• Practice
  • About 20 min
  • 5 practice questions

Properties of inverses and adjoints, orthogonal matrices and cryptography

Use the properties of A⁻¹ and adj A (reversal law, double inverse, |adj A| = |A|ⁿ⁻¹ and more), recognise orthogonal matrices from rotation of axes, and encode and decode messages with a matrix.

If something looks wrong, please tell us.

By the end of this lesson you can

  • Use |A⁻¹| = 1/|A|, (Aᵀ)⁻¹ = (A⁻¹)ᵀ and (λA)⁻¹ = (1/λ)A⁻¹.
  • Apply the cancellation laws, the reversal law (AB)⁻¹ = B⁻¹A⁻¹ and the law of double inverse.
  • Use the results on adj A: |adj A| = |A|ⁿ⁻¹, adj(adj A) = |A|ⁿ⁻²A, adj(λA) = λⁿ⁻¹ adj A and adj(AB) = (adj B)(adj A).
  • Test whether a matrix is orthogonal and use A⁻¹ = Aᵀ for it.
  • Encode and decode a message with a non-singular matrix and its inverse.

Once the inverse is defined, a family of rules lets us work with inverses and adjoints without recomputing them from scratch. These rules are favourite board-exam material, and they also explain two applications in the Tamil Nadu textbook: rotating coordinate axes (orthogonal matrices) and sending secret messages (cryptography).

Properties of the inverse

Let A be non-singular of order n, so AA⁻¹ = A⁻¹A = Iₙ. Then:

  • |A⁻¹| = 1/|A|. Taking determinants of AA⁻¹ = Iₙ gives |A||A⁻¹| = 1.
  • (Aᵀ)⁻¹ = (A⁻¹)ᵀ. Transposing AA⁻¹ = Iₙ and using (PQ)ᵀ = QᵀPᵀ gives (A⁻¹)ᵀAᵀ = Iₙ, and similarly Aᵀ(A⁻¹)ᵀ = Iₙ.
  • (λA)⁻¹ = (1/λ)A⁻¹ for a non-zero scalar λ, because (λA)((1/λ)A⁻¹) = AA⁻¹ = Iₙ.
  • Left cancellation: if AB = AC then B = C (pre-multiply by A⁻¹). Right cancellation: if BA = CA then B = C (post-multiply by A⁻¹).
  • Reversal law: if A and B are non-singular of the same order, AB is non-singular and (AB)⁻¹ = B⁻¹A⁻¹, because (AB)(B⁻¹A⁻¹) = A(BB⁻¹)A⁻¹ = AA⁻¹ = Iₙ.
  • Double inverse: A⁻¹ is non-singular and (A⁻¹)⁻¹ = A.

Cancellation needs A to be non-singular. Take A = [1 1; 1 1] (singular), B = [1 0; 0 0] and C = [0 1; 1 −1]. Then AB = [1 0; 1 0] and AC = [1·0 + 1·1, 1·1 + 1·(−1); 1·0 + 1·1, 1·1 + 1·(−1)] = [1 0; 1 0], so AB = AC although B ≠ C.

Worked example 1: checking the reversal law

Let A = [2 1; 1 1] and B = [1 2; 0 1]. Both have determinant 1, so A⁻¹ = [1 −1; −1 2] and B⁻¹ = [1 −2; 0 1]. Working: AB = [2·1 + 1·0, 2·2 + 1·1; 1·1 + 1·0, 1·2 + 1·1] = [2 5; 1 3], with |AB| = 6 − 5 = 1, so (AB)⁻¹ = [3 −5; −1 2]. Now B⁻¹A⁻¹ = [1·1 + (−2)(−1), 1·(−1) + (−2)·2; 0·1 + 1·(−1), 0·(−1) + 1·2] = [3 −5; −1 2]. They agree. But A⁻¹B⁻¹ = [1 −3; −1 4], which is different: the order really does reverse.

Properties of the adjoint

For a non-singular matrix A of order n, from adj A = |A|A⁻¹ we obtain:

Results on adj A (A non-singular of order n)
ResultReason in one line
(adj A)⁻¹ = adj(A⁻¹) = (1/|A|)AInvert adj A = |A|A⁻¹; or replace A by A⁻¹ in adj A = |A|A⁻¹
|adj A| = |A|ⁿ⁻¹Take determinants in A(adj A) = |A|Iₙ: |A||adj A| = |A|ⁿ
adj(adj A) = |A|ⁿ⁻²AApply A(adj A) = |A|I to the matrix adj A, then pre-multiply by A
|adj(adj A)| = |A|^((n−1)²)Take determinants of the previous line
adj(λA) = λⁿ⁻¹ adj Aadj(λA) = |λA|(λA)⁻¹ = λⁿ|A|(1/λ)A⁻¹
(adj A)ᵀ = adj(Aᵀ)Transpose adj A = |A|A⁻¹ and use |Aᵀ| = |A|
adj(AB) = (adj B)(adj A)|AB|(AB)⁻¹ = |B|B⁻¹·|A|A⁻¹

Two consequences are worth remembering. If A is non-singular of odd order, |adj A| = |A|ⁿ⁻¹ with n − 1 even, so |adj A| is positive. If A is of order 3, then |adj A| = |A|², so |A| = ±√|adj A|, and from adj(adj A) = |A|A we get A = ±(1/√|adj A|) adj(adj A) and A⁻¹ = ±(1/√|adj A|) adj A.

Worked example 2: finding A from adj A

A 3 × 3 matrix A has adj A = [6 −5 1; −6 8 −2; 2 −3 1]. Find A. Working: |adj A| = 6(8·1 − (−2)(−3)) − (−5)((−6)·1 − (−2)·2) + 1((−6)(−3) − 8·2) = 6(2) + 5(−2) + 1(2) = 4. Next find the cofactors of B = adj A: B₁₁ = 2, B₁₂ = 2, B₁₃ = 2, B₂₁ = 2, B₂₂ = 4, B₂₃ = 8, B₃₁ = 2, B₃₂ = 6, B₃₃ = 18. Transposing, adj(adj A) = [2 2 2; 2 4 6; 2 8 18]. So A = ±(1/√4)[2 2 2; 2 4 6; 2 8 18] = ±[1 1 1; 1 2 3; 1 4 9]. (Compare lesson 1: this is the matrix we started from, or its negative.)

Orthogonal matrices and rotation of axes

Rotate the x- and y-axes about the origin through an angle θ. If a point P has old coordinates (x, y) and new coordinates (X, Y), trigonometry on the figure gives x = X cos θ − Y sin θ and y = X sin θ + Y cos θ, that is [x; y] = W[X; Y] with W = [cos θ −sin θ; sin θ cos θ]. Multiplying out shows WWᵀ = [cos²θ + sin²θ, 0; 0, sin²θ + cos²θ] = I₂, so W⁻¹ = Wᵀ and the new coordinates are X = x cos θ + y sin θ, Y = −x sin θ + y cos θ. This transformation is used in computer graphics.

A square matrix A is orthogonal if AAᵀ = AᵀA = I. Equivalently, A is non-singular and A⁻¹ = Aᵀ.

— Orthogonal matrix

Worked example 3: rotation and an orthogonal test

(i) The axes are rotated through 60°. Find the new coordinates of P(2, 2√3). Working: cos 60° = 1/2 and sin 60° = √3/2, so X = 2(1/2) + 2√3(√3/2) = 1 + 3 = 4 and Y = −2(√3/2) + 2√3(1/2) = −√3 + √3 = 0. So P is (4, 0) in the new system: it lies on the new X-axis, and the distance OP = √(4 + 12) = 4 is unchanged, as a rotation should leave it.

(ii) Is A = (1/3)[1 2 2; 2 1 −2; 2 −2 1] orthogonal? Working: AAᵀ has (i, j) entry equal to (row i)·(row j)/9. Row 1·row 1 = (1 + 4 + 4)/9 = 1, row 1·row 2 = (2 + 2 − 4)/9 = 0, row 1·row 3 = (2 − 4 + 2)/9 = 0, row 2·row 2 = (4 + 1 + 4)/9 = 1, row 2·row 3 = (4 − 2 − 2)/9 = 0, row 3·row 3 = (4 + 4 + 1)/9 = 1. So AAᵀ = I₃: A is orthogonal and A⁻¹ = Aᵀ (here Aᵀ = A, since A is symmetric).

Matrices in cryptography

Cryptography is communication that keeps a message hidden from others. Encryption turns plain text into an unreadable coded form; decryption turns it back. Sender and receiver share a secret key. One key is a non-singular encoding matrix A: the sender post-multiplies each block of the message by A, and the receiver post-multiplies each coded block by the decoding matrix A⁻¹.

Sending a message with a matrix key

  1. Convert letters to numbers A = 1, B = 2, ..., Z = 26 and a blank space = 0.
  2. Cut into row matrices Blocks of the key's order; pad the last block with zeros.
  3. Encode Post-multiply each block by the encoding matrix A and send.
  4. Decode The receiver post-multiplies each block by A⁻¹.
  5. Read the message Turn the numbers back into letters.

Worked example 4: encoding and decoding MATHS

Use the encoding matrix A = [2 1; 1 1], with |A| = 1 and A⁻¹ = [1 −1; −1 2]. Working: M = 13, A = 1, T = 20, H = 8, S = 19, so the blocks are [13 1], [20 8], [19 0] (one zero added to complete the last block). Encode: [13 1]A = [13·2 + 1·1, 13·1 + 1·1] = [27 14]; [20 8]A = [40 + 8, 20 + 8] = [48 28]; [19 0]A = [38 19]. The coded message is [27 14] [48 28] [38 19]. Decode: [27 14]A⁻¹ = [27 − 14, −27 + 28] = [13 1]; [48 28]A⁻¹ = [48 − 28, −48 + 56] = [20 8]; [38 19]A⁻¹ = [38 − 19, −38 + 38] = [19 0]. Reading 13, 1, 20, 8, 19, 0 gives MATHS.

Key takeaways: inverse and transpose both reverse the order of a product; cancellation is allowed only by a non-singular matrix; |adj A| = |A|ⁿ⁻¹ and adj(adj A) = |A|ⁿ⁻²A; an orthogonal matrix has A⁻¹ = Aᵀ, like the rotation matrix; and a non-singular matrix with its inverse works as a key for coding messages.

Check your understanding

  1. 1. A is a 3 × 3 matrix with |A| = 4. What is |adj A|?

    1. 4
    2. 64
    3. 16
    4. 12
    Show answer

    Answer: C. |adj A| = |A|ⁿ⁻¹ with n = 3, so |adj A| = 4² = 16.

  2. 2. A and B are non-singular matrices of the same order. Which expression equals (AB)⁻¹?

    1. A⁻¹B⁻¹
    2. BA
    3. (BA)⁻¹
    4. B⁻¹A⁻¹
    Show answer

    Answer: D. By the reversal law (AB)⁻¹ = B⁻¹A⁻¹, since (AB)(B⁻¹A⁻¹) = A(BB⁻¹)A⁻¹ = I. (BA)⁻¹ equals A⁻¹B⁻¹, which is different in general because matrix multiplication is not commutative.

  3. 3. A is a square matrix of order 3. Which is equal to adj(3A)?

    1. 9 adj A
    2. 3 adj A
    3. 27 adj A
    4. (1/3) adj A
    Show answer

    Answer: A. adj(λA) = λⁿ⁻¹ adj A. With λ = 3 and n = 3, adj(3A) = 3² adj A = 9 adj A.

  4. 4. Which of these matrices is orthogonal?

    1. [1 1; 0 1]
    2. [0 1; −1 0]
    3. [2 0; 0 2]
    4. [1 1; 1 −1]
    Show answer

    Answer: B. For [0 1; −1 0], AAᵀ = [0 1; −1 0][0 −1; 1 0] = [1 0; 0 1] = I₂. For [2 0; 0 2], AAᵀ = 4I₂. For [1 1; 1 −1] the rows are perpendicular but each has length √2, so AAᵀ = 2I₂. For [1 1; 0 1], AAᵀ = [2 1; 1 1].

  5. 5. With encoding matrix A = [2 1; 1 1] (so A⁻¹ = [1 −1; −1 2]), a coded block [43 24] is received. What was the original block?

    1. [19 5], that is the letters SE
    2. [5 19], that is the letters ES
    3. [110 67]
    4. [24 43]
    Show answer

    Answer: A. Decode by post-multiplying by A⁻¹: [43 24][1 −1; −1 2] = [43·1 + 24·(−1), 43·(−1) + 24·2] = [19, 5]. 19 is S and 5 is E. [110 67] is what you get by multiplying by A again, which encodes instead of decoding.

Privacy · Terms · Refunds · Delivery