Ising Model
The Ising model describes binary spins $z_i=\pm1$ with energy $E(z)=\sum J_{ij}z_iz_j+\sum h_iz_i$; its quantum version is a Hamiltonian diagonal in $Z$, with added $X$ terms in the transverse-field model.
The Ising model is a model from statistical physics for the magnetism of materials: each site $i$ has a spin $z_i\in\{+1,-1\}$, and the energy is $$E(z)=\sum_{i0$ is an antiferromagnetic interaction that favours opposite spins, $J_{ij}<0$ is a ferromagnetic interaction that favours aligned spins, and $h_i$ is a longitudinal field that favours one direction. Books differ in the sign of the terms, so the convention should always be stated.
The quantum version replaces $z_i$ with the operator $Z_i$: $H=\sum J_{ij}Z_iZ_j+\sum h_iZ_i$. This Hamiltonian is diagonal in the computational basis, so its eigenstates are the bit strings and its eigenvalues are $E(z)$, with $Z|0\rangle=+|0\rangle$, i.e. $|0\rangle\leftrightarrow z=+1$. So finding the ground state is a purely classical optimisation problem, but it is NP-hard for general graphs, which is what makes it a common language for optimisation problems.
The transverse-field Ising model adds terms that do not commute with $Z$: $$H=-J\sum_{\langle i,j\rangle}Z_iZ_j-h\sum_iX_i.$$ The Hamiltonian is now non-diagonal, its ground state is in general an entangled superposition, and it becomes a genuinely quantum model in which quantum phase transitions are studied. This model is the basis of quantum annealing, where the transverse term is used to create quantum fluctuations, and it is also a common test problem for VQE.
The link to optimisation: Max-Cut corresponds to an antiferromagnetic model without a field: minimising $\sum w_{ij}z_iz_j$ is equivalent to maximising the number of edges whose ends are opposite. Other problems need field terms and constraints imposed with penalty terms. In the lab, the "ising" mode shows the energy of every bit string for a small graph.
An important note: on graphs with "frustration", such as the antiferromagnetic triangle, not all interactions can be satisfied at once, so the ground state becomes degenerate, which is one source of difficulty in these problems.
Before using any Ising formulation in the lab or in another reference, it is advisable to check the convention by computing the energy of two simple strings by hand and comparing it with what is shown. The size of the solution space is easy to compute: $n$ spins mean $2^n$ configurations, so exhaustive enumeration quickly becomes impractical; $n=50$ gives more than $10^{15}$ configurations.
Mathematical formula
Worked example
- An antiferromagnetic triangle: $H=Z_0Z_1+Z_1Z_2+Z_0Z_2$. For 000: $E=3$. For any string with one differing bit, such as 001: $E=-1-1+1=-1$. The three terms cannot all be $-1$ at once (frustration), so the minimum energy is $-1$, for six states.
- The transverse-field model for two qubits: $H=-Z_0Z_1-h(X_0+X_1)$ with $h=0.5$. Diagonalising the $4\times4$ matrix gives the eigenvalues $\{-\sqrt2,-1,1,\sqrt2\}$, so the ground-state energy is $-\sqrt{1+4h^2}=-\sqrt2\approx-1.414$, lower than the $-1$ given by the best computational basis state (00 or 11): the transverse term lowers the energy through superposition.
Test your understanding
The quiz needs JavaScript.
Related topics
References
- Andrew Lucas (2014). Ising formulations of many NP problems. Frontiers in Physics 2:5 doi:10.3389/fphy.2014.00005
- Tadashi Kadowaki; Hidetoshi Nishimori (1998). Quantum annealing in the transverse Ising model. Phys. Rev. E 58(5):5355–5363 doi:10.1103/PhysRevE.58.5355
- Michael A. Nielsen; Isaac L. Chuang (2010). Quantum Computation and Quantum Information (10th Anniversary Edition). Cambridge University Press doi:10.1017/CBO9780511976667The standard reference for quantum computing; the Crossref record gives 2012 for the Cambridge Core online edition.
