Unitary Quantum Cellular Automata for Density Classification

TL;DR

This work asks whether a unitary, number-conserving quantum cellular automaton can solve the 1D density classification task under local rules. It uses a two-layer, partitioned PUQCA with gates $W_0$ and $W_1$ and optimizes their continuous parameters via a genetic algorithm, redefining the success criterion through measurement probabilities at a fixed time. The authors demonstrate near-perfect density classification for multiple system sizes and identify a classically simulable regime—via an SU(2) constraint—that preserves high performance while enabling efficient classical computation. The analysis connects to Dicke-state structures and fermionic mappings, offering a conceptual path toward analytic constructions and potential relevance to quantum error correction decoders. Overall, the results show that unitary QCAs can approximate or achieve DCT-like behavior under suitable probabilistic criteria, with practical implications for quantum information processing and low-latency quantum-classical interfaces.

Abstract

We investigate the density classification task (DCT) -- determining the majority bit in a one-dimensional binary lattice -- within a quantum cellular automaton (CA) framework. While there is no one-dimensional two-state, radius $r \geq 1$, deterministic CA with periodic boundary conditions that solves the DCT perfectly, we explore whether a unitary quantum model can succeed. We employ the Partitioned Unitary Quantum Cellular Automaton (PUQCA), a number-conserving model, and, via evolutionary search, find solutions to the DCT where the success condition is stipulated in terms of measurement probabilities rather than convergence to fixed-point configurations. Finally, we identify a classically simulable regime of the PUQCA in which we find rules that solve the DCT at fixed system sizes and analyze their performance.

Figures (3)

• Figure 1: Quantum-circuit representation of a one-dimensional PUQCA. Each time step is $\mathcal{E}(W_{1}, W_{0})$: first apply the two-qubit gate $W_{0}$ on even bonds, then $W_{1}$ on odd bonds. Layers are drawn as continuing above and below to indicate repetition, and periodic boundary conditions connect the first and last qubits.
• Figure 2: These plots show the result of the quantum classifier solution for 8 cells, see \ref{['tb:res1']}, with 3 excitations after 4 time steps. The $y$-axis represents the probability of measuring an excitation at each cell, indexed on the $x$-axis. In (a), the initial state is $\ket{b} = \ket{10110000}$, while in (b), the state is $\ket{b'} = T^2 \ket{b} = \ket{00101100}$, which is related to $\ket{b}$ by a two-step cyclic translation.
• Figure 3: Here are the plots for the quantum classifier with the rule given in Table \ref{['tab:fitmult']} with 14 cells and 8 particles after 7 time steps. Likewise, in the case of 8 qubits, the values in the bars are the probabilities to measure one excitation in a given cell, $x$-axis. In (a) our initial state is given by $\ket{b}=\ket{11010111100100}$ and in (b) we have the state $\ket{b'}=\ket{11010111110000}$.

Theorems & Definitions (16)

• Definition 1
• Definition 2
• Definition 3
• Lemma 4.1: Translation covariance of PUQCA-based density classification
• proof
• Theorem 4.2: Dicke state solutions to the DCT
• proof
• Lemma 6.1
• Lemma 6.2
• Lemma 6.3: Linear evolution under local quadratic Hamiltonians