Every dot is a spin, up or down. Every line is a rule. Some bonds want their two spins to match, some want them to differ. Flip spins to satisfy as many rules as you can and reach the ground state, the lowest energy arrangement. It starts easy. Then you hit a shape where you cannot please everyone at once. That feeling is frustration.
The board is a Hamiltonian. It is a stack of small local rules, and each one scores two spins. The lowest total score is the ground state, and finding it is the whole game.
An odd loop of opposite wanting bonds can never be fully happy. Some bond always loses. The rules are simple, but there is no perfect answer. That is why the big lattices get hard.
Let the spins go quantum, so they can be superposed and entangled. Now even estimating the ground energy is QMA complete, the quantum cousin of NP complete.
Does even a rough estimate stay this hard? That is the quantum PCP conjecture. Nobody has proved it, and that is why this corner stays alive.
Push the disorder and cool it down. The low energy states reorganise in two famous ways. Drag the sliders.
Below the clustering transition, the set of low energy states breaks into many clusters. Each one is tiny and sits far from the rest. Local moves get trapped inside a single cluster, so simple dynamics stall and the system stops exploring. You see this in real glasses and in random problems like SAT and graph colouring.
The walker only makes small local moves, so it stays inside the cluster it started in. The barriers between clusters are too high to cross, and the ground state lives in a different cluster (in teal), so the walker never reaches it. Random restarts drop it into a random cluster, but with this many the odds of hitting the right one are about 1 in 6. Push the slider up. When the clusters grow exponentially, restarting stops working.
The Gibbs measure splits into many pure states, and they organise into a hierarchy. On the left,
that hierarchy is an ultrametric tree. On the right, the overlap q between two
independent replicas stops being one fixed number. Its distribution P(q) widens from a single
spike at qEA into a whole spread of values down to small q. That
function is Parisi's order parameter.
Hunting for ground states is not a toy. The same math shows up everywhere.
Routing, scheduling, and MAX CUT all map straight onto spins like these. Quantum annealers such as D Wave hunt for exactly these minima.
The resting state of a real magnet or material is the ground state of its Hamiltonian. Predict one and you predict the other.
Hopfield networks store memories as energy minima. Recalling one is just rolling downhill into the nearest valley.
Ground state energies drive chemistry and materials design. That is one of the main reasons to build a quantum computer at all.
The anneal button heats the system so spins can jump out of bad spots, then slowly cools it until everything settles low. That is simulated annealing. Real quantum annealers do the same thing, but they tunnel through barriers instead of hopping over them.