Mathwords logoMathwords

Equivalent Matrix — Definition, Formula & Example

An equivalent matrix is a matrix that can be turned into another matrix of the same size using a mix of invertible row operations and invertible column operations. Two matrices of the same size are equivalent exactly when they have the same rank.

Two m×nm \times n matrices AA and BB over a field are equivalent, written A∼BA \sim B, if there exist an invertible m×mm \times m matrix PP and an invertible n×nn \times n matrix QQ such that B=PAQB = PAQ. This is an equivalence relation, and A∼BA \sim B if and only if rank⁡(A)=rank⁡(B)\operatorname{rank}(A) = \operatorname{rank}(B). Each class contains exactly one canonical form [Ir000]\begin{bmatrix} I_r & 0 \\ 0 & 0 \end{bmatrix}, where rr is the rank. Equivalence differs from row equivalence (B=PAB = PA) and from similarity (B=P−1APB = P^{-1}AP, square matrices only).

Key Formula

B=PAQB = PAQ
Where:
  • A,BA, B = the two m×nm \times n matrices being compared
  • PP = an invertible m×mm \times m matrix that records the row operations
  • QQ = an invertible n×nn \times n matrix that records the column operations

Worked Example

Problem: Show that A=[1224]A = \begin{bmatrix} 1 & 2 \\ 2 & 4 \end{bmatrix} and B=[3000]B = \begin{bmatrix} 3 & 0 \\ 0 & 0 \end{bmatrix} are equivalent. Then decide whether they are row equivalent.
Step 1: Compare sizes and ranks. Both matrices are 2×22 \times 2. Row 2 of AA is twice row 1, so AA has rank 1. BB has one nonzero row, so it also has rank 1. Equal size and equal rank mean the two matrices are equivalent.
rank⁡(A)=rank⁡(B)=1\operatorname{rank}(A) = \operatorname{rank}(B) = 1
Step 2: Exhibit PP and QQ explicitly. PP performs R2→R2−2R1R_2 \to R_2 - 2R_1 and then R1→3R1R_1 \to 3R_1. QQ performs C2→C2−2C1C_2 \to C_2 - 2C_1. Since det⁡P=3\det P = 3 and det⁡Q=1\det Q = 1, both matrices are invertible.
[30−21][1224][1−201]=[3600][1−201]=[3000]\begin{bmatrix} 3 & 0 \\ -2 & 1 \end{bmatrix}\begin{bmatrix} 1 & 2 \\ 2 & 4 \end{bmatrix}\begin{bmatrix} 1 & -2 \\ 0 & 1 \end{bmatrix} = \begin{bmatrix} 3 & 6 \\ 0 & 0 \end{bmatrix}\begin{bmatrix} 1 & -2 \\ 0 & 1 \end{bmatrix} = \begin{bmatrix} 3 & 0 \\ 0 & 0 \end{bmatrix}
Step 3: Test for row equivalence by comparing reduced row echelon forms. The two forms differ, so no sequence of row operations alone turns AA into BB. The column operation is essential.
rref⁡(A)=[1200]≠[1000]=rref⁡(B)\operatorname{rref}(A) = \begin{bmatrix} 1 & 2 \\ 0 & 0 \end{bmatrix} \neq \begin{bmatrix} 1 & 0 \\ 0 & 0 \end{bmatrix} = \operatorname{rref}(B)
Answer: AA and BB are equivalent, because B=PAQB = PAQ with P=[30−21]P = \begin{bmatrix} 3 & 0 \\ -2 & 1 \end{bmatrix} and Q=[1−201]Q = \begin{bmatrix} 1 & -2 \\ 0 & 1 \end{bmatrix}. They are not row equivalent.

Why It Matters

Equivalence classifies every m×nm \times n matrix by a single number, its rank. In a linear algebra course, this shows that rank is the only property preserved when you change bases independently in the domain and in the codomain. Rank decompositions such as the SVD, used in data compression, build on this same idea.

Common Mistakes

Mistake: Treating equivalent, row equivalent, and similar as the same relation.
Correction: Row equivalence (B=PAB = PA) allows only row operations. Similarity (B=P−1APB = P^{-1}AP) uses the same matrix on both sides. Equivalence (B=PAQB = PAQ) allows independent PP and QQ, so it is the weakest of the three. Some introductory texts also say "equivalent" when they mean row equivalent, so check which definition your course uses.

Related Terms

  • Matrix — The rectangular array that gets transformed by row and column operations
  • Augmented Matrix — Row operations on it preserve solutions; column operations do not
  • Coefficient Matrix — Its rank fixes which equivalence class it belongs to
  • Column of a Matrix — Column operations, recorded by Q, act on these
  • Compatible Matrices — P, A, and Q must be compatible for PAQ to exist
  • Row Operations — Invertible row operations are used to turn one matrix into another
  • Rank (Matrix) — The one number that decides whether two matrices are equivalent
  • Similar Matrices — A stricter relation, B = P⁻¹AP, that uses one change of basis on square matrices