{"id":"20599220-36c4-4a14-bbd4-b227765c6ede","arxiv_id":"2412.14893","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Quantum polyspectral analysis of QPC current reveals a hidden third Markov state in double quantum dot switching, and proves higher-order waiting-time distributions add no information for three-state two-output models.","lead":"This paper tests a newer 'quantum polyspectra' method on current noise from a quantum point contact that watches a double quantum dot. It finds that the apparently two-level switching signal really needs three states to explain it, and shows that ordinary waiting-time statistics cannot tell which three-state model is correct.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim of a hidden third state is load-bearing on the Sec. II background subtraction: if low-frequency environmental noise is charge-configuration-dependent, the narrow residual features in Fig. 3 could be mistaken for a third Markov state.","rationale":"The reader's weakest assumption is the same background-subtraction premise, and I agree it is the point where the central claim is least secure. The strongest parts of the paper, the factorization theorem for three-state two-output WTDs (App. D) and the demonstration that the four minimal three-state models share identical output statistics (App. C), are mathematical and unaffected by experimental systematics. The experimental conclusion, however, rests on distinguishing a genuine third Markov state from environmental fluctuations that are not captured by the single-electron background. Since the manuscript itself flags that environmental noise is not included in the error estimates, this is a self-acknowledged gap rather than an invented objection. The proposed reanalysis of unsubtracted data with a background fluctuator would settle whether the residual narrow polyspectral features are intrinsic. This does not change the reader's CONDITIONAL verdict: the paper should report that test or temper the headline claim. No change to the verdict is needed.","tokens_in":19739,"tokens_out":13827,"duration_ms":121157,"concrete_test":"Reanalyze the raw (pre-subtraction) measurement trace: fit a two-state Markov model augmented by one independent two-level background fluctuator (or a free low-frequency Lorentzian noise term) to the measured second-, third-, and fourth-order polyspectra, and compare AIC with the three-state models of Table I. If the background-enhanced two-state fit attains an AIC within roughly 4 of the three-state fits, the narrow zero-frequency features no longer force a third Markov state; if the three-state model still wins, the background-subtraction concern is refuted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The experimental identification of a hidden third state rests on the premise that, after subtracting a background spectrum taken with a single electron in the stationary (1,0) configuration (Sec. II), all remaining non-two-level statistics in the QPC current originate from the double-dot dynamics. During the actual measurement the dot alternates between (2,0) and (1,1), so the electrostatic environment and the operating point of the QPC differ from the single-electron case. A slow non-Gaussian fluctuator, for example a charge trap active only in the two-electron configuration, would survive the subtraction and produce narrow Lorentzian features in S(2), S(3), and S(4) at zero frequency, exactly where Fig. 3 shows the peaks that rule out a two-state model. The paper's own error estimates exclude 'environmental noise apart from a general white background noise' (Sec. IV), so this possibility is not safeguarded. The analytic factorization proofs in Apps. C and D are independent of this issue and appear sound; the concern is specifically that the QPS/AIC model selection could be selecting a background artifact rather than a dot state. This does not invalidate the WTD evidence for a biexponential high-level dwell time, but it makes the headline 'revealed hidden third state' conditional on the background assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes time-resolved QPC current traces from a gate-defined GaAs double quantum dot tuned near the (2,0)-(1,1) degeneracy. The authors apply their quantum polyspectral (QPS) method to the raw current and fit second-, third-, and fourth-order spectra with models of multi-state Markov dynamics under continuous measurement. They find that a two-state random telegraph model is insufficient and that four structurally different three-state Markov models with two output levels fit the spectra with equal AIC values. They interpret the two high-current Markov states as the singlet and triplet configurations of the (2,0) charge state, i.e., a hidden third state. In addition, the paper proves that for a general three-state Markov process with two output levels, the single-time WTDs already determine all multi-time WTDs (which factorize as products of single-time WTDs), and it proves that four specific reduced three-state models share identical WTDs. The paper also compares QPS with a waiting-time analysis and finds consistent transition rates.","tokens_in":19951,"tokens_out":6306,"duration_ms":49449,"significance":"The WTD factorization theorem (App. D) and the non-uniqueness theorem (App. C) are clean, parameter-free analytic results that clarify the information content of waiting-time statistics for three-state two-output Markov models; these are likely to be of independent interest to the quantum transport and continuous-measurement communities. The open-source software (SignalSnap, QuantumCatch, MarkovAnalyzer) and the detailed AIC comparison are strengths: the fits are reproducible and the derivation steps in Apps. C and D are explicit and internally consistent. The comparison with the earlier five-state model of Maisi et al. is a useful caution about over-parameterization. However, the experimental identification of the 'hidden third state' is conditional on the background subtraction and on equilibrium assumptions; if those hold, the method is a noteworthy assumption-light tool for state-structure inference.","major_comments":[{"comment":"The central 'hidden third state' claim is load-bearing on the background subtraction procedure described in Section II ('Background spectra had been subtracted that were recorded with the same system gate settings but with only a single electron loaded'). The paper's own caveat in Section IV that environmental noise beyond white background noise is not included in the error estimates leaves open the possibility that a slow non-Gaussian fluctuator, active only when the dot is in the two-electron configuration, survives the subtraction and produces the narrow zero-frequency features in S(2), S(3), and S(4) shown in Fig. 3. Because the AIC model selection is based on those spectra, the assignment of a third Markov state rather than a background artifact requires either a control measurement with a known two-state system or an independent check that the residual narrow features vanish when the dot is held in a two-electron configuration without inter-dot tunneling. Without such a test, the headline claim is conditional; this should be addressed or explicitly qualified.","section":"Section II and Section IV"},{"comment":"The identification of the hidden state as triplet states of the (2,0) configuration relies on additional assumptions not tested by the data. The four three-state models in Table I have identical output statistics (App. C), and the data alone cannot distinguish them. The paper excludes Models 2 and 3 only by invoking thermodynamic equilibrium (Section V), and the singlet-triplet assignment further assumes spin-flip scattering within the (1,1) state and equal transition rates within a Markov state. These assumptions are physically plausible but not measured; the reported rates such as gamma_21 = 185 Hz (Model 1) versus gamma_20 = 173 Hz (Model 4) are therefore model-dependent. The conclusion 'reveals hidden excited triplet states' should be softened to state that the data are consistent with a third Markov state, and that the triplet interpretation is one possible assignment under equilibrium assumptions.","section":"Section V and Table I"}],"minor_comments":[{"comment":"The phrase 'the statistics of a three-state Markov model is fully described without multi-time waiting-time distributions' is confusing; it should say that the statistics are fully described by the single-time WTDs, since multi-time WTDs factorize into products of single-time WTDs and thus add no information.","section":"Abstract"},{"comment":"Reference [53] gives the title 'SignalSnap Toolbox' but the URL points to MarkovAnalyzer; the title should be corrected to 'MarkovAnalyzer Toolbox'.","section":"References"},{"comment":"The identities in (D8), which underlie the factorization theorem, are said to be found 'via an explicit calculation for the general three-state Markov model using computer algebra' without showing the intermediate expressions; providing those expressions or a derivation sketch would improve verifiability.","section":"Appendix D"},{"comment":"The caption states 'Thicker arrows represent larger rates' without a quantitative scale; specifying the numerical values or adding a scale bar would make the figure self-contained.","section":"Figure 4 caption"},{"comment":"The notation for the reduced fourth-order spectrum, 'S^(4)(omega_1, omega_2, -omega_1)', is incomplete because a fourth-order polyspectrum has three independent frequency arguments plus the sum condition; please specify explicitly which slice (including the fourth frequency) is used.","section":"Section IV"},{"comment":"The statement 'The general three-state system results in an equally precise fit, however, while introducing two unnecessary parameters immediately visible in the AIC value, which increases by four' could be clarified by noting that a four-parameter increase in AIC corresponds to exactly the 2k term for two extra parameters, assuming unchanged RSS.","section":"Section IV"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's abstract and conclusion present the hidden-state claim more strongly than the current evidence supports: the background subtraction and the equilibrium assumptions are load-bearing but not yet validated. The WTD factorization and non-uniqueness theorems are valuable independent contributions that would remain significant even if the experimental claim is downgraded to 'consistent with' a third Markov state. I would encourage the authors to rebalance the framing of the title and abstract accordingly. The heavy reliance on the authors' own QPS software and prior formalism is not itself a problem, but it strengthens the case for including an independent control or a re-analysis with a different background subtraction before the central claim is accepted at face value."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful core here is the App. D proof: for any three-state Markov model with two output levels, all multi-time waiting-time distributions factor into products of single-time WTDs. That is a clean, parameter-free analytic result, and it explains why the four minimal models in Table I are statistically indistinguishable. The paper also does something right that is easy to underrate: it compares quantum polyspectra and WTD fits on the same 89 s trace, with error bars from 250 simulations, and shows the two methods agree within three standard deviations. The software is public, the fitting procedure is described, and the authors openly state that the general three-state model cannot converge to a unique rate set. That is honest work.\n\nThe soft spots are real but not fatal. The main one is the headline claim of having \"revealed a hidden third state.\" What the data actually establish is that a two-state model fits badly and that the four three-state models fit equally well. The physical assignment of the third Markov state to triplet states in (2,0) needs equilibrium assumptions and a choice between Models 1 and 4; the paper argues for that choice, but it is not forced by the statistics. The stress-test concern about the background subtraction is worth taking seriously, and it is a genuine experimental weakness: the background was taken with one electron in (1,0), while the actual measurement alternates between (2,0) and (1,1), so any charge-configuration-dependent low-frequency noise would survive the subtraction. The paper's own Sec. IV admits environmental noise beyond white background noise is not in the error estimates. That said, the biexponential WTD of the (2,0) level is visible directly in Fig. 2, independent of polyspectra, so the evidence for a second state associated with the high level does not rest entirely on the subtraction. I would not call the central claim load-bearing in the sense that removing the background issue would collapse the paper; it would weaken the triplet interpretation and the model selection, but the WTD data still show two timescales.\n\nThe four-state model search is underreported — \"larger models result in even higher AIC values\" without showing them — and the App. C parameterization is described as \"guessing guided by Table I,\" which is fine for a proof but thin for a model-selection claim. Minor.\n\nWho is this for? Experimentalists working on QD charge detection and people interested in what higher-order statistics can and cannot tell you about hidden states. The App. D result deserves citation on its own. I would send it to a serious referee, and if I were the referee I would ask for the four-state search details and a more careful statement of what the third state is versus what the fit requires.","headline":"A careful experimental QPS analysis with a genuinely new analytic result about WTD factorization, but the 'hidden third state' headline overreaches what a non-unique model fit can establish.","tokens_in":20571,"tokens_out":692,"would_cite":true,"duration_ms":8431,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Higher-order spectra of the detector current reveal a hidden third Markov state in a double quantum dot that appears to switch between only two levels, and waiting-time analysis provably cannot distinguish the alternative three-state…","keywords":["quantum polyspectra","quantum dot dynamics","hidden Markov state","waiting-time distributions","quantum point contact","stochastic master equation","information criterion","telegraph noise"],"falsifier":"Repeat the same double dot measurement at a temperature low enough that $k_BT$ is small compared with the inferred singlet-triplet splitting $\\Delta E\\approx 200\\text{--}300\\,\\mu\\mathrm{eV}$, or apply a magnetic field that separates the triplet states; if the hidden state is a thermally excited spin configuration, the narrow zero-frequency features in the second- and third-order polyspectra and the double-exponential tail of the high-level waiting-time distribution should disappear, leaving a single Lorentzian and a mono-exponential distribution. If they persist, the background subtraction has left a charge-configuration-dependent environmental contribution and the third state is an artifact.","tokens_in":19483,"feed_emoji":"⚛️","tokens_out":13680,"duration_ms":78995,"temperature":0.7,"pith_summary":"This paper claims that the seemingly two-level switching of a double quantum dot monitored by a quantum point contact is actually driven by three Markov states, and that higher-order correlation spectra of the raw detector current, called quantum polyspectra, can reveal that hidden state without identifying individual quantum jumps. The authors fit second-, third-, and fourth-order spectra to candidate models and use an information criterion to show that four different three-state models with very different transition rates fit equally well, implying that the (2,0) charge configuration must be split into two Markov states, most plausibly a singlet and a triplet spin configuration. They prove that for any three-state Markov model with two output levels, every multi-time waiting-time distribution factorizes into single-time waiting-time distributions, so waiting-time analysis cannot distinguish among these models. Both methods recover the same rates with similar accuracy, but the polyspectra method works from the raw current and stays usable at low signal-to-noise, where jump-based full counting statistics falters. If this is right, the result offers a route to detecting hidden dynamics in quantum dot arrays and challenges earlier five-state interpretations of similar data.","feed_headline":"Polyspectra reveal a hidden third state in quantum dot switching","feed_subtitle":"Higher-order current correlations expose dynamics waiting-time analysis cannot distinguish, even in noisy weak measurements.","key_machinery":"The load-bearing objects are the quantum polyspectra $S^{(n)}_z(\\omega_1,\\ldots,\\omega_n)$ of the detector current, defined through $n$th-order cumulants of the Fourier-transformed signal, and their analytic expressions from the stochastic master equation in terms of the Liouvillian $\\mathcal{L}$ and the measurement superoperator $\\mathcal{A}$. The measurement operator that lumps the high output level onto two states is $A=I_{\\mathrm{low}}|0\\rangle\\langle 0|+I_{\\mathrm{high}}(|1\\rangle\\langle 1|+|2\\rangle\\langle 2|)$, so the (2,0) configuration is represented by two Markov states that share the same detector output. The proof that waiting-time analysis cannot go further rests on the identity that for any three-state two-output Markov model, $J_{\\mathrm{up}}e^{\\mathcal{L}_0\\tau}J_{\\mathrm{down}}\\rho_0 = w_{\\mathrm{low}}(\\tau)J_{\\mathrm{up}}\\rho_0$ and the analogous relation for down jumps, which iterates to the factorization $w_{i_n,\\ldots,i_1}(\\tau_n,\\ldots,\\tau_1)=\\prod_{k=1}^n w_{i_k}(\\tau_k)$. Model selection is done by an information criterion, which is what picks out the four minimal three-state models.","core_discovery":"The central claim is that the measured QPC current statistics are exactly those of a three-state Markov model with two output levels, in which the low level corresponds to the (1,1) charge configuration and the high level corresponds to two distinct Markov states within the (2,0) configuration, assigned to the singlet and triplet spin states. A simple two-state model cannot reproduce the second-, third-, and fourth-order spectra, which contain both broad and narrow features, whereas four three-state models with only four nonzero transition rates achieve the same minimal information-criterion score. These models have significantly different transition rates yet yield the same waiting-time distributions and the same polyspectra, because for such a model the identity $w_{i_n,\\ldots,i_1}(\\tau_n,\\ldots,\\tau_1)=\\prod_{k=1}^n w_{i_k}(\\tau_k)$ holds: all multi-time waiting-time distributions are products of single-time ones. Waiting-time analysis therefore cannot identify which model is correct. The polyspectra method and the waiting-time method produce consistent rate estimates within error, but the polyspectra method does not require jump detection and remains accurate at low signal-to-noise. The paper further reinterprets an earlier five-state model for a similar double dot, arguing that the same data can be identically described by three Markov states with modified rates.","pith_inferences":["The same kind of model degeneracy is likely to appear whenever a measurement has fewer output levels than Markov states; treating output-level multiplicity as an additional constraint could be a general strategy for finding minimal state-space models in other hidden-Markov experiments, quantum or classical.","The factorization identity suggests a no-go result for any measurement that lumps exactly two Markov states into one output level: no waiting-time or counting statistics can resolve rates within the lumped level; only noise that couples to the internal dynamics, such as polyspectra, can.","A direct experimental test would be to tune temperature or magnetic field so the triplet is depopulated: the slow component's weight should follow a thermal factor $3\\exp(-\\Delta E/k_BT)$; if narrow spectral features persist under those conditions, the hidden state is environmental rather than spin-related.","Applying the same polyspectra analysis to the earlier five-state dataset for a similar device would show whether that model's extra states are required by the data or are an artifact of model choice."],"forward_implications":["For a three-state Markov model with two output levels, all higher-order waiting-time distributions are redundant: the single-time distributions already determine the full output statistics, so any two models with the same single-time WTDs are experimentally indistinguishable by jump-based analysis.","The four three-state models found in the experiment give identical polyspectra and identical model-selection scores, so the transition rates cannot be uniquely determined from the measurement alone; physical arguments such as bidirectionality and steady-state probabilities must be used to select between them.","Because the polyspectra method uses the raw detector current rather than detected jumps, it can be applied in weak-measurement or low-signal-to-noise regimes where full counting statistics fails.","The two-level telegraph noise with alternating periods of rapid switching and no switching is the fingerprint of a hidden Markov state within the high-current (2,0) configuration, most likely the singlet-triplet splitting, meaning spin dynamics inside the dot are visible in the charge detector statistics.","The earlier five-state model proposed for a similar double quantum dot system need not be correct: the same data can be identically described by a three-state model, so conclusions drawn from the five-state model are not uniquely supported by waiting-time statistics."],"supporting_citations":[{"why":"Supplies the analytic quantum polyspectra expressions up to fourth order from the stochastic master equation that the paper fits to the measured QPC current.","marker":"[36]"},{"why":"Introduces the quantum polyspectra estimation method and the detector-output model used to generate the model spectra.","marker":"[37]"},{"why":"Presents the earlier five-state interpretation of a similar double dot that the paper argues is not uniquely supported and can be reduced to three Markov states.","marker":"[44]"},{"why":"Provides the Liouvillian-based recipe for waiting-time distributions that underlies the proof of identical WTDs and the factorization identity.","marker":"[59]"},{"why":"Defines polyspectra via higher-order cumulants, the spectral quantities central to the method.","marker":"[46]"},{"why":"Supplies the information criterion used to select the minimal three-state models among the candidates.","marker":"[55]"},{"why":"Introduces the stochastic master equation formalism for continuous measurement from which the model spectra are derived.","marker":"[39]"}],"fun_headline_variants":["Quantum polyspectra unmask hidden third state in dot dynamics","Hidden third state in quantum dots exposed by polyspectra","Polyspectra find hidden state that waiting-time analysis misses","Quantum correlators expose extra state in quantum dot switching"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that the noise recorded with only one electron in the dots is identical to the environmental noise present during the two-electron measurement, so that after subtracting that background the remaining narrow spectral peaks reflect dot dynamics and not noise that changes with the charge configuration.","fun_headline_variants_meta":{"raw":{"variants":["Quantum polyspectra unmask hidden third state in dot dynamics","Hidden third state in quantum dots exposed by polyspectra","Polyspectra find hidden state that waiting-time analysis misses","Quantum correlators expose extra state in quantum dot switching"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000593,"raw_usage":{"total_tokens":2850,"prompt_tokens":1089,"completion_tokens":1761,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":705,"completion_tokens_details":{"reasoning_tokens":1693}},"tokens_in":705,"tokens_out":1761,"duration_ms":9305,"temperature":1.0,"reasoning_tokens":1693,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:49:32.905205+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the same double dot measurement at a temperature low enough that $k_BT$ is small compared with the inferred singlet-triplet splitting $\\Delta E\\approx 200\\text{--}300\\,\\mu\\mathrm{eV}$, or apply a magnetic field that separates the triplet states; if the hidden state is a thermally excited spin configuration, the narrow zero-frequency features in the second- and third-order polyspectra and the double-exponential tail of the high-level waiting-time distribution should disappear, leaving a single Lorentzian and a mono-exponential distribution. If they persist, the background subtraction has left a charge-configuration-dependent environmental contribution and the third state is an artifact.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the analytic quantum polyspectra expressions up to fourth order from the stochastic master equation that the paper fits to the measured QPC current."},{"cited_title":"Sifft, A","cited_arxiv_id":null,"evidence_quote":"Introduces the quantum polyspectra estimation method and the detector-output model used to generate the model spectra."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Presents the earlier five-state interpretation of a similar double dot that the paper argues is not uniquely supported and can be reduced to three Markov states."},{"cited_title":"Brandes, Waiting times and noise in single parti- cle transport, Annalen der Physik 520, 477 (2008), https://onlinelibrary.wiley.com/doi/pdf/10.1002/andp.20085200707","cited_arxiv_id":null,"evidence_quote":"Provides the Liouvillian-based recipe for waiting-time distributions that underlies the proof of identical WTDs and the factorization identity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines polyspectra via higher-order cumulants, the spectral quantities central to the method."},{"cited_title":"Akaike, A new look at the statistical model identification, IEEE Transactions on Automatic Control19, 716 (1974)","cited_arxiv_id":null,"evidence_quote":"Supplies the information criterion used to select the minimal three-state models among the candidates."},{"cited_title":"Jacobs and D","cited_arxiv_id":null,"evidence_quote":"Introduces the stochastic master equation formalism for continuous measurement from which the model spectra are derived."}],"review_version":1}