Matrix Product Dimensions And Conformability

Reading the Dimensions of a Product

Before calculating a matrix product, check whether the multiplication is permitted. The test is a comparison of dimensions: the number of columns in the first matrix must equal the number of rows in the second. When that condition holds, the product takes its row count from the first matrix and its column count from the second. These two checks—whether the product exists and what size it has—are called conformability and product-dimension checks.

Throughout, matrices have entries in a common field \(\mathbb{F}\), such as \(\mathbb{R}\) or \(\mathbb{C}\). If \(\mathbf{A}\) has \(m\) rows and \(n\) columns, write \(\mathbf{A}\in\mathbb{F}^{m\times n}\). The first number is always the row count, and the second is always the column count. Thus a matrix in \(\mathbb{F}^{m\times n}\) has \(m\) rows, each with \(n\) entries, and \(n\) columns, each with \(m\) entries.

A compact way to keep track is to place the dimension labels beside the matrices:

The repeated \(n\) is the required match. It is not part of the result size; the two unpaired dimensions, \(m\) and \(q\), give the result size. If the inner dimensions differ, the product is not defined. In that case there is no resulting matrix whose dimensions can be reported.

Why the Dimension Test Works

Matrix multiplication is defined by the row-by-column rule: each position in the proposed product is formed from a row of the first matrix and a column of the second. A row of \(\mathbf{A}\in\mathbb{F}^{m\times n}\) has \(n\) entries. A column of \(\mathbf{B}\in\mathbb{F}^{p\times q}\) has \(p\) entries. The row and column can be paired entry by entry only if these lengths match. This explains why the inner dimensions must be equal. The product has one position for each choice of a row from \(\mathbf{A}\) and a column from \(\mathbf{B}\), giving \(m\) rows of positions and \(q\) columns of positions.

This reasoning establishes the dimension criterion without requiring any calculation of the entries. The next topic develops the entry formula; here, the goal is to determine whether a product is permitted and, if so, its shape.

Proof. By the definition of matrix multiplication, forming an entry of \(\mathbf{A}\mathbf{B}\) requires pairing a row of \(\mathbf{A}\) with a column of \(\mathbf{B}\). Each such row has \(n\) entries, while each such column has \(p\) entries. The row-by-column operation is defined exactly when the two lists have the same length, which is the condition \(n=p\). This proves both necessity and sufficiency. If the operation is defined, there is a product position for every choice of a row of \(\mathbf{A}\) and a column of \(\mathbf{B}\). There are \(m\) choices of the first kind and \(q\) choices of the second kind, so the product has \(m\) rows and \(q\) columns. \(\square\)

The phrase “if and only if” matters here. Matching inner dimensions guarantees that the product is defined, and a defined product requires matching inner dimensions. The result-size rule applies only after that condition has been met.

Worked Dimension Checks

Changing the order changes which dimensions must match. A product can be defined in one order and not the other, as in the example. The dimension test alone does not say that two defined products have the same entries or are equal; it answers only whether each proposed multiplication is permitted and what size its output would have.

Tracking Dimensions Through a Chain

For a sequence of matrices, apply the two-matrix test at each multiplication in the grouping that is written. For example, in \((\mathbf{A}\mathbf{B})\mathbf{C}\), first determine the size of \(\mathbf{A}\mathbf{B}\). Then use that size as the left factor’s dimensions when checking the multiplication by \(\mathbf{C}\). This is dimension bookkeeping for a specified grouping; it makes no claim about changing the grouping.

Proof. For \(k=2\), the result is the matrix product dimension test: the product is defined exactly when \(n_1=m_2\), and its size is \(m_1\times n_2\). Suppose the claim holds for a left-associated product of \(k-1\) matrices. That partial product is defined exactly when \(n_i=m_{i+1}\) for \(i=1,\dots,k-2\), and, when defined, has size \(m_1\times n_{k-1}\). Multiplying it by \(\mathbf{A}_k\in\mathbb{F}^{m_k\times n_k}\) is defined exactly when its column count \(n_{k-1}\) equals \(m_k\). This adds precisely the remaining condition. When it holds, the result has the partial product’s row count \(m_1\) and \(\mathbf{A}_k\)’s column count \(n_k\), hence size \(m_1\times n_k\). Induction proves the claim for every \(k\ge2\). \(\square\)

Common Mistake: Matching the Wrong Dimensions

A frequent error is to compare the outer dimensions when deciding whether a product is defined. For instance, if \(\mathbf{X}\in\mathbb{R}^{2\times3}\) and \(\mathbf{Y}\in\mathbb{R}^{4\times2}\), someone might notice that the outer numbers are both \(2\) and conclude that \(\mathbf{X}\mathbf{Y}\) is defined with size \(2\times2\). That calculation is wrong: the dimensions are \((2\times3)(4\times2)\), so the inner numbers are \(3\) and \(4\), and \(3\ne4\). Therefore \(\mathbf{X}\mathbf{Y}\) is not defined.

The correction is to write the dimensions in order and identify the middle pair before naming a result size:

Only after the middle numbers match should the outside numbers be read as the product dimensions. Keeping these checks in order prevents the common mistake of assigning an output size to a product that does not exist.

Why Conformability Is Useful

Dimension checks are a quick way to catch an impossible operation before doing any arithmetic. They also help organize longer calculations: the dimensions of each intermediate result determine whether the next multiplication is available. When working with a system represented by matrices, for example, tracking dimensions helps distinguish a matrix from a vector and ensures that a proposed product has a meaningful shape. The check is necessary, but it does not guarantee a particular numerical value or any further algebraic property.

Use this sequence whenever a product is proposed:

1. Read each matrix size as rows by columns.
2. Compare the column count of the left matrix with the row count of the right matrix.
3. If they differ, stop: the product is undefined.
4. If they match, report the row count of the left matrix and the column count of the right matrix.
5. For a chain, repeat the test using the size of each intermediate product.