Post-selected Criticality in Measurement-induced Phase Transitions

arXiv:2603.15744 · quant-ph, cond-mat.dis-nn, cond-mat.stat-mech, cond-mat.str-el · Submitted 2026-03-16 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Post-selected Criticality in Measurement-induced Phase Transitions".

Mira: Information-theoretic phase transitions, such as measurement-induced phase transitions (MIPT),

Kai: First, who's behind it and why it matters.

Paper summary: Kai: So, we're diving into this paper, "Post-selected Criticality in Measurement-induced Phase Transitions," which looks like it's tackling a really interesting topic in how quantum systems behave when you start monitoring them. Mira, what's the core idea they are pushing here?

Mira: Well, the main thesis of this paper is that measurement-induced phase transitions, or MIPT, have a specific way they change depending on whether you use standard Born-rule post-selection or forced measurements. The authors claim that explicit post-selection fundamentally alters the universality class by reweighting trajectories that would normally be very rare.

Lev: That sounds deep, Mira; if this is true, it suggests that how we condition our measurements changes the fundamental physics of the transition itself, which is a big deal for error correction applications where we often rely on specific measurement outcomes.

Kai: Exactly. It's not just a detail; it's about how the system organizes itself based on what we select post-measurement. Mira, can you elaborate on what this means in practical terms for quantum dynamics?

Mira: When they compare the post-selected MIPT to standard Born-rule MIPT, they find some distinct scaling behavior. Specifically, they report a correlation-length exponent of ν ≈ two point one for the post-selected case, which is larger than what you'd expect from the standard MIPT <ref:2603.15744#pg0>. They also found an effective central charge ceff ≈ −zero point four for this post-selected version, which is negative, unlike the positive value seen in the Born-rule MIPT at k = one <ref:2603.15744#pg0>.

Lev: A negative central charge is interesting from an error correction standpoint; it usually points toward some sort of structure or boundary condition that's different from the standard thermal phase, and I wonder if this negativity implies a specific constraint on the noise that could be exploited in certain protocols <ref:2603.15744#pg0>.

Kai: That's what makes me curious about the models they use to prove this; they're looking at two main setups, the Post-selected Random Quantum Circuit or P-RQC, and the Random Tensor Network or RTN <ref:2603.15744#pg1>. Mira, how do these different models help illustrate their point about trajectory reweighting?

Mira: The P-RQC model involves two-qubit gates from the Haar distribution arranged in a one-dimensional bricklayer geometry with forced measurements on some qubits with probability p that project them onto zero⟩ <ref:2603.15744#pg1>. Meanwhile, the RTN involves a two-dimensional network of local tensors where the effective bond dimension is tuned by tracking the second R´enyi mutual information chi n=two across an edge <ref:2603.15744#pg2>. These setups let them map out how these transitions manifest in different physical architectures, which is crucial for checking if this reweighting effect is general.

Paper summary: Lev: From a hardware perspective, the P-RQC setup sounds like something you could potentially realize on a superconducting circuit platform where you can implement controlled gates and insert specific measurement operators at certain points <ref:2603.15744#pg1>. I'm thinking about the physical constraints of implementing those forced measurements reliably in a real system.

Kai: Right, and the results they extract from these models show some critical parameters, like (P-RQC) = zero point two four(one) and chi c (RTN) = zero point nine(nine), which are used to determine the critical points <ref:2603.15744#pg1>. These numbers help ground the theoretical claims in actual parameter regimes for these specific models.

Mira: And the scaling relations they derive, such as nu RQC = two point one(five) and nu RTN = two point two(five), are what really highlight the shift in universality class compared to the standard MIPT <ref:2603.15744#pg0>. They also found that this post-selected MIPT shares its universality class with the entanglement transition of Random Tensor Networks, which is a significant connection because it ties information theory to tensor network physics, where contracting the network reveals the area-law versus volume-law behavior <ref:2603.15744#pg2>.

Lev: Connecting it to RTNs is powerful because those networks are often used as simplified models for more complex quantum states we encounter in simulations, and if this critical property holds there too, it means the physical principle is robust across different mathematical frameworks <ref:2603.15744#pg1>.

Kai: I'm really interested in the discussion about qutrits versus qubits; they found that moving to qutrits restores the transition when varying the weak-measurement strength p w, but for qubits, p w drifts to zero as system size increases <ref:2603.15744#pg1>. That hints at a strong dependence on the onsite Hilbert space dimension in how these transitions behave.

Mira: That dependence is key, because it shows that the randomness within individual realizations isn't always necessary for MIPT criticality in these specific post-selected weak-measurement settings when you use qutrits <ref:2603.15744#pg1>. Furthermore, they found that the translationally invariant RTN with three-state physical legs exhibits an entanglement transition as the effective bond dimension is tuned <ref:2603.15744#pg2>.

Lev: If we consider running this on actual hardware, the finding that nu > two suggests that the system might actually be robust against static disorder, which is something we always look for in realizing stable quantum states for error correction <ref:2603.15744#pg0>. But what about the limitation they admit?

Kai: They mentioned that the major difference in critical properties compared to the standard MIPT originates from this reweighting of trajectories, but they also noted that finite-size effects aren't perfectly captured at these system sizes, or maybe there is a more subtle difference between those two models <ref:2603.15744#pg1>. That's something we have to keep in mind when trying to translate this into a lab setting.

Paper summary: Mira: They also pointed out that p c is larger after post-selection than it was in the Born-rule MIPT, which they suggest could enhance the volume-law phase's ability to retain quantum information <ref:2603.15744#pg0>. The Casimir form for their free energy density, F(L, t)/A = f(L) = f(L = infinity) - pi c eff/6L squared, really formalizes how this change in the phase is described thermodynamically <ref:2603.15744#pg0>.

Lev: That Casimir form gives us a clear thermodynamic description of the transition's nature, which is exactly what we need to connect these abstract scaling exponents back to measurable quantities in a physical experiment like this <ref:2603.15744#pg0>.

Kai: So, to wrap up the summary of "Post-selected Criticality in Measurement-induced Phase Transitions," the paper argues that post-selection doesn't just slightly nudge the transition; it fundamentally changes its nature by reweighting trajectories, leading to new universality classes characterized by different exponents and central charges.

Mira: The implications are significant because this work links information theory directly to entanglement transitions in tensor networks, suggesting a deep connection between how information spreads and the underlying structure of quantum states <ref:2603.15744#pg2>.

Lev: For error correction, if we can understand how post-selection alters the critical behavior, it might inform us about designing measurement sequences that are more robust against certain types of noise or disorder when implementing fault-tolerant operations <ref:2603.15744#pg0>.

Kai: It’s exciting to see how these abstract theoretical concepts map onto concrete models like the P-RQC and RTN, giving us tangible benchmarks to look for in experimental setups <ref:2603.15744#pg1>.

Mira: Indeed, the study confirms that nu > two suggests robustness against static disorder, which is a positive result when comparing it to the Born-rule MIPT which flows towards infinite randomness <ref:2603.15744#pg0>.

Lev: We have a lot of work ahead, though; translating these findings into error correction protocols requires figuring out how to handle the specific constraints imposed by these post-selected states when designing actual quantum gates and measurements <ref:2603.15744#pg1>.

Kai: That's where the experimentalists come in, figuring out if we can actually build something that exhibits these critical properties under controlled conditions <ref:2603.15744#pg0>.

Mira: We have a solid theoretical framework now showing how conditioning measurements impacts phase transitions across different models, which opens up new avenues for understanding information flow in disordered systems <ref:2603.15744#pg2>.

Lev: This paper provides a strong foundation for thinking about how measurement outcomes dictate the accessible physical phases of a quantum system, which is something vital for the next generation of quantum computation research <ref:2603.15744#pg0>.

Conclusion: Kai: So, we've been walking through the technical details of this paper on measurement-induced phase transitions and how post-selection fundamentally reweights those quantum trajectories. Now, Mira and I want to bring it back to the big picture with a look at what this whole paper is actually about.

Mira: This work focuses on how explicit post-selection in measurement experiments alters the nature of these phase transitions by affecting which quantum paths are considered valid or probable. It’s essentially showing that conditioning your measurements changes the physics you observe.

Lev: From my side, I'm thinking about what this means for building error correction circuits; if we can control how post-selection affects the critical behavior, it could give us new ways to stabilize our physical systems against noise.

Kai: Right, and looking at the title 'Post-selected Criticality in Measurement-induced Phase Transitions,' it really captures that core idea—how the act of selecting a measurement outcome dictates what kind of phase transition we see. The authors are showing that this isn't just a small tweak; it shifts the entire universality class based on how those trajectories are weighted.

Mira: Exactly, and their comparison between post-selected and standard Born-rule MIPT results in different scaling exponents, like that correlation length exponent nu about two point one, which is distinct from the standard result they're comparing against. That difference is where the theoretical meat of the paper lies.

Lev: That distinction in scaling exponents is exactly what I need to see when thinking about whether a specific error correction scheme will hold up on real hardware; those exponents tell us a lot about stability under disorder.

Kai: And the connection they draw between this post-selected transition and the entanglement transition found in Random Tensor Networks is quite interesting, suggesting a deeper link between measurement physics and information structure in these models.

Mira: That's a big assumption, but it's what makes this paper compelling; linking MIPT to RTN entanglement transitions suggests that these concepts aren't isolated but are part of a larger information-theoretic picture of how quantum states organize themselves.

Lev: It’s fascinating because RTNs offer a different mathematical framework than the circuit models they tested, so seeing them share the same critical properties strengthens the idea that this phenomenon is inherent to certain types of complex quantum systems.

Kai: So, in simple terms, this paper demonstrates that how we choose what we measure post-experiment doesn't just give us more data; it fundamentally changes the underlying physics governing how a quantum system transitions between different states.

Mira: It’s about realizing that the measurement process itself is an active part of defining the quantum phase transition, rather than just passively observing it.

Lev: And this has direct implications for error correction because understanding these altered critical points could guide us in designing measurement strategies that are more resilient to noise during the computation.

Kai: It really opens up a new way to think about the experimental side of things, moving beyond just "does it work?" to "how does *this specific way* of measuring affect what phase we get?".

Mira: And with these results, I'm eager to see how this information theory perspective can inform our understanding of complex, disordered quantum systems we encounter in nature.

Lev: That leads perfectly into the next part where we look at how these theoretical findings translate into practical constraints for building scalable quantum hardware.

Department of Physics and Astronomy, Louisiana State University · Department of Theoretical Physics, University of Geneva · Department of Physics, University of Massachusetts Amherst · City College of New York, City University of New York · The Pennsylvania State University

quant-ph, cond-mat.dis-nn, cond-mat.stat-mech, cond-mat.str-el

Submitted: 2026-03-16

Updated: 2026-10-02

Comments: 5+3 pages, 3+6 figures

Journal ref: Phys. Rev. Lett. 137, 130403 (2026)

DOI: 10.1103/gx2q-9xz1

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 78/100

The gist: Information-theoretic phase transitions, such as measurement-induced phase transitions (MIPT), characterize the robustness of quantum dynamics to local monitoring and are naturally formulated in

Key concepts

Measurement-Induced Phase Transitions (MIPT)
These are quantum phase transitions where the system's properties change based on local monitoring or measurement. The paper examines how forcing specific measurements (post-selection) modifies the nature and critical behavior of these transitions.
Universality Class
This refers to a set of critical exponents that describe the universal behavior of a physical system near a phase transition. The study shows that post-selection changes this class, linking it to Random Tensor Networks, which have their own distinct critical properties.
Post-selection Reweighting
When researchers select specific measurement outcomes (post-selection), they effectively change the probability of observing certain system trajectories. This reweighting process is key; it shifts the critical behavior away from standard Born-rule MIPT and towards a new universality class.
Random Tensor Networks (RTN)
RTNs are a model used to study entanglement transitions in quantum systems. The paper finds that post-selected MIPT shares its universality class with RTNs, implying that the transition behavior is naturally interpreted through the network's structure.

Terminology

Summary

Information-theoretic phase transitions, such as measurement-induced phase transitions (MIPT), characterize the robustness of quantum dynamics to local monitoring and are naturally formulated in terms of trajectories conditioned on typical measurement outcomes, which are naively accessible only through post-selection. This work implements forced measurements to investigate how explicit post-selection alters the nature of the transition, finding that it fundamentally reweights trajectories that are otherwise rare and shifts the universality class.

Key Findings on Universality Class Alteration

The study demonstrates that post-selection fundamentally alters the universality class by reweighting trajectories that are otherwise rare. Specifically, when comparing post-selected MIPT to standard Born-rule MIPT, the results show:

  1. A correlation-length exponent ν ≈ 2.1, which is larger than that of the standard MIPT.

  2. A negative effective central charge ceff ≈ −0.4.

Furthermore, the authors confirm that this post-selected MIPT shares its universality class with the entanglement transition of Random Tensor Networks (RTN). This connection is significant because in RTNs, "the informationtheoretic interpretation of the entanglement transition is natural: in the area-law phase, contracting the network is efficient, whereas in the volume-law phase it becomes exponentially hard in the system size."

Model Implementations and Critical Parameters

The research numerically investigates two primary models:

  1. The Post-selected Random Quantum Circuit (P-RQC): This model consists of two-qubit gates, Ui,i+1, drawn from the Haar distribution and arranged in a one-dimensional bricklayer geometry with randomly interspersed forced measurements with probability p that project the measured qubits onto 0⟩.

  2. The Random Tensor Network (RTN): This involves a two-dimensional network of local tensors Tµ,ν,κ,λ, where tensor entries are drawn from a Gaussian distribution. The effective bond dimension is tuned by tracking the second R´enyi mutual information χn=2 of the two-legged state λ⟩ across an edge.

The critical points extracted from scaling relations are:

** (P-RQC) = 0.24(1) and χc (RTN) = 0.9(9). The correlation length exponents are νRQC = 2.1(5) and νRTN = 2.2(5). These exponents suggest that Harris’ criterion implies that the transition could be stable against static disorder. Furthermore, the authors find that for a single two-site unitary circuit with post-selected weak measurements, moving to qutrits restores the transition as the weak-measurement strength pw is varied, suggesting a strong dependence on the onsite Hilbert-space dimension. The translationally invariant RTN with three-state physical legs exhibits an entanglement transition as the effective bond dimension is tuned. In contrast, for qubits, the weak-measurement crossover pw drifts to 0 with increasing system size. This implies that randomness within individual realizations is not strictly necessary for MIPT criticality in these post-selected weak-measurement settings when using qutrits. The effective central charges are found to be ceff,RQC = −0.40(6) and ceff,RTN = −0.35(2). The negativity of ceff is attributed to the fact that c(k) = 0 at both k = 0 and k = 1 due to trivial partition functions in these limits, and for the Born-rule MIPT, the effective central charge at k = 1... is positive. The post-selected model's free energy density follows the Casimir form: F(L, t)/A = f(L) = f(L = ∞) − πceff/6L squared. Finally, the study extracts spatial correlation exponents ηRQC = 0.17(2) and ηRTN = 0.16(2). The conclusion is that the major differences in critical properties as compared to the standard MIPT (see Table I) originate from a reweighting of trajectories. The authors also note that pc is larger after post-selection than it was in the Born-rule MIPT, suggesting that even naive selection of trajectories could enhance the volume-law phase’s ability to retain quantum information. The paper also confirms that for the translationally invariant RQC with qubits, we do not observe a MIPT, but this transition reappears for qutrits. The critical exponents are consistent between the models, though scaling functions appear to differ. This suggests either that finite-size effects are not well captured at these system sizes or that there may be a more subtle difference between the two models. The paper concludes by noting that "ν > 2 indicates that the system may be robust against static disorder, unlike the Born-rule MIPT which flows to infinite randomness.

Improvements for AI systems

As a fastidious and diligent researcher, I have analyzed this paper, Post-selected Criticality in Measurement-induced Phase Transitions, focusing on its findings related to measurement-induced phase transitions (MIPT) and their connection to Random Tensor Networks (RTN).

The paper provides fundamental insights into how explicit post-selection alters the universality class of quantum phase transitions. The core takeaway is that post-selection fundamentally changes the critical exponents, leading to a distinct universality class characterized by a negative effective central charge.

Here are specific improvements for AI systems based on this research:

) Improved AI System Capabilities:

The insights from this paper can be leveraged to develop next-generation quantum machine learning (QML) and quantum simulation algorithms, specifically targeting systems where local monitoring or measurement is a feature rather than an external perturbation.

  1. Enhanced Robustness in Quantum Simulation (Post-Selection Aware):

  2. Novel Phase Transition Detection Algorithms:

  3. Improved Quantum Neural Network Training Stability:

) Specific Improvements and Applications:

  1. The system can be trained to identify and navigate quantum states exhibiting MIPT behavior more reliably by incorporating post-selection constraints into the training objective function, mimicking the reweighting of trajectories discussed in Section 1. The improved AI system could perform state preparation or optimization in noisy environments where specific measurement outcomes are desired (e.g., optimizing a circuit's output conditioned on a specific low-error measurement result).

  2. The AI can be designed to detect the onset of an MIPT in complex quantum circuits or tensor network simulations by monitoring correlation functions (like tripartite mutual information, Section 3) and looking for the characteristic scaling collapse described by the critical exponents (e.g., extracting a correlation length exponent ν ≈ 2.1). This allows for rapid identification of critical points in disordered quantum models, potentially accelerating materials discovery simulations where phase transitions are key phenomena.

  3. The system can be utilized to design more stable and scalable Quantum Neural Networks (QNNs) by understanding the role of local monitoring. The paper suggests that for certain systems (like qutrit circuits), the criticality is independent of randomness within individual realizations, implying a pathway to designing QNN architectures where local monitoring doesn't necessarily destabilize the learning process. This would lead to QNNs that are more robust against stochastic gate errors or noise, potentially requiring fewer classical resources for error correction.

  4. The AI can be employed in Purification-Based state estimation algorithms. By using the local order parameter derived from tracing out an ancilla (Section 3), the system can estimate the purification time or purification capability of a quantum state under weak monitoring, leading to more accurate and efficient quantum tomography protocols for large systems where full state tomography is infeasible due to Hilbert space growth.

  5. For large-scale quantum data processing, the AI can learn to map data onto models exhibiting the RTN universality class (which shares the same post-selected MIPT universality class), allowing it to efficiently compress and analyze highly entangled quantum states by leveraging the known scaling behavior of these networks.

Sources

Related papers