Two-sided unimodular equivalence of matrices #
Matrices related by L * A * R with L and R in SL. Nothing here assumes a Smith normal
form or a divisibility chain along a diagonal, so these facts sit below that theory rather
than inside it, and hold over an arbitrary finite index type. The corresponding statement for
merely invertible factors is Matrix.GeneralLinearGroup.inv_mul_mul_inv_of_mul_mul_eq, in
TauCeti.LinearAlgebra.Matrix.GeneralLinearGroup.Equivalence.
Main results #
Matrix.prod_eq_det_of_mul_mul_eq_diagonal: the product of a diagonalisation's diagonal entries is the determinant.Matrix.exists_SL_mul_mul_eq_of_mul_mul_eq: two matrices carried to a common value bySL-transformations are themselvesSL-equivalent.Matrix.exists_eq_diagonal_mul_iff: a matrixAwithdet A = ∏ i, d ia non-zero-divisor lies indiagonal d * SLexactly whend idivides every entry of thei-th row ofA.
The product of a diagonalisation's diagonal entries is the determinant. Both SL
factors have determinant 1, so taking determinants through L * A * R = diagonal d leaves
the product of the diagonal.
No divisibility chain is assumed, so d need not be the invariant factors.
Matrices sharing an SL-transform are SL-equivalent. If L_A A R_A = L_B B R_B
then L_B⁻¹ L_A and R_A R_B⁻¹ carry A to B.
Pure group algebra: nothing is assumed about the common value, which need not be diagonal. Callers holding two diagonalisations with equal diagonals compose them into this single hypothesis.
Right SL-cosets of a diagonal matrix. If det A = ∏ i, d i is a left non-zero-divisor,
then A = diagonal d * g for some g ∈ SL exactly when d i divides every entry of the i-th
row of A.