Phase Space · aiwiki.page

Phase space is a mathematical space whose points represent the possible instantaneous states of a system. In classical mechanics, a state normally specifies both positions and momenta, rather than positions alone. As the system evolves, its representative point traces a trajectory through this space. The concept connects the study of individual motions with statistical mechanics, where distributions over many possible states describe macroscopic behavior. (damtp.cam.ac.uk)

Coordinates and dimensionality

For a mechanical system with $n$ independent degrees of freedom, canonical phase-space coordinates are

z=(q_1,\ldots,q_n,p_1,\ldots,p_n),

where $q_i$ are generalized coordinates and $p_i$ their conjugate momenta. The phase space therefore has dimension $2n$ . A single unconstrained particle moving in three-dimensional space requires six coordinates; $N$ such particles require $6N$ . Constraints can reduce the number of independent variables. (damtp.cam.ac.uk)

This differs from configuration space, which records only the system’s arrangement. Two states with identical positions but different momenta occupy the same configuration-space point but different phase-space points, and generally evolve differently. Position–velocity coordinates are also useful, particularly for equations originally expressed in terms of accelerations. More broadly, a dynamical system may use any variables sufficient to specify its state, without dividing them into positions and momenta. (damtp.cam.ac.uk)

Evolution and Hamiltonian structure

In Hamiltonian mechanics, evolution is generated by a classical Hamiltonian $H(q,p,t)$ :

\dot q_i=\frac{\partial H}{\partial p_i}, \qquad \dot p_i=-\frac{\partial H}{\partial q_i}.

These differential equations assign a direction of motion to each phase-space point. For standard mechanical systems, the Hamiltonian represents total energy. If it has no explicit time dependence, its value is conserved, so a trajectory remains on a constant-energy surface rather than exploring the entire phase space. (live.ocw.mit.edu)

The geometric formulation treats phase space as a manifold equipped with a symplectic structure. In canonical coordinates, its defining two-form can be written

\omega=\sum_{i=1}^{n}dq_i\wedge dp_i.

This structure pairs coordinates with momenta and determines Hamiltonian evolution. Symplectic geometry concerns this structure rather than ordinary distances or angles. Canonical transformations preserve it, allowing different coordinate descriptions of the same dynamics. (damtp.cam.ac.uk)

Phase portraits and examples

A phase portrait displays trajectories for different initial conditions, usually in a two-dimensional phase plane. It reveals qualitative behavior without requiring an explicit solution for every initial state. Stationary states appear as fixed points, while periodic motions appear as closed trajectories. Stability theory examines whether nearby trajectories remain close to, approach, or move away from a stationary state or orbit. (math.mit.edu)

For a one-dimensional harmonic oscillator,

H(q,p)=\frac{p^2}{2m}+\frac{kq^2}{2},

where $m$ is the mass and $k$ the spring constant. Positive-energy trajectories are ellipses in the $(q,p)$ plane, surrounding the equilibrium at the origin. Each ellipse represents repeated conversion between kinetic energy and potential energy. With linear damping, trajectories instead approach the equilibrium; in the underdamped case they spiral inward. (math.mit.edu)

Nonlinear systems can exhibit more complex structures. Chaotic dynamics involves sensitive dependence on initial conditions despite deterministic equations. A Poincaré section reduces the visualization problem by recording successive intersections with a selected surface, or by sampling periodically driven motion at a fixed driving phase. Such representations help distinguish periodic behavior from more complicated motion. A low-dimensional projection can conceal state variables and need not preserve all features of the complete trajectory. (live.ocw.mit.edu)

Conservation of phase-space volume

Liouville’s theorem states that Hamiltonian evolution preserves phase-space volume. In canonical coordinates, the volume element is

d\Gamma=\prod_{i=1}^{n}dq_i\,dp_i.

A moving region may stretch, fold, or change shape, but its volume remains constant. This result also holds for explicitly time-dependent Hamiltonians: volume preservation does not require energy conservation. (damtp.cam.ac.uk)

If an ensemble has probability density $\rho(q,p,t)$ , its density remains constant along Hamiltonian trajectories, giving Liouville’s equation $d\rho/dt=0$ . This does not mean that density at a fixed location is unchanging. Dissipative systems generally do not preserve the corresponding phase-space volume and may concentrate trajectories near attracting states. (damtp.cam.ac.uk)

Statistical interpretation

Statistical mechanics replaces an exactly specified microscopic state with a probability distribution over accessible states. Equilibrium ensembles impose different restrictions or statistical weights. For an isolated equilibrium system, the microcanonical description assigns equal weight to accessible states at the prescribed energy. In a canonical ensemble, states receive weights proportional to $e^{-\beta H}$ , where $\beta=1/(k_{\mathrm B}T)$ . (damtp.cam.ac.uk)

For $N$ identical, structureless particles in the classical regime, the partition function is

Z_N=\frac{1}{N!h^{3N}}\int e^{-\beta H(q,p)}\,d\Gamma.

Here $h$ is the Planck constant. The factor $h^{3N}$ supplies the semiclassical state-counting scale, while $N!$ corrects for indistinguishability. Phase-space integration thereby links microscopic dynamics to equilibrium quantities such as entropy and free energy. (damtp.cam.ac.uk)

Quantum phase-space descriptions

In quantum mechanics, a state cannot generally be interpreted as a point with simultaneously definite position and momentum. Phase-space formulations nevertheless remain possible. The Wigner function represents a quantum state using position–momentum variables, but is a quasiprobability rather than an ordinary joint probability density. It can take negative values associated with quantum interference; physically valid representations must still satisfy the constraints imposed by quantum-state positivity and the uncertainty principle. (personal.utdallas.edu)