Two-dimensional tensor networks could speed up complex quantum calculations now

Difficulties in constructing accurate shallow circuits hinder preparing complex quantum states for computation. A method using projected entangled pair states (PEPS), an extension of tensor networks, now prepares these initial states more effectively than previous approaches based on Matrix Product States which struggled in higher dimensions. Projected entangled pair states (PEPS) are two-dimensional tensor networks that offer advantages over earlier one-dimensional methods; those methods had difficulty representing entanglement across larger areas.

The team also developed a more efficient way to calculate key values within this PEPS system, making it better suited for use on existing quantum computers. Consider setting up the starting configuration of qubits like arranging pieces at the beginning of a chess game, more intricate arrangements require increasingly sophisticated planning.

Similarly, Matrix Product States (MPS) and PEPS can be viewed as blueprints for building structures with LEGOs: MPS allows simpler, linear designs while PEPS enables far more elaborate two-dimensional constructions. The team devised a way to efficiently calculate key values within this system, making it better suited for use on existing hardware, and combined this with gradient optimisation, an iterative process akin to adjusting knobs on a radio until you get the clearest signal.

Gradient Optimisation Enables Fivefold Increase in Simulated Quantum Entanglement Scale

David Awschalom and colleagues achieved a fivefold increase in quantum state size prepared through simulation compared to prior PEPS methods, accurately simulating entanglement beyond approximately 10 qubits was previously computationally impossible. The advancement combines local disentanglement with gradient optimisation applied to two-dimensional projected entangled pair states, or PEPS, alongside an efficient procedure for calculating gradients within these complex systems. Approximating area law entanglement states is now more effective than using one-dimensional Matrix Product States, which struggle with higher dimensional representations.

Finer control over circuit construction is enabled by iteratively refining initial state preparation through optimisation akin to tuning a radio receiver until achieving optimal signal clarity. Now extending beyond the previous limit of approximately 10 qubits, the scale of quantum states prepared via simulation has significantly increased.

Superorthogonal canonical form PEPS factorises node tensors and uses diagonal matrices related to entanglement as approximations when correlations are tree-like; this allows scientists efficiently calculate how changes in circuit parameters affect the overall system for precise optimisation. By optimising parametrised circuits layer by layer using algorithms like L-BFGS, researchers minimise bond Rényi entropy, effectively ‘tuning’ the quantum state preparation from a given PEPS configuration.

Loop size impacts optimisation efficiency of entangled state preparation

While preparing complex quantum states with two-dimensional projected entangled pair states offers improvements over one-dimensional methods, scaling remains an open question; numerical simulations suggest promise but haven’t yet bridged the gap to practical hardware implementation. Certain lattice structures featuring smaller loops present particular computational hurdles due to increased resource demands during optimisation, as acknowledged by the authors. This limitation highlights that efficiently representing entanglement within specific geometries continues to challenge even advanced tensor network techniques and suggests alternative approaches might prove necessary for tackling particularly intricate systems.

Despite these challenges, some lattice designs demanding significant resources, this work represents a vital step forward in preparing complex quantum states. It could unlock more powerful simulations of currently intractable systems, though further refinements are needed to tackle all geometries efficiently.

Future development will focus on unlocking previously inaccessible simulations; scientists have created an improved method for initial state preparation by combining two-dimensional projected entangled pair states (blueprints allowing complex arrangements of qubits) with iterative refinement via gradient optimisation. This approach overcomes limitations inherent in previous one-dimensional techniques which struggled to represent entanglement across larger areas and opens questions regarding optimal lattice structures for future development, the team is now investigating how different network topologies affect computational cost and scalability.

This research demonstrated a quantum state preparation method using two-dimensional projected entangled pair states combined with gradient optimization. The technique addresses challenges faced by earlier methods that used one-dimensional approaches when preparing complex, highly connected quantum states. By optimizing these states based on their geometry, researchers achieved more accurate approximations of initial conditions for quantum computation. The authors are currently exploring how varying network arrangements impact the efficiency and scale of this process.

👉 More information
🗞 Efficient computation of PEPS gradients for optimized quantum state preparation on 2D topologies
✍️ Illia Lukin, Vladyslav Bohun and Maciej Koch-Janusz (Affiliation: Haiqu)
🧠 ArXiv: https://arxiv.org/abs/2611.17158

Stay current

See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.

Avatar of Muhammad Rohail T.

Latest Posts by Muhammad Rohail T.: