Order parameters and symmetry-breaking patterns

For Ising spins $s_i=\pm1$, uniform magnetization per site averages the spin signs. It detects alignment but misses a perfect alternating antiferromagnet, whose positive and negative contributions cancel. A staggered average includes a sign assigned to the two sublattices and then detects alternating order. The correct observable depends on the ordering wavevector and lattice structure. A vanishing poorly chosen average is not evidence that the sample is disordered.

An order parameter may be scalar, vector, complex, or tensor valued. Nematic orientational order, for example, identifies a direction without distinguishing its two ends, so a simple polar vector is inappropriate. The variable must transform correctly under the physical symmetries before one writes a free-energy expansion for it. Symmetry breaking is powerful but does not classify every phase: some topological distinctions require other observables rather than a local symmetry-breaking expectation value.

Call the zero-field symmetry group $G$ and the subgroup preserving a chosen ordered state $H$. In an Ising ferromagnet, spin reversal exchanges the two magnetized states, so a selected state preserves only the identity within that reversal group. The identity subgroup is not the empty set. For an isotropic n-component vector order parameter, selecting one direction reduces O(n) to O(n−1), with a sphere of equivalent directions.

This state manifold describes static degeneracy. Slow spatial changes of direction can cost only gradient energy when a global continuous symmetry is broken, producing soft fluctuations. It does not guarantee an independent propagating mode for every broken generator in every dynamical system. Spatial dimension, interaction range, gauge constraints, and conservation laws matter. In particular, short-range two-dimensional systems at positive temperature need special care before concluding that a continuous orientation acquires true long-range order.

For finitely many Ising spins at positive temperature, the partition function is a finite sum of positive Boltzmann weights and is analytic for real finite couplings and field. At exactly zero field with symmetric boundaries, every configuration and its reversed partner have equal weight. Their magnetizations cancel in the ensemble mean. A sharply bimodal distribution or large second moment can still reveal strong collective organization in that finite sample.

The thermodynamic limit can make reversal between macroscopically distinct states effectively singular and allows the ordering of limits in the spontaneous-magnetization definition. Removing the field first preserves the symmetric mean at every finite size. Selecting one branch first yields a symmetry-broken phase when the model supports it. Long observation times and preparation also influence what a finite laboratory sample appears to do, but those practical effects should be distinguished from a mathematical nonanalyticity of its infinite-system free energy.