Process-tensor approach to full counting statistics of charge transport in quantum many-body circuits

This paper introduces a numerical tensor-network method based on the process tensor to compute full counting statistics of charge transport in interacting one-dimensional quantum systems, successfully benchmarking the approach on the XXZ spin chain to recover known transport exponents and confirm the breakdown of Kardar-Parisi-Zhang universality in higher-order cumulants at the isotropic point.

The Gist

The Big Picture: Counting the "Unseen" Traffic

Imagine a massive, infinite highway made of quantum particles (like tiny, invisible cars) moving back and forth. In physics, we often want to know: How many cars crossed a specific toll booth in the middle of the highway over a long period?

In the quantum world, this isn't just about counting; it's about understanding the fluctuations. Sometimes 10 cars cross, sometimes 0, sometimes 100. The pattern of these numbers (the "Full Counting Statistics") tells us if the traffic is flowing smoothly (ballistic), getting stuck in jams (diffusive), or doing something weird in between (superdiffusive).

The Problem:
Calculating these statistics for a quantum system is incredibly hard. It's like trying to predict the exact number of cars that will cross a bridge in a city where every car is also a ghost, can be in two places at once, and is constantly interacting with every other car. Traditional computer methods run out of memory almost immediately because the "entanglement" (the spooky connection between particles) grows too fast.

The Solution:
The authors, Hari Kumar Yadalam and Mark T. Mitchison, invented a new "mathematical telescope" called the Process-Tensor Approach. Instead of trying to simulate the entire infinite highway at once, they focus only on the "toll booth" (the interface) and treat the rest of the highway as a "background environment."

---

The Core Idea: The "Influence Matrix"

Think of the highway as a giant, noisy ocean. You are standing on a small raft (the interface) in the middle. You want to know how much water (charge) flows past your raft.

You can't see the whole ocean, but you can feel the waves hitting your raft. The authors realized that all the complex history of the ocean's waves hitting your raft can be compressed into a single, manageable "memory card."

• The Old Way: Try to simulate the entire ocean wave-by-wave. (Too much data! The computer crashes.)
• The New Way: Create a "Process Tensor." This is a mathematical object that summarizes how the ocean influences your raft. It's like a "weather forecast" for the raft that only cares about the past waves that actually matter.

They represent this "memory card" using a Matrix Product State (MPS). Imagine this as a chain of linked paperclips. Each paperclip holds a tiny bit of information about the past. If the chain gets too long, the computer can't hold it.

The Secret Sauce: The "Normalization" Trick

Here is the clever part. When you compress a long chain of paperclips (the MPS) to make it fit in memory, you usually have to throw away some links. In standard physics, throwing away links often breaks the math, making the probabilities add up to more than 100% or less than 0%. That's like saying there's a 110% chance of rain.

The authors developed a special compression algorithm that acts like a "smart filter."

• Analogy: Imagine you are packing a suitcase for a trip. You have to throw away some clothes to fit. Usually, you might accidentally throw away your passport.
• Their Trick: They designed a rule that says, "No matter what you throw away, you must keep the passport." In physics terms, they ensure that the "No-Intervention" probability (the chance that nothing weird happens) stays exactly 1. This keeps the math physically valid even when they cut corners to save memory.

What They Found: The Three Types of Traffic

They tested their method on a famous model called the XXZ Spin Chain (a quantum version of a magnetic highway). They looked at three different "traffic regimes" based on how the cars interact:

1. Ballistic Traffic (The Highway):

   • What it is: Cars zoom through without stopping.
   • The Result: The fluctuations are "Gaussian" (a nice, bell-shaped curve). It's predictable. If you know the average, you know the rest.
   • Analogy: A clear day on the M1. Traffic flows smoothly; the number of cars crossing is very consistent.

2. Diffusive Traffic (The Gridlock):

   • What it is: Cars bump into each other, get stuck, and move slowly.
   • The Result: The fluctuations are wildly non-Gaussian. The bell curve is broken. You get huge spikes where thousands of cars suddenly move, followed by long periods of nothing.
   • Analogy: A chaotic city center during rush hour. You can't predict the flow; it's all about sudden, massive jams and bursts of movement.

3. Superdiffusive Traffic (The KPZ Anomaly):

   • What it is: A weird middle ground (at the "isotropic point").
   • The Result: They found that the famous Kardar-Parisi-Zhang (KPZ) theory, which scientists thought explained this "weird middle ground," actually breaks down when you look at the highest levels of detail (higher-order statistics).
   • Analogy: It's like a river that flows faster than a walk but slower than a sprint, but with a secret twist: the ripples on the surface don't follow the standard rules of fluid dynamics. The authors proved that the "standard rulebook" for this type of traffic is incomplete.

Why This Matters

This paper is a breakthrough for two reasons:

1. It works where others fail: Previous methods crashed when trying to simulate these systems for long times. This new "Process Tensor" method allows them to simulate much longer times, revealing the true nature of the traffic.
2. It changes the theory: By proving that the KPZ universality class breaks down for high-order statistics, they are telling physicists that our current understanding of how quantum matter moves is missing a piece of the puzzle.

In summary: The authors built a new, efficient way to count quantum cars on an infinite highway. They created a "smart suitcase" that holds the history of the traffic without overflowing, and they discovered that the traffic rules we thought we knew are actually more chaotic and interesting than we imagined.

Technical Summary

1. Problem Statement

The paper addresses the challenge of computing Full Counting Statistics (FCS) for conserved charge transport in interacting, one-dimensional (1D) quantum many-body systems far from equilibrium.

• Context: While average currents are often insufficient to characterize transport universality classes (e.g., distinguishing between ballistic, diffusive, and superdiffusive regimes), FCS provides a complete statistical description via the probability distribution of transferred charge.
• Challenge: FCS is defined by a multi-time measurement protocol (measuring charge at initial and final times), making it a non-standard observable. Traditional weak-coupling approximations fail in strongly interacting, isolated systems. Existing numerical methods, such as evolving Matrix Product Operators (MPOs), are limited by the rapid growth of operator entanglement, restricting simulations to short timescales or specific integrable limits.
• Goal: Develop a scalable numerical tensor-network method to compute FCS in 1D quantum circuits with $U(1)$ symmetry, capable of handling non-Markovian memory effects and long-time dynamics.

2. Methodology

The authors introduce a novel algorithm based on the Process Tensor (also known as the influence functional or influence matrix) formalism, represented as a Temporal Matrix Product State (MPS).

A. Process Tensor Formulation

• Interface as Open System: The 1D lattice is partitioned at an interface (site $j=0$). The transport of charge across this interface is treated as an open quantum system problem where the "system" is the interface and the "environment" consists of the semi-infinite left and right halves of the lattice.
• Process Tensors: The influence of the environment on the interface is encoded in process tensors. These tensors describe the multi-time correlations and memory effects of the environment.
• Vectorized Representation: The moment-generating function (MGF), $Z(\lambda, n)$, which generates the charge cumulants, is expressed as the contraction of these process tensors with a "tilted" unitary operator at the interface. The counting field $\lambda$ modifies the unitary gate at the interface, while the bulk gates remain unchanged.

B. Light-Cone Folding and Temporal MPS

• Light-Cone Structure: Due to the unitarity and unitality of the brickwork circuit gates, an exact light-cone structure emerges. This allows the infinite lattice network to be folded into a finite, diamond-shaped tensor network.
• Temporal MPS Representation: Instead of evolving the state in time (which increases spatial entanglement), the authors represent the left and right process tensors as Temporal MPSs.
  • The bond dimension ($\chi$) of the temporal MPS quantifies the temporal entanglement (environmental memory).
  • For integrable systems and certain chaotic systems, this temporal entanglement grows slowly (sublinearly), making the MPS representation efficient.

C. Normalization-Preserving Truncation

• The Challenge: Standard MPS compression (via Singular Value Decomposition) truncates small singular values to reduce bond dimension. However, this generally violates the no-intervention normalization condition (i.e., the process tensor must sum to 1 if no measurement is performed), leading to unphysical results for the generating function ($Z(0, n) \neq 1$).
• The Solution: The authors develop a specialized truncation scheme inspired by density matrix truncation:
  1. They exploit the gauge freedom of the MPS to transform the basis.
  2. They perform a QR decomposition on the "temporal environments" to isolate a specific basis vector that contributes to the normalization.
  3. Truncation is applied only to the complementary block of the matrix, leaving the normalization vector intact.
  4. This ensures that the truncated process tensor exactly preserves the physical normalization condition at every time step.

3. Key Contributions

1. Novel Algorithm: Introduction of a process-tensor-based MPS algorithm specifically designed for computing FCS in interacting quantum circuits.
2. Normalization-Preserving Truncation: A rigorous mathematical scheme to compress temporal MPSs without violating the physical normalization of the process tensor, a critical requirement for accurate FCS.
3. Efficiency: The method shifts the computational bottleneck from operator entanglement (limiting MPO methods) to temporal entanglement. The authors argue that temporal entanglement grows more slowly in many regimes, allowing access to longer timescales.
4. Benchmarking: Successful application to the infinite-temperature XXZ spin-1/2 brickwork circuit, a standard model for studying non-equilibrium transport.

4. Results

The method was applied to study magnetization transport in the XXZ circuit across three regimes defined by the anisotropy parameter $\Delta = J'/J$:

• Ballistic Regime ($|\Delta| < 1$):

  • Observed ballistic transport with exponent $z \approx 1$.
  • Fluctuations are asymptotically Gaussian (kurtosis and sextosis converge to zero).
  • The MGF follows a large-deviation form $Z(\lambda, n) \sim e^{-nF(\lambda)}$.

• Isotropic Point ($\Delta = 1$):

  • Observed superdiffusive transport with exponent $z \approx 3/2$, consistent with the Kardar-Parisi-Zhang (KPZ) universality class.
  • Breakdown of KPZ Universality: While the variance follows KPZ scaling, higher-order cumulants (kurtosis, sextosis) do not converge to the Tracy-Widom or Baik-Rains distributions predicted by standard KPZ theory. This confirms recent findings that KPZ universality breaks down for higher-order statistics in this model.
  • The MGF exhibits a self-similar scaling form $Z(\lambda, n) = Z(\lambda n^{1/2z})$.

• Diffusive Regime ($\Delta > 1$):

  • Observed diffusive transport with exponent $z \approx 2$.
  • Fluctuations are highly non-Gaussian, with kurtosis and sextosis growing monotonically.
  • The MGF also follows the self-similar scaling form, indicating a breakdown of the large-deviation principle.

• Convergence: The method showed excellent convergence for the second cumulant (variance) up to circuit depths of $n=100$. Higher-order cumulants required larger bond dimensions ($\chi \sim 2^{11}$) and converged more slowly, particularly in the diffusive regime.

5. Significance and Outlook

• Theoretical Insight: The work provides strong numerical evidence for the anomalous nature of transport in integrable quantum circuits, specifically the breakdown of standard universality classes (like KPZ) when considering higher-order fluctuations.
• Methodological Advancement: By demonstrating that process tensors can be efficiently truncated while preserving normalization, the paper opens the door to studying non-Markovian open quantum systems and current fluctuations in strongly correlated systems that were previously inaccessible to standard tensor-network time-evolution methods.
• Future Directions: The authors suggest that this approach could be extended to continuous-time dynamics (via Trotterization), dissipative systems, and systems with out-of-equilibrium initial conditions. They also propose that systematic coarse-graining of the process tensor could further extend the accessible timescales for hydrodynamic studies.

In summary, this paper establishes a powerful new numerical framework for analyzing the full statistical distribution of charge transport in quantum many-body systems, revealing deep anomalies in transport universality that average currents alone cannot detect.