Probing the memory of a superconducting qubit environment

Nicolas Gosling, Denis Bénâtre, Nicolas Zapata, Paul Kugler, Mitchell Field, Sumeru Hazra, Simon Günzler, Thomas Reisinger, Martin Spiecker, Mathieu Féchant, Ioan M. Pop · 2026

Read Probing the memory of a superconducting qubit environment by Nicolas Gosling, Denis Bénâtre, Nicolas Zapata, Paul Kugler, Mitchell Field, Sumeru Hazra, Simon Günzler, Thomas Reisinger, Martin Spiecker, Mathieu Féchant, Ioan M. Pop (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-11889.txt.

Introduction

Gosling and co-authors show how to expose long-lived two-level systems hiding in a superconducting qubit's environment. Fault-tolerant designs assume the environment forgets instantly, a Born-Markov picture. In practice, stray two-level systems polarize during use and remember past qubit states. The team fits Solomon equations to quantum jump traces left by thermal fluctuations. That fit separates the qubit's coupling to the memory-keeping defects from its coupling to the standard bath. Sweeping the qubit frequency reveals a peak for each long-lived defect.

Catalog record

CREATORNicolas Gosling, Denis Bénâtre, Nicolas Zapata, Paul Kugler, Mitchell Field, Sumeru Hazra, Simon Günzler, Thomas Reisinger, Martin Spiecker, Mathieu Féchant, Ioan M. Pop
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29625
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.11889

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Probing the memory of a superconducting qubit environment Nicolas Gosling, Denis Bénâtre, Nicolas Zapata, Paul Kugler, Mitchell Field, Sumeru Hazra, Simon Günzler, Thomas Reisinger, Martin Spiecker, Mathieu Féchant, Ioan M. Pop Abstract Achieving fault tolerance with superconducting quantum processors requires qubits to operate within the regime of threshold theorems based on the Born-Markov approximation. This approximation, which models dissipation as constant energy decay into a memoryless environment, breaks down when qubits couple to long-lived two-level systems (TLSs) that become polarized during operation and retain memory of past qubit states. Here, we show that non-Poissonian quantum jump traces carry the information required to distinguish long-lived TLSs from the standard Markovian bath. By fitting the Solomon equations to measured quantum jumps dynamics arising naturally due to thermal fluctuations, we can disentangle the coupling of the qubit to the two environments. Sweeping the qubit frequency reveals distinct peaks, each associated with a TLS that outlives the qubit, providing a handle to understand their microscopic origin. Figure 1: Non-Markovian qubit dynamics induced by coupling to a long-lived TLS. (a) Schematics of the coupling of the qubit to both a TLS and a thermal bath. The coupling rates of the qubit to the TLS and the thermal bath are \mathrm{\Gamma_{qt}} and \mathrm{\Gamma_{q}} , respectively. When the coupling rate of the TLS to the thermal bath \mathrm{\Gamma_{t}} is smaller or comparable to \mathrm{\Gamma_{qt}} , the TLS becomes polarizable and we refer to it as a long-lived TLS. (b) Quantum jumps correlations from Solomon equation based simulation (dashed line) and from measured data (orange). As expected for a single photon emitter within the Born-Markov approximation (black dashed line), the \mathrm{g_{2}} correlation function (see Eq. 1 ) shows anti-bunching of quantum jumps at short time scales and tends exponentially to uncorrelated jumps. In contrast, in the presence of a long-lived TLS, the correlation function shows quantum jump bunching persisting longer than the T\mathrm{{}_{1}} of the qubit (orange). (c) IQ quadrature of 5000 single-shot readouts of the fluxonium qubit, plotted in the complex plane. The radius of the circles indicate 2 standard deviations for the distribution of the points. (d) Representative and cropped sample of the I quadrature over a 3 ms time trace, with the corresponding extracted state indicated by the continuous line. Orange shaded areas highlight three examples of 0.5 ms long time traces, post selected after an energy relaxation event and contributing to the trace averaging. (e,f) Qubit dynamics shown as averaged trajectories conditioned on ground-state selection (green) and jump-down events (orange). The data are well described by fits to the single-TLS Solomon model (eq. 2 ). (f) Zoom-in and re-scaling on the averaged dynamics after the ground state selection scenario, showing deviations from the expected exponential decay. While advances in circuit design and materials have extended superconducting qubit lifetimes into the millisecond regime 21 ; 28 ; 3 ; 23 , improved quantum control and measurement fidelity have revealed new decoherence phenomena that elude characterization by conventional T\mathrm{{}_{1}} energy relaxation times. Measurements of non exponential qubit decay 20 ; 12 , non-Poissonian quantum jumps 31 ; 25 , and hyperpolarization of the qubit environment 24 ; 19 ; 35 , point to the existence of a memory effect in qubit dissipation. This violates the Markov approximation and underscores why T\mathrm{{}_{1}} alone is an insufficient metric. An immediate practical consequence for error correction algorithms is the appearance of time correlations in error syndromes, for example, due to a syndrome qubit flagging the same error multiple times. Note that these temporal correlations are different from the spacial correlation of several qubits on the same chip experiencing bit-flip errors simultaneously due to ionizing radiation impacts in the device substrate 27 ; 30 ; 5 ; 32 ; 28 ; 10 ; 29 . The existence of a memory in the qubit environment has recently been associated with so-called long-lived two-level systems (TLSs) that exceed by orders of magnitude the qubit lifetime 24 ; 15 ; 19 . In this work, we leverage the temporal correlation in the qubit’s quantum jump dynamics to quantify and disentangle the coupling to long-lived TLSs from a Markovian background. In contrast to previous pump–probe experiments 24 ; 35 , this method involves no qubit manipulation. Moreover, continuous monitoring of the qubit via quantum jumps eliminates the need for reset times between measurement cycles. Using quantum jump correlation spectroscopy at different qubit frequencies and electric field biases we uncover a resolved spectrum of electric field susceptible TLS memories, on top of a mostly featureless Markovian environment. Figure 2: Unraveling of the hidden TLS dynamics. Qubit (a) and hidden TLS (b) populations for the qubit at \mathrm{\Delta}t = 0 in the ground state (green) or immediately after a jump down event (orange). The filled markers show measurements. Crosses show populations calculated from quantum jumps simulations, using the \mathrm{\Gamma_{qt}} , \mathrm{\Gamma_{q}} and \mathrm{\Gamma_{t}} extracted from the Solomon equations fit of the measured \mathrm{P_{q}} for \mathrm{\Delta t} > 0. The dashed line shows populations from quantum jumps simulations assuming a Born-Markov environment. Figure 3: Electric field susceptibility of the long-lived TLSs . (a) Single electric field correlation spectroscopy vs qubit frequency at 0 V/m. The Markovian environment of the qubit is nearly featureless, as shown by the qubit relaxation rate \Gamma_{q} (b), while the qubit-TLS decay rate \Gamma_{qt} shows distinct peaks susceptible to electric field bias (c). Insets show the qubit dynamics at the same frequency f_{q}=$2.045\text{\,}\mathrm{G}\mathrm{H}\mathrm{z}$ preformed under two different electric fields, corresponding to a mostly Markovian environment ( \circ ) and a long-lived TLS in resonance ( \diamond ), respectively. To aid the discussion in the main text the long-lived TLSs are labeled A-H. A defining feature of a quantum two-level system is that after emitting a photon, it cannot emit another until it is re-excited. In the case of a continuously monitored qubit this implies an antibunching of quantum jumps. Yet, while superconducting qubits have been shown to behave as emitters of uncorrelated single-photons 13 ; 4 , the presence of even a single long-lived TLS (see Fig. 1 a) in their environment gives rise to bunching of quantum jumps 26 . To quantify this we use the \mathrm{g_{2}} correlation function: g_{2}=\frac{\sum_{t}P(\downarrow,t|\downarrow,t-\tau)}{P(\downarrow)}, (1) where P(\downarrow,t|\downarrow,t-\tau) denotes the conditional probability of a jump down at a time t, given a jump down at time t-\tau , and P(\downarrow) is the unconditional decay probability. For an ideal single-photon emitter, g_{2}(\tau) starts at zero and exponentially approaches unity on the energy relaxation timescale T_{1} (dashed line in Fig. 1 (b)). Experimentally, to extract g_{2}(\tau) for a superconducting qubit, we obtain a quantum jump trace by performing stroboscopic readout every 6 \mathrm{\upmu s} (cf. Fig. 1 (c,d)), identify all downward jump events, average the subsequent time traces and normalize them by P(\downarrow) . Figure 1 (b) shows results obtained on a fluxonium qubit (For information on the device see Appendix A ), biased at f_{q}=2\,\mathrm{GHz} near the half-flux sweet spot. Although we observe the expected antibunching at short \tau , for \tau exceeding the qubit relaxation time T_{1} we instead find bunching: the photon emission probability exceeds its steady-state value. The bunching of quantum jumps over times scales much longer than the qubit’s T\mathrm{{}_{1}} is indicative of the

Read the complete text