Depth-reduced quantum simulation of Lindbladian dynamics via adaptive variational quantum trajectory compression

arXiv:2607.09051 · quant-ph · Submitted 2026-07-10 · 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: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Depth-reduced quantum simulation of Lindbladian dynamics via adaptive variational quantum trajectory compression".

Mira: Here's the gist of the paper. Quantum simulation of open quantum systems in the noisy intermediate-scale quantum (NISQ) era is hindered by the non-unitary nature of dissipative dynamics and the limited quantum resources available on near-term quantum processors.

Kai: In this work, we propose a resource-efficient algorithm for simulating Lindbladian dynamics on NISQ devices.

Mira: Adjoint Channel Derivation The authors first derive a compact and stable mixed-unitary adjoint channel that approximates the target dissipative dynamics and enables ancilla-free implementation through trajectory sampling.

Kai: For Pauli dissipations, they focus on deriving a more stable adjoint channel representation that eliminates the need for the additional classical post-processing used in the previous formulation.

Title and authors: Kai: So, we're looking at this paper, "Depth-reduced quantum simulation of Lindbladian dynamics via adaptive variational quantum trajectory compression." Basically, they're tackling the problem that simulating open quantum systems with noise on near-term quantum computers is really hard because the dynamics aren't unitary.

Mira: Right. The core idea here is to propose a way to simulate these dissipative dynamics efficiently using an algorithm that uses fewer resources, specifically on noisy intermediate-scale quantum devices. It aims to solve the problem of non-unitary evolution without needing extra qubits for every single trajectory, which was a big sticking point before.

Lev: From my side, I'm interested in how this maps onto actual hardware constraints. If we can simulate this without auxiliary qubits, that drastically cuts down the overhead we have to manage on superconducting circuits or whatever platform Kai is working with.

Kai: Exactly. The paper proposes a resource-efficient algorithm for simulating Lindbladian dynamics on NISQ devices by using something called an adjoint channel derivation and then combining it with adaptive trajectory compression. It's about making the simulation practical for the current hardware landscape.

Mira: And what makes this approach promising is that they derive a compact and stable mixed-unitary adjoint channel that approximates the target dissipative dynamics, which lets them do everything without needing auxiliary qubits for trajectory sampling, which is a big win over previous methods.

Lev: That stability in the derivation is key for me. When we talk about real hardware, any method that requires classical post-processing to correct deviations adds latency and complexity that just eats up our limited coherence time.

Kai: The paper focuses on Pauli dissipations, and they show that by using this adjoint channel representation, they can simulate the dynamics without those extra steps, which means we can stick to the qubits we already have.

Mira: They define the total number of Kraus operators in this new representation as being proportional to D plus one times m, where D is related to Pauli operators and m is something else. It's a compact scaling that makes sense for implementation complexity.

Lev: The scaling itself tells me how much circuit depth we are looking at. If the total number of Kraus operators scales with D plus one times m, we need to see how large m gets in practice for these open systems.

Kai: Then they introduce an adaptive variational quantum trajectory compression framework to tackle that circuit depth issue. This involves training a depth-adaptive parameterized quantum circuit to approximate repeated Trotterized Hamiltonian simulation operators.

Title and authors: Mira: The idea is that instead of running many long, complex unitary circuits for each step in the trajectory, you train shallow parameterized quantum circuits to approximate those s-step Hamiltonian evolutions, which are denoted as U of theta sub s.

Lev: Training shallow circuits to approximate longer evolution is always a trade-off. On real hardware, training time itself becomes a resource drain that we have to account for before we even start the simulation runs.

Kai: They quantify this approximation quality using phase-insensitive fidelity defined through the Hilbert-Schmidt inner product, which is a standard way to measure how close two quantum states are in terms of their overlap.

Mira: And they discuss a few ways to evaluate that training objective. One is the Hilbert-Schmidt test which needs two n qubit registers, but another approach is a basis-sampled loss function.

Lev: A basis-sampled loss function sounds more hardware friendly because it avoids needing those extra two registers for the full Hilbert space overlap calculation, which would be tough on current NISQ machines.

Kai: That's what they suggest. This basis-sampled loss function is defined by taking a loss function L B of theta equal to one minus the fidelity between the trained circuit and the true s step Hamiltonian evolution on a finite set of input states.

Mira: It’s a surrogate for training those compressed blocks, and it avoids auxiliary qubit requirements associated with Hilbert-Schmidt tests because it only checks agreement on a specific set of basis states.

Lev: Testing agreement on a finite set is definitely less demanding than trying to measure the full Hilbert space overlap, which is what I usually run into when trying to verify approximations in real experiments.

Kai: For constructing the target dynamics, they look at two strategies: direct training and an indirect iterative training strategy. The iterative strategy trains those s step operators to approximate a longer evolution by building them up from shorter blocks.

Mira: That iterative approach allows you to build longer coherent blocks progressively from shorter trained blocks, which is useful when the total simulation time you need is longer than the maximum compression length s max they can handle.

Lev: Building up incrementally sounds like it might save us a lot of computation if we can manage the training efficiently enough that we don't waste time retraining too many times.

Kai: They also mention symmetry-aware sampling as a way to improve training efficiency by restricting the sampled basis states to the physically relevant target subspace, which is particularly helpful for systems like the dissipative quantum XY model.

Title and authors: Mira: That makes sense because in those specific models, like when particle number is conserved and the initial states are in a specific sector, you don't need to explore every possible basis state randomly.

Lev: Restricting the search space sounds good for efficiency, but we have to make sure that the physically relevant subspace they restrict sampling to actually captures all the important physics of the dynamics they are trying to simulate.

Kai: So, let's look at what this means overall for simulation and resource reduction. They ran numerical simulations on typical dissipative models, like a melting domain-wall state in a one-dimensional XY chain, and they showed it works and reduces resources.

Mira: Specifically, the direct replacement strategy they tested provided the largest resource reduction while still preserving the imbalance dynamics of that system. At time t equals ten, direct replacement lowered both single qubit and two qubit gate counts by about forty three percent on average.

Lev: Forty three percent is significant when you're dealing with NISQ devices where every gate operation costs real time and energy. That reduction in gate count is what makes the difference between running a simulation for ten steps or being completely unable to run it at all.

Kai: So, combining these structured adjoint-channel representations with the variational trajectory compression framework gives them a practical route toward open system quantum simulation on near term hardware. It's a solid combination.

Mira: I think the real strength is that this framework successfully marries the physically structured adjoint channel derivation with the variational trajectory compression to provide that practical route for open system quantum simulation on near term quantum hardware.

Lev: For me, it means we have a blueprint for how to translate these theoretical concepts into something you can actually run on a physical machine right now, even if the fidelity of the approximation is still what we need to push further.

Kai: So, to wrap up on this paper about "Depth-reduced quantum simulation of Lindbladian dynamics via adaptive variational quantum trajectory compression," it successfully combines physically structured adjoint-channel representations with variational trajectory compression to provide a practical route toward opensystem quantum simulation on near term quantum hardware.

Mira: It shows how we can get away from the auxiliary qubit requirements of older methods by using this new channel structure and then compressing the necessary evolution steps adaptively.

Lev: We should keep an eye on how they extend this to more general dissipative models or higher-order channel constructions because that's where the real challenge lies for practical application.

The paper's summary: Kai: So, this paper is about taking the really messy math of open quantum systems, which is non-unitary evolution, and finding a way to simulate it efficiently on the kind of hardware we have right now.

Mira: Right. They're basically tackling how to run simulations for things like noise and dissipation without needing a massive number of extra qubits just to track every possible path the system could take.

Lev: I mean, the big deal here is they found a way to do this using something called an adjoint channel derivation, which gives them a compact, stable approximation of the messy dynamics.

Kai: That's right. They build this mixed-unitary approximation that lets them use trajectory sampling without needing those extra classical post-processing steps that used to be required.

Mira: It makes the implementation cleaner because they get away with ancilla-free simulation, which is a huge win for NISQ devices where qubit count is always tight.

Lev: But then you have this problem—the circuit depth gets insane when you try to implement all those sampled trajectories directly.

Kai: That’s where the second part of their work comes in, and it’s called adaptive variational quantum trajectory compression. They train a shallow parameterized quantum circuit to approximate those long, repeated Hamiltonian evolution operators.

Mira: So instead of running a really deep circuit repeatedly for each step in the trajectory, you use this shallower trained circuit to stand in for the complex evolution.

Lev: And they show how they evaluate how good that approximation is using two different methods. One is a Hilbert-Schmidt test that requires two extra qubit registers, but the other one is a basis-sampled loss function.

Kai: That basis-sampled loss function sounds much more hardware friendly because it only checks agreement on a specific set of input states, avoiding those extra registers we don't want to use.

Mira: It’s basically a shortcut for training those compression blocks, letting you test the approximation without needing two extra registers for the full Hilbert space overlap.

Lev: And they also have this iterative training strategy where they build up longer evolution blocks from shorter ones, which is important when your total simulation time is longer than what one block can handle.

Kai: They even suggest using symmetry-aware sampling to narrow down the basis states they test, which makes the training much faster for systems like those in the dissipative quantum XY chain.

Mira: The result they show on their physical model—the melting domain-wall state in a one-dimensional XY chain—is that this direct replacement strategy cuts single qubit and two qubit gate counts by about forty three percent at time ten.

Lev: Forty three percent is substantial when you're counting every gate operation that costs time and energy on a real machine, so that kind of reduction is definitely worth looking at.

Kai: So, the main point here is combining these physically structured adjoint channels with the adaptive compression framework gives a practical path for simulating open systems on near-term quantum hardware.

Mira: It moves us closer to being able to run simulations that look more like real, noisy physical processes instead of just abstract unitary ones.

Lev: I mean, this is a blueprint for how we can translate these theoretical ideas into something you can actually test and run on a physical machine right now.

Kai: The next thing we should look at is how they extend this to more general models, like higher-order channel constructions, because that's where the real challenge for practical applications lies.

The paper's improvements: Kai: So, we're talking about the specific tweaks they suggest to make this whole simulation method even better than it is right now.

Mira: They focused on improving the training part of that variational compression framework, specifically how you evaluate that approximation quality.

Lev: They looked at two ways to test the accuracy of those compressed evolution blocks: the Hilbert-Schmidt test and a basis-sampled loss function.

Kai: The key improvement they highlight is ditching the Hilbert-Schmidt test because it needs those extra two qubit registers for testing, which we don't want.

Mira: They argue that the basis-sampled loss function is much better because it lets you test agreement on a finite set of input states, avoiding that extra resource drain.

Lev: That makes sense from a hardware standpoint; fewer resources needed for verification means less overhead when you actually run the simulation on superconducting qubits.

Kai: Plus, they discussed an iterative training strategy where you build up longer evolution blocks gradually from shorter ones, which helps if your total simulation time is longer than what one block can handle.

Mira: That builds a kind of hierarchy for the training process, letting you tackle long times without needing one single massive trained circuit upfront.

Lev: And they brought in symmetry-aware sampling as another way to make training more efficient, restricting the sampled states to the physically relevant subspace for specific models.

Kai: That means when you're simulating something like a particle number conserving system, you don't waste time exploring states that aren't actually part of the physics.

Mira: It’s about making sure your training time is used efficiently by not wasting effort on irrelevant quantum states.

Lev: I wonder how much training time it takes for this iterative approach to converge before we get a useful result on real hardware, that's the next question for me.

Conclusion: Kai: So we're wrapping up this talk on "Depth-reduced quantum simulation of Lindbladian dynamics via adaptive variational quantum trajectory compression." Basically, this paper shows how we can simulate noisy, open quantum systems using fewer qubits and less circuit depth than before.

Mira: It really proves that combining the adjoint channel structure with the adaptive compression framework gives us a practical toolkit for NISQ devices.

Lev: I'm glad they showed how to reduce those gate counts by about forty three percent on their physical model; that kind of resource saving is what we need when building things that actually work.

Kai: That reduction in gates means we can run longer simulations, which is crucial for getting meaningful results on noisy hardware.

Mira: The main implication is that we don't have to be limited to very short simulation times if we use this adaptive compression method correctly.

Lev: It’s a practical route, but the caveat I always bring up is that the fidelity of that approximation still depends on how well you train the variational circuit, which takes time itself.

Kai: Exactly, so training efficiency remains a key challenge for applying this to more complex physical problems.

Mira: Moving forward, we need to see how they extend these ideas to more general dissipative models or higher-order channel constructions because that’s where the real application potential is.

Lev: I'm also curious if they can show how this framework handles systems that have even more complex noise structures than the simple Pauli dissipations they used in their test case.

Institute of Artificial Intelligence, Hefei Comprehensive National Science Center · Laboratory of Quantum Information, University of Science and Technology of China · Anhui Province Key Laboratory of Quantum Network · Origin Quantum Computing Technology (Hefei) Company

quant-ph

Submitted: 2026-07-10

Updated: 2026-11-06

Comments: 18 pages, 17 figures

License: http://creativecommons.org/publicdomain/zero/1.0/

Importance score: 89/100

The gist: Quantum simulation of open quantum systems in the noisy intermediate-scale quantum (NISQ) era is hindered by the non-unitary nature of dissipative dynamics and the limited quantum resources available…

Key concepts

Adjoint Channel Derivation
This technique creates a compact and stable mixed-unitary approximation of the target dissipative dynamics. It is more robust than previous methods because it eliminates the need for classical correction steps, enabling ancilla-free simulation using trajectory sampling.
Adaptive Trajectory Compression
This framework trains shallow parameterized quantum circuits (PQCs) to approximate repeated Trotterized Hamiltonian evolution operators. By training these blocks adaptively, the overall circuit depth needed to simulate long trajectories is drastically reduced.
Basis-Sampled Loss Function
Instead of complex tests requiring many qubits, this loss function measures agreement between the PQC and the target unitary on a finite set of input states. This provides a hardware-friendly way to train the compressed evolution blocks without needing auxiliary registers.

Terminology

Summary

Quantum simulation of open quantum systems in the noisy intermediate-scale quantum (NISQ) era is hindered by the non-unitary nature of dissipative dynamics and the limited quantum resources available on near-term quantum processors. p. 0

Where the paper says it

Simulation of Lindbladian dynamics via adaptive variational quantum trajectory compression Huan-Yu Liu,1, ∗ Cheng Xue,1 Yun-Jie Wang,2, 3 Xi-Ning Zhuang,4 Chao Wang,4 Yu-Chun Wu,2, 3, 1, † Zhao-Yun Chen,1, ‡ and Guo-Ping Guo2, 3, 1, 4 1 Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, Hefei, Anhui, 230088, China 2Laboratory of Quantum Information, University of Science and Technology of China, Hefei, Anhui, 230026, China 3Anhui Province Key Laboratory of Quantum Network, University of Science and Technology of China, Hefei 230026, China 4Origin Quantum Computing Technology (Hefei) Co. , Hefei, Anhui, 230026, China (Dated: July 13, 2026) Quantum simulation of open quantum systems in the noisy intermediate-scale quantum (NISQ) era is hindered by the non-unitary nature of dissipative dynamics and the limited quantum resources available on near-term quantum processors.

In this work, we propose a resource-efficient algorithm for simulating Lindbladian dynamics on NISQ devices. p. 0
Where the paper says it

, Hefei, Anhui, 230026, China (Dated: July 13, 2026) Quantum simulation of open quantum systems in the noisy intermediate-scale quantum (NISQ) era is hindered by the non-unitary nature of dissipative dynamics and the limited quantum resources available on near-term quantum processors. In this work, we propose a resource-efficient algorithm for simulating Lindbladian dynamics on NISQ devices.

Adjoint Channel Derivation

The authors first derive a compact and stable mixed-unitary adjoint channel that approximates the target dissipative dynamics and enables ancilla-free implementation through trajectory sampling. For Pauli dissipations, they focus on deriving a more stable adjoint channel representation that eliminates the need for the additional classical post-processing used in the previous formulation. p. 1

Where the paper says it

However, hardware noise and finite-sampling errors may be amplified during the post-processing stage, which can degrade the accuracy and stability of the method. To address this issue, we derive a more stable adjoint channel representation that eliminates the need for the additional classical post-processing used in the previous formulation.

This channel provides a mixed-unitary approximation to the target dissipative dynamics and can be simulated without auxiliary qubits. p. 2
Where the paper says it

1(b), A (𝜌) = ∑︁ 𝐷 𝑘=0 𝑝𝑘𝑈𝑘 𝜌𝑈† 𝑘 , (6) where 𝑝0 = 𝑍 𝑍 + Γ𝛿𝑡 , 𝑝𝑘>0 = 𝛾𝑘 𝛿𝑡 𝑍 + Γ𝛿𝑡 , 𝑈0 = 𝑒 −𝑖𝐻Δ𝑡 , 𝑈𝑘>0 = 𝑃𝑘 . (7) This channel provides a mixed-unitary approximation to the target dissipative dynamics and can be simulated without auxiliary qubits.

The multi-step composite channel can be written as A m(rho) = ∑ k p k U k rho U† k, where the total number of Kraus operators in this representation scales as (D + 1) m. p. 2
Where the paper says it

, 𝑘𝑚), and 𝑝𝒌 = Ö𝑚 𝑖=1 𝑝𝑘𝑖 , 𝑈𝒌 = 𝑈𝑘𝑚𝑈𝑘𝑚−1 · · ·𝑈𝑘1 . (9) The total number of Kraus operators in this representation scales as (𝐷 + 1) 𝑚, making a direct summation over all trajectories computationally intractable for large 𝑚.

Adaptive Trajectory Compression Framework

To reduce the circuit depth required for implementing the sampled trajectories, an adaptive variational quantum trajectory compression framework is introduced. This framework involves training a depth-adaptive parameterized quantum circuit to approximate repeated Trotterized Hamiltonian simulation operators. p. 0

Where the paper says it

In this framework, a depth-adaptive parameterized quantum circuit is trained to approximate repeated Trotterized Hamiltonian simulation operators, which are then used to replace repeated unitary segments appearing in the sampled trajectories. Importantly, the training procedure can also be performed without auxiliary qubits.

The key idea is to train shallow PQCs to approximate the s-step Hamiltonian evolution, U(θs) ≈ (U0) s = e(-iHΔt s). p. 2
Where the paper says it

The key idea is to train shallow PQCs to approximate repeated Hamiltonian-evolution blocks and then insert the trained circuits into sampled trajectories. Specifically, for each 𝑠 ∈ {1, 2, · · · , 𝑠max}, we train a PQC 𝑈(𝜽𝑠) to approximate the 𝑠-step Hamiltonian evolution, 𝑈(𝜽𝑠) ≈ (𝑈0) 𝑠 = 𝑒 −𝑖𝐻Δ𝑡 𝑠 .

This approximation is quantified by the phase-insensitive fidelity defined through the Hilbert-Schmidt inner product (HSP). p. 2
Where the paper says it

Specifically, for each 𝑠 ∈ {1, 2, · · · , 𝑠max}, we train a PQC 𝑈(𝜽𝑠) to approximate the 𝑠-step Hamiltonian evolution, 𝑈(𝜽𝑠) ≈ (𝑈0) 𝑠 = 𝑒 −𝑖𝐻Δ𝑡 𝑠 . (11) Since a global phase does not affect expectation values of the form Tr[𝑈 𝜌𝑈†𝑂], the approximation between two unitary operators can be quantified by the phase-insensitive fidelity defined through the Hilbert-Schmidt inner product (HSP), 𝐹HS(𝑈, 𝑉) = 1 𝑑 ⟨𝑈, 𝑉⟩HS 2 = 1 𝑑 Tr[𝑈 †𝑉] 2 , (12) where 𝑑 = 2 𝑛 is the Hilbert-space dimension.

Training Objectives and Strategies

The authors discuss two possible ways to evaluate this training objective:

  1. The Hilbert-Schmidt test (HST) which requires two n-qubit registers. p. 3

    Where the paper says it

    The first is to use the Hilbert-Schmidt test (HST) to estimate the HSP and thereby the loss function. This method requires two 𝑛-qubit registers; see Appendix A for details.

  2. A basis-sampled loss function L B(θ) = 1 − F B(U(θ), (U0) s), where the agreement on each basis state is Px i = ⟨x iU(θ)†((U0) s)x i⟩/2. p. 3

    Where the paper says it

    Given a basis-state set B = {|𝑥𝑖⟩}𝐵 𝑖=1 , the agreement on each basis state is 𝑃𝑥𝑖 = ⟨𝑥𝑖 |𝑈(𝜽) † (𝑈0) 𝑠 |𝑥𝑖⟩ 2 , (14) which can be measured by the quantum circuit shown in Fig. Then the basis-sampled fidelity becomes 𝐹B (𝑈(𝜽), (𝑈0) 𝑠 ) = 1 𝐵 ∑︁ 𝐵 𝑖=1 𝑃𝑥𝑖 .

    The authors suggest this basis-sampled loss provides a hardware-friendly surrogate for training the compressed Hamiltonian-evolution blocks. p. 3
    Where the paper says it

    (15) The corresponding loss function is 𝐿B (𝜽) = 1 − 𝐹B (𝑈(𝜽), (𝑈0) 𝑠 ) . (16) This basis-sampled loss provides a hardware-friendly surrogate for training the compressed Hamiltonian-evolution blocks.

For target construction, two strategies are considered: direct training and an indirect iterative training strategy. p. 4

Where the paper says it

The price one pays is that approximation errors may accumulate through the iterative construction. Thus, the choice between direct and indirect training should be made according to the target simulation time, the available circuit depth, and the desired accuracy.

The iterative strategy trains U(θs) to approximate (U0) s = (U0)(s-1)U0. p. 4
Where the paper says it

To avoid using a deep target circuit during training, we further introduce an indirect iterative training strategy. We first train a shallow circuit 𝑈(𝜽1) to approximate a single nojump step 𝑈(𝜽1) ≈ 𝑈0.

Simulation and Resource Reduction

Numerical simulations on typical dissipative models demonstrate the correctness of the proposed method and its capability for quantum-resource reduction. The paper simulates a physically motivated open-system transport problem: the melting of a domain-wall state in a dissipative one-dimensional XY chain. p. 4

Where the paper says it

Overall, the proposed adaptive variational compression framework converts frequently occurring repeated Hamiltonian-evolution blocks into shallow trainable circuits, thereby reducing the depth of sampled trajectories while preserving the ancilla-free structure of the adjoint-channel simulation. SIMULATING THE DISSIPATIVE QUANTUM 𝑋𝑌 MODEL We next apply the adjoint channel and adaptive trajectorycompression strategy to simulate a physically motivated opensystem transport problem: the melting of a domain-wall state in a dissipative one-dimensional 𝑋𝑌 chain.

The direct replacement strategy provides the largest resource reduction while preserving the imbalance dynamics. p. 6
Where the paper says it

The direct library gives the largest resource reduction while preserving the imbalance dynamics. At 𝑡 = 10, direct replacement lowers both the single-qubit and two-qubit gate counts by about 43% on average.

At t = 10, direct replacement lowers both the single-qubit and two-qubit gate counts by about 43% on average. p. 6
Where the paper says it

The direct library gives the largest resource reduction while preserving the imbalance dynamics. At 𝑡 = 10, direct replacement lowers both the single-qubit and two-qubit gate counts by about 43% on average.

Conclusion and Outlook

The proposed framework successfully combines physically structured adjoint-channel representations with variational trajectory compression to provide a practical route toward opensystem quantum simulation on near-term quantum hardware. The results suggest that combining physically structured adjoint-channel representations with variational trajectory compression provides a practical route toward opensystem quantum simulation on near-term quantum hardware. p. 7

Where the paper says it

Understanding and optimizing this balance between algorithmic error and hardware-induced error is an important direction for future work. Overall, our results suggest that combining physically structured adjoint-channel representations with variational trajectory compression provides a practical route toward opensystem quantum simulation on near-term quantum hardware.

This framework may be further extended to more general dissipative models, higher-order channel constructions, and application-oriented simulations of realistic many-body open quantum systems. p. 7
Where the paper says it

Overall, our results suggest that combining physically structured adjoint-channel representations with variational trajectory compression provides a practical route toward opensystem quantum simulation on near-term quantum hardware. This framework may be further extended to more general dissipative models, higher-order channel constructions, and application-oriented simulations of realistic many-body open quantum systems.

The authors conclude that the direct replacement strategy provides the best balance between accuracy and compression in this test. p. 6
Where the paper says it

(d) Two-qubit gate counts of individual trajectories at 𝑡 = 10. rect strategy provides the best balance between accuracy and compression in this test.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Adaptive variational quantum trajectory compression framework allows a depth-adaptive parameterized quantum circuit to be trained to approximate repeated Trotterized Hamiltonian simulation operators, which are then used to replace repeated unitary segments appearing in the sampled trajectories. This reduces the circuit depth while preserving the sampled-channel structure and can be performed without auxiliary qubits.

  2. The adjoint channel derivation provides a compact and stable mixed-unitary approximation of dissipative dynamics, enabling ancilla-free implementation through trajectory sampling, which is more stable than previous formulations that required classical post-processing to correct deviation.

  3. The basis-sampled loss function provides a hardware-friendly surrogate for training the compressed Hamiltonian evolution blocks by testing agreement between the PQC and the target unitary on a finite set of input states, avoiding auxiliary qubit requirements associated with Hilbert-Schmidt tests.

  4. The iterative training strategy for compressing Hamiltonian evolution blocks allows building up longer coherent blocks progressively from shorter trained blocks, mimicking a procedure where longer coherent blocks are built progressively from shorter trained blocks. This is useful when the target simulation time requires more than the maximum compression length smax.

  5. Symmetry-aware sampling can improve training efficiency by restricting the sampled basis states to the physically relevant target subspace, which is beneficial in simulations like the dissipative quantum XY model where particle number is conserved and initial states lie in a specific sector.

Abstract

Quantum simulation of open quantum systems in the noisy intermediate-scale quantum (NISQ) era is hindered by the non-unitary nature of dissipative dynamics and the limited quantum resources available on near-term quantum processors, namely restricted qubit counts and noise-limited circuit depths. In this work, we propose an efficient quantum algorithm for simulating Lindbladian dynamics on NISQ devices. For open quantum systems with Pauli dissipations, we first derive a compact and stable mixed-unitary adjoint channel that approximates the target dissipative dynamics and enables ancilla-free implementation through trajectory sampling. To further reduce the circuit depth required for implementing the sampled trajectories, we introduce an adaptive variational quantum trajectory compression framework. In this framework, a depth-adaptive parameterized quantum circuit is trained to approximate repeated Hamiltonian simulation operators, which are then used to replace repeated unitary segments appearing in the sampled trajectories. Importantly, the training procedure can also be performed without auxiliary qubits. Numerical simulations of the dissipative quantum XY model and the interacting disordered XXZ chain demonstrate the accuracy and reduced execution gate counts of the proposed algorithm. Our results provide a practical route toward ancilla-free and depth-reduced simulation of open quantum systems on near-term quantum hardware.

Sources

Related papers