Author: Anastasis Kratsios
Abstract
It is shown that the Hochschild Cohomological dimension of an associative algebra is not an upper-semi continuous function, showing the semi-continuity theorem is no longer valid for non-commutative algebras. A family of $\mathbb{C}$-algebras parameterized by $\mathbb{C}$ exhibits this behavior: all but one of which has Hochschild cohomological dimension 2, and the other having Hochschild cohomological dimension 1.
1. Introduction
In algebraic geometry and homological algebra, the behavior of "dimension" under deformation is a central theme. A fundamental property often sought in such contexts is upper semi-continuity. Intuitively, if one considers a family of mathematical objects varying continuously according to a parameter, the dimension of a "special" or "limit" member of that family should be at least as large as the dimension of the "generic" members. This intuition holds for the topological dimension of manifolds and the Krull dimension of commutative rings.
However, the transition from commutative to non-commutative algebraic geometry often breaks these established intuitions. This paper addresses whether the Hochschild cohomological dimension (HCDim)—a critical measure of the homological complexity of an algebra—retains upper semi-continuity in the non-commutative setting. By constructing a specific counter-example, we demonstrate that HCDim can indeed "jump down" at a limit point, proving that the semi-continuity theorem does not extend to general associative algebras.
2. Construction of the Algebra Family
To demonstrate the failure of upper semi-continuity, we define a family of $\mathbb{C}$-algebras $A_a$ parameterized by a complex scalar $a \in \mathbb{C}$. The algebras are defined as quotients of the free algebra on two generators:
$$A_a := \mathbb{C} \langle x, y \rangle / (xay - ayx - x)$$
This family allows us to examine the homological properties of the algebra as the parameter $a$ approaches the origin.
3. Analysis of the Generic Case ($a \neq 0$)
For any non-zero value of the parameter $a$, the algebras $A_a$ are isomorphic. This can be demonstrated via a simple scaling transformation. By defining a new variable $y' = ay$, the defining relation becomes:
$$xy' - y'x - x = 0$$
This is the standard commutation relation for the universal enveloping algebra of the unique non-abelian two-dimensional Lie algebra. Because $A_a \cong A_1$ for all $a \neq 0$, their homological properties are identical.
Using a projective resolution and the Poincaré-Birkhoff-Witt (PBW) Theorem, it can be rigorously shown (as detailed in Lemma 1.3) that the Hochschild cohomological dimension of these generic members is exactly 2:
$$HCDim(A_a) = 2, \quad \forall a \in \mathbb{C} \setminus \{0\}$$
4. The Singular Collapse ($a = 0$)
As the parameter $a$ reaches the limit point $0$, the structure of the algebra undergoes a fundamental shift. Substituting $a=0$ into the defining relation $(xay - ayx - x)$, the non-commutative terms vanish:
$$x(0)y - (0)yx - x = 0 \implies -x = 0$$
Consequently, the generator $x$ is eliminated from the algebra entirely. The resulting algebra $A_0$ is isomorphic to the free algebra on one generator $y$, which is simply the commutative polynomial ring in one variable:
$$A_0 \cong \mathbb{C}[y]$$
Applying the Hochschild-Kostant-Rosenberg (HKR) Theorem or Van den Bergh duality, we find that the dimension of a polynomial ring in $n$ variables is $n$. Thus:
$$HCDim(A_0) = 1$$
5. Conclusion and Implications
The results of this construction are summarized in Theorem 1.2. We have a family where:
- $HCDim(A_a) = 2$ for $a \neq 0$
- $HCDim(A_0) = 1$
Since $\lim_{a \to 0} HCDim(A_a) = 2 > HCDim(A_0)$, the function is not upper semi-continuous. In a well-behaved geometric setting, the dimension at the limit point would be expected to be $\ge 2$.
This finding marks a definitive distinction between commutative and non-commutative algebraic geometry. While the paper focuses on this specific counter-example, it opens questions regarding which sub-categories of non-commutative algebras might still satisfy semi-continuity and whether other dimensional measures, such as the Gelfand-Kirillov dimension, exhibit similar "jumps."
Summary updated