1.2.29 Tensor Factor Definition
In tensor algebra, a tensor factor is a building block used to construct multidimensional arrays through tensor products.
Tensor Factor Definition is the characterization of each individual vector space that participates in a tensor product construction, together with the corresponding role that each such space, or an element drawn from it, plays inside a simple tensor. In an expression of the form , the spaces and are called the tensor factors of the product, and in a simple tensor , the individual vectors and are likewise referred to as its factors.
Formal Definition
Given a tensor product of vector spaces
over a common field , each vector space for is called a tensor factor of the product space. Correspondingly, an element of the product that can be written in the simple, or decomposable, form
with for each , is said to have tensor factors , one drawn from each of the participating spaces. The order of the factors matters, since the tensor product distinguishes from whenever these lie in different tensor product spaces, and , even though the two spaces are canonically isomorphic.
Distinguishing Space-Level and Vector-Level Factors
The term tensor factor is used at two related levels, and care is needed to distinguish them.
Space-Level Factors
At the level of the ambient product space, the tensor factors are the vector spaces themselves, which determine the structure and dimension of the resulting tensor product space. For example, in , the two space-level tensor factors are and its dual , and this pairing of factors is what defines the space of type tensors on .
Vector-Level Factors
At the level of an individual simple tensor, the tensor factors are the specific vectors chosen from each space that combine to produce that element. Two different simple tensors built from the same space-level factors, such as and , generally represent different elements of whenever their vector-level factors differ.
Non-Uniqueness of Factor Decomposition
A key subtlety of tensor factors is that the decomposition of a simple tensor into its factors is not unique, since the bilinearity relations defining the tensor product allow scalars to be redistributed between factors without changing the resulting tensor:
for any nonzero scalar . Consequently, the individual factors and of a simple tensor are not themselves canonically determined by the tensor alone; only their combined product , together with the one-dimensional spans they generate, is an invariant of the element.
Factors of Non-Simple Tensors
Not every element of a tensor product space possesses a single set of tensor factors, since general elements are finite sums of simple tensors rather than simple tensors themselves:
Such a tensor is said to have tensor rank if is the smallest number of simple terms needed to express it, and it is called entangled or non-decomposable when , meaning it has no single well-defined set of tensor factors and must instead be described as a sum of several factor pairs. This distinction between decomposable and non-decomposable elements is central to applications of tensor factor structure outside pure algebra, most notably in the description of composite quantum systems, where a decomposable state corresponds to an unentangled configuration and a non-decomposable one corresponds to an entangled configuration.
Role Within Tensor Algebra
Identifying the tensor factors of a space or an element is a basic organizational tool throughout tensor algebra. The type of a tensor is defined directly in terms of how many of its space-level factors are copies of and how many are copies of the dual , operations such as contraction act on a specified pair of factors, and index notation assigns one index to each tensor factor so that a component can be read off directly against the corresponding factor of the tensor product space. Understanding tensor factors as the atomic building blocks of both the ambient space and its elements is therefore a prerequisite for describing how tensors are constructed, decomposed, and manipulated in coordinates.