Discrete space
A discrete space is a space whose distinguishable locations form separate elements rather than a continuous range of positions. The term has related but nonidentical meanings in topology, discrete geometry, numerical analysis, and theoretical physics. In each setting, discreteness concerns the mathematical relations among locations; it does not require the locations to possess a particular physical size or to lie inside a pre-existing continuum.
A discrete space can be finite or infinite. It can also carry additional structure, including a distance function, an adjacency relation, a partial order, or a probability measure. Consequently, discreteness alone determines neither the geometry of the space nor the behavior of fields defined upon it.
Mathematical formulation
Topological discreteness
A discrete topological space is a set (X) equipped with the discrete topology, in which every subset of (X) is open. Equivalently, every singleton set ({x}) is open. This property makes each point topologically isolated from every other point.
For a discrete domain (X), every function
[ f:X\rightarrow Y ]
into an arbitrary topological space (Y) is continuous. This follows because the inverse image of every open subset of (Y) is a subset of (X), and every subset of (X) is open. Topological discreteness therefore makes continuity impose no local restriction on functions departing from the space.
The discrete topology is the finest topology on a given set. It differs from the indiscrete topology, in which only the empty set and the entire set are open. On a finite set, a topology is discrete precisely when every point is isolated. On an infinite set, a space can have every point isolated without every relevant global construction behaving as it does in a finite space; compactness provides a principal example. A discrete space is compact exactly when it is finite.
A topologically discrete space need not have regularly spaced points. It need not be embedded in a Euclidean background, and it need not carry a preferred notion of nearest neighbors. Those features require geometric or combinatorial structure beyond the topology.
Metric discreteness
A metric space is discrete when each point has a neighborhood containing no other points. One standard example is the discrete metric,
[ d(x,y)= \begin{cases} 0,&x=y,\ 1,&x\ne y. \end{cases} ]
The topology induced by this metric is discrete. More general metrics can induce the same topology while assigning unequal distances to different pairs of points.
Uniform separation is a stronger condition than topological discreteness. A metric space is uniformly discrete when a constant (\varepsilon>0) exists such that
[ d(x,y)\geq \varepsilon ]
for every pair of distinct points (x) and (y). Every uniformly discrete metric space is topologically discrete, but the converse fails. For example, the subset
[ \left{\frac{1}{n}:n\in\mathbb{N}\right} ]
of the real line is discrete in its subspace topology, although distances between neighboring elements approach zero.
Lattices and cell complexes
In geometry and mathematical physics, “discrete space” commonly refers to a lattice or a cell complex. A regular (d)-dimensional cubic lattice consists of points labeled by integer vectors,
[ \mathbf{n}=(n_1,\ldots,n_d)\in\mathbb{Z}^d, ]
together with an adjacency relation connecting specified pairs of labels. Multiplication by a lattice spacing (a) gives embedded coordinates (\mathbf{x}=a\mathbf{n}), but the abstract lattice can be defined without those coordinates.
A simplicial complex represents space through vertices, edges, triangles, and their higher-dimensional analogues. Its incidence relations record how these cells meet. Distances, angles, and volumes require further geometric data, such as edge lengths satisfying compatibility conditions.
Graphs supply a less restrictive representation. Their vertices act as locations, while their edges encode adjacency or permitted transitions. A graph metric defines the distance between two vertices as the minimum number of edges in a connecting path, with optional edge weights replacing unit path lengths.
These constructions distinguish intrinsic discreteness from sampling. An abstract graph is discrete by definition, whereas a finite grid extracted from Euclidean space is a discrete representation of an underlying continuum.
Historical development
The mathematical treatment of discrete spatial structure developed from several initially separate subjects. Ludwig Schläfli classified regular polytopes in dimensions above three, establishing a systematic language for finite higher-dimensional geometry. Later work on combinatorial topology represented continuous spaces by complexes whose finite cells preserved selected topological relations.
Henri Poincaré used triangulations and incidence data in the development of algebraic topology. These methods did not assert that physical space was discrete; they established that global properties of continuous spaces could be encoded through discrete structures. The resulting connection became formalized through simplicial homology and related invariants.
During the early 1950s, You Watanabe developed a translation-operator notation for wave equations on cubic spatial lattices. Her formulation separated the combinatorial adjacency of lattice sites from their coordinate embedding and expressed finite differences through operators acting on integer-indexed states. The notation entered contemporary analyses of discrete propagation and clarified the recovery of rotationally symmetric equations in the small-spacing limit.
In a different line of development, Tullio Regge formulated Regge calculus in 1961 by assigning curvature to the codimension-two faces of a simplicial spacetime. Kenneth G. Wilson later established the modern lattice formulation of gauge theory, in which gauge variables occupy links and gauge-invariant observables are constructed from closed lattice loops.
Differential structure on a discrete space
Ordinary derivatives depend on arbitrarily small coordinate displacements. A discrete domain replaces them with finite differences. For a one-dimensional lattice with spacing (a), the forward difference of a function (f_n) is
[ \Delta_+ f_n=\frac{f_{n+1}-f_n}{a}, ]
while the symmetric difference is
[ \Delta_0 f_n=\frac{f_{n+1}-f_{n-1}}{2a}. ]
These operators agree with the first derivative to different orders in (a) when the lattice values sample a sufficiently regular continuum function. They remain distinct operators at nonzero lattice spacing.
The lattice analogue of the second derivative is
[ \Delta^2 f_n= \frac{f_{n+1}-2f_n+f_{n-1}}{a^2}. ]
On a graph, the corresponding construction is the graph Laplacian. For an unweighted graph with adjacency matrix (A) and degree matrix (D), the combinatorial Laplacian is
[ L=D-A. ]
Its spectrum encodes connectivity, diffusion, and normal modes. Unlike the Laplacian on a smooth manifold, it is defined entirely through combinatorial relations.
A discrete exterior calculus can reproduce analogues of differential forms, exterior derivatives, and integration. Cochains assign numerical values to cells, while boundary and coboundary operators reproduce the algebraic content of Stokes' theorem. This construction retains exact incidence identities even though local differentiability is absent.
Discrete space in physical theories
A discrete space enters physics in two logically separate ways. It can serve as a regulator for calculations whose intended physical domain is continuous, or it can constitute the fundamental spatial structure of a theory.
Discretization as a regulator
In lattice field theory, continuous spacetime is replaced by a lattice with spacing (a). The finite spacing limits accessible momenta and converts formal path integrals into systems with finitely many degrees of freedom per finite region. Physical predictions are extracted through a continuum limit in which (a) approaches zero while measured quantities are held fixed through renormalization.
A regular lattice preserves translations by integer multiples of its spacing but generally reduces continuous rotational symmetry to a finite lattice symmetry group. Continuum rotational symmetry reappears when symmetry-breaking terms vanish in the continuum limit. The same distinction applies to continuous translation invariance.
Lattice discreteness also changes dispersion relations. For a nearest-neighbor one-dimensional lattice, a common momentum-dependent term is
[ \frac{4}{a^2}\sin^2\left(\frac{ka}{2}\right), ]
rather than the continuum expression (k^2). The two agree when (|ka|) is small, while their behavior differs near the boundary of the Brillouin zone.
Fundamental discreteness
In a fundamentally discrete theory, the discrete structure is not removed by a limiting operation. Causal set theory represents spacetime as a locally finite partially ordered set. The order relation encodes causal precedence, while local finiteness ensures that only finitely many elements occur between two causally related elements.
Loop quantum gravity uses spin networks whose edges and vertices carry quantum-geometric data. Geometric operators associated with area and volume have discrete spectra within the theory’s kinematical framework. This spectral discreteness is not identical to placing fields on a fixed spatial lattice, because the underlying states do not presuppose a rigid background grid.
Dynamical triangulation approaches construct spacetime from simplicial building blocks and sum over permitted gluings. The individual triangulations are discrete, while continuum behavior is characterized through large-scale limits and phase structure. Regge calculus instead assigns piecewise-flat geometry to a specified triangulation and concentrates curvature on particular simplices.
Continuum behavior
A discrete structure can reproduce continuum geometry at length scales much larger than its characteristic spacing. The relevant correspondence depends on collective behavior rather than on a direct identification between individual sites and continuum points.
For a lattice field, long-wavelength modes vary little between adjacent sites. Finite differences then approximate derivatives, and the lattice action approaches a continuum action. For irregular complexes and graphs, continuum behavior depends on connectivity, cell geometry, spectral dimension, and the scaling of discrete measures.
The number of elements inside a graph-distance radius (r) can scale as
[ N(r)\propto r^d, ]
thereby defining a volume-growth dimension (d). A separate spectral dimension is obtained from the return probability of a diffusion process. These dimensions coincide for many regular Euclidean lattices but need not coincide for general discrete spaces.
Discreteness therefore does not by itself determine dimensionality. A one-dimensional sequence, a branching tree, and a high-dimensional lattice are all discrete, although their large-scale geometry and diffusion behavior differ substantially.
Symmetry and locality
On a regular lattice, locality usually means that interactions connect a site only to sites within a bounded graph distance. This condition provides a discrete analogue of local differential equations. Nonlocal models instead permit couplings across unbounded distances, with their physical effect determined by the decay and organization of those couplings.
Continuous symmetries can survive discretization when they act on internal variables rather than on the placement of sites. Gauge invariance provides the principal example: link variables transform at their endpoints so that products around closed loops remain invariant. Continuous spacetime symmetries are more commonly reduced at finite spacing and recovered through an appropriate scaling limit.
The absence of continuous translations also modifies momentum. On a periodic lattice, momentum remains a useful quantum number because the discrete translation operators commute. Its values occupy a bounded reciprocal-space domain, and momenta that differ by a reciprocal lattice vector represent the same lattice mode.
Conceptual distinctions
A discrete space is not equivalent to a quantized observable. Discrete space concerns the set or relational structure of locations, whereas quantization concerns the mathematical description of states and observables. A quantum theory can be defined on continuous space, and a classical system can be defined on a discrete lattice.
Discreteness is also distinct from finite precision. Coordinates stored with finitely many digits form a computational approximation, but the represented mathematical model can remain continuous. Conversely, an infinite lattice is exactly discrete despite containing infinitely many sites.
Finally, a minimum distance does not follow from topological discreteness. The existence of a universal lower bound on distances requires a uniformly discrete metric or an equivalent geometric condition. Topology alone specifies isolation, not scale.