Global Dimension¶
A ring's global dimension is the supremum of projective-resolution lengths across all modules on a specified side.
Core Idea¶
The global dimension of a ring asks for a uniform length of projective resolutions across its modules. A module \(M\) has projective dimension \(d\) if \(d\) is the least length of an exact resolution of \(M\) by projective modules; if no finite such resolution exists, its projective dimension is infinite. Fix the left or right module category for a noncommutative ring. The corresponding global dimension is the supremum of the projective dimensions of all modules in that category. If no finite uniform bound exists, the global dimension is infinity.[1][2]
This invariant does not measure the size of the ring's underlying set, and it is not simply the projective dimension of a favored module. It measures the worst required length of a specific kind of algebraic replacement. A field has global dimension zero because every module over it is a vector space and hence free. For \(k[x_1,\ldots,x_n]\) over a field, the global dimension is \(n\). The latter result is a theorem arising from regularity and dimension facts, not the definition of global dimension.[3][4]
Structural Signature¶
- Ring and side: a ring \(R\), with left \(R\)-modules or right \(R\)-modules fixed when order of multiplication matters.
- Projective resolution: an exact sequence \(\cdots \to P_1\to P_0\to M\to0\) of projective modules ending at \(M\).
- Modulewise minimum: \(\operatorname{pd}_R(M)\), the least finite resolution length or infinity.
- Categorywide supremum: \(\operatorname{gldim}(R)=\sup_M\operatorname{pd}_R(M)\), over every module on the selected side.
- Conditional comparison: under additional ring hypotheses, homological length may coincide with a geometric dimension or characterize regularity.
Condensed: choose ring/side → minimize each module's projective resolution → take the all-modules supremum.[1][2]
Sig role-phrases: fixed ring and module side; exact projective resolution; least modulewise length; supremum over every module; conditional regular-local comparison.
What It Is Not¶
- Not Krull dimension by definition. Krull dimension uses chains of prime ideals. Equality is a conditional theorem, for example for commutative Noetherian regular local rings.[3]
- Not weak dimension. That neighboring invariant uses flat resolutions; projective and flat are not interchangeable in the definition.
- Not the dimension of one module. One module may have a short resolution while another needs longer, so a single witness gives at best a lower bound.
- Not the length of any chosen resolution. Projective dimension uses the shortest permitted finite length.
- Not automatically side-free for noncommutative rings. Left and right module categories must be distinguished. Auslander proved their global dimensions equal when the ring is both left- and right-Noetherian; that theorem does not assert unrestricted equality.[2]
- Not a statement that every regular Noetherian ring has finite global dimension. The Stacks Project explicitly distinguishes regularity from a uniform finite bound when dimension is unbounded.[3]
Scope of Application¶
In homological algebra, global dimension expresses a bound on how many projective stages a ring's modules can require. The definition applies to unital rings, with left/right conventions made explicit where needed. An equivalent finite-bound test can be phrased using vanishing of sufficiently high \(\operatorname{Ext}\) groups; it is the uniform quantification over modules that matters.[1][2]
In commutative local algebra, let \((R,\mathfrak m,\kappa)\) be a Noetherian local ring. The Stacks Project proves that \(\kappa\) having finite projective dimension, \(R\) having finite global dimension, and \(R\) being regular local are equivalent. In that case the global dimension equals \(\dim R\), the Krull dimension, and the dimension of \(\mathfrak m/\mathfrak m^2\) as a \(\kappa\)-vector space. Those hypotheses do substantive work: a rule about a local Noetherian ring is not a general identity between two different definitions.[3]
For polynomial rings over a field, \(k[x_1,\ldots,x_n]\) is Noetherian, regular and of Krull dimension \(n\), and the Stacks Project explicitly proves its global dimension is \(n\). An attaining module is the residue field at \(\mathfrak m=(x_1,\ldots,x_n)\): after localization at \(\mathfrak m\), the regular local ring has embedding dimension \(n\), so its residue field needs projective dimension at least \(n\). If that residue field had a shorter projective resolution over the polynomial ring, localization would give an impermissibly shorter one locally. For a field \(k\), \(n=0\), and the result is visible directly because every \(k\)-module is free.[4][3]
Clarity¶
There are two quantifiers. Within each module, minimize over its projective resolutions. Then maximize in the supremum sense across all modules. Reversing or omitting either quantifier changes the question. A long resolution of one module is not proof that its projective dimension is long if a shorter projective resolution exists; a short resolution of one module says nothing about a ring-wide upper bound.
For a noncommutative ring, state “left global dimension” or “right global dimension.” Auslander's equality theorem requires the ring to be both left- and right-Noetherian; it does not make the side distinction dispensable for arbitrary rings.[2]
Manages Complexity¶
The invariant replaces a potentially unbounded collection of module-specific homological problems with one ring-level bound. Dimension zero says that every module is projective. A finite value says that every module can be resolved within that many projective steps, even though the modules themselves may be very different. The compression is powerful only when the quantifier remains visible; checking a convenient family of modules needs a theorem to justify that it controls all modules.[1]
Abstract Reasoning¶
Start with a ring and its module side. For an arbitrary module, ask whether a finite projective resolution exists and what minimal length is possible. To prove \(\operatorname{gldim}R\le n\), obtain an \(n\)-step projective resolution for every module on that side, or invoke a valid equivalent criterion. To prove \(\operatorname{gldim}R\ge n\), exhibit a module whose projective dimension is at least \(n\). If no finite uniform bound works, the answer is infinity.[1]
Comparison to geometry is a separate step. In the commutative Noetherian local setting, regularity licenses the equality with Krull dimension. In an arbitrary commutative Noetherian ring, regular localizations of unbounded dimensions do not supply one finite global bound. The Stacks Project states this distinction expressly.[3]
Knowledge Transfer¶
The projective-resolution method transfers among module categories and algebraic settings as long as the ring, module side and projective objects are preserved. Results about regular local rings help analyze localized commutative problems. They should not be transplanted into noncommutative or non-Noetherian settings as if “dimension” had one meaning everywhere. The adjacent weak dimension can guide comparison, but switching from projective to flat resolutions changes the measured structure.
Examples¶
A field¶
Every module over a field \(k\) is a vector space with a basis, hence free and projective. Each has projective dimension zero, so the supremum is zero.[1]
Mapped back: every module → zero-step projective resolution → ring-wide supremum zero.
A polynomial ring with an attaining module¶
Take \(R=k[x,y]\), \(\mathfrak m=(x,y)\), and \(M=R/\mathfrak m\cong k\). The two-variable Koszul sequence \(0\to R\xrightarrow{f\mapsto(-yf,xf)}R^2\xrightarrow{(a,b)\mapsto xa+yb}R\to k\to0\) is an exact free resolution, giving \(\operatorname{pd}_R(k)\le2\). Its maps compose to zero because \(x(-yf)+y(xf)=0\); exactness follows from the regular sequence \(x,y\) in the polynomial ring. At the local ring \(R_{\mathfrak m}\), the images of \(x,y\) form a basis of \(\mathfrak mR_{\mathfrak m}/(\mathfrak mR_{\mathfrak m})^2\), so Stacks Lemma 10.110.3 gives \(\operatorname{pd}_{R_{\mathfrak m}}(k)\ge2\). Localization cannot lengthen a projective resolution, hence \(\operatorname{pd}_R(k)\ge2\) as well. This residue field attains length 2, while Stacks Proposition 10.114.2 gives the all-module upper bound \(\operatorname{gldim}R=2\).[3][4]
Mapped back: chosen commutative ring/side \(R=k[x,y]\), an explicit module \(k\), a length-two projective resolution, a local lower-bound proof that no shorter resolution exists, and the theorem giving the uniform all-modules upper bound.
Nonregular Noetherian local ring¶
If a commutative Noetherian local ring is not regular, the equivalence in the Stacks Project rules out finite global dimension. A single module with a short resolution would not reverse that conclusion; the all-modules bound is the issue.[3]
Mapped back: failed regularity under the local Noetherian hypotheses → no finite uniform projective bound.
One short-lived module¶
Finding a projective module \(P\) over a ring proves \(\operatorname{pd}_R(P)=0\), but it does not prove that the ring's global dimension is zero: other modules may not be projective.
Mapped back: modulewise minimum established; categorywide supremum still open.
Structural Tensions¶
No universal intrinsic opposed-cost tension belongs to this ring invariant. “One module versus all modules” is its defining quantifier order; “homological versus geometric dimension” distinguishes different invariants; “left versus right” specifies a noncommutative category. These are necessary typing and proof checks, not choices with two competing benefits. A particular computational strategy for estimating global dimension may trade proof strength against effort, but no such universal cost pair follows from the invariant's definition.
Structural–Framed Character¶
Global dimension lies near the formal-structural end of the structural–framed spectrum. Once a ring, module side and projective module category are fixed, the invariant is a supremum of precisely defined minimal resolution lengths; \(\infty\) is a legitimate value. Evaluative language such as “regular” or “simple” may motivate its use, but the value itself is not a judgment of mathematical quality. Human mathematical practice matters in choosing left versus right for noncommutative rings, deciding which theorem's hypotheses are met and selecting a witness or proof technique. These are conventions and proof obligations around an exact object, not permission to replace projective with flat resolutions.
The concept comes from homological algebra and ring theory. Its vocabulary travels across module categories only when projectives, exactness and the all-objects supremum are preserved. Importing “global dimension” into an arbitrary organization's longest workflow would be a metaphor; independently constructing the same invariant for a new ring/module side is recognition of the mathematical structure. Its character: an exact category-relative homological invariant with strict quantifier and carrier requirements, whose geometric equalities depend on explicitly stated ring hypotheses.[1][3][2]
Structural Core vs. Domain Accent¶
The portable reasoning skeleton has two quantifiers: minimize a permitted resolution length for each object, then take the supremum across all objects in a category. The domain-bound mechanism supplies exactly what those words mean here—\(R\)-modules on a chosen side, exact sequences, projective terms and minimal homological length. A longest path in a project plan may share the min/sup form but lacks modules and projective resolutions; flat resolutions instead define Weak Dimension, not this entry.
The named invariant fails the prime bar because its exact value and theorems depend on that algebraic substrate, not merely the portable quantifier pattern. Weak Dimension is a flat-resolution neighbor and cannot serve as a strict genus without erasing the changed class of admissible resolutions. Gelfand–Kirillov and Krull dimensions measure other structures; no immediate strict parent was established. A future prime for “worst minimal resolution complexity” would need nonmodule settings with a justified notion of resolution, invariant and transfer theorem. The present field and polynomial examples remain within one homological family, so they do not establish that wider identity.
Instantiates / Related Primes¶
Global dimension has no broader abstraction in the catalog yet. Weak Dimension is a close neighbor, but it measures the complexity of flat rather than projective resolutions. Gelfand–Kirillov Dimension and geometric notions of dimension address different growth or chain phenomena; the shared word “dimension” does not make them the same concept.
Neighborhood in Abstraction Space¶
Global Dimension sits in a sparse region of the domain-specific corpus (89th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2551 abstractions)
Nearest neighbors
- Cohomological dimension — 0.83
- Projective variety — 0.80
- Weak dimension — 0.80
- Perfect ring — 0.79
- Finite morphism — 0.79
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
Projective Dimension of a Module: one module's shortest projective resolution; global dimension takes the supremum over all modules. Weak Dimension: flat-resolution analogue. Krull Dimension: length of chains of prime ideals, conditionally equal in the regular local Noetherian case. Gelfand–Kirillov Dimension: algebraic growth invariant, not homological resolution length.
References¶
[1] The Stacks Project, Algebra §10.109, “Rings of finite global dimension”, especially Definitions 10.109.2 and 10.109.10 and the finite-bound criteria. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g
[2] Auslander, “On the Dimension of Modules and Algebras (III): Global Dimension,” Nagoya Mathematical Journal 9 (1955): 67–77, p. 67 for the general left/right definitions and Ext criterion; p. 70, Theorem 4 and Corollary 5 for the sided Noetherian comparison. registry ↩a ↩b ↩c ↩d ↩e ↩f
[3] The Stacks Project, Algebra §10.110, “Regular rings and global dimension”, especially Proposition 10.110.5 and the warning before Lemma 10.110.8. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i
[4] The Stacks Project, Algebra §10.114, “Dimension of finite type algebras over fields”, polynomial-ring dimension and regularity. registry ↩a ↩b ↩c