What Is the Transpose of a Matrix?
The transpose of an m×n matrix A is the n×m matrix Aᵀ obtained by turning rows into columns: the (i, j) entry of Aᵀ equals the (j, i) entry of A. Geometrically, the transpose reflects the matrix across its main diagonal.
For example:
A = [[1, 2, 3], → Aᵀ = [[1, 4],
[4, 5, 6]] [2, 5],
[3, 6]]
A is 2×3; Aᵀ is 3×2.
Key transpose properties:
- (Aᵀ)ᵀ = A — transposing twice returns the original
- (A + B)ᵀ = Aᵀ + Bᵀ
- (kA)ᵀ = k·Aᵀ
- (AB)ᵀ = BᵀAᵀ — order reverses
- det(Aᵀ) = det(A)
- rank(Aᵀ) = rank(A)
The reversal property (AB)ᵀ = BᵀAᵀ connects transpose to matrix multiplication — the order of factors reverses just as it does for inverses. The equality det(Aᵀ) = det(A) means you can use the determinant calculator on either a matrix or its transpose and always get the same result.
Symmetric and Skew-Symmetric Matrices
A square matrix A is symmetric if Aᵀ = A. This means A[i][j] = A[j][i] for all i, j — the matrix is identical to its reflection. Symmetric matrices arise constantly in linear algebra: covariance matrices in statistics, stiffness matrices in finite element analysis, and the matrix XᵀX in least squares.
A square matrix is skew-symmetric (or antisymmetric) if Aᵀ = −A. The diagonal entries of a skew-symmetric matrix must all be zero (since A[i][i] = −A[i][i] implies A[i][i] = 0). Cross-product matrices in 3D mechanics are skew-symmetric.
Every square matrix can be decomposed as the sum of a symmetric and a skew-symmetric part: A = (A + Aᵀ)/2 + (A − Aᵀ)/2.
Transpose and Inner Products
For column vectors u and v in ℝⁿ, the dot product equals uᵀv (treating u and v as n×1 matrices). This notation connects matrix multiplication and inner products cleanly.
The matrix product AᵀA is always symmetric and positive semi-definite. It appears in the normal equations for least squares (XᵀXβ = Xᵀy), in the computation of the covariance matrix, and in singular value decomposition.
For an orthogonal matrix Q (one whose columns form an orthonormal basis), Qᵀ = Q⁻¹. This makes orthogonal matrices especially easy to invert — just transpose. Rotation matrices and reflection matrices are orthogonal.
How to Transpose a Matrix By Hand
Write the rows of A as the columns of Aᵀ. Equivalently, write the columns of A as the rows of Aᵀ. For a square matrix, this is equivalent to reflecting across the main diagonal (top-left to bottom-right).
For small matrices, you can read off entries directly: Aᵀ[row i][col j] = A[row j][col i]. For a 3×4 matrix A, Aᵀ is 4×3 — the entry in row 2, column 3 of Aᵀ is the entry in row 3, column 2 of A.
Applications of the Transpose
Least squares regression. The normal equations (XᵀX)β = Xᵀy give the best-fit coefficients. The transpose ensures XᵀX is square and symmetric.
Orthogonal transformations. Rotation matrices R satisfy Rᵀ = R⁻¹. So to undo a rotation, just transpose — no matrix inversion needed.
Covariance matrices. The sample covariance matrix Σ = (1/n)XᵀX is symmetric, a consequence of the transpose relationship.
Graph theory. For a directed graph with adjacency matrix A, Aᵀ is the adjacency matrix of the reversed graph.
Image processing. 2D discrete cosine transforms (used in JPEG compression) rely on matrix transpose as part of the separable computation.
The Four Fundamental Subspaces
Every m×n matrix A defines four fundamental subspaces — Gilbert Strang calls these "the heart of linear algebra." Two involve A and two involve Aᵀ:
- Column space of A (image/range): all vectors Ax. Dimension = rank(A).
- Null space of A (kernel): all x with Ax = 0. Dimension = n − rank(A) (nullity).
- Row space of A = Column space of Aᵀ: span of A's rows. Dimension = rank(A).
- Left null space = Null space of Aᵀ: all y with Aᵀy = 0. Dimension = m − rank(A).
Key orthogonality: the null space of A is orthogonal to the row space of A; the left null space is orthogonal to the column space. These four subspaces pair up and fill ℝⁿ and ℝᵐ completely. To find all four, compute RREF(A) with the RREF calculator: pivot rows give the row space, free variable columns indicate null space directions.
Orthogonal Matrices: When Aᵀ = A⁻¹
A square matrix Q is orthogonal if its columns form an orthonormal set — each column has unit length, and all pairs of distinct columns are perpendicular. The defining property QᵀQ = I means Qᵀ = Q⁻¹: to invert an orthogonal matrix, just transpose it.
Rotation, reflection, and permutation matrices are all orthogonal. For a 2D rotation by angle θ:
Q = [[cos θ, −sin θ], [sin θ, cos θ]]
Qᵀ = [[cos θ, sin θ], [−sin θ, cos θ]] (rotation by −θ)
QQᵀ = [[1,0],[0,1]] = I ✓
In 3D graphics (OpenGL, WebGL, DirectX), view and model matrices are products of rotation and reflection matrices — all orthogonal. Inverting them (needed to move between coordinate spaces) requires only transposing, not full Gaussian elimination. The matrix inverse calculator handles general inversion; for orthogonal matrices, this transpose calculator gives the inverse directly.
Positive Semi-Definite Matrices: AᵀA and Applications
The product AᵀA is always symmetric and positive semi-definite: for any vector x, xᵀ(AᵀA)x = |Ax|² ≥ 0. This product appears throughout applied mathematics:
- Least squares regression: Normal equations AᵀAβ = Aᵀy give the best-fit line or hyperplane.
- Covariance matrix: Σ = (1/n)XᵀX (after centering) is symmetric by this property.
- Gram matrix: K[i][j] = xᵢᵀxⱼ; central to kernel methods and support vector machines.
- SVD: The eigenvalues of AᵀA are the squared singular values of A.
Symmetry of AᵀA follows from (AᵀA)ᵀ = AᵀAᵀᵀ = AᵀA. Use the matrix transpose calculator above to verify the symmetry of AᵀA for any A you enter.
The Gram-Schmidt Process and QR Decomposition
Gram-Schmidt converts any linearly independent set of vectors into an orthonormal basis using the transpose (via dot products as projections):
- Normalize the first vector: u₁ = v₁/|v₁|.
- For each subsequent vᵢ, subtract its projections: wᵢ = vᵢ − Σⱼ₌₁ⁱ⁻¹ (vᵢ·uⱼ)uⱼ, then normalize uᵢ = wᵢ/|wᵢ|.
This produces the QR decomposition A = QR, where Q has orthonormal columns (Qᵀ = Q⁻¹ for square Q) and R is upper triangular. In fact R = QᵀA — the transpose appears directly. QR decomposition underlies the QR algorithm for computing eigenvalues and is the standard numerical method for solving least squares problems. Use the dot product calculator to compute each projection coefficient in the Gram-Schmidt steps.
Frequently Asked Questions
What does it mean to transpose a matrix?
Transposing swaps rows and columns. The (i,j) entry of Aᵀ is the (j,i) entry of A. An m×n matrix becomes an n×m matrix.
Is the transpose the same as the inverse?
Not in general. For orthogonal matrices, Aᵀ = A⁻¹, but this is a special property. For a general matrix, Aᵀ and A⁻¹ are unrelated.
What is a symmetric matrix?
A square matrix where A = Aᵀ, meaning A[i][j] = A[j][i] for all i, j. It looks the same when reflected across the main diagonal.
Does transposing change the rank?
No. rank(Aᵀ) = rank(A). The row space of A equals the column space of Aᵀ and vice versa.
Does transposing change the determinant?
No. det(Aᵀ) = det(A). This means a matrix is singular if and only if its transpose is singular.
What is a Hermitian matrix?
A complex generalization of a symmetric matrix: Aᴴ = A, where Aᴴ is the conjugate transpose (transpose and complex-conjugate each entry). Real symmetric matrices are a special case.
Can I transpose a non-square matrix?
Yes. The transpose of an m×n matrix is always n×m. This calculator handles any rectangular matrix.
Is this free?
Yes, completely free with no account required.