (4). However, the protection against phase-flip errors is still present, and we explicitly demonstrated the enhancement in fidelity of the teleportation by postselecting states without detected errors. Thus, our results also indicate the feasibility of phase error protection in MZM-based topological quantum computing. Longer chains are expected to enhance the phase error protection, but will make the states more susceptible to FIG. 4: Tomography of the final teleported state after using bit-flip errors due to a higher probability of an error oc- the error syndrome measurements. The initial state prepared curring somewhere on the chain. Thus while we do not on qubit 2 is (a) 0 , (b) 1 , (c) + , (d) , (e) + i , (f) expect that the spin encoding benefits from larger N, i . Frames show| i ideal| teleportationi | i states,|−i colored| i bars | − i in the original fermion model we expect that the logical shows the experimentally determined state. states should have an enhanced protection. In addition to the error detection performed here, with the addition of a topological gap to energetically separate the logi- passively improve the fidelity of the circuit by discarding cal space from the error space, errors could be actively any results where errors have occurred [45, 46, 48–51]. suppressed, further improving the error protection. Figure 3 shows the teleportation fidelities using the er- The authors thank the Laboratory of Microfabrication, ror syndrome measurements on qubits 1, 3, 5, 7. A mea- University of Science and Technology of China, Institute surement of the state 1 on any of these qubits signals | i of Physics CAS, and National Center for Nanoscience that at least one phase flip error has occurred, such that and Technology for supporting the sample fabrication. it is removed from the data set. All results on the ancilla The authors also thank QuantumCTek Co., Ltd., for qubit 8 are included. We observe that the fidelities of supporting the fabrication and the maintenance of room- the telportation improve significantly for all states, with temperature electronics. Funding: This research was an average of 84.60 0.11% for the six states. This fur- ± supported by the National Key Research and Develop- ther increases the average fidelity beyond the classical ment Program of China (Grants No. 2017YFA0304300), bound by over 163 standard deviations. The final tele- NSFC (Grants No. 11574380, No. 11905217), the ported state for the error syndrome improved state is Chinese Academy of Science and its Strategic Prior- tomographically reconstructed using a maximum likeli- ity Research Program (Grants No. XDB28000000), hood estimator of the density matrix and shown in Fig. the Science and Technology Committee of Shanghai 4. We see that the states of the six input states are very Municipality, Shanghai Municipal Science and Tech- well-reproduced, demonstrating that the teleportation is nology Major Project (Grant No.2019SHZDZX01), and being performed correctly over the six mutually unbiased Anhui Initiative in Quantum Information Technolo- basis states. gies. T.B. is supported by the National Natural Sci- In summary, we have performed a quantum simulation ence Foundation of China (61571301,D1210036A); the of the teleportation of a qubit encoded as the MZM states NSFC Research Fund for International Young Scientists of the Kitaev chain. The teleportation circuit is per- (11650110425, 11850410426); NYU-ECNU Institute of formed entirely using braiding operations of the MZMs, Physics at NYU Shanghai; the Science and Technology including the quantum gates for classical correction. In Commission of Shanghai Municipality (17ZR1443600, our teleportation circuit we were careful to be faithful 19XD1423000); the China Science and Technology Ex- to the braiding process of the MZMs in the sense that change Center (NGA-16-001); and the NSFC-RFBR Col- no gate simplifications were performed in the circuit Fig. laborative grant (81811530112). H.-L. H. is supported 2(c). This demonstrates via an equivalent spin encod- by the Open Research Fund from State Key Laboratory ing that a non-trivial quantum circuit can be performed of High Performance Computing of China (Grant No. using a topological quantum computing sequence using 201901-01), NSFC (Grants No. 11905294), and China braids. In addition, numerous demonstrations of telepor- Postdoctoral Science Foundation. 6

H.-L. H., M. N. and F. L. contributed equally to this [31] J.-S. Xu, K. Sun, J. K. Pachos, Y.-J. Han, C.-F. Li, and work. G.-C. Guo, Sci. Adv. 4, eaat6533 (2018). [32] C.-Y. Lu, W.-B. Gao, O. Gühne,X.-Q. Zhou, Z.-B. Chen, and J.-W. Pan, Phys. Rev. Lett. 102, 030502 (2009). [33] J. Pachos, W. Wieczorek, C. Schmid, N. Kiesel, R. Pohlner, and H. Weinfurter, New J. Phys. 11, 083010 (2009). † Electronic address: [email protected] [1] P. W. Shor, in Proceedings of 37th Conference on Foun- [34] Y.-P. Zhong, D. Xu, P. Wang, C. Song, Q. Guo, W. Liu, dations of Computer Science (IEEE, 1996) pp. 56–65. K. Xu, B. Xia, C.-Y. Lu, S. Han, et al., Phys. Rev. Lett. [2] D. Gottesman, Phys. Rev. A 57, 127 (1998). 117, 110501 (2016). [3] A. M. Steane, Nature 399, 124 (1999). [35] H.-N. Dai, B. Yang, A. Reingruber, H. Sun, X.-F. Xu, [4] D. Aharonov and M. Ben-Or, arXiv:quant-ph/9906129 Y.-A. Chen, Z.-S. Yuan, and J.-W. Pan, Nat. Phys. 13, (1999). 1195 (2017). [5] J. Preskill, in Introduction to quantum computation and [36] C. Song, D. Xu, P. Zhang, J. Wang, Q. Guo, W. Liu, information (World Scientific, 1998) pp. 213–269. K. Xu, H. Deng, K. Huang, D. Zheng, et al., Phys. Rev. [6] D. Gottesman, in Quantum information science and its Lett. 121, 030502 (2018). contributions to mathematics, Proceedings of Symposia [37] C. Liu, H.-L. Huang, C. Chen, B.-Y. Wang, X.-L. Wang, in Applied Mathematics, Vol. 68 (2010) pp. 13–58. T. Yang, L. Li, N.-L. Liu, J. P. Dowling, T. Byrnes, et al., [7] S. J. Devitt, W. J. Munro, and K. Nemoto, Rep. Prog. Optica 6, 264 (2019). Phys. 76, 076001 (2013). [38] Y. Tserkovnyak and D. Loss, Phys. Rev. A 84, 032333 [8] E. T. Campbell, B. M. Terhal, and C. Vuillot, Nature (2011). 549, 172 (2017). [39] A. Mezzacapo, J. Casanova, L. Lamata, and E. Solano, [9] A. Y. Kitaev, Ann. Phys. 303, 2 (2003). New J. Phys. 15, 033005 (2013). [10] M. Freedman, A. Kitaev, M. Larsen, and Z. Wang, Bull. [40] C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, Am. Math. Soc. 40, 31 (2003). A. Peres, and W. K. Wootters, Phys. Rev. Lett. 70, [11] J. Preskill, Lecture notes for physics 219 (2004). 1895 (1993). [12] C. Nayak, S. H. Simon, A. Stern, M. Freedman, and [41] M. H. Freedman, M. Larsen, and Z. Wang, Commun. S. D. Sarma, Rev. Mod. Phys. 80, 1083 (2008). Math. Phys. 227, 605 (2002). [13] J. K. Pachos, Introduction to topological quantum com- [42] A. Stern and N. H. Lindner, Science 339, 1179 (2013). putation (Cambridge University Press, 2012). [43] X.-C. Yao, T.-X. Wang, H.-Z. Chen, W.-B. Gao, A. G. [14] S. D. Sarma, M. Freedman, and C. Nayak, NPJ Quan- Fowler, R. Raussendorf, Z.-B. Chen, N.-L. Liu, C.-Y. Lu, tum Inf. 1, 15001 (2015). Y.-J. Deng, et al., Nature 482, 489 (2012). [15] V. Lahtinen and J. K. Pachos, SciPost Phys. 3 (2017). [44] D. Nigg, M. Mueller, E. A. Martinez, P. Schindler, [16] A. Y. Kitaev, Phys.-Uspekhi 44, 131 (2001). M. Hennrich, T. Monz, M. A. Martin-Delgado, and [17] J. Alicea, Rep. Prog. Phys. 75, 076501 (2012). R. Blatt, Science 345, 302 (2014). [18] M. Leijnse and K. Flensberg, Semicond. Sci. Technol. 27, [45] J. Kelly, R. Barends, A. G. Fowler, A. Megrant, E. Jef- 124003 (2012). frey, T. C. White, D. Sank, J. Y. Mutus, B. Campbell, [19] C. Beenakker, Annu. Rev. Condens. Matter Phys. 4, 113 Y. Chen, et al., Nature 519, 66 (2015). (2013). [46] A. D. Córcoles, E. Magesan, S. J. Srinivasan, A. W. [20] R. t. Lutchyn, E. Bakkers, L. P. Kouwenhoven, Cross, M. Steffen, J. M. Gambetta, and J. M. Chow, P. Krogstrup, C. Marcus, and Y. Oreg, Nat. Rev. Mater. Nat. Commun. 6, 6979 (2015). 3, 52 (2018). [47] N. Ofek, A. Petrenko, R. Heeres, P. Reinhold, Z. Leghtas, [21] R. Willett, J. Eisenstein, H. Störmer,D. Tsui, A. Gos- B. Vlastakis, Y. Liu, L. Frunzio, S. Girvin, L. Jiang, sard, and J. English, Phys. Rev. Lett. 59, 1776 (1987). et al., Nature 536, 441 (2016). [22] S. D. Sarma, M. Freedman, and C. Nayak, Phys. Rev. [48] N. M. Linke, M. Gutierrez, K. A. Landsman, C. Figgatt, Lett. 94, 166802 (2005). S. Debnath, K. R. Brown, and C. Monroe, Sci. Adv. 3, [23] R. L. Willett, L. N. Pfeiffer, and K. West, Proc. Natl. e1701074 (2017). Acad. Sci. 106, 8853 (2009). [49] M. Takita, A. W. Cross, A. Córcoles,J. M. Chow, and [24] R. Willett, C. Nayak, K. Shtengel, L. Pfeiffer, and J. M. Gambetta, Phys. Rev. Lett. 119, 180501 (2017). K. West, Phys. Rev. Lett. 111, 186401 (2013). [50] S. Rosenblum, P. Reinhold, M. Mirrahimi, L. Jiang, [25] V. Mourik, K. Zuo, S. M. Frolov, S. Plissard, E. P. L. Frunzio, and R. Schoelkopf, Science 361, 266 (2018). Bakkers, and L. P. Kouwenhoven, Science 336, 1003 [51] R. Harper and S. T. Flammia, Phys. Rev. Lett. 122, (2012). 080504 (2019). [26] R. M. Lutchyn, J. D. Sau, and S. D. Sarma, Phys. Rev. [52] C. K. Andersen, A. Remm, S. Lazar, S. Krinner, Lett. 105, 077001 (2010). N. Lacroix, G. J. Norris, M. Gabureac, C. Eichler, and [27] Y. Oreg, G. Refael, and F. von Oppen, Phys. Rev. Lett. A. Wallraff, Nat. Phys. , 1 (2020). 105, 177002 (2010). [53] C. Nayak and F. Wilczek, Nucl. Phys. B 479, 529 (1996). [28] R. A. Sola and L. Kouwenhoven, Phys. Today 73, 44 [54] D. A. Ivanov, Phys. Rev. Lett. 86, 268 (2001). (2020). [55] M. A. Nielsen and I. Chuang, “Quantum computation [29] H.-L. Huang, D. Wu, D. Fan, and X. Zhu, Sci. China and quantum information,” (2002). Inf. Sci. 63, 1 (2020). [56] M. Gong, M.-C. Chen, Y. Zheng, S. Wang, C. Zha, [30] J.-S. Xu, K. Sun, Y.-J. Han, C.-F. Li, J. K. Pachos, and H. Deng, Z. Yan, H. Rong, Y. Wu, S. Li, et al., Phys. G.-C. Guo, Nat. Commun. 7, 13194 (2016). Rev. Lett. 122, 110501 (2019). [57] H.-L. Huang, Y. Du, M. Gong, Y. Zhao, Y. Wu, C. Wang, 7

S. Li, F. Liang, J. Lin, Y. Xu, et al., arXiv:2010.06201 [68] W. Pfaff, B. J. Hensen, H. Bernien, S. B. van Dam, M. S. (2020). Blok, T. H. Taminiau, M. J. Tiggelman, R. N. Schouten, [58] S. Massar and S. Popescu, Phys. Rev. Lett. 74, 1259 M. Markham, D. J. Twitchen, et al., Science 345, 532 (1995). (2014). [59] L. Vaidman, L. Goldenberg, and S. Wiesner, Phys. Rev. [69] J.-G. Ren, P. Xu, H.-L. Yong, L. Zhang, S.-K. Liao, A 54, R1745 (1996). J. Yin, W.-Y. Liu, W.-Q. Cai, M. Yang, L. Li, et al., [60] M. Grassl, T. Beth, and T. Pellizzari, Phys. Rev. A 56, Nature 549, 70 (2017). 33 (1997). [70] A. Furusawa, J. L. Sørensen, S. L. Braunstein, C. A. [61] S. P. Jordan, E. Farhi, and P. W. Shor, Phys. Rev. A Fuchs, H. J. Kimble, and E. S. Polzik, Science 282, 706 74, 052322 (2006). (1998). [62] D. Bouwmeester, J.-W. Pan, K. Mattle, M. Eibl, H. We- [71] Q. Zhang, A. Goebel, C. Wagenknecht, Y.-A. Chen, infurter, and A. Zeilinger, Nature 390, 575 (1997). B. Zhao, T. Yang, A. Mair, J. Schmiedmayer, and J.-W. [63] D. Boschi, S. Branca, F. De Martini, L. Hardy, and Pan, Nat. Phys. 2, 678 (2006). S. Popescu, Phys. Rev. Lett. 80, 1121 (1998). [72] H. Krauter, D. Salart, C. Muschik, J. M. Petersen, [64] M. Riebe, H. Häffner,C. Roos, W. Hänsel,J. Benhelm, H. Shen, T. Fernholz, and E. S. Polzik, Nat. Phys. 9, G. Lancaster, T. Körber, C. Becher, F. Schmidt-Kaler, 400 (2013). D. James, et al., Nature 429, 734 (2004). [73] X.-L. Wang, X.-D. Cai, Z.-E. Su, M.-C. Chen, D. Wu, [65] M. Barrett, J. Chiaverini, T. Schaetz, J. Britton, L. Li, N.-L. Liu, C.-Y. Lu, and J.-W. Pan, Nature 518, W. Itano, J. Jost, E. Knill, C. Langer, D. Leibfried, 516 (2015). R. Ozeri, et al., Nature 429, 737 (2004). [74] Y.-H. Luo, H.-S. Zhong, M. Erhard, X.-L. Wang, L.- [66] X.-S. Ma, T. Herbst, T. Scheidl, D. Wang, C. Peng, M. Krenn, X. Jiang, L. Li, N.-L. Liu, C.-Y. S. Kropatschek, W. Naylor, B. Wittmann, A. Mech, Lu, A. Zeilinger, and J.-W. Pan, Phys. Rev. Lett. 123, J. Kofler, E. Anisimova, et al., Nature 489, 269 (2012). 070505 (2019). [67] M. Baur, A. Fedorov, L. Steffen, S. Filipp, M. Da Silva, and A. Wallraff, Phys. Rev. Lett. 108, 040502 (2012). Supplemental Material: Emulating quantum teleportation of a Majorana zero mode qubit

1, 2, 3, 4, 5, 1, 2, 3, 1, 2, 3 1, 2, 3 He-Liang Huang, ∗ Marek Naro˙zniak, ∗ Futian Liang, ∗ Youwei Zhao, Anthony D. Castellano, Ming Gong,1, 2, 3 Yulin Wu,1, 2, 3 Shiyu Wang,1, 2, 3 Jin Lin,1, 2, 3 Yu Xu,1, 2, 3 Hui Deng,1, 2, 3 Hao Rong,1, 2, 3 6, 7, 8 1, 2, 3 4, 8, 9, 5 1, 2, 3, 1, 2, 3 Jonathan P. Dowling, Cheng-Zhi Peng, Tim Byrnes, Xiaobo Zhu, † and Jian-Wei Pan 1Hefei National Laboratory for Physical Sciences at the Microscale and Department of Modern Physics, University of Science and Technology of China, Hefei 230026, China 2Shanghai Branch, CAS Center for Excellence in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China 3Shanghai Research Center for Quantum Sciences, Shanghai 201315, China 4New York University Shanghai, 1555 Century Ave, Pudong, Shanghai 200122, China 5Department of Physics, New York University, New York, NY 10003, USA 6Hearne Institute for Theoretical Physics, Department of Physics and Astronomy, Louisiana State University, Baton Rouge, Louisiana 70803, USA 7Hefei National Laboratory for Physical Sciences at Microscale and Department of Modern Physics, University of Science and Technology of China, Hefei, Anhui 230026, China 8NYU-ECNU Institute of Physics at NYU Shanghai, 3663 Zhongshan Road North, Shanghai 200062, China 9State Key Laboratory of Precision Spectroscopy, School of Physical and Material Sciences, East China Normal University, Shanghai 200062, China (Dated: December 10, 2020)

I. MAJORANA MODES IN THE KITAEV According to the deﬁnition, Majorana fermions are CHAIN purely real

In this section we provide a brief review of Majorana γn,σ = γn,σ† , (S6) modes in the Kitaev chain. We refer the reader to several excellent reviews for further details [1–5]. where σ `, r . They share similarities with standard fermions∈ with { regard} to their anti-commutation property

γ , γ = 2δ δ . (S7) A. Deﬁnition of Majorana fermions { n,σ n0,σ0 } nn0 σσ0 However, unlike standard fermions which obey the Pauli Consider a set of N fermions, which can be described 2 2 exclusion principle cn = (cn† ) = 0, Majorana fermions standard fermionic anticommutation relations are their own anti-particle and we have

cn, cn0 = 0 2 { } γn,σ = 1. (S8) cn, c† = δnn , (S1) { n0 } 0

where δnn0 is the Kronecker delta. We may rewrite the B. Delocalized fermions operators for creating and annihilating a fermion on site n in terms of two Majorana fermions in following way Under this formalism, it appears the concept of Majo- 1 rana fermions is just an algebraic manipulation. The c = (γ + iγ ) (S2) n n,` n,r interesting aspect of utilizing the Majorana operators arXiv:2009.07590v2 [quant-ph] 9 Dec 2020 2 1 arises when we construct other types of fermions that are cn† = (γn,` iγn,r). (S3) 2 − not necessarily the physical fermions cn. Following the form of the fermion operators shown in (S3), new delocal- These equations can be solved for γ and γ resulting ` r ized fermions can be defined using any pair of Majorana with definitions of Majorana fermions in terms of a single modes fermion 1 γ = c + c† (S4) f = (γ + iγ ) (S9) n,` n n p 2 n,σ m,ν γn,r = icn + icn† . (S5) 1 − f † = (γ iγ ), (S10) p 2 n,σ − m,ν where σ, ν `, r . Here ∈ { } These three authors contributed equally ∗ p (n, σ, m, ν) (S11) †Electronic address: [email protected] → 2 is a pairing label between two Majorana modes labeled p=1 p=2 by (n, σ) and (m, ν). The fermion operator fp is con- a structed from two Majorana modes, which are poten- γ ℓ γ r γ ℓ γ r γ ℓ γ r γ ℓ γ r tially at different physical sites n = m, hence we call this 1, 1, 2, 2, 3, 3, 4, 4, a delocalized fermion. A particular6 pair p always involves two different Majorana modes, such that (n, σ) = (m, ν), p=3 p=4 meaning that a pair with both n = m and σ =6 ν is not b allowed. It is possible however to have a pairing such γ 1,ℓ γ 1,r γ 2,ℓ γ 2,r γ 3,ℓ γ 3,r γ 4,ℓ γ 4,r that n = m but σ = ν, or n = m but σ = ν. The former 6 6 is exactly the case of physical fermions as shown in (S3). p=1 p=2 p=3 p=4 Given a set of N fermions, and hence 2N Majorana p=4 modes, let us fix a particular pairing configuration la- c beled by (S11). Various examples of Majorana pairings γ γ γ are shown in Fig. S1. When establishing a pairing con- 1,ℓ 1,r γ 2,ℓ γ 2,r γ 3,ℓ γ 3,r γ 4,ℓ 4,r figuration, Majorana modes are never used twice, such that for different pairs p = p0, the underlying Majoranas p=1 p=2 p=3 are all different. Under these6 conditions, the anticommutation relations of the delocalized fermions (S10) can be evaluated as FIG. S1: Various Majorana mode pairing configurations. (a) Random, (b) topologically trivial, and (c) Kitaev chain pair- 1 ings are shown. f , f = (δ δ + iδ δ + iδ δ δ δ ) { p p0 } 2 nn0 σσ0 nm0 σν0 n0m σ0ν − mm0 νν0 = 0 (S12) a E where the pairing label p0 (n0, σ0, m0, ν0). The Kro- 2t → necker delta functions simplify in (S12) because if p = p0 it implies that t

(n, σ) = (n0, σ0) = (m, ν) = (m0, ν0), (S13) 6 0 but if p = p0 then it implies that 6 b E (n, σ) = (n0, σ0) = (m, ν) = (m0, ν0). (S14) 6 6 6 t Similarly we can evaluate

1 0 fp, f † = (δnn δσσ iδnm δσν + iδn mδσ ν + δmm δνν ) { p0 } 2 0 0 − 0 0 0 0 0 0

= δpp0 (S15) FIG. S2: Energy spectrum of Majorana pairing Hamiltonian which shows that the delocalized fermions are fermion for N = 2 fermions. (a) Topologically trivial Hamiltonian operators as claimed. (S17) as shown in Fig. S1(b). (b) Kitaev Hamiltonian (S24) as shown in Fig. S1(c). The Majorana mode labels are suppressed, and the occupancy of the Majorana modes are de- noted by the red bars. C. Majorana pairing Hamiltonian

To enforce a particular pairing conﬁguration of Majo- where t is some energy constant. The eigenstates of this rana modes, we must energetically stabilize the fermions Hamiltonian are given by that are deﬁned by (S10). For example, the Hamilto- N nian to enforce the regular fermion pairing (Fig. S1(b)) j j , . . . , j = (c† ) n 0 (S18) is given by | 1 N i n | i n=1 Y N where jn 0, 1 labels the occupancy of the nth H = t cn† cn (S16) fermion. The∈ energies { } of these states are n=1 X t N N = (1 + iγ γ ), (S17) E = t j . (S19) 2 n,` n,r n n=1 n=1 X X 3

This can be rewritten in the Majorana language, where and the energy spectrum is a fermion occupancy of the nth site means that the un- N 1 derlying Majorana modes are both occupied. Fig. S2(a) − E = t jn. (S27) shows the spectrum for the example of N = 2. n=1 Similarly, for the delocalized fermions we can define a X This does not involve the occupancy label j , which Majorana pairing Hamiltonian according to N explicitly shows the double degeneracy. An example of N the spectrum of the Kitaev Hamiltonian is shown in Fig. S2(b). H = t fp†fp, (S20) p=1 X where p runs over all N Majorana pairs, for example that D. Majorana zero modes defined in Fig. S1(a). The eigenstates are again defined by the occupancy of the new fermions The doubly degenerate ground states of the Kitaev Hamiltonian (S24) have zero energy and form a pair of N orthogonal states. Let us as usual take the ground state jp j1, . . . , jN = (f †) 0 , (S21) with the absence of any fermions (S23) or (S25) by 0 . | i p | i | i p=1 Then fermion operator (S25) then transforms this ground Y state into its degenerate pair, where the Majoranas on the where 0 is the ground state of (S20), and jp 0, 1 la- | i ∈ { } end of the chain are occupied fN† 0 . These two states are bels the occupancy of the pth pair. The energy spectrum used as the logical qubit states| ofi the quantum compu- is again given by tation, where

N 0 0 | Li ≡ | i E = t jp. (S22) 1L fN† 0 . (S28) p=1 | i ≡ | i X For simplicity we denote in the main text f = fN , which The Kitaev chain [6] is a particular example for the is the annihilation operator for the edge states of the Ki- pairing configuration given in Fig. S1(c). In this case, taev Hamiltonian. The Majorana modes γ1,` and γN,r the fermions are defined as have zero energy and hence are called Majorana zero modes (MZMs). 1 f = (γ + iγ ), (S23) One Kitaev chain therefore encodes one logical qubit’s n 2 n,r n+1,` worth of information. In order to have multiple logical qubits, then multiple Kitaev chains are required. La- where n [1,N 1] for this operator. Here a Majorana ∈ − belling the Majorana mode labelled by (n, σ) on the mth mode in the right box on site n is paired with another in (m) chain as γ , the Hamiltonian for the multiple chain the left box of site n + 1. The Kitaev chain Hamiltonian n,σ case then reads is then N 1 t − (m) N 1 H = (1 + iγ(m)γ ) − 2 n,r n+1,` H = t f f n=1 n† n X n=1 N 1 X t − (m) (m) N 1 = (1 c †c(m) c(m)†c t − 2 − n+1 n − n n+1 = (1 + iγ γ ). (S24) n=1 2 n,r n+1,` X n=1 (m) (m) (m) † (m)† X + cn cn+1 + cn+1 cn ), (S29) Importantly, this Hamiltonian does not involve the Ma- which up to a constant energy offset is the Hamiltonian jorana pairing of the delocalized fermion corresponding (1) in the main text. Here we denote the fermion annihi- to lation operator on the nth site of the mth Kitaev chain (m) 1 as cn . f = (γ + iγ ). (S25) N 2 1,` N,r The MZMs on the mth Kitaev chain then occur on the first and last Majorana sites and we define the operator This means that this fermion costs zero energy to excite, 1 f (m) = (γ(m) + iγ(m)) (S30) and makes every state in the spectrum of (S24) doubly 2 1,` N,r degenerate, including the ground state. As before, the which destroys a MZM on the mth chain. Here hence- eigenstates of (S24) are forth use the notation N γ(m) γ(m) j ` 1,` j , . . . , j = (f †) n 0 , (S26) ≡ | 1 N i n | i (m) (m) n=1 γr γN,r . (S31) Y ≡ 4

The full set of 2M logical states are built up by apply- as shown in Fig. S3. The M Kitaev chains each with ing the creation operator f (m)† on the ground state 0 N fermions are arranged in a larger chains, in ascending contain zero Majorana modes. | i order. We label the spin operators from 1 to NM, the total number of fermions and spins in the mapping. In this case, the MZM can be transformed to spin variables II. BRAIDING MAJORANA ZERO MODES according to

pN N − The MZMs are an example of non-Abelian anyons be- (p) z x γ` = σi σpN N+1 − cause their interchange causes a non-trivial effect on the i=1 ! Y ground state manifold. In this section we derive the effect pN 1 − of braiding of the Majorana zero modes on two Kitaev (p) z y γr = σi σpN , (S35) chains. It is sufficient to consider two Kitaev chains be- i=1 ! cause we will consider the braiding of two MZMs to be Y the elementary process. The two MZMs can originate where p is the chain index. In the calculations below, we from the same Kitaev chain, or one MZM each from two only consider two chains, and hence it is convenient to Kitaev chains. This gives a total of 6 possible braidings explicitly write the spin mapped MZM operators of two MZMs, since each chain has two MZMs. (1) x γ` = σ1 N 1 (1) − z y z z y A. Braiding operator γr = σk σN = σ1 . . . σN 1σN ! − kY=1 N Braiding two zero modes γn,σ and γm,ν in a clockwise (2) z x z z x direction can be achieved by applying the operator [1, 7, γ` = σk σN+1 = σ1 . . . σN σN+1 ! 8] kY=1 2N 1 π − γn,σ γm,ν 1 (2) z y z z y B(n,σ)(m,ν) = e 4 = (1 + γn,σγm,ν ). (S32) γr = σk σ2N = σ1 . . . σ2N 1σ2N (S36) √2 ! − kY=1 It is apparent that this performs a braiding operation via The logical states on which Jordan-Wigner trans- the transformation formed operators act are defined explicitly in a following B γ B† = γ way (n,σ)(m,ν) n,σ (n,σ)(m,ν) − m,ν 1 B(n,σ)(m,ν)γm,ν B(†n,σ)(m,ν) = γn,σ. (S33) 0L = ( + + + ) (S37) | i √2 | ··· i |− · · · −i When applied on the ground state manifold of the Ki- 1 taev chains, the braiding operators realize unitary oper- 1L = ( + + ) (S38) | i √2 | ··· i − |− · · · −i ations on the MZM states. Consider for the purposes of this section that there are M = 2 Kitaev chains, such that the logical states are C. Derivation of the six braiding gates in Fig. S4

00L 0 | i ≡ | i (1) (1) 1. γ` γr braid: √Z1 gate 10 f (1)† 0 | Li ≡ | i (2) 01L f † 0 We can express this braiding operation in terms of | i ≡ | i spin operators by applying Jordan-Wigner transforma- (1) (2) 11L f †f † 0 , (S34) tion (S36) | i ≡ | i where 0 is again the state with zero Majorana modes π γ(1)γ(1) B = e 4 ` r (S39) everywhere.| i The purpose of the following section will be (1,`),(1,r) 1 to derive the eﬀect of various braiding operators on the = (1 + γ(1)γ(1)) (S40) logical space of states (S34). √2 ` r 1 x z z y = (1 + σ1 σ1 . . . σN 1σN ) (S41) √2 − B. Spin representation 1 y z z y = (1 iσ1 σ2 . . . σN 1σN ). (S42) √2 − − For each braiding operator acting on logical space, an equivalent spin operator can be derived acting on cor- This is the braiding operator in the spin representation. responding physical space. This is done using Jordan- To see the eﬀect of this braiding operator in the logical Wigner transformation to transform the Majorana vari- space, we operate the above state on the spin represen- ables to spin variables. We consider a layout of spins tation of the logical states (S38). For compactness let us 5

(1) (1) (1) (1) (1) (1) (2) (2) (2) (2) (2) (2) (3) (3) (3) (3) (3) (3) γ 1,ℓ γ 1,r γ 2,ℓ γ 2,r γ 3,ℓ γ 3,r γ 1,ℓ γ 1,r γ 2,ℓ γ 2,r γ 3,ℓ γ 3,r γ 1,ℓ γ 1,r γ 2,ℓ γ 2,r γ 3,ℓ γ 3,r

c(1)c (1)c (1)c (2) c(2) c(2) c(3)c (3) c(3) 1 2 3 1 2 3 1 2 3

σ1 σ 2 σ 3 σ4 σ 5 σ 6 σ7 σ 8 σ 9 chain 1 chain 2 chain 3

FIG. S3: Example of the labelling convention between Majorana modes, fermions, and spin operators for the case of M = 3 chains each with N = 3 fermions.

first define the non-trivial part of the braiding operator formation (S36) into the braiding operator as π γ(1)γ(2) B(1,`),(2,`) = e 4 ` ` y z z y Γ = iσ1 σ2 . . . σN 1σN . (S43) 1 (1) (2) − − = (1 + γ γ ) √2 ` ` Applying Γ on the logical states we find 1 x z z x = (1 + σ1 σ1 . . . σN σN+1) i √2 Γ 0L = ( + + + ) (S44) | i √2 |− · · · −i | ··· i 1 y z z x = (1 iσ1 σ1 . . . σN σN+1) (S51) =i 0 (S45) √2 − | Li This is the braiding operator in the spin representation. and To examine the effect on the logical states, we again define the non-trivial part of the above operator as i Γ 1L = ( + + ) (S46) Γ = iσyσz . . . σz σx . (S52) | i − √2 | ··· i − |− · · · −i − 1 1 N N+1 = i 1 . (S47) Using the relations (S48), we can evaluate − | Li Γ 00 = 11 Here we used the fact that | Li | Li Γ 01 = 10 | Li | Li σx = Γ 10 = 01 |±i ±|±i | Li −| Li σy = i Γ 11 = 00 . (S53) |±i ∓ |∓i | Li −| Li σz = . (S48) |±i |∓i where we used the explicit expansions Since we can write 1 00L = ( + + + + + + + | i √2 | ··· i | ··· i | ··· i |− · · · −i 1 B(1,`),(1,r) = (1 + Γ), (S49) + + + ) √2 |− · · · −i | ··· i |− · · · −i |− · · · −i 1 01L = ( + + + + + + it then follows that | i √2 | ··· i | ··· i − | ··· i |− · · · −i + + + ) 1 iπ/4 |− · · · −i | ··· i − |− · · · −i |− · · · −i B(1,`),(1,r) 0L = (1 + i) 0L = e 0L 1 | i √2 | i | i 10L = ( + + + + + + + | i √2 | ··· i | ··· i | ··· i |− · · · −i 1 iπ/4 B(1,`),(1,r) 1L = (1 i) 1L = e− 0L . (S50) + + ) | i √2 − | i | i − |− · · · −i | ··· i − |− · · · −i |− · · · −i 1 11L = ( + + + + + + . This corresponds to the √Z1 operator. | i √2 | ··· i | ··· i − | ··· i |− · · · −i + + + ) − |− · · · −i | ··· i |− · · · −i |− · · · −i (S54) 2. γ(1) γ(2) braid: √Y X gate ` ` 1 2 Applying the braiding operator 1 For braiding involving more than one chain we apply B(1,`),(2,`) = (1 + Γ) (S55) the same method, substituting the Jordan-Wigner trans- √2 6 then gives

1 B 00 = ( 00 + 11 ) (1,`),(2,`)| Li √ | Li | Li 1 2 B 00L = ( 00L + B0( + + + + (S66) 1 | i √2 | i | ··· i | ··· i B(1,`),(2,`) 01L = ( 01L + 10L ) + + + ) | i √2 | i | i | ··· i |− · · · −i + + + ) 1 |− · · · −i | ··· i B(1,`),(2,`) 10L = ( 10L 01L ) + ) | i √2 | i − | i |− · · · −i |− · · · −i 1 1 = ( 00L i( (S67) B(1,`),(2,`) 11L = ( 11L 00L ). (S56) √2 | i − |− · · · −i |− · · · −i | i √2 | i − | i + + ) − |− · · · −i | ··· i + + ) − | ··· i |− · · · −i This corresponds to the √Y X gate. + + + + + ) 1 2 | ··· i | ··· i 1 = ( 00L + i 11L ) (S68) √2 | i | i 1 B 01L = ( 01L + B0( + + + + (S69) | i √2 | i | ··· i | ··· i (1) (2) √ + + ) 3. γ` γr braid: Y1Y2 gate − | ··· i |− · · · −i + + + ) |− · · · −i | ··· i ) − |− · · · −i |− · · · −i 1 = ( 01L i( (S70) √2 | i − − |− · · · −i |− · · · −i π (1) (2) 4 γ` γr + + ) B(10 ,`),(2,r) = e (S57) − |− · · · −i | ··· i + + + ) 1 (1) (2) | ··· i |− · · · −i = (1 + γ` γr ) (S58) + + + + + ) √2 | ··· i | ··· i 1 1 x z z z z x = ( 01L i 10L ) (S71) = (1 + σ1 σ1 . . . σN σ1 . . . σ2N 1σ2N ) √ √2 − 2 | i − | i (S59) 1 B 10 = ( 10 + B0( + + + + (S72) | Li √ | Li | ··· i | ··· i 1 y z z x 2 = (1 iσ1 σN+1 . . . σ2N 1σ2N ) (S60) + + + ) √2 − − | ··· i |− · · · −i + + ) 1 − |− · · · −i | ··· i = (1 + B0 ) (S61) ) √2 (1,`),(2,r) − |− · · · −i |− · · · −i 1 = ( 10L i( (S73) √2 | i − − |− · · · −i |− · · · −i + + + ) We demonstrate the correctness of this operator ap- |− · · · −i | ··· i + + ) plied to the logical space and replicating the following. − | ··· i |− · · · −i + + + + + ) | ··· i | ··· i 1 = ( 10L i 01L ) (S74) √2 | i − | i 1 Y1Y2 00L = ( 00L + i 11L ) (S62) 1 | i √ | i | i B 11L = ( 11L + B0( + + + + (S75) 2 | i √2 | i | ··· i | ··· i p 1 + + ) Y1Y2 01L = ( 01L i 10L ) (S63) − | ··· i |− · · · −i | i √2 | i − | i + + ) − |− · · · −i | ··· i p 1 + ) Y1Y2 10L = ( 10L i 01L ) (S64) |− · · · −i |− · · · −i | i √2 | i − | i 1 p = ( 11L i( (S76) 1 √2 | i − − |− · · · −i |− · · · −i Y1Y2 11L = ( 11L i 00L ) (S65) + + ) | i √2 | i − | i − |− · · · −i | ··· i p + + ) − | ··· i |− · · · −i + + + + ) − | ··· i | ··· i Now we apply B(10 ,`),(2,r) to states (S54) and we repli- 1 = ( 11L + i 00L ) (S77) cate (S62-S65) as follows. √2 | i | i y z z x Let B = B(1,`),(2,r) and B0 = iσ1 σN+1 . . . σ2N 1σ2N ) (S78) − − 7

(1) (2) 4. γr γ` braid: √X1X2 gate 1 B 00L = ( 00L + B0( + + + + (S88) | i √2 | i | ··· i | ··· i + + + ) | ··· i |− · · · −i + + + ) |− · · · −i | ··· i + ) |− · · · −i |− · · · −i 1 = ( 00L + i( + + + + (S89) √2 | i | ··· i | ··· i + + ) π (1) (2) − | ··· i |− · · · −i 4 γr γ` B(10 ,r),(2,`) = e (S79) + + ) − |− · · · −i | ··· i 1 (2) + ) = (1 + γ(1)γ ) (S80) |− · · · −i |− · · · −i √2 r ` 1 = ( 00L + i 11L ) (S90) 1 z z y z z x √2 | i | i = (1 + σ1 . . . σN 1σN σ1 . . . σN σN+1) √2 − 1 B 01L = (+B0( + + + + (S91) (S81) | i √2 | ··· i | ··· i 1 + + + ) = (1 + iσx σx ) (S82) | ··· i |− · · · −i √ N N+1 + + ) 2 − |− · · · −i | ··· i 1 ) = (1 + B0 ) (S83) − |− · · · −i |− · · · −i √2 (1,r),(2,`) 1 = ( 01L + i 10L ) (S92) √2 | i | i 1 = ( 01L + i( + + + + (S93) √2 | i | ··· i | ··· i + + + ) | ··· i |− · · · −i + + ) − |− · · · −i | ··· i ) − |− · · · −i |− · · · −i We demonstrate the correctness of this operator ap- 1 plied to the logical space and replicating the following. = ( 01L + i 10L ) (S94) √2 | i | i 1 B 10L = ( 10L + B0( + + + + (S95) | i √2 | i | ··· i | ··· i + + + ) | ··· i |− · · · −i + + ) − |− · · · −i | ··· i ) − |− · · · −i |− · · · −i 1 = ( 10L + i( + + + + (S96) 1 √2 | i | ··· i | ··· i X1X2 00L = ( 00L + i 11L ) (S84) + + ) | i √2 | i | i − | ··· i |− · · · −i + + + ) p 1 |− · · · −i | ··· i X1X2 01L = ( 01L + i 10L ) (S85) ) | i √2 | i | i − |− · · · −i |− · · · −i p 1 1 = ( 10L + i 01L ) (S97) X1X2 10L = ( 10L + i 01L ) (S86) | i √2 | i | i √2 | i | i p 1 1 B 11L = ( 11L + B0( + + + + (S98) X1X2 11L = ( 11L + i 00L ) (S87) | i √2 | i | ··· i | ··· i | i √2 | i | i + + ) p − | ··· i |− · · · −i + + ) − |− · · · −i | ··· i + ) |− · · · −i |− · · · −i 1 = ( 11L + i( + + + + (S99) √2 | i | ··· i | ··· i + + + ) | ··· i |− · · · −i + + + ) Now we apply B0 to states (S54) and we repli- |− · · · −i | ··· i (1,r),(2,`) + ) cate (S84-S87) as follows. |− · · · −i |− · · · −i 1 = ( 11L + i 00L ) (S100) √2 | i | i x x Let B = B and B0 = iσ σ (1,r),(2,`) − N N+1 (S101) 8

(1) (2) 5. γr γr braid: √X1Y2 gate

1 B 00L = ( 00L + B0( + + + + (S111) | i √2 | i | ··· i | ··· i + + + ) | ··· i |− · · · −i + + + ) |− · · · −i | ··· i + ) |− · · · −i |− · · · −i 1 π (1) (2) = ( 00 ( + + (S112) 4 γr γr L B(10 ,r),(2,r) = e (S102) √2 | i − | ··· i |− · · · −i + + + + ) 1 (1) (2) − | ··· i | ··· i = (1 + γr γr ) (S103) ) √2 − |− · · · −i |− · · · −i + + + ) 1 y z z x |− · · · −i | ··· i = (1 + σN σN . . . σ2N 1σ2N ) (S104) 1 √ − 2 = ( 00L + 11L ) (S113) √2 | i | i 1 x z z x = (1 + iσ1 σN+1 . . . σ2N 1σ2N ) (S105) 1 √ − 2 B 01L = ( 01L + B0( + + + + (S114) 1 | i √2 | i | ··· i | ··· i = (1 + B0 ) (S106) + + ) √ (1,r),(2,r) − | ··· i |− · · · −i 2 + + + ) |− · · · −i | ··· i ) − |− · · · −i |− · · · −i 1 = ( 01L ( + + (S115) √2 | i − | ··· i |− · · · −i + + + + + ) | ··· i | ··· i ) − |− · · · −i |− · · · −i + + ) We demonstrate the correctness of this operator ap- − |− · · · −i | ··· i 1 plied to the logical space and replicating the following. = ( 01L 10L ) (S116) √2 | i − | i 1 B 10L = ( 10L + B0( + + + + (S117) | i √2 | i | ··· i | ··· i + + + ) | ··· i |− · · · −i + + ) − |− · · · −i | ··· i ) − |− · · · −i |− · · · −i 1 = ( 10L ( + + (S118) 1 √2 | i − − | ··· i |− · · · −i X1Y2 00L = ( 00L + 11L ) (S107) + + + + + ) | i √2 | i | i | ··· i | ··· i ) p 1 − |− · · · −i |− · · · −i X1Y2 01L = ( 01L 10L ) (S108) + + + ) | i √2 | i − | i |− · · · −i | ··· i p 1 1 X1Y2 10L = ( 10L + 01L ) (S109) = ( 10L + 01L ) (S119) | i √2 | i | i √2 | i | i p 1 1 X Y 11 = ( 11 00 ) (S110) B 11L = ( 11L + B0( + + + + (S120) 1 2 L √ L L | i √2 | i | ··· i | ··· i | i 2 | i − | i + + ) p − | ··· i |− · · · −i + + ) − |− · · · −i | ··· i + ) |− · · · −i |− · · · −i 1 = ( 11L ( + + (S121) √2 | i − | ··· i |− · · · −i + + + + + ) | ··· i | ··· i + ) Now we apply B to states (S54) and we repli- |− · · · −i |− · · · −i (10 ,r),(2,r) + + + ) cate (S107-S110) as follows. |− · · · −i | ··· i 1 = ( 11L 00L ) (S122) √2 | i − | i x z z x Let B = B(1,r),(2,r) and B0 = iσ1 σN+1 . . . σ2N 1σ2N (S123) − 9

(2) (2) 6. γ` γr braid: √Z2 gate fermion sites 1 and N. By well-separating the ends of the chain, we can suppress such logical errors, and hence sup-

The √Z2 gate represents same braid as √Z1 the dif- press Z errors. Looking at this in the spin language we ference is all indices are shifted by N which is the length have of single logical qubit. y z z y Z = σ1 σ2 . . . σN 1σN . (S128) − This still possesses the delocalized form that is desired III. ERROR PROTECTION OF MAJORANA for error protection, since there must be an interaction ZERO MODES IN THE KITAEV CHAIN between sites 1 and N. In fact, it is arguably even bet- ter protected than the fermionic form, since it involves In this section we explain why the logical MZM states z strings of σn operators arising from the Jordan-Wigner of the Kitaev chain are protected against errors, and the transformation. Thus, we expect that for logical Z er- nature of the error protection after the Jordan-Wigner rors, the spin encoding should still offer protection. In spin mapping. z our case, physical σn errors can be detected and removed To understand the nature of the error protection of the from the final results via postselection as explained in the MZM states in the original fermion Hamiltonian (Eq. (1) main text. This results in an improved teleportation fi- in the main text), let us revisit Kitaev’s original paper delity, as it removes single qubit phase errors that may [9] introducing the model. In Ref. [9], it is explained why occur during the teleportation circuit. information encoded in this system is protected. The sug- gestion in this paper is to use delocalized fermion degrees of freedom (consisting of MZMs) as the logical qubits. IV. MODIFIED TELEPORTATION CIRCUIT Specifically, these are given by (S28) in our notation. It is then explained that in order to protect the quantum Our version of the teleportation circuit uses slightly information, we must protect against both logical X and different gates to the standard version of teleportation, Z errors. We look at both of these in turn. such as that given in Ref. [10], where Hadamard and For the logical X errors, in terms of the delocalized CNOT gates are used. Instead of these gates, we based fermions, Majoranas, and original fermions, this can be our teleportation on the √X1X2 gate, which is imple- written respectively as mented by braiding the right-most MZM from the first chain with the left-most MZM of the second chain. This X = f + f † = γ = c + c†. (S124) N N 1,` 1 1 is most convenient type of gate because for the spin- We see that in terms of the underlying fermions, such mapped representation, this does not involve high order errors require either the loss or gain of a single fermion. spin operations to be performed. For example, Eq. (S82) Such terms are charge non-conserving, and even under in the Supplemental Material is a second order operation, a superconducting Hamiltonian, are fermion parity non- while Eq. (S60) in the Supplementary involves a product conserving (since BCS terms have two fermion annihila- of many spin operators. tion/creation operators). This means that they are un- As with the standard teleportation circuit, there are likely to naturally occur. Meanwhile, in the spin version primarily three steps: (i) preparation of an entangled of the Hamiltonian, these types of errors can occur since qubit between Alice and Bob; (ii) measurement of Alice’s after the Jordan-Wigner mapping we have qubits in the Bell basis; (iii) classical correction at Bob, conditioned on Alice’s measurement outcome. In our cir- x X = c1 + c1† = σ1 . (S125) cuit, the entanglement in (i) is prepared using a logical √X1X2 gate. The Bell measurement (ii) is performed by In the transverse Ising model, this appears as a magnetic combining an entangling operation √X2X3 with a mea- field in the x-direction, which has no reason not to be surement in the local basis. Finally, as explained in the present. The lack of protection is also evident from Eq. main text, the classical correction is performed by ap- (4) in the main text, where applying a σx-error gives 1 plying two √Z3 gates to perform a Z3 correction, two √X X gates with an ancilla qubit set to X = 1 to x 1 3 4 4 σ1 0L = ( + + ) = 1L , (S126) perform the X correction. | i √2 | i − | − −i | i 3 Working in the logical space, the entanglement prepa- which is a logical error. We thus see that the spin map- ration step produces the state ping renders the logical bits susceptible to X errors. Now let us look at Z errors. Here we may evaluate E 23 = X2X3 00L 23 | i | i p1 Z = 1 2f † f = ( 00L 23 + i 11L 23). (S129) − N N √2 | i | i = c1cN + c† c1 + c†cN + c† c†. (S127) N 1 N 1 Logical qubit 1 is meanwhile prepared in the state The point made in Ref. [9] is that such a logical er- ψ = α 0 + β 1 . (S130) ror consists of physically delocalized interactions between | i1 | Li1 | Li1 10

Braiding

KC1 KC2 MZM exchange (1) (1) (2) (2) r r Logical Z Y X Y Y Operation 1 1 2 1 2

Braiding

MZM exchange

Logical X X X Y Z Operation 1 2 1 2 2

FIG. S4: Majorana modes and their braiding operations. The six possible braiding operations for two Kitaev chains (KC), and the eﬀect in terms of the logical states. The left- and right-most Majorana Zero Mode (MZM) on chains 1 and 2 are labeled by (1,2) x y z γ`,r respectively. We denote the Pauli operators for the underlying physical qubits by σ , σ , σ and the higher level logical operators by X,Y,Z.

Measurement outcome Correction V. QUANTUM GATES

00L Z3 | i12 01L X3 | i12 10L X3Z3 | i12 In our approach, we execute the teleportation circuit 11L I3 | i12 as shown in Fig. 2 of the main text by following the same TABLE S1: Classical correction required for the teleportation steps as that followed in a topological quantum computa- protocol. tion. All the steps of the quantum teleportation are performed by successive braiding operations and measurements. Each of the braiding operations are performed by applying the corresponding unitary operations as de-

Now applying the √X1X2 gate we have rived in the previous section. Since our superconducting quantum processor is composed of spins, rather than real anyons, we perform the corresponding unitary operation X X ψ E that achieves the same operation to the braid. 1 2| i1| i23 p1 = (α 000L + iα 011L + β 100L + iβ 111L 2 | i | i | i | i In this section we provide the details on how these + iα 110L α 101L + iβ 010L β 001L ) operations are translated to physical qubit operations in | i − | i | i − | i (S131) Fig. 2(c) of the main text. From this ﬁgure it can be seen 1 that the only gates that are required are the X X , √Z, = ( 00 (α 0 β 1 ) + i 01 (β 0 + α 1 ) i j 2 | Li | Li − | Li | Li | Li | Li encoder, and decoder circuits. We show that the gate de- p 00L (β 0L α 1L ) + i 01L (α 0L + β 1L )). compositions as shown in Fig. 2(c) reproduce these op- − | i | i − | i | i | i | (S132)i erations. We derive these for Kitaev chains are of length N = 2, according to our implementation. We derive in this section the gates for the encoding and decoding A measurement in the logical basis on the ﬁrst two qubits operations which produce the states in terms of the spin- collapses the state to four outcomes, and leaves logical mapped MZM ground states of the Kitaev chain. Finally, qubit 3 in one of four possible states, as can be seen from we also comment on the gates that are performed on the (S132). These can be corrected to the original state by fourth ancilla qubit which helps to perform the X3 clas- applying the operations as summarized in Table S1. sical correction. 11

A. Logical √X1X2 braiding gate The above relation was derived for chain 1, but more generally for a chain of length N = 2, the operations are From (S82) we see that the desired braiding operator applied on the two spins comprising the chain. Let us ξ ξ acting on the physical qubits for the case N = 2 is more generally denote σa as the left-most site and σb as the right-most site, where ξ x, y, z . ∈ { } π x x Analogously to the previous section, we modify (S134) B(1,r)(2,`) = exp i σ σ , (S133) 4 2 3 into the correct form by applying single qubit gates and performing a σx-rotation. The braiding gate is then The above relation was derived between chain 1 and chain i π σy σy 2, but more generally, the operations are applied on the B(1,`)(1,r) = e 4 a b (S141) right-most spin of the ﬁrst chain and the left-most spin of i π σx i π σx i π σz σz i π σx i π σx ξ = e 4 a e 4 b e 4 a b e− 4 a e− 4 b (S142) the second chain. Let us more generally denote σa as the π π x π x π z π z π x π x ξ i 4 i 4 σa i 4 σb i 4 σa i 4 σb i 4 σa i 4 σb right-most site of the ﬁrst chain and σb as the left-most = e− e e e e CZabe− e− site of chain 2, where ξ x, y, z . (S143) ∈ { } On our superconducting quantum processor, the nat- i π x π x π z π z π x π x π = e− 4 R ( )R ( )R ( )R ( )CZ R ( )R ( ), urally available gates are CZ and single qubit unitary a 2 b 2 a 2 b 2 ab a − 2 b − 2 oeprations. Hence rather than decompose our operations (S144) into elementary CNOT gates, we perform decompositions with preference of using CZ gate instead. The CZ gate where in the last line we rewrote the gates in terms of between qubits i and j can be decomposed as (S139). The above expression gives the gate decomposition in Fig. 2(c) of the main text. i π π z π z π z z CZ = e 4 exp i σ exp i σ exp i σ σ . ij − 4 i − 4 j 4 i j (S134) C. Encoder circuit

By removing the single qubit gates and rotating the ba- In this section we derive the encoder quantum circuit, sis of the interaction, we can thus produce the desired deﬁned as the unitary operation that achieves the follow- braiding gate (S133). The above relation was derived for ing chain 1, but more generally for a chain of length N = 2, the operations are applied on the two spin comprising the Uenc 0 (α 0 + β 1 ) = α 0L + β 1L , (S145) | i | i | i | i | i chain. Let us more generally denote σξ as the left-most a where site and σξ as the right-most site, where ξ x, y, z . b ∈ { } 1 The braiding gate is then 0L = ( ++ + ) | i √2 | i |−−i i π σxσx B(1,r)(2,`) = e 4 a b (S135) 1 1L = ( ++ ). (S146) i π σy i π σy i π σz σz i π σy i π σy = e 4 a e 4 b e 4 a b e− 4 a e− 4 b (S136) | i √2 | i − |−−i i π i π σy i π σy i π σz i π σz i π σy i π σy = e− 4 e 4 a e 4 b e 4 a e 4 b CZabe− 4 a e− 4 b The encoder circuit shown in Fig. 2(c) corresponds to the (S137) operator i π y π y π z π z π y π y π = e− 4 R ( )R ( )R ( )R ( )CZ R ( )R ( ), Uenc = H2CZ12H1H2. (S147) b 2 b 2 a 2 b 2 ab a − 2 b − 2 (S138) We show explicitly this achieves (S145) according to the steps below where in the last line we have rewritten the qubit operations in terms of rotation angles on the Bloch sphere Uenc 0 (α 0 + β 1 ) = H2CZ12(α ++ + β + )) | i | i | i | i | −i 1 ξ ξ = H2CZ12 α( 00 + 01 + 10 + 11 ) Rj (θ) = exp iσj θ/2 . (S139) 2 | i | i | i | i + β( 00 01 + 10 11 )) (S148) | i − | i | i − | i where ξ x, y, z . The above expression gives the gate 1 ∈ { } = H α( 00 + 01 + 10 11) decomposition in Fig. 2(c) of the main text. 2 2 | i | i | i − | i + β(00 01 + 10 + 11 )) (S149) | i − | i | i | i 1 B. Logical √Z1 braiding gate = α( 00 + 11 ) + β( 01 + 10 )) (S150) √2 | i | i | i | i 1 From (S42) we see that the desired braiding operator = (α( ++ + ) + β( ++ ))) (S151) acting on the physical qubits for N = 2 √2 | i |−−i | i − |−−i =α 0 + β 1 , (S152) π L L B = exp i σyσy . (S140) | i | i (1,`)(1,r) 4 1 2 as desired. 12

D. Decoder circuit and error detection a fourth ancilla qubit in the state with eigenvalue X4 = +1. Then applying the logical √X3X4 gate twice, we Similarly, we also need to be able to perform reverse accomplish the X3 gate. operation, where the input state is the a two qubit MZM The state with eigenvalue X4 = +1 is in terms of phys- encoded state α 0 +β 1 ), and output the unencoded ical qubits | Li | Li qubit state. This is of course the inverse of (S145) and 1 given by +L = ( 0L + 1L ) | i √2 | i | i U = U † = ++ , (S157) dec enc | i = H1†H2†CZ12† H2† according to (S146). This could be prepared using the = H1H2CZ12H2, (S153) encoder of the previous section, but a simpler way is sim- ply to apply two Hadamard gates since the Hadamard and CZ operations are Hermitian. As discussed in the main text, when a phase flip occurs +L = H1H2 00 . (S158) on the logical states (S146), the states transform as | i | i The only operation that is applied to logical ancilla z z 1 qubit 4 is the braiding operation √X3X4, which (S157) 0˜L = σ 0L = σ 0L = ( + + + ) | i 1 | i 2 | i √2 |− i | −i is an eigenstate of. Hence it should remain unchanged z z 1 after each braiding operation. 1˜L = σ 1L = σ 1L = ( + + ). (S154) | i 1 | i − 2 | i √2 |− i − | −i Finally, the state is decoded using Udec. We use the decoding operation here because we would like to de- We now show that decoding a state with a single phase tect any phase flip errors that may have inadvertently flip error results in a 1 on the first qubit, which allows occurred on these qubits. Without any errors, the state one to detect the error.| i after the decoding is z Specifically, we consider that a σ1 error occurs on the U + = 0 + (S159) output state (S152) such that we have the state dec| Li | i| i z σ (α 0L + β 1L ) = α 0˜L + β 1˜L according to (S152). A measurement of the second qubit 1 | i | i | i | i z 1 here in the σ eigenbasis gives 0 and 1 with 0.5 prob- = [(α + β) + + (α β) + ] ability each. Rather than obtaining| i a| randomi result, it √2 |− i − | −i is more informative to measure in a basis such that any (S155) deviations from the ideal case can easily detected. For this reason we use the modified decoder corresponding to Applying the decoder operation then gives U = H CZ H (S160) U (α 0˜ + β 1˜ ) dec0 1 12 2 dec | Li | Li 1 such that instead the final state is = H1H2CZ12[(α + β) 0 + (α β) +1 ] √2 |− i − | i U 0 + = 0 0 . (S161) 1 dec| Li | i| i = H1H2CZ12[(α + β)( 00 10 ) + (α β)( 01 + 11 ] 2 | i − | i − | i | i In this way the error detection can be still performed in 1 = H1H2[(α + β)( 00 10 ) + (α β)( 01 11 ] a consistent way, and deviations from the ideal result of 2 | i − | i − | i − | i 0 on the second qubit can be easily detected. =H H (α + + β ) | i 1 2 |− i |−−i = 1 (α 0 + β 1 ). (S156) | i | i | i VI. SIMULATION RESULTS OF MAJORANA We this see that the decoder the errored state produces a TELEPORTATION state 1 on the first qubit as claimed. A phase flip on the | i second qubit gives similar results, except that β β. To numerically test our teleportation circuit we simu- → − lated the gate evolutions as given in Fig. 2 of the main text, including gate errors and dephasing effects. We E. Ancilla qubit model both errors by applying random gates that simulate the effect of the noise. In order to match the experi- We finally comment on the gate operations performed mental results we begin by tuning our numerical param- on the fourth ancilla qubit. As explained in the main eters to fixed values provided by characterization of the text, the only role of this is to facilitate the X3 classical experiment. correlation required in the teleportation circuit. Since the To simulate the gate errors, we assume that the Hamil- braiding operations of Fig. S4 does not provide a single tonians that implement the gate are performed correctly, qubit X gate, we can perform this instead by preparing but there is some randomness in the time of the pulse. 13

The time that the pulse is applied is drawn from a Gaus- where the projectors are sian distribution, and the fidelity of the simulation is calculated for each pulse duration according to Πc =I c1 c1 I c2 c2 ⊗ | ih | ⊗ ⊗ | ih | I I I I (S166) N ⊗ ⊗ ⊗ ⊗ 1 2 Πm,k = m1 m1 I m2 m2 I f X S = 1 Rx(π + ξxπ) 0 (S162) | ih | ⊗ ⊗ | ih | ⊗ 2 N |h | | i| n m3 m3 I m4 m4 k k (S167) X ⊗ | ih | ⊗ ⊗ | ih | ⊗ | ih | 1 N f = 1 R (ξ π)CZ 1+ 2 (S163) Here the index c runs over all classical correction out- CZS N |h −| z CZ 12 | i| n comes, and m runs over the the measurements over the X syndrome measurements, k representing the measure- for the X/2 and CZ12 gates respectively. Here Rx,z are ment of the ancilla qubit, which plays no role in the com- single qubit rotation operators, and for the CZ12 gate putation thus no post-selection is defined on its measured the random phase is applied on the target qubit. The value. We note that the state (S165) is unnormalized due gate times ξ are chosen from a Gaussian distribution with to the projectors acting on it. 2 2 mean zero and variance σ , i.e. ξx (0, σg ) and ξCZ We now explain how the teleportation fidelities are cal- 2 ∼ N ∼ culated from the state including the gate and dephasing (0, σCZ). The parameters to tune are the standard N 2 errors. First consider the case when no error syndrome deviations for random sampling: σCZ for a two qubit 2 measurements are made (NS). Given all possible clas- gate and σg for a single qubit gate. The tuned values for each qubit are provided in the Table S4. sical corrections, all possible error detecting qubit out- Dephasing is also simulated in the same way by intro- comes and all possible ancilla qubit outcomes, that have ducing a set of random Gaussian pulses in the middle been evaluated, for nth random draw we can prepare for of the processing. Again, as in case of gate error, de- a traced out density matrix corresponding to the tele- phasing error is characterized by a variance parameter. ported qubit. We denote the variance of the randomly applied dephas- 2 2 ρNS = Tr ( ψf ψf ) ing as σd. Appropriate values for the variance σd are 6,n 1,2,3,4,5,7,8 c,m,k,n c,m,k,n c k m calculated from the experimental dephasing times T2∗ as ED X X X (S168) given in Table S5. In order to adapt those experimentally obtained quantities to act in the numerical simulation we convert them to dimensionless units by applying For the case that error syndrome measurements are made normalization and multiplying by a common phenomeno- (ES), we fix the outcomes of the odd numbered qubits to outcome zero m1 = m2 = m3 = m4 = 0 logical constant cd which accounts for the overall amount of decoherence in the system and is shared among all the ES f f qubits to preserve the individual proportions resulting ρ6,n = Tr1,2,3,4,5,7,8( ψc,m=0,k,n ψc,m=0,k,n ). c k from experimentally measured T ∗. The value of c is ED 2 d X X (S169) calibrated to match the final simulated teleportation fidelities to experimentally obtained corresponding values. We note that the above is an unnormalized state because The teleportation fidelity is calculated as follows. The the full set of measurements are not used. initial state is a state to be teleported initialized on qubit By averaging over a large number of random draws to Q2, simulate the effects of gate errors and decoherence, and applying appropriate normalization we get the fidelity of Ψ = 0 ψ 0 0 0 0 0 0 | 0i | i ⊗ | i ⊗ | i ⊗ | i ⊗ | i ⊗ | i ⊗ | i ⊗ | i the teleported state ψ (S164) | i 1 N ψ ρ ψ where ψ 0 , 1 , + , , +i , i is the state f = h | 6,n| i. (S170) | i ∈ {| i | i | i |−i | i |− i} S N tr(ρ ) to be teleported. From this initial state, we calculate n=1 6,n the fidelity by applying the unitary teleportation circuit X Un(σd, σg, σCZ) with random gate errors and random de- The denominator is present to account for the case that phasing with standard deviations σg, σCZ and σd. The the state is unnormalized. index n represents the nth random draw. To the resulting We calculated the fidelity for both cases, with and state we apply a classical correction circuit U c(σ , σ ). without error detection, for all input state ψ . The n g CZ | i The calculation is repeated for all possible classical cor- numerical values we obtained compared against exper- rections and all possible measurements of error detecting imental values after averaging over N = 2000 random qubits. This is performed by applying a series of projec- runs are provided in the Table S3. The overall features tors, which gives the final state of the form of fidelity profile matches the experiment and in average among all the input states, the error detected fidelity is 2 f c above the 3 threshold. We observed the closest match ψc,m,k,n = Πm,kUn(σg, σCZ)ΠcUn(σd, σg, σCZ) Ψ0 | i of the fidelities for the constant cd = 0.15. Generally E (S165) the theoretically calculated fidelities are higher than the

Simulated and experimental fidelities compared tal fidelities, as well as the errors and the averages are 1.0 6.67% 4.35% 3.41% 13.75% 7.06% provided in the Table S3. The same data is visualized in 16.0% 6.1% form of a bar chart on Fig. S5. 9.59% 26.98% 0.8 1.41% 10.81% 0.0% 13.51% 14.47% VII. EXPERIMENTAL DETAILS 0.6 Our superconducting quantum processor has 12 Fidelities 0.4 frequency-tunable transmon qubits of the Xmon vari- ety. In our experiments, the eight qubits (Fig. 1(a)) are chosen from the quantum processor. The processor has 0.2 qubits lying on a 1D chain, and the qubits are capaci- tively coupled to their nearest neighbors. Each qubit has Simulation No error detection Error detection a microwave drive line (XY), a fast flux-bias line (Z) and 0.0 |0> |1> |+> |-> |+i> |-i> AVG a readout resonator. All readout resonators are coupled to a common transmission line for state readout. The single-qubit rotation gates are implemented by driving FIG. S5: Bar chart visualization of experimental and numer- the XY control lines, and the CZ gate is implemented ical fidelity from Table S3, for cases with and without er- by driving the Z line using the “fast adiabatic” method. ror detection. Chart includes percent errors to compare how The performances of the eight qubits we chosen in our closely simulation matches the experiment. The horizontal experiment are listed in Table S5. 2 dashed line is indicating the 3 threshold. During running the quantum circuits, we have performed the tomography measurement on the initial state ψ 2 on qubit 2 that we prepared for teleportation (see experimentally obtained values. We attribute this to the |Fig.i S6), and the fidelities of six initial states are 0.9998, fact that measurement errors are not taken into account 0.9998, 0.9982, 0.9997, 0.9999, and 0.9989. in our simulation. We expect that this will further reduce In addition, we also performed the tomography mea- the overall fidelities. surement on the final teleported state that before using The detailed values of the simulated and experimen- the error syndrome measurements (see Fig. S7).

[1] M. Leijnse and K. Flensberg, Semiconductor Science and [6] A. Y. Kitaev, Annals of Physics 303, 2 (2003). Technology 27, 124003 (2012). [7] C. Nayak and F. Wilczek, Nuclear Physics B 479, 529 [2] J. Preskill, in Introduction to quantum computation and (1996). information (World Scientiﬁc, 1998), pp. 213–269. [8] D. A. Ivanov, Physical review letters 86, 268 (2001). [3] C. Nayak, S. H. Simon, A. Stern, M. Freedman, and S. D. [9] A. Y. Kitaev, Phys.-Uspekhi 44, 131 (2001). Sarma, Reviews of Modern Physics 80, 1083 (2008). [10] M. A. Nielsen and I. Chuang, Quantum computation and [4] J. K. Pachos, Introduction to topological quantum com- quantum information (2002). putation (Cambridge University Press, 2012). [5] V. Lahtinen and J. K. Pachos, SciPost Physics 3 (2017). 15

Description Experiment Simulation

Teleportation without error detection fNE fNS

Teleportation with error detection fEE fES

X/2 gate ﬁdelity f X f X 2 E 2 S CZ gate ﬁdelity fCZE fCZS

TABLE S2: Fidelity notation to assign a dedicated symbol to a ﬁdelity value corresponding to particular scenario.

0 1 + +i i AVG | i | i | i |−i | i |− i fNE 0.66 0.74 0.74 0.76 0.73 0.63 0.71

fNS 0.66 0.66 0.64 0.65 0.8 0.8 0.7 Error (%) 0.0 10.81 13.51 14.47 9.59 26.98 1.41

fEE 0.75 0.82 0.9 0.92 0.88 0.8 0.85

fES 0.87 0.87 0.96 0.96 0.91 0.91 0.91 Error (%) 16.0 6.1 6.67 4.35 3.41 13.75 7.06

TABLE S3: Teleportation ﬁdelity for a set of input states and average between all those states, calculated with and without error detection, compared to the experimental ﬁdelity for same input states by calculating the percent error.

Qubit Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 AVG 2 σd 0.11407 0.05426 0.11841 0.03014 0.15 0.05764 0.11334 0.06969 0.08844 2 σg 0.016 0.017 0.017 0.018 0.014 0.017 0.013 0.014 0.01575

f X 0.9994 0.9993 0.9993 0.9992 0.9995 0.9993 0.9996 0.9995 0.9994 2 S f X 0.9994 0.9993 0.9993 0.9992 0.9995 0.9993 0.9996 0.9995 0.9994 2 E 2 σCZ 0.08287 0.075524 0.0729 0.0757 0.10285 0.0528 0.056 0.074092

fCZS 0.9832 0.9861 0.987 0.9861 0.9744 0.9932 0.9923 0.986

fCZE 0.983 0.986 0.987 0.986 0.974 0.993 0.992 0.986

TABLE S4: Numerical calibration of qubits to match the experimental performance. Includes gate ﬁdelity of of each qubit, the standard deviation of random error used to reproduce the eﬀect of dephasing.

Qubit Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 AVG

ω10/2π (GHz) 5.066 4.18 5.01 4.134 5.08 4.22 5.132 4.19 -

T1 (µs) 35.2 31.69 35.23 31.01 25.79 27.98 34.79 28.94 31.32

T2∗ (µs) 4.73 2.25 4.91 1.25 6.22 2.39 4.7 2.89 3.67

f00 0.980 0.952 0.981 0.949 0.923 0.896 0.915 0.912 0.939

f11 0.865 0.866 0.905 0.887 0.863 0.858 0.888 0.873 0.876 X/2 gate ﬁdelity 0.9994 0.9993 0.9993 0.9992 0.9995 0.9993 0.9996 0.9995 0.9994 CZ gate ﬁdelity 0.983 0.986 0.987 0.986 0.974 0.993 0.992 0.986

TABLE S5: Performance of qubits. ω10 is idle points of qubits. T1 and T2∗ are the energy relaxation time and dephasing time, respectively. f00 (f11) is the possibility of correctly readout of qubit state in 0 ( 1 ) after successfully initialized in 0 ( 1 ) state. X/2 gate fidelity and CZ gate fidelity are single and two-qubit gate fidelities| i | obtainedi via performing randomized| i benchmarking.| i 16

(a) (b)

(e) (f)

FIG. S6: Tomography of the initial state ψ 2. The initial state prepared on qubit 2 is (a) 0 , (b) 1 , (c) + , (d) , (e) + i , (f) i . | i | i | i | i |−i | i | − i 17

(a) (b)

(e) (f)

FIG. S7: Tomography of the ﬁnal teleported state before using the error syndrome measurements. The initial state prepared on qubit 2 is (a) 0 , (b) 1 , (c) + , (d) , (e) + i , (f) i . Frames show ideal teleportation states, colored bars shows the experimentally| i determined| i state.| i |−i | i | − i