Real-time pseudo entropy and modular-Hamiltonian correlations

This paper investigates the short-time behavior of real-time pseudo entropy, demonstrating that its initial imaginary and real responses are fundamentally governed by the symmetrized covariance and commutator, respectively, between the physical Hamiltonian and the modular Hamiltonian, thereby revealing pseudo entropy as a time-oriented modular response rather than a mere branch artifact.

The Gist

Imagine you have a perfectly sealed, frictionless box containing a complex machine. Inside, everything is moving in perfect harmony. If you watch the whole box, nothing ever gets "messier" or "more random"; the total order is perfectly preserved. This is how a closed quantum system works: it's reversible, and no entropy (disorder) is created in the grand scheme.

But, what if you only look at one small gear inside that machine?

This paper explores what happens when we zoom in on just a tiny part of a quantum system and watch it evolve over time. The author introduces a new way of measuring "disorder" for this small part, called Pseudo Entropy.

Here is the breakdown of the paper's ideas using everyday analogies:

1. The "Time-Traveling Snapshot" (Pseudo Entropy)

Usually, to measure how messy a system is, you take a snapshot of it right now. But this paper uses a special tool called a Transition Matrix.

Imagine you take a photo of a dancer at the start of a routine (Time 0) and another photo of them at a later moment (Time $t$).

• Standard Entropy looks at just the second photo and asks, "How messy is this pose?"
• Pseudo Entropy looks at the relationship between the first photo and the second. It asks, "How does the transition from the start pose to the end pose look?"

Because this tool connects two different moments in time, it can produce a number that isn't just a simple "amount of mess." It produces a complex number (a number with a real part and an imaginary part). Think of this like a compass: the "real" part tells you the distance, but the "imaginary" part tells you the direction.

2. The Main Discovery: The "Imaginary Arrow"

The paper's biggest finding is about what happens in the very first split-second after the system starts moving.

The author found that the "imaginary part" of this new entropy isn't just a mathematical glitch or a weird side effect. It is a real, measurable arrow of time.

• The Analogy: Imagine a river flowing. If you drop a leaf in, it moves downstream.
  • The Real Part of the entropy change is like the leaf getting wet or the water getting turbulent (it depends on how the water swirls).
  • The Imaginary Part is the direction the leaf is drifting. It tells you, "This is forward in time."

The paper proves that this "direction" (the imaginary part) is generated by a specific relationship between two things:

1. The Engine (Physical Hamiltonian): The force driving the time evolution (the river's current).
2. The Map (Modular Hamiltonian): The internal structure or "memory" of the specific part of the system you are watching (the shape of the riverbed).

If the engine and the map are "correlated" (they work together in a specific way), the system immediately generates this time-arrow signal. It's like the system saying, "I am moving forward because my internal structure is reacting to the engine."

3. The "Real" Part vs. The "Imaginary" Part

The paper separates the response into two distinct behaviors:

• The Imaginary Response (The Arrow): This happens even if the system is perfectly symmetrical. It is driven by how much the "engine" and the "map" are covariant (how they move together). It's the primary signal that time is passing.
• The Real Response (The Change): This part only happens if the "engine" and the "map" clash (if they don't commute). It's like two gears grinding against each other. If they are perfectly aligned, this part doesn't change immediately; it only grows slowly over time.

4. Testing the Theory

The author didn't just do math on paper; they tested this idea in three different ways:

• A Simple Toy Model: They used a system with just two "qubits" (quantum bits) to show that the math works perfectly.
• A Chain of Spins (Ising Model): They simulated a long chain of magnets. They found that near a "critical point" (where the magnets are on the verge of flipping from one state to another), this "time-arrow" signal becomes very strong. It's like the system is most sensitive to the flow of time right when it's about to change its mind.
• A "Ghost" System (Non-Hermitian): They looked at systems where energy isn't perfectly conserved (like a system losing energy to the air). They showed that even in these "ghostly" systems, the same rule applies, though the math gets a bit wilder (like the compass needle spinning wildly near a magnetic storm).

5. Why This Matters (Without Overhyping)

The paper clarifies a confusing point in physics: Where does the "Arrow of Time" come from?

In a closed universe, time is reversible. But if you zoom in on a small piece, you see a direction. This paper says that this direction isn't just a result of us forgetting information (coarse-graining) later on. It is built into the amplitude (the quantum probability wave) itself, right from the very first instant.

The "Imaginary Part" of this entropy is the universe's way of whispering, "I am moving forward," before any actual disorder (heat, mess) has even had a chance to build up. It is a microscopic, quantum-level "time orientation" generated by the correlation between how a system moves and how it is structured.

In short: The paper discovered that if you look closely enough at a quantum system, the very first moment it starts moving reveals a hidden "time arrow" (the imaginary part of pseudo entropy) that is caused by the system's internal structure reacting to the forces driving it.

Technical Summary

Technical Summary: Real-time pseudo entropy and modular-Hamiltonian correlations

Problem Statement
The paper addresses the microscopic origin of the thermodynamic arrow of time in closed quantum systems. While unitary evolution preserves the total von Neumann entropy, irreversibility typically emerges through subsystem selection, decoherence, or coarse graining. Standard formulations of entropy production rely on probabilistic distinguishability (e.g., relative entropy) between forward and backward processes. The authors investigate whether a more fundamental, amplitude-level entropic quantity exists that carries a time orientation before the emergence of a real-valued second-law inequality. They focus on pseudo entropy, a complex-valued generalization of entanglement entropy defined via a reduced transition matrix rather than a reduced density matrix, to probe this question in the context of real-time unitary evolution.

Methodology
The authors define the real-time pseudo entropy $S_A(t, 0)$ for a subsystem $A$ using a transition matrix $\tau(t, 0)$ constructed from an initial pure state $|\Psi\rangle$ and its unitary time evolution $|\Psi(t)\rangle = e^{-iHt}|\Psi\rangle$:
$$\tau(t, 0) = \frac{|\Psi(t)\rangle\langle\Psi|}{\langle\Psi|\Psi(t)\rangle}, \quad \tau_A(t, 0) = \text{Tr}_{\bar{A}} \tau(t, 0)$$
$$S_A(t, 0) = -\text{Tr}_A [\tau_A(t, 0) \log \tau_A(t, 0)]$$
The study proceeds by performing a short-time expansion ($t \to 0$) of this quantity. The authors utilize the known first-law-like variation of pseudo entropy, relating the first variation of the entropy to the modular Hamiltonian $K_A = -\log \rho_A$ of the initial reduced state. They decompose the resulting linear response into real and imaginary parts, analyzing the correlation between the physical Hamiltonian $H$ and the modular Hamiltonian $K_A$.

The methodology includes:

1. Analytical Derivation: Expanding the transition matrix and applying trace-logarithm variations to derive the leading-order behavior.
2. Solvable Models: Exact solutions in a Schmidt-diagonal model (where $H$ and the initial state share an eigenbasis) and a non-commuting two-qubit model.
3. Many-Body Simulations: Numerical exact diagonalization of transverse-field Ising chain quenches to study finite-size effects and critical behavior.
4. Non-Hermitian Extension: Formulating a one-sided biorthogonal transition matrix for non-Hermitian systems and analyzing the analytic structure of pseudo entropy across $PT$-symmetry breaking.

Key Contributions and Results

1. Universal Short-Time Expansion: The paper establishes that for an arbitrary Hermitian Hamiltonian, the real-time pseudo entropy admits a universal short-time expansion:
   $$S_A(t, 0) = S_A(0) - it \langle K_A (H - \langle H \rangle) \rangle + O(t^2)$$
   where $\langle \cdot \rangle$ denotes the expectation value in the initial state.

2. Covariance/Commutator Decomposition: The central result is the separation of the initial response into distinct physical mechanisms:

   • Imaginary Part (Time-Oriented Response): The initial imaginary response is governed by the symmetrized covariance between the physical Hamiltonian and the modular Hamiltonian:
     $$\frac{d}{dt} \text{Im } S_A(t, 0) \bigg|_{t=0} = -\frac{1}{2} \langle \{ K_A - \langle K_A \rangle, H - \langle H \rangle \} \rangle$$
     The authors argue this is not a mere artifact of logarithmic branch cuts but a genuine "time-oriented modular response" generated by the correlation between microscopic time evolution and the subsystem's information structure.
   • Real Part (Commutator Response): The initial real response is governed by the commutator of the two operators:
     $$\frac{d}{dt} \text{Re } S_A(t, 0) \bigg|_{t=0} = \frac{1}{2i} \langle [K_A, H] \rangle$$
     This implies the real part has no linear response if $[K_A, H]$ has a vanishing expectation value (e.g., in time-reversal symmetric setups with real initial data).

3. Thermal and Solvable Limits:

   • In a Schmidt-diagonal model, the response is resummed to all orders using a complex tilted distribution.
   • Thermal pseudo entropy is recovered as a special case (thermofield double states), where the imaginary response is proportional to the thermal energy variance, consistent with previous literature.

4. Many-Body and Critical Behavior:

   • Numerical results for the transverse-field Ising chain confirm that the imaginary response is quantitatively governed by the modular covariance.
   • The authors define a modular susceptibility $\chi_A(h) = \text{Re} \langle K_A (V - \langle V \rangle) \rangle$ (where $V = \partial_h H$). They demonstrate that this susceptibility peaks near the critical region of the Ising model, suggesting the imaginary pseudo-entropy slope probes many-body critical correlations.

5. Non-Hermitian Extension:

   • The framework is extended to non-Hermitian systems using a one-sided biorthogonal transition matrix.
   • In $PT$-unbroken regimes, the response parallels the Hermitian case with biorthogonal expectation values.
   • In $PT$-broken regimes, the analytic structure shifts from hyperbolic (thermal-like) to trigonometric (branch-sensitive), with singularities arising from exceptional points.

Significance and Claims
The paper claims to provide a microscopic characterization of an oriented entropic response at the amplitude level. The authors emphasize that the imaginary part of real-time pseudo entropy is not merely a mathematical artifact of non-Hermitian eigenvalues or branch cuts; rather, it is a physical quantity generated by the correlation between the system's dynamics ($H$) and its subsystem information structure ($K_A$).

The authors position this result as a precursor to conventional entropy production. They argue that while the total von Neumann entropy is conserved in unitary dynamics, the oriented pseudo-entropy response distinguishes forward and backward transition matrices at the level of reduced amplitudes. This quantity serves as a "phase-sensitive precursor" to probabilistic irreversibility, which only emerges after additional coarse graining or measurement protocols convert amplitudes into probabilities.

The work also suggests a potential holographic interpretation: the leading imaginary part of complex extremal areas in holographic pseudo entropy may correspond to a boundary modular covariance with the real-time Hamiltonian, offering a bridge between real-time pseudo entropy and gravitational dynamics.

Limitations and Future Directions
The paper acknowledges that the connection to ordinary entropy production requires specifying a coarse-graining map, which is not derived here. The finite-size scaling of the modular susceptibility near criticality is observed but not rigorously established due to system size limitations. The authors identify the derivation of second-order response formulae in terms of modular-flow correlators and the explicit mapping to conventional entropy production in open systems as open problems for future work.