---
title: T-Geometries in Mathematics & Physics
url: https://www.emergentmind.com/topics/t-geometries
type: topic
---

# T-Geometries in Mathematics & Physics

Searching arXiv for recent and relevant papers on “T-geometries” and closely related usages.
“T-geometries” is not a single standardized concept across mathematics and theoretical physics. In the contemporary literature, the label is used for several technically distinct constructions: transitional geometries interpolating among constant-curvature planes [1411.5801]; rank-three incidence geometries with a canonical triality obtained from triangle complexes of rank-two geometries [2504.06025]; Kähler geometries of T-model type in supergravity and inflationary model building [2502.00636]; teleparallel geometries characterized by vanishing curvature with nonzero torsion and/or nonmetricity [2303.17812]; and, in coarse geometry, discrete metric spaces with Geometric Property (T), sometimes informally described as “T-geometries” [1311.6197], [2312.00893]. The term also appears in algebraic geometry through T-varieties with effective torus actions [1102.5760], and in string-theoretic constructions involving T-duality and non-abelian twists [2311.14285], [1810.08093]. The plurality of these usages makes context essential: the common “T” may stand for transition, triality, T-model, teleparallel, torus, or T-duality, rather than denoting a single invariant mathematical object.

## 1. Terminological scope and competing usages

The expression “T-geometries” is used in at least three broad ways.

First, in differential and classical geometry, A’Campo and Papadopoulos develop “transitional geometry,” a one-parameter family of constant-curvature geometries connecting spherical, Euclidean, and hyperbolic planes continuously through a curvature parameter \(k\) [1411.5801]. Here the “T” is explicitly transitional.

Second, in incidence geometry, the recent triangle-complex construction associates to a rank-two incidence system \(\Gamma\) a rank-three pregeometry \(\Delta(\Gamma)\) with a canonical triality \(\tau\), and the surrounding discussion explicitly treats these as geometries with trialities [2504.06025]. In this setting, “T-geometries” may be read as triality-geometries.

Third, in mathematical physics, the label appears in several unrelated ways: T-model Kähler geometries in \(\alpha\)-attractor supergravity [2502.00636]; teleparallel geometries with \(R^A{}_B=0\) [2303.17812]; and T-duality-generated supergravity backgrounds or twisted geometries [2311.14285], [1810.08093].

A separate, operator-algebraic usage concerns Geometric Property (T) for discrete metric spaces. One source explicitly presents “T-geometries” as discrete metric spaces with Geometric Property (T), together with associated Kazhdan projections in maximal uniform Roe algebras [2312.00893]. This is consistent with the broader coarse-geometric treatment of Geometric Property (T) as a strong spectral-gap condition on representations of the translation algebra \(C_u[X]\) [1311.6197].

This terminological heterogeneity implies that the phrase has no universal definition. A plausible implication is that encyclopedia treatment is best organized by domain rather than by forcing a false unification.

## 2. Transitional geometries of constant curvature

In transitional geometry, one studies a family \(E_k\) parameterized by real curvature \(k\), with spherical geometry for \(k>0\), Euclidean geometry for \(k=0\), and hyperbolic geometry for \(k<0\) [1411.5801]. The ambient model is defined in \(\mathbb R^3\) via
\[
q_k(x,y,z)=k\,x^2+k\,y^2+z^2,
\]
with \(G_k\) the identity component of the orthogonal group of \(q_k\). For \(k>0\), \(G_k\cong \mathrm{SO}(3)\) acts transitively on \(S_k=\{q_k=1\}\); for \(k<0\), \(G_k\cong \mathrm{SO}(2,1)\) acts on the two-sheeted hyperboloid \(q_k=1\); and for \(k=0\), after restricting to coherent elements, \(G_0\) becomes the Euclidean motion group \(\mathbb R^2\rtimes \mathrm{SO}(2)\) [1411.5801].

A distinctive feature is the group-theoretic definition of points and lines. Points in \(E_k\) are maximal compact abelian subgroups of \(G_k\), equivalently involutions \(s_p\in G_k\) of order two. A line is a maximal subset \(L\subset E_k\) such that for any \(p,p'\in L\), the product \(s_p s_{p'}\) lies in an abelian one-parameter subgroup of \(G_k\). In this framework, any two points lie on a unique line, and for \(k\neq 0\) lines correspond to great circles or geodesics on \(S_k\) [1411.5801].

The metric structure is normalized so that the Euclidean case arises as a genuine limit. For \(k\neq 0\), one first defines an angular distance on \(S_k\) from the polar bilinear form \(\beta_k\), and then introduces the normalized distance
\[
d_k(P,Q)=
\begin{cases}
\displaystyle\frac{1}{\sqrt{k}\,\arccos\!\bigl(\beta_k(P,Q)\bigr)},&k>0,\\[1em]
\sqrt{(x_1-x_2)^2+(y_1-y_2)^2},&k=0,\\[0.5em]
\displaystyle\frac{1}{\sqrt{-k}\,\operatorname{arccosh}\!\bigl(-\beta_k(P,Q)\bigr)},&k<0.
\end{cases}
\]
The family \(d_k\) varies continuously as \(k\to 0\), and the Euclidean metric is recovered by Taylor expansion [1411.5801].

The same continuity principle extends to triangles, angles, trigonometric identities, and area. The unified functions \(\Sin_k\) and \(\Cos_k\) satisfy curvature-dependent sine and cosine laws valid in all three regimes, and the area satisfies the Gauss–Bonnet-type formula
\[
\Area_k(\Delta)=\frac{A+B+C-\pi}{k},\qquad (k\neq 0),
\]
with Euclidean area arising as the \(k\to 0\) limit [1411.5801]. This formulation is significant because Euclidean geometry is not treated as an exceptional case but as a smooth zero-curvature member of a single family.

## 3. Triality geometries from triangle complexes

For a rank-two incidence geometry
\[
\Gamma=(\mathcal P\sqcup\mathcal L,\star,t),
\]
the triangle-complex construction defines a rank-three pregeometry
\[
\Delta(\Gamma)=(X,\star_\Delta,t_\Delta)
\]
with type set \(\{1,2,3\}\) and element set
\[
X=\{(p,L,i)\mid p\in\mathcal P,\;L\in\mathcal L,\;p\star L,\;i\in\{1,2,3\}\}
\]
[2504.06025]. Thus \(X\) consists of three copies of the flags of \(\Gamma\). The incidence relation is defined cyclically by
\[
(p,L,i)\star_\Delta (p',L',i+1\!\!\mod 3)
\quad\Longleftrightarrow\quad
p\neq p',\quad
\Res_\Gamma(L)\cap\Res_\Gamma(L')=\{p\}.
\]
Equivalently, \((p,L,i)\) is joined to \((q,M,i+1)\) precisely when \(L\) and \(M\) meet in exactly the point \(p\) and \(q\neq p\) [2504.06025].

The canonical triality is the cyclic shift
\[
\tau(p,L,i)=(p,L,i+1\!\!\mod 3),
\]
which is a correlation of type the 3-cycle \((1,2,3)\) and satisfies \(\tau^3=\mathrm{Id}\) [2504.06025]. The construction therefore produces rank-three systems with an intrinsic order-three symmetry exchanging types.

Several structural criteria govern when \(\Delta(\Gamma)\) is an incidence geometry with strong transitivity properties. If \(\Delta(\Gamma)\) is an incidence geometry, then the gonality of \(\Gamma\) is at most \(3\). In particular, a flag in \(\Delta(\Gamma)\) can be completed to a chamber if and only if every point-line pair in \(\Gamma\) lies in a triangle [2504.06025]. For a thick finite linear space \(\Gamma\) of gonality \(3\), flag-transitivity of \(\Delta(\Gamma)\) under \(\Aut(\Delta(\Gamma))\) is equivalent to transitivity of \(\Aut(\Gamma)\) on
\[
T(\Gamma)=\{(p,q,r):p,q,r\text{ distinct non-collinear points}\}
\]
[2504.06025].

Thickness and residual connectedness admit precise criteria. \(\Delta(\Gamma)\) is thick if and only if every line of \(\Gamma\) has at least \(3\) points and every point lies on at least \(3\) lines. Moreover, \(\Delta(\Gamma)\) admits a duality if and only if \(\Gamma\) does. Residual connectedness holds for planes \(\PG(2,q)\), \(\AG(2,q)\), and for unitals \(\UH(q)\) with \(q=2\) or \(4\), but fails in certain higher-dimensional projective or affine spaces and in complete graphs \(K_n\) because some rank-two residues split into multiple components [2504.06025].

The classification theorem stated for the flag-transitive linear spaces under consideration identifies exactly when \(\Delta(\Gamma)\) is firm, residually connected, and flag-transitive: this occurs precisely for
\[
\PG(2,q),\;q\ge 2;\qquad
\AG(2,q),\;q\ge 3;\qquad
\UH(q),\;q=2\text{ or }4
\]
[2504.06025]. In particular, \(\Delta(\AG(2,q))\) yields an infinite family of thick, flag-transitive, residually connected geometries with triality but no duality, described there as the first such infinite family [2504.06025].

Two standard examples illustrate the scale of the construction.

| Base geometry \(\Gamma\) | Size of \(\Delta(\Gamma)\) | Structural features |
|---|---:|---|
| \(\PG(2,q)\) | \(3(q^2+q+1)(q+1)\) elements | thick, FT, RC, triality and dualities |
| \(\AG(2,q)\), \(q\ge 3\) | \(3q^2(q+1)\) elements | thick, FT, RC, triality but no duality |

For planes, all rank-two residues of \(\Delta(\Gamma)\) have point-diameter \(4\), line-diameter \(4\), and gonality \(3\), so the Buekenhout diagram is a triangle with edge-labels \(\{3,4,4\}\) [2504.06025]. This makes the triangle-complex T-geometries a concrete bridge between rank-two linear spaces and higher-rank incidence structures with triality.

## 4. T-model Kähler geometries in supergravity and inflation

In supergravity model building, “T-geometries” can denote Kähler geometries of T-model type. A recent realization appears in Starobinsky-like inflation within supergravity, where the inflaton sector is built from Kähler potentials parameterizing hyperbolic geometries known from T-model inflation [2502.00636].

In the gauge-singlet case, the total Kähler potential is
\[
K(\Phi,\bar\Phi,S,\bar S)=K_2(S,\bar S)+K_T(\Phi,\bar\Phi)+K_d(\Phi,\bar\Phi),
\]
with
\[
K_2=N_S\ln[1+|S|^2/N_S],
\]
\[
K_T=-N\ln F_T(\Phi,\bar\Phi),\qquad
F_T=\frac{1-|\Phi|^2}{\sqrt{(1-\Phi^2)(1-\bar\Phi^2)}},
\]
\[
K_d=-(n/2)\ln(1+\Phi)-(n/2)\ln(1+\bar\Phi)
\]
[2502.00636]. The piece \(K_T\) is exactly the Kähler potential of the Poincaré disk of radius \(\sqrt N\), realizing the symmetric space \(\mathrm{SU}(1,1)/\mathrm{U}(1)\) with metric
\[
g_{\Phi\bar\Phi}=K_{T,\Phi\bar\Phi}=\frac{N}{(1-|\Phi|^2)^2}.
\]
The associated constant holomorphic sectional curvature is
\[
R_K=-g^{\Phi\bar\Phi}\partial_\Phi\partial_{\bar\Phi}\ln g_{\Phi\bar\Phi}=-\frac{2}{N},
\]
and in \(\alpha\)-attractor language one defines \(\alpha=2/|R_K|=N\) [2502.00636].

The superpotential is constrained by symmetry. For a gauge-singlet inflaton,
\[
W=S\,\Phi^{n/2}\qquad (n\text{ even}),
\]
with an \(R\)-symmetry \(R(S)=1\), \(R(\Phi)=0\) forcing linearity in \(S\), and a global \(\mathrm U(1)_X\) with \(Q_X(S)=-1\), \(Q_X(\Phi)=2/n\) forbidding other couplings. The holomorphic terms in \(K_d\) break \(\mathrm U(1)_X\) softly [2502.00636]. In the gauge non-singlet case,
\[
W=S[(2\bar\Phi\Phi)^{n/4}-M^2],\qquad M\ll 1,
\]
with \(\Phi,\bar\Phi\) carrying opposite charges under a gauge \(\mathrm U(1)_X\) [2502.00636].

The paper emphasizes exact and softly broken shift symmetry. The T-part of \(K\) depends on
\[
X\equiv \tanh(\phi/\sqrt{2N}),\qquad \Phi=Xe^{i\theta},
\]
and is invariant under the real shift \(\delta\,\mathrm{Re}\,X=\mathrm{const.}\), while the imaginary direction has an exact shift symmetry \(\theta\to \theta+\mathrm{const.}\) in the absence of \(K_d\) [2502.00636]. The additional term \(K_d\), and in the non-singlet case also the mass term \(M^2\), violate this shift symmetry mildly and generate a nontrivial inflaton dependence in the denominator of the scalar potential.

Along the inflationary trajectory \(S=0\), one finds in the gauge-singlet case
\[
V_I(\Phi)=\lambda^2\Phi^n(1+\Phi)^{-n},
\qquad
\frac{d\hat\phi}{d\Phi}=\frac{\sqrt{2N}}{1-\Phi^2},
\]
so that \(\Phi=\tanh(\hat\phi/\sqrt{2N})\) and
\[
V_I(\hat\phi)=\lambda^2[\tanh^n(\hat\phi/\sqrt{2N})][1+\tanh(\hat\phi/\sqrt{2N})]^{-n}
\]
[2502.00636]. For \(n\ll N\), the spectral observables satisfy
\[
n_s\simeq 1-\frac{2}{N}+O(1/N^2),\qquad
r\simeq \frac{12}{N^2}+\frac{4n}{N^2}+O(1/N^3),
\]
while the pure T-model limit \(n=0\) yields
\[
n_s\simeq 1-\frac{2}{N},\qquad r\simeq \frac{12}{N^2}
\]
[2502.00636].

The geometric interpretation given there places these constructions squarely in the class of \(\alpha\)-attractors. In the gauge-singlet case, the inflaton sector is a single \(\mathrm{SU}(1,1)/\mathrm{U}(1)\) disk of curvature \(-2/N\). In the gauge non-singlet case, the Kähler sector can instead realize either \(\mathrm{SU}(1,1)/\mathrm{U}(1)\times \mathrm{SU}(1,1)/\mathrm{U}(1)\) or \(\mathrm{SU}(2,1)/\mathrm{SU}(2)\times \mathrm{U}(1)\), with scalar curvatures
\[
R_K=-\frac{8}{N}+\frac{2}{N_S},
\qquad
R_K=-\frac{6}{N}+\frac{2}{N_S}
\]
respectively [2502.00636]. The article further suggests a taxonomy of “\(T_n\)” geometries labeled by pole order \(n\) and curvature parameter \(N\), where \(n\) controls the shift-symmetry breaking through the factor \((1+\Phi)^{-n}\) and \(N\) fixes the Kähler radius [2502.00636]. Since this nomenclature is presented as a possible labeling scheme rather than a universally adopted standard, it is best regarded as local to that work.

## 5. Teleparallel geometries

In differential geometry and gravitational theory, teleparallel geometries are sometimes informally grouped under “T-geometries.” These are metric-affine geometries on a manifold \(M\) equipped with an independent metric tensor \(g\) and affine connection \(\nabla\), or equivalently a coframe \(\{e^A\}\) and \(GL(n,\mathbb R)\)-connection 1-form \(\omega^A{}_B\) [2303.17812]. The fundamental objects are the nonmetricity 1-form,
\[
Q_{AB}:=-\tfrac12\,Dg_{AB}=\tfrac12(-dg_{AB}+\omega_{AB}+\omega_{BA}),
\]
the torsion 2-form,
\[
T^A:=De^A=de^A+\omega^A{}_B\wedge e^B,
\]
and the curvature 2-form,
\[
{R^A}{}_B:=D\omega^A{}_B=d\omega^A{}_B+\omega^A{}_C\wedge\omega^C{}_B
\]
[2303.17812].

The connection decomposes uniquely into Levi-Civita and distortion pieces,
\[
\omega^A{}_B=\widetilde\omega^A{}_B+K^A{}_B+q^A{}_B+Q^A{}_B,
\]
where \(\widetilde\omega\) is the Riemannian connection, \(K\) is the contortion, \(q\) is the antisymmetric part of nonmetricity, and \(Q\) is the symmetric part of \(\omega_{AB}\) [2303.17812]. Metric-affine geometries are then classified by which of \(Q_{AB}\), \(T^A\), and \(R^A{}_B\) vanish. The teleparallel branches are:

| Geometry | Vanishing tensors | Nonvanishing tensors |
|---|---|---|
| Metric teleparallel | \(Q=0,\;R=0\) | \(T\neq 0\) |
| Symmetric teleparallel | \(T=0,\;R=0\) | \(Q\neq 0\) |
| General teleparallel | \(R=0\) | \(Q\neq 0,\;T\neq 0\) |

The defining constraint of the general teleparallel case is simply
\[
R^A{}_B=0
\]
[2303.17812]. This permits gauge choices in which the connection becomes pure gauge. In the Weitzenböck gauge one sets \(\omega^a{}_b\equiv 0\), obtaining \(Q_{ab}=0\), \(R^a{}_b=0\), and \(T^a=de^a\neq 0\). In the coincident gauge one sets \({\omega^\alpha}{}_\beta\equiv 0\) in a coordinate frame, giving \(T^\alpha=0\), \(R^\alpha{}_\beta=0\), and \(Q_{\alpha\beta}=-\tfrac12\,dg_{\alpha\beta}\neq 0\). In a mixed gauge for general teleparallelism one can set \(\omega^A{}_B\equiv 0\), so \(Q_{AB}=-\tfrac12\,dg_{AB}\neq 0\) and \(T^A=de^A\neq 0\) [2303.17812].

A major methodological result is the recasting of a Riemannian geometry into teleparallel form by a \(GL(n)\) transformation of the frame. Starting from \(\{M,g,\widetilde\omega^A{}_B(g)\}\), one chooses \(L^A{}_B(x)\) and transforms
\[
e'^A=L^A{}_B e^B,\qquad
\omega'^A{}_B=L^A{}_C\,\widetilde\omega^C{}_D\,(L^{-1})^D{}_B+L^A{}_C d(L^{-1})^C{}_B.
\]
Fixing \(\omega'^A{}_B\equiv 0\), the original Riemannian curvature is then encoded entirely in torsion and nonmetricity of the new frame [2303.17812]. This is not merely a formal restatement: the paper develops a general parity-even quadratic Lagrangian with \(T^2\), \(Q^2\), and \(QT\) sectors and derives the corresponding field equations in differential-form language [2303.17812].

The same work presents exact solutions in \(n=2,3,4\), including two-dimensional static metrics, three-dimensional BTZ-type configurations, and four-dimensional Kerr–de Sitter and Reissner–Nordström-like solutions recast in teleparallel form [2303.17812]. The significance of these constructions lies in the relocation of gravitational degrees of freedom from curvature to torsion and nonmetricity, with applications both in gravity and in continuum descriptions of defects in materials.

## 6. Geometric Property (T) as a coarse-geometric notion

In coarse geometry, “T-geometries” can refer to discrete metric spaces with Geometric Property (T). Let \(X\) be a discrete metric space of bounded geometry, and let \(C_u[X]\) denote the algebraic uniform Roe algebra consisting of uniformly bounded, finite-propagation matrices on \(X\) [1311.6197]. A vector \(\xi\) in a representation \(\pi:C_u[X]\to \mathcal B(\mathcal H)\) is invariant if for every partial translation \(v\),
\[
\pi(v)\xi=\pi(vv^*)\xi
\]
[2312.00893]. Denoting by \(\mathcal H^\pi\) the invariant subspace, Geometric Property (T) requires a uniform obstruction to almost-invariant vectors orthogonal to \(\mathcal H^\pi\).

One formulation states that \(X\) has Geometric Property (T) if for every controlled generating set \(E\subset X\times X\), there exists \(c>0\) such that for every representation \(\pi\) and every unit vector \(\xi\in (\mathcal H^\pi)^\perp\), there is a partial translation \(v\) with \(\mathrm{supp}(v)\subset E\) satisfying
\[
\|(\pi(v)-\pi(vv^*))\xi\|\ge c\,\|\xi\|
\]
[2312.00893]. An earlier equivalent formulation uses the orthogonal complement of the constant-vector subspace and measures \(\|(vv^*-v)\xi\|\) [1311.6197]. This is a coarse analogue of Kazhdan’s Property (T).

The theory admits a spectral-gap characterization via the combinatorial Laplacian \(\Delta^E\). For a controlled set \(E\), Geometric Property (T) is equivalent to the existence of \(c_E>0\) such that the maximal spectrum satisfies
\[
\sigma_{\max}(\Delta^E)\subset \{0\}\cup [c_E,\infty)
\]
for every controlled \(E\), or equivalently for some generating \(E\) [1311.6197]. For disjoint unions of finite connected graphs of uniformly bounded degree, this is stronger than ordinary expander behavior: Geometric Property (T) implies \(\inf_n h(X_n)>0\), but not conversely [1311.6197].

The property is a coarse invariant: if \(X\) and \(Y\) are coarsely equivalent, then \(X\) has Geometric Property (T) if and only if \(Y\) does [1311.6197]. It also interacts sharply with amenability. For an infinite connected bounded-geometry graph, Geometric Property (T) holds if and only if the graph is non-amenable [1311.6197]. For box spaces of residually finite groups, Geometric Property (T) is equivalent to the underlying group having Kazhdan’s Property (T) [1311.6197].

A later refinement characterizes Geometric Property (T) through a Kazhdan projection in the maximal uniform Roe algebra \(C^*_{u,\max}(X)\): there exists a projection \(P=P^2=P^*\) such that in every representation \(\pi\), \(\pi(P)\) is exactly the orthogonal projection onto \(\mathcal H^\pi\) [2312.00893]. This projection lies in the norm-closure of a positive normalized cone in the unistochastic subalgebra \(\mathcal A_X\), and decomposes over coarse components. If \(X_n\) is a finite coarsely connected component, the componentwise Kazhdan projection is the rank-one matrix
\[
[Q_n(P)]_{x,y}=\frac1{|X_n|}.
\]
If \(X_n\) is infinite and coarsely connected, then \(Q_n(P)=0\) [2312.00893]. This behavior marks a substantial difference from group \(C^*\)-algebra Kazhdan idempotents.

The main misconception in this area is terminological: Geometric Property (T) concerns spectral gaps in Roe-algebra representations and should not be conflated with teleparallel, toric, or T-duality geometries merely because they also carry a “T.”

## 7. Related T-prefix geometries and reasons for ambiguity

Several further research areas use a T-prefix in ways adjacent to, but distinct from, the preceding notions.

In algebraic geometry, a T-variety is a normal variety \(X\) endowed with an effective action of an algebraic torus \(T\cong (\mathbb K^*)^k\), with complexity \(n-k\) when \(n=\dim X\) [1102.5760]. The modern language of polyhedral divisors and divisorial fans classifies affine and global T-varieties, recovers toric varieties as the complexity-zero case, and organizes invariant divisors, intersection theory, singularities, Cox rings, deformations, and polarizations [1102.5760]. These are not usually called “T-geometries,” but the overlap in naming can generate confusion.

In supergravity and string theory, “T-geometries” may also appear in connection with T-duality. One line of work studies geometries with twisted spheres and non-abelian T-dualities, where one starts with metrics carrying \(\mathrm{SO}(k+1)\times \mathrm{SO}(k+1)\) isometries, applies a non-abelian generalization of a TsT or spectral-flow twist, and then performs NATD to generate new backgrounds [2311.14285]. Another line embeds Abelian and non-Abelian T-duals of the Brandhuber–Oz \(AdS_6\) solution into the D’Hoker–Gutperle–Uhlemann classification of \(AdS_6\times S^2\) geometries with 7-branes, with the Abelian dual living on an annulus and the non-Abelian dual on the upper half-plane [1810.08093]. Here the “T” is unambiguously duality-theoretic.

Finally, exceptional generalized geometry introduces structures connected to non-Abelian and Poisson–Lie T-duality through the Exceptional Drinfeld Algebra, a Leibniz-algebraic generalization of the Drinfeld double in M-theory [2006.12452]. Although this concerns generalized U-dualities rather than T-geometries in a narrow sense, it further illustrates the breadth of T-prefixed geometric nomenclature in current research.

Taken together, these examples show that “T-geometries” is a context-dependent umbrella label rather than a single discipline-wide term. In current usage, the most coherent meanings are transitional geometries in constant-curvature geometry [1411.5801], triangle-complex geometries with triality [2504.06025], T-model Kähler geometries in inflationary supergravity [2502.00636], teleparallel geometries with vanishing curvature [2303.17812], and coarse metric spaces with Geometric Property (T) [1311.6197], [2312.00893]. The technical content of any occurrence is therefore fixed not by the letter “T” alone, but by the surrounding mathematical program.

Source: https://www.emergentmind.com/topics/t-geometries