Skip to content
TheChatArchiveThe Chat Archive
GLMGLM-5 / Pony Alpha· unknownHugging Face Frontier Reasoning Traces

Show that the determinant of a product of matrices equals the product of determinants.

  1. User

    Show that the determinant of a product of matrices equals the product of determinants.

  2. GLM Assistant

    The user wants a proof that det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B).

    1. Understand the Request:

    • Statement: det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B) for square matrices AA and BB of size n×nn \times n.
    • Target Audience: Likely a student or someone interested in linear algebra. The proof should be clear, rigorous enough, but accessible.
    • Possible Approaches:
      • Permutation definition: Using the Leibniz formula for determinants. This is very rigorous but computationally heavy and often messy to follow.
      • Elementary matrices: Breaking down the matrix operations. This is a very common and intuitive approach in introductory linear algebra courses.
      • Eigenvalues: det⁡(A)=∏λi\det(A) = \prod \lambda_i. This is elegant but requires AA and BB to be diagonalizable or working over an algebraically closed field (though the result holds generally). It also relies on properties of eigenvalues of products, which are usually taught after determinants.
      • Multilinearity/Exterior Algebra: The most "high-level" elegant proof using the properties of alternating multilinear maps. det⁡(AB)(e1∧⋯∧en)=(AB)e1∧⋯∧(AB)en=A(Be1∧⋯∧Ben)=det⁡(A)det⁡(B)(e1∧⋯∧en)\det(AB)(e_1 \wedge \dots \wedge e_n) = (AB)e_1 \wedge \dots \wedge (AB)e_n = A(Be_1 \wedge \dots \wedge Be_n) = \det(A)\det(B)(e_1 \wedge \dots \wedge e_n). This might be too abstract.

    2. Selecting the Best Approach:

    • Approach A (Permutation): Good for strict definition-based courses.
    • Approach B (Elementary Matrices): Good for intuition regarding row operations. It connects det⁡\det to invertibility.
      • Case 1: AA is singular (not invertible). Then ABAB is singular. LHS = 0, RHS = 0. Holds.
      • Case 2: AA is invertible. AA is a product of elementary matrices Ek…E1E_k \dots E_1.
      • Show det⁡(EX)=det⁡(E)det⁡(X)\det(EX) = \det(E)\det(X) for any elementary EE and any square XX.
      • Apply repeatedly.
    • Approach C (Properties of Multilinear Maps): A nice middle ground using the definition of determinant as a unique function satisfying specific properties (multilinear, alternating, normed).

    3. Developing the Proof (Drafting - Method: Elementary Matrices + Invertibility): This is usually the standard textbook proof.

    • Step 1: Consider the function f(B)=det⁡(AB)det⁡(A)f(B) = \frac{\det(AB)}{\det(A)}. (Wait, this assumes det⁡(A)≠0\det(A) \neq 0. Better to start with invertibility cases).
    • Case 1: AA is not invertible.
      • det⁡(A)=0\det(A) = 0.
      • If AA is not invertible, then ABAB is not invertible (since if ABAB had an inverse CC, then (AB)C=I  ⟹  A(BC)=I(AB)C = I \implies A(BC) = I, so AA would have an inverse).
      • Therefore det⁡(AB)=0\det(AB) = 0.
      • 0=0⋅det⁡(B)0 = 0 \cdot \det(B). Equation holds.
    • Case 2: AA is invertible.
      • Any invertible matrix can be written as a product of elementary matrices: A=E1E2…EkA = E_1 E_2 \dots E_k.
      • We need to show that for an elementary matrix EE and any square matrix MM, det⁡(EM)=det⁡(E)det⁡(M)\det(EM) = \det(E)\det(M).
      • Actually, it's easier to use the effect of row operations on the determinant.
        • If EE swaps rows, det⁡(EM)=−det⁡(M)\det(EM) = -\det(M). We know det⁡(E)=−1\det(E) = -1. So det⁡(EM)=det⁡(E)det⁡(M)\det(EM) = \det(E)\det(M).
        • If EE multiplies a row by kk, det⁡(EM)=kdet⁡(M)\det(EM) = k\det(M). We know det⁡(E)=k\det(E) = k. So det⁡(EM)=det⁡(E)det⁡(M)\det(EM) = \det(E)\det(M).
        • If EE adds a multiple of a row to another, det⁡(EM)=det⁡(M)\det(EM) = \det(M). We know det⁡(E)=1\det(E) = 1. So det⁡(EM)=det⁡(E)det⁡(M)\det(EM) = \det(E)\det(M).
      • Apply this recursively:
        • det⁡(AB)=det⁡((E1…Ek)B)=det⁡(E1(E2…EkB))\det(AB) = \det((E_1 \dots E_k)B) = \det(E_1 (E_2 \dots E_k B)).
        • By induction/stacking: det⁡(AB)=det⁡(E1)det⁡(E2)…det⁡(Ek)det⁡(B)\det(AB) = \det(E_1)\det(E_2)\dots\det(E_k)\det(B).
        • Also det⁡(A)=det⁡(E1…Ek)=det⁡(E1)…det⁡(Ek)\det(A) = \det(E_1 \dots E_k) = \det(E_1)\dots\det(E_k).
        • Therefore, det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B).

    4. Developing the Proof (Drafting - Method: Permutation/Leibniz Formula):

    • Maybe just outline this or put it in a "formal" section. It's messy to type out fully without looking like a wall of symbols.
    • det⁡(C)=∑σ∈Snsgn(σ)∏i=1nci,σ(i)\det(C) = \sum_{\sigma \in S_n} \text{sgn}(\sigma) \prod_{i=1}^n c_{i, \sigma(i)}.
    • Let C=ABC = AB. Then cik=∑jaijbjkc_{ik} = \sum_{j} a_{ij} b_{jk}.
    • Substitute, expand, rearrange sums... it works, but it's painful to read.

    5. Refining the Strategy: I will provide the proof based on Elementary Matrices and Invertibility. It connects the algebraic manipulation to the geometric/row-operation intuition. It's the most pedagogical approach for general audiences.

    • Structure:
      1. State the theorem.
      2. Handle the Singular Case (AA is not invertible). Show 0=00=0.
      3. Handle the Invertible Case.
        • Decompose AA into elementary matrices.
        • Show the property holds for a single elementary matrix (det⁡(EB)=det⁡(E)det⁡(B)\det(EB) = \det(E)\det(B)).
        • Extend to the product of elementary matrices.
      4. Conclude.

    6. Drafting the Content (Mental Check):

    • Singular Case: If AA is singular, columns are linearly dependent. Columns of ABAB are linear combinations of columns of AA. So columns of ABAB are linearly dependent. det⁡(AB)=0\det(AB) = 0. det⁡(A)=0\det(A) = 0. Equality holds.
    • Invertible Case:
      • A=E1E2…EkA = E_1 E_2 \dots E_k.
      • det⁡(AB)=det⁡(E1…EkB)\det(AB) = \det(E_1 \dots E_k B).
      • We need the lemma: Row operations on BB multiply determinant by the scalar associated with the row operation.
        • Lemma 1: Swapping rows -> times -1.
        • Lemma 2: Scaling row -> times scalar.
        • Lemma 3: Adding multiple of row -> times 1.
      • These lemmas correspond exactly to det⁡(E)\det(E) for the elementary matrix EE.
      • So det⁡(EB)=det⁡(E)det⁡(B)\det(EB) = \det(E)\det(B).
      • By induction det⁡(E1…EkB)=det⁡(E1)…det⁡(Ek)det⁡(B)\det(E_1 \dots E_k B) = \det(E_1)\dots\det(E_k)\det(B).
      • Since det⁡(A)=det⁡(E1)…det⁡(Ek)\det(A) = \det(E_1)\dots\det(E_k), we are done.

    7. Alternative "Shortcut" Proof using Multilinearity: Define DA(B)=det⁡(AB)D_A(B) = \det(AB). Show DAD_A is alternating and multilinear. DAD_A is a multiple of determinant. DA(I)=det⁡(AI)=det⁡(A)D_A(I) = \det(AI) = \det(A). So DA(B)=det⁡(A)det⁡(B)D_A(B) = \det(A)\det(B). Critique: This is elegant but requires assuming the uniqueness theorem of determinants (that a normalized alternating multilinear map is unique). I will mention this as an "elegant" approach but stick to the elementary matrix one as the primary explanation because it's more constructive.

    8. Final Polish: Make sure notation is clear (n×nn \times n matrices). Use bolding for steps.

    Let's write out the Elementary Matrix proof. Theorem: det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B). Proof: Case 1: det⁡(A)=0\det(A) = 0. AA is singular. Ax=0Ax=0 has non-trivial solution. Then (AB)y=A(By)(AB)y = A(By). If AA is singular, the column space of ABAB is contained in the column space of AA. Since rank(A)<n\text{rank}(A) < n, rank(AB)<n\text{rank}(AB) < n. ABAB is singular. det⁡(AB)=0\det(AB)=0. 0=0⋅det⁡(B)0 = 0 \cdot \det(B). True.

    Case 2: det⁡(A)≠0\det(A) \neq 0. AA is invertible. A=E1E2…EkA = E_1 E_2 \dots E_k where EiE_i are elementary matrices. Need to show det⁡(EX)=det⁡(E)det⁡(X)\det(EX) = \det(E)\det(X). This relies on the effect of pre-multiplication on the determinant.

    • Swap rows: det changes sign. det⁡(E)=−1\det(E) = -1.
    • Scale row: det scales. det⁡(E)=k\det(E) = k.
    • Add row: det unchanged. det⁡(E)=1\det(E) = 1. In all 3 cases, det⁡(EB)=det⁡(E)det⁡(B)\det(EB) = \det(E)\det(B). Extend via induction.

    This looks solid. I will present this clearly.

    Self-Correction during drafting: Some might find the singular proof regarding column spaces slightly abstract. I can stick to the simple logic: If AA is singular, it's not invertible. If AA is not invertible, ABAB cannot be invertible (because if (AB)−1(AB)^{-1} existed, then A(B(AB)−1)=IA(B(AB)^{-1}) = I, implying AA has an inverse). Thus det⁡(AB)=0\det(AB) = 0. This is simpler.

    Final plan:

    1. Intro.
    2. Case 1: Singular (using invertibility argument).
    3. Case 2: Invertible (using elementary matrices).
    4. Summary. To prove that the determinant of a product of matrices equals the product of their determinants, i.e., det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B), we can use a proof that splits the problem into two cases: one where the matrix AA is singular (not invertible) and one where AA is invertible.

    Let AA and BB be n×nn \times n square matrices.

    Case 1: AA is Singular (Not Invertible)

    If AA is singular, then by definition, det⁡(A)=0\det(A) = 0.

    We need to show that det⁡(AB)=0\det(AB) = 0.

    1. Since AA is singular, the linear transformation defined by AA is not onto (it does not map the vector space onto itself). Consequently, the columns of AA are linearly dependent.
    2. The product matrix ABAB consists of linear combinations of the columns of AA. Specifically, the columns of ABAB are formed by multiplying AA by the column vectors of BB.
    3. Since the columns of AA are linearly dependent, any linear combination of them will also be linearly dependent. Therefore, the columns of ABAB are linearly dependent.
    4. A matrix with linearly dependent columns has a determinant of 00.
    5. Therefore, det⁡(AB)=0\det(AB) = 0.

    Since det⁡(A)=0\det(A) = 0 and det⁡(AB)=0\det(AB) = 0, the equation holds: det⁡(AB)=0=0⋅det⁡(B)=det⁡(A)det⁡(B)\det(AB) = 0 = 0 \cdot \det(B) = \det(A)\det(B)


    Case 2: AA is Invertible

    If AA is invertible, we can use the properties of Elementary Matrices. Any invertible matrix AA can be decomposed into a product of elementary matrices: A=E1E2…EkA = E_1 E_2 \dots E_k where each EiE_i is an elementary matrix corresponding to a specific row operation.

    The determinant of an elementary matrix depends on the operation:

    • Row Swap (PijP_{ij}): det⁡(Pij)=−1\det(P_{ij}) = -1.
    • Row Scaling (Di(c)D_i(c)): det⁡(Di(c))=c\det(D_i(c)) = c.
    • Row Addition (Tij(c)T_{ij}(c)): det⁡(Tij(c))=1\det(T_{ij}(c)) = 1.

    We know that performing a row operation on a matrix MM multiplies the determinant of MM by the determinant of the corresponding elementary matrix. That is, for any elementary matrix EE and square matrix MM: det⁡(EM)=det⁡(E)det⁡(M)\det(EM) = \det(E)\det(M)

    We can now prove the theorem for the invertible case using induction or recursive substitution:

    1. Consider the determinant of the product ABAB: det⁡(AB)=det⁡((E1E2…Ek)B)\det(AB) = \det((E_1 E_2 \dots E_k)B)
    2. Group the matrices as (E1(E2…EkB))(E_1 (E_2 \dots E_k B)). Using the property of elementary matrices, we can pull E1E_1 out: det⁡(AB)=det⁡(E1)det⁡((E2…Ek)B)\det(AB) = \det(E_1) \det((E_2 \dots E_k)B)
    3. Repeat this process for E2E_2, then E3E_3, and so on: det⁡(AB)=det⁡(E1)det⁡(E2)…det⁡(Ek)det⁡(B)\det(AB) = \det(E_1) \det(E_2) \dots \det(E_k) \det(B)
    4. We also know that the determinant of a product of matrices is the product of the determinants (for elementary matrices). So, for AA itself: det⁡(A)=det⁡(E1E2…Ek)=det⁡(E1)det⁡(E2)…det⁡(Ek)\det(A) = \det(E_1 E_2 \dots E_k) = \det(E_1)\det(E_2)\dots\det(E_k)
    5. Substitute this back into the equation from step 3: det⁡(AB)=(det⁡(E1)det⁡(E2)…det⁡(Ek))det⁡(B)\det(AB) = (\det(E_1)\det(E_2)\dots\det(E_k)) \det(B) det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B)

    Conclusion

    Since the theorem holds for both the singular case and the invertible case, we conclude that for any two n×nn \times n matrices AA and BB: det⁡(AB)=det⁡(A)det⁡(B)\det(AB) = \det(A)\det(B)

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.