Random-Bond AKLT States

• Random-bond AKLT states are a class of quantum spin states constructed by introducing randomness into local bond configurations, resulting in statistically robust and topologically nontrivial properties.
• Their formulation employs ergodic probability measures and tensor-network techniques to model exponential correlation decay and gapless spectral features in disordered spin chains.
• These states serve as universal resources for measurement-based quantum computation, maintaining operational power despite disorder and the absence of a uniform spectral gap.

A random-bond AKLT (Affleck–Kennedy–Lieb–Tasaki) state generalizes the canonical AKLT construction by introducing disorder or randomness in the local bond degrees of freedom, leading to an ensemble of quantum spin states with statistical and topological properties controlled by the underlying random variables. These states naturally arise in the context of matrix product states (MPS) with random parameters, projected entangled pair states (PEPS) with randomly chosen Bell states on the virtual bonds, and as unique ground states of certain frustration-free, disordered parent Hamiltonians. Random-bond AKLT states exhibit robust features such as exponentially decaying correlations and nontrivial disorder-protected $\mathbb{Z}_2$ indices, and are universal resources for measurement-based quantum computation (MBQC) on suitable lattices (Roon et al., 9 Jul 2025, Guo et al., 6 Feb 2026). The following sections detail their mathematical construction, stochastic and tensor-network formulations, spectral properties, operational significance, and topological invariants.

1. Mathematical Formulation and Probability Structure

The formalism of random-bond AKLT states begins with a probability space $(\Omega, \mathbb{P})$, often constructed as $\Omega = K^\mathbb{Z}$, where $K$ is a compact parameter space such as $[0,\pi/4]$ and $\mathbb{P} = \bigotimes_{j\in\mathbb{Z}}\mu$ is the product measure of i.i.d. random variables $\omega = (\omega_j)_{j\in\mathbb{Z}}$ distributed according to a fixed probability law $\mu$ on $K$. The bilateral shift $\theta:\Omega \to \Omega$ defined by $(\theta\omega)_j = \omega_{j+1}$ renders the ergodic structure and underpins translation covariance.

The physical system is modeled as a spin chain with quasi-local algebra $$\mathcal{A}_\mathbb{Z} = \bigotimes_{j \in \mathbb{Z}} M_3$$, corresponding to spin-1 degrees of freedom at each site. States are weak*-continuous maps $$\omega \mapsto \psi_\omega \in S(\mathcal{A}_\mathbb{Z})$$ satisfying the covariance symmetry $\psi_\omega \circ \tau_k = \psi_{\theta^k \omega}$, with $\tau_k$ the translation automorphism by $k$-units. This construction replaces strict translation invariance with statistical invariance under the ergodic shift (Roon et al., 9 Jul 2025).

2. Tensor-Network Description and Ensemble Specification

A prototypical random-bond AKLT state is assembled in two main steps:

• Random Bell Bond Assignment: For a finite graph $G = (V, E)$, each vertex $v$ of degree $z_v$ is equipped with $z_v$ virtual qubits (spin-½), one per incident edge. For every edge $e = \{v, w\}$, a Bell state $\ket{B_e}$ is drawn independently at random from the set $$\{\ket{\Phi^+}, \ket{\Phi^-}, \ket{\Psi^+}, \ket{\Psi^-}\}$$, with weights given by a probability distribution $p_b$ over the four Bell types.
• Symmetrization (Projection): At each vertex $v$, a projection $P_v$ onto the symmetric (total spin $S = z_v/2$) subspace of its $z_v$ qubits is applied. For a configuration $\{B_e\}$ of Bell states:

$$\ket{\mathrm{AKLT}_{\rm RB}(\{B_e\})} = \left(\bigotimes_{v\in V} P_v\right) \left(\bigotimes_{e\in E} \ket{B_e}\right).$$

The PEPS ensemble consists either of mixtures (density matrices) or random pure states, depending on whether averaging is performed (Guo et al., 6 Feb 2026).

In the translation-invariant chain setting, the random parameter $\omega_j$ on bond $j$ selects the AKLT isometry $V_{\omega_j}$ or the bond Kraus operators $A_i(\omega_j)$ used in the MPS representation:

$$E_{a, \omega}(b) = \sum_{i,j = -,0,+}\langle i|a|j\rangle\, A_j(\omega)\, b\, A_i(\omega)^*.$$

The MPS on interval $[m, n]$ for disorder configuration $\omega$ is given by:

$$\psi_\omega(a_{i_m} \otimes \cdots \otimes a_{i_n}) = \frac{1}{2} \mathrm{Tr}\left[ A_{i_n}(\theta^{n-1}\omega)\cdots A_{i_m}(\theta^{m}\omega) \right].$$

This construction yields finitely correlated, ergodic MPS with random, site-dependent bond structure (Roon et al., 9 Jul 2025).

3. Spectral Properties and Parent Hamiltonians

Random-bond AKLT states $\nu_\omega$ are the unique frustration-free ground states of random, nearest-neighbor parent Hamiltonians:

$$H(\omega) = \sum_{j\in\mathbb{Z}} h_{j,j+1}(\omega), \quad h_{j,j+1}(\omega) = \text{projection onto } \left[G_{j,j+1}(\omega)^\perp\right],$$

where $G_{j,j+1}(\omega)$ is the span of products of Kraus operators $A_i(\omega_j)\otimes A_k(\omega_{j+1})$ with $i,k\in \{-,0,+\}$ (Roon et al., 9 Jul 2025).

A central feature is the characteristic bulk spectral gap closure ("gaplessness"). By constructing long rare regions with nearly uniform $\omega_j$ close to $0$, one shows that the spectrum of the GNS Hamiltonian $H_\omega$ is $\theta$-invariant and, almost surely, there is no uniform lower bound to the gap. This follows from the contradiction between observed exponential clustering in most regions and the persistence of long-range correlations in rare uniform regions, using variants of the Nachtergaele–Sims exponential clustering theorem (Roon et al., 9 Jul 2025).

4. Correlation Functions and Decay

Despite the absence of a bulk gap, random-bond AKLT states display exponentially decaying two-point correlations for generic disorder realizations. For spin observables $S_i^\alpha$ and $S_j^\beta$ ($\alpha, \beta \in \{x, y, z\}$), the connected correlation function is defined by:

$$C_{\alpha\beta}(i,j; \omega) = \nu_\omega(S_i^\alpha S_j^\beta) - \nu_\omega(S_i^\alpha)\,\nu_\omega(S_j^\beta),$$

and satisfies the exponential decay bound:

$$|C_{\alpha\beta}(i,j;\omega)| \leq \|S^\alpha\|\|S^\beta\| \cdot e^{-|j-i|/\xi(\omega)},$$

where the correlation length $\xi(\omega) = -1/\ln|\lambda_2(\omega)|$ is controlled by the second eigenvalue $\lambda_2(\omega)$ of the transfer operator $T_\omega(b) = \sum_i A_i(\omega)^* b A_i(\omega)$ (Roon et al., 9 Jul 2025).

This suggests that random disorder does not induce algebraic or power-law correlations even when the gap is closed, provided the disorder is generic and not tuned to enhance rare regions over macroscopically large scales.

5. Topological and Symmetry-Protected Invariants

Random-bond AKLT states are invariant under time-reversal symmetry, i.e., $\nu_\omega \circ \Theta = \nu_\omega$ almost surely, with $\Theta$ acting as $S_j^\alpha \mapsto -S_j^\alpha$ on spin-1 sites and conjugation by $Z = i\sigma^y$ on ancilla spaces. On the level of the MPS, this translates to Kraus operator symmetry $Z A_\pm(\omega) Z^* = -A_\mp(\omega)$ and $Z A_0(\omega) Z^* = A_0(\omega)$.

A disorder-robust $\mathbb{Z}_2$-valued topological index, the Tasaki index $\mathscr{O}(\nu_\omega)$, is defined as the asymptotic sign of the expectation of the Affleck–Lieb twist operator $T_L$ over intervals $[-L, L+1]$:

$$T_L = \bigotimes_{x=-L}^{L+1} \exp\left(-2\pi i\,\frac{x+L}{2L+1}\, S^z_x\right), \quad \mathscr{O}(\nu_\omega) = \lim_{L\to\infty}\mathrm{sgn}\left(\nu_\omega(T_L)\right).$$

For the random-bond ensemble, almost surely $\mathscr{O}(\nu_\omega) = -1$, confirming nontrivial topological phase and robustness to disorder (Roon et al., 9 Jul 2025).

6. Measurement-Based Quantum Computational Universality

Random-bond AKLT states constructed on trivalent (or more generally, MBQC-universal) lattices retain their quantum computational power. Upon performing local POVMs at each vertex (e.g., $\{F_{v,x}, F_{v,y}, F_{v,z}\}$ in the spin-3/2 case), the post-measurement state is locally equivalent to an encoded random graph state on a derived effective graph $G(\alpha)$. The protocol:

1. Fuse all edges incident on vertices with identical POVM outcomes into multi-edges.
2. Delete edges with even multiplicity (modulo-2 rule).
3. The resulting simple graph $G(\alpha)$ determines the encoded graph state structure (Guo et al., 6 Feb 2026).

Edge occupation probability $p_\text{eff}$ in the random-bond case differs negligibly ($O(10^{-3})$) from the uniform singlet-bond AKLT ensemble, with frustration effects exponentially suppressed in the cycle length. For the honeycomb lattice, $p_\text{eff} \approx 2/3$ and the percolation threshold is $p_c \approx 0.652$, ensuring a spanning cluster and hence MBQC universality in the thermodynamic limit (Guo et al., 6 Feb 2026).

This establishes:

• Universality of the random-bond AKLT ensemble for MBQC with probability approaching 1 in the large-system limit.
• Equivalence of randomly decorated AKLT states—after fusion measurement (which converts decorated states to random-bond AKLT states)—in computational power to the original singlet-bond AKLT construction.

7. Context, Variants, and Generalizations

Random-bond AKLT states encompass both 1D chains and higher-dimensional settings and include as special cases both translation-invariant and decorated AKLT models. In the context of topological quantum phases, such states provide examples of SPT phases robust to translation symmetry breaking and local random disorder. The techniques developed for random-bond AKLT constructions form the foundation for generalized studies of disordered matrix product and PEPS states on arbitrary graphs, including their ergodic, spectral, and computational properties (Roon et al., 9 Jul 2025, Guo et al., 6 Feb 2026).

A plausible implication is that the random-bond formalism provides a bridge between condensed matter models with topological order and practical quantum architectures where stochasticity in state preparation or fabrication is unavoidable, while maintaining essential quantum computational and topological functionalities.