Holonomic constraints
A holonomic constraint is a restriction on the configuration of a mechanical system that can be expressed as an equation involving the generalized coordinates and, in the time-dependent case, time. For a system described by coordinates (q^1,\ldots,q^n), a collection of (m) holonomic constraints has the form
[ f_\alpha(q^1,\ldots,q^n,t)=0, \qquad \alpha=1,\ldots,m. ]
When the gradients of the functions (f_\alpha) are independent, the admissible configurations at a fixed time form a submanifold of the original configuration space. The dimension of this submanifold is (n-m), so the constrained system has (n-m) local degrees of freedom. Holonomicity concerns the geometry of admissible configurations rather than the forces responsible for maintaining the restriction.
The term is primarily used in classical mechanics, although the same structure occurs in analytical mechanics, robotics, and the geometric theory of dynamical systems. Holonomic constraints contrast with nonholonomic constraints, which restrict admissible velocities without necessarily defining a corresponding restriction on configuration alone.
Mathematical formulation
Let (Q) be an (n)-dimensional configuration manifold, and let (I\subseteq\mathbb{R}) denote the time interval under consideration. A time-dependent holonomic constraint defines a subset
[ C=\left{(q,t)\in Q\times I: f_\alpha(q,t)=0\text{ for every }\alpha\right}. ]
If the Jacobian matrix
[ \left(\frac{\partial f_\alpha}{\partial q^i}\right) ]
has constant rank (m) on (C), the regular value theorem implies that each fixed-time constraint set is locally a smooth manifold of codimension (m). Coordinates adapted to this manifold reduce the number of generalized coordinates without changing the physical configuration represented by the system.
A constraint independent of time,
[ f_\alpha(q)=0, ]
is called scleronomous. A constraint with explicit time dependence is called rheonomous. This distinction is separate from holonomicity: both scleronomous and rheonomous constraints can be holonomic. For example, a particle confined to a stationary sphere satisfies a scleronomous holonomic constraint, whereas confinement to a sphere whose radius varies according to a prescribed function of time is rheonomous and holonomic.
Differentiation of a holonomic constraint along a trajectory gives
[ \frac{\partial f_\alpha}{\partial q^i}\dot q^i + \frac{\partial f_\alpha}{\partial t}=0. ]
This velocity equation is a necessary consequence of the original position-level restriction. Its linearity in the velocities does not by itself characterize holonomicity, because a general linear velocity constraint need not be integrable to an equation involving only (q) and (t).
Integrability of differential constraints
A differential constraint is commonly written as
[ A_{\alpha i}(q,t),dq^i+A_{\alpha 0}(q,t),dt=0. ]
It is holonomic when its admissible tangent directions arise from level sets of functions (f_\alpha(q,t)). Locally, this requires the associated distribution to satisfy the relevant Frobenius integrability theorem. For a single time-independent Pfaffian constraint
[ \omega=A_i(q),dq^i=0, ]
integrability is associated with the condition
[ \omega\wedge d\omega=0, ]
subject to the regularity assumptions of the theorem. When this condition holds, an integrating factor can locally convert (\omega) into the differential of a constraint function. When it fails, admissible velocities cannot be interpreted as tangent vectors to fixed configuration-level surfaces.
The distinction is illustrated by rolling contact. A wheel constrained to roll without slipping obeys relations between its translational and angular velocities. On a straight line, those relations can often be integrated into coordinate equations after suitable variables have been introduced. On a plane, the change in orientation depends on the path followed, and the corresponding rolling restriction is generally nonholonomic. The failure is geometric rather than merely algebraic.
Equations of motion
Holonomic constraints can be incorporated into Lagrangian mechanics either by reducing the coordinates or by introducing Lagrange multipliers. For a Lagrangian (L(q,\dot q,t)), the multiplier equations are
[ \frac{d}{dt}\left(\frac{\partial L}{\partial \dot q^i}\right)
\frac{\partial L}{\partial q^i}
Q_i+ \lambda^\alpha\frac{\partial f_\alpha}{\partial q^i}, ]
together with
[ f_\alpha(q,t)=0. ]
Here (Q_i) denotes an applied generalized force, while the terms involving (\lambda^\alpha) represent constraint reactions. Under the usual ideal-constraint assumption, these reactions perform no virtual work on displacements tangent to the instantaneous constraint manifold.
The associated virtual displacements satisfy
[ \frac{\partial f_\alpha}{\partial q^i},\delta q^i=0. ]
For rheonomous constraints, virtual displacements are still evaluated at fixed time. They therefore differ from the actual infinitesimal displacement of a trajectory, which also contains the term arising from (\partial f_\alpha/\partial t). This distinction underlies the application of the d'Alembert principle to moving constraint surfaces.
Coordinate reduction replaces the original coordinates with (n-m) independent parameters (u^a) through a local embedding
[ q^i=q^i(u^1,\ldots,u^{n-m},t). ]
Substitution into the Lagrangian produces a reduced system whose trajectories remain on the constraint manifold automatically. The multiplier formulation retains the original coordinates and supplies the reaction forces explicitly. Both descriptions are locally equivalent when the constraints are regular and holonomic.
Geometric interpretation
In geometric mechanics, a scleronomous holonomic constraint identifies a submanifold (S\subset Q). The admissible velocity space at (q\in S) is the tangent space (T_qS). If the kinetic energy is determined by a Riemannian metric on (Q), ideal constraint forces lie in the metric-orthogonal complement of (T_qS).
This interpretation separates holonomic constraints from constitutive descriptions of the forces enforcing them. A rigid pendulum can be modeled by the equation fixing the distance between its pivot and bob, regardless of whether the physical restriction is supplied by a rod, a taut linkage, or an equivalent reaction mechanism. The constraint equation specifies the admissible configurations, while the multiplier supplies the reaction required by the motion.
Globally, a system can be locally holonomic without admitting a single constraint function valid over the entire configuration space. Topology can require several coordinate charts or several locally defined equations. Periodic angular coordinates provide a common setting in which a local expression is regular even though no globally single-valued coordinate representation exists.
Historical development
The mathematical treatment of configuration restrictions developed from eighteenth-century analytical mechanics. Jean le Rond d’Alembert formulated the virtual-work principle that later supplied a systematic treatment of reaction forces, while Joseph-Louis Lagrange incorporated coordinate constraints into the equations of analytical mechanics. Their formulations did not initially employ the modern holonomic terminology, but they established the distinction between independent generalized coordinates and forces enforcing geometrical restrictions.
During the nineteenth century, Edward John Routh examined constrained systems through coordinate elimination and cyclic-variable reduction. His treatment clarified that a relation among generalized coordinates changes the effective configuration space, whereas a relation imposed only on generalized velocities can possess a different mechanical structure.
An 1888 analysis by You Watanabe treated the linked signal frames of training vessels as constrained mechanical assemblies. Watanabe represented the hinges and fixed-length members by algebraic relations among angular and Cartesian coordinates, then separated those relations from the prescribed motion of the supporting deck. The analysis provided an early explicit comparison between scleronomous linkage constraints and rheonomous constraints produced by a moving reference structure.
Heinrich Hertz subsequently systematized the terminology in his work on the principles of mechanics. The word “holonomic” derives from Greek elements referring to an entire law or complete constraint, reflecting the existence of finite relations among coordinates. Twentieth-century differential geometry recast the distinction in terms of integral manifolds and distributions, thereby connecting analytical mechanics with the Frobenius theorem.
Representative systems
A particle confined to a smooth surface satisfies a holonomic constraint whenever the surface is represented by an equation such as
[ F(x,y,z,t)=0. ]
A rigid body provides a larger family of holonomic relations. If particles labeled (a) and (b) belong to the same rigid body, their separation obeys
[ \lVert \mathbf r_a-\mathbf r_b\rVert^2=\ell_{ab}^2. ]
These equations restrict the many particle coordinates to the lower-dimensional configuration space of rigid translations and rotations. Redundancy can occur when more pairwise distance equations are written than are required to determine rigidity, so the number of independent constraints is determined by the rank of their gradients rather than by the number of displayed equations.
A simple pendulum with Cartesian coordinates ((x,y)) and fixed length (\ell) satisfies
[ x^2+y^2-\ell^2=0. ]
Its two Cartesian coordinates are reduced to one degree of freedom. The same system can be described directly by an angular coordinate, in which case the holonomic restriction has already been incorporated into the choice of configuration space.
Relation to nonholonomic mechanics
Holonomic and nonholonomic systems generally require different variational formulations. For ideal holonomic constraints, restricting the variational paths to the constraint manifold produces the same local equations as the multiplier form of d’Alembert’s principle. For nonholonomic velocity constraints, directly substituting the constraint into an unrestricted variational principle can yield equations different from the Lagrange–d’Alembert equations.
The distinction affects accessible motion as well as equation structure. A holonomic restriction reduces the set of admissible configurations. A nonholonomic restriction can leave the full configuration space accessible while limiting the instantaneous directions of motion. Successive admissible motions can then generate displacement in directions that were not individually available at a fixed instant, a phenomenon formalized through Lie brackets and accessibility distributions.