Quantum computing with spin networks
A. Landahl,M. Christandl,N. Datta,A. Ekert
2004-01-01
Abstract:We present a class of spin networks that act as perfect quantum wires—quantum states transfer across these networks with unit fidelity in a time proportional to the network’s size. No couplings need to be switched on and off during this process. We prove that N-spin hypercube networks with equal-strength nearest-neighbor Heisenberg and XY couplings have a maximal perfect communication distance of 2 log 3N. We show how to engineer coupling strengths in linear Heisenberg and XY spin chains to efficiently achieve perfect state transfer over arbitrarily long distances. Finally, we show how to augment these linear spin chain networks with more complex, but static, interactions to achieve universal quantum computation. Quantum circuits are built from quantum wires and quantum gates. How these components are technologically realized depends very much on the application. For example, wires might take the form of “flying qubit” photons for long-distance quantum communication whereas they might take the form of “bucket brigade” optical lattice modulations for quantum memory. For applications in which qubits are packed at high density, it is worthwhile to investigate the extent to which direct qubit-qubit interactions can be harnessed to technologically realize quantum wires and quantum gates. A useful model to consider in this context is a spin network—a collection of N qubits with nearest-neighbor interactions describable by a graph. Recently, equal-strength Heisenberg and XY linear spin chain networks have been shown to act as quantum wires with high, but non-unit fidelity [1]. In this paper, we examine spin networks that can serve as perfect , i.e., unit fidelity, quantum wires. We also show that these perfect quantum wires can be augmented to perform perfect quantum gates. A spin network is described by a graph G with verticesV(G) representing spins and edges E(G) representing interactions between them. The adjacency matrixcorresponding to this graph is Ai j (G) := { 1 if (i, j) ∈ E(G) 0 otherwise. (1) We label the two vertices in V(G) between which we wish to consider perfect state transfer A andB. Consider spin networks whose interactions are time-independent Hamiltonians. In the subsequent, we consider specifically anXY coupling HG = 1 2 ∑ (i, j)∈E(G) [ σ i σ x j +σ y i σ y j ] , (2) whereσx i ,σ y i andσ z i are the Pauli matrices acting on the i th qubit; later we consider other interactions. Using the (nonstandard) identification between spin states and qubits |0〉 ≡ |↓〉 and|1〉 ≡ |↑〉, we denote the ground state of an XY spin network by|0〉= |0A00· · ·00B〉. Although the Hilbert space HG of theN-spin network has dimension 2 N, state transfer dynamics occurs completely within the N-dimensional subspace SG spanned by the basis vectors |n〉 with n = 1, . . . ,N, corresponding to spin configurations in which all spins are ‘down’ apart from just one spin at the vertex n which is ‘up’. This is because HG commutes with the total spin operator in the z-direction, forcing total spin in thez-direction to be conserved. When restricted to this subspace, HG|SG is represented by an N×N matrix that isidentical to the adjacency matrix A(G), Eq. (1), of the underlying graph G. Because of this, the time evolution of the network in the SG subspace is a continuous-time quantum walk on G [2]. The question we are interested in is: When will the quantum walk propagate from A to B with unit fidelity? To answer this, we need to compute the probability amplitude F(t) = 〈N |e−itHG |1〉 . (3) Perfect state transfer is obtained for times t for which |F(t)|= 1. For theXY linear spin chain, one can compute F(t) explicitly by diagonalizing the Hamiltonian or the corresponding adjacency matrix. The eigenstates and the corresponding eigenvalues are ̃ |k〉= √ 2 N+1 N ∑ n=1 sin ( πkn N+1 ) |n〉 , Ek =−2 cos kπ N+1 , (4)