Home / Courses / GATE Physics / Practice / Hermitian and unitary matrices, diagonalisation

Hermitian and unitary matrices, diagonalisation

Mathematical methods · 78 questions and templates for GATE Physics, in order from warm-up to top rank. Check each answer, then read the full solution.

New to this topic? Read the lesson first: Hermitian and unitary matrices, diagonalisation.

Loading questions… (they need JavaScript; nothing you type is sent anywhere)

Written questions with solutions

The same questions as a list, so you can print them or read without the practice box. Templates (fresh numbers) appear only in the practice box above.

  1. MCQWarm-up

    Which of these numbers can be an eigenvalue of a unitary matrix?
    1. 2
    2. (1+i)/2
    3. 1+i
    4. 0
    Answer and solution

    Answer: B

    If Uv→=λv→ then ‖Uv→‖=|λ|‖v→‖, and a unitary matrix preserves lengths, so |λ|=1. Only (1+i)/2=eiπ/4 has modulus 1; |1+i|=2.

    Why people get this wrong: 1+i has the right angle but the wrong length.

  2. NATWarm-up

    A 4×4 Hermitian matrix H has eigenvalues 3,−1,2 and 0. The value of tr(H2) is ______.
    Answer and solution

    Answer: 14

    H2 has eigenvalues λ2: 9,1,4,0. The trace is their sum, 14.

    Why people get this wrong: Squaring the trace, (3−1+2+0)2=16, is not the trace of the square.

  3. MCQWarm-up

    For A=(1ii1), which statement is true?
    1. A is Hermitian
    2. A is anti-Hermitian
    3. A/2 is unitary
    4. A is nilpotent
    Answer and solution

    Answer: C

    A†=(1−i−i1), which is neither A nor −A. But AA†=(1+1−i+ii−i1+1)=2I, so (A/2)(A/2)†=I: unitary. A is symmetric, not Hermitian; for complex matrices the two are different.

    Why people get this wrong: Symmetric (AT=A) is not Hermitian (A†=A) once entries are complex.

  4. MCQWarm-up

    If U is unitary and H is Hermitian, the matrix U†HU is
    1. unitary
    2. Hermitian with eigenvalues the complex conjugates of those of H
    3. always diagonal
    4. Hermitian with the same eigenvalues as H
    Answer and solution

    Answer: D

    (U†HU)†=U†H†U=U†HU: Hermitian. Since U†=U−1 it is similar to H, so it has the same eigenvalues (real, so conjugating changes nothing). It is diagonal only for the right choice of U.

    Why people get this wrong: A unitary change of basis keeps the spectrum; only one special U makes the matrix diagonal.

  5. MCQWarm-up

    Two Hermitian matrices can be diagonalised by the same unitary matrix if and only if they
    1. commute
    2. have the same trace
    3. are multiples of each other
    4. are both invertible
    Answer and solution

    Answer: A

    If A=UD1U† and B=UD2U†, then AB=UD1D2U†=UD2D1U†=BA because diagonal matrices commute. Conversely, commuting Hermitian matrices share a complete orthonormal set of eigenvectors. In quantum mechanics: commuting observables have a common eigenbasis.

    Why people get this wrong: Being multiples of each other is enough but far from necessary.

  6. NATWarm-up

    A 2×2 anti-Hermitian matrix (A†=−A) has trace i and one eigenvalue 3i. The imaginary part of its other eigenvalue is ______.
    Answer and solution

    Answer: -2

    The eigenvalues add to the trace: 3i+λ=i, so λ=−2i, imaginary part −2. (Anti-Hermitian eigenvalues are purely imaginary, since iA is Hermitian.)

    Why people get this wrong: Answering +2 by dropping the sign.

  7. MCQWarm-up

    The eigenvalues of (250−3) are
    1. 2 and 5
    2. 2 and −3
    3. 5 and −3
    4. −1 and 5
    Answer and solution

    Answer: B

    For a triangular matrix, det⁡(A−λI)=(2−λ)(−3−λ), so the eigenvalues are the diagonal entries, 2 and −3. Check: their sum −1 is the trace and their product −6 is the determinant.

    Why people get this wrong: The off-diagonal 5 does not appear among the eigenvalues of a triangular matrix.

  8. MCQStandard

    The eigenvalues of the permutation matrix P=(010001100) are (ω=e2πi/3)
    1. 1,1,1
    2. 1,ω,ω2
    3. 1,−1,0
    4. 0,0,0
    Answer and solution

    Answer: B

    Moving each entry one place round three times brings it back, so P3=I and every eigenvalue obeys λ3=1. P is unitary with trace 0 and determinant 1; the cube roots of unity 1,ω,ω2 add to 0 and multiply to 1, as needed. (Directly: det⁡(P−λI)=1−λ3.)

    Why people get this wrong: 1,1,1 would need P=I (a unitary matrix with only eigenvalue 1 is the identity).

  9. NATStandard

    With nˆ=13(1,2,2) and σ→ the Pauli matrices, the value of det⁡(2I+nˆ⋅σ→) is ______.
    Answer and solution

    Answer: 3

    (nˆ⋅σ→)2=|nˆ|2I=I and the trace is 0, so its eigenvalues are +1 and −1. Then 2I+nˆ⋅σ→ has eigenvalues 3 and 1, and determinant 3.

    Why people get this wrong: Writing out the matrix works too, but the eigenvalue route takes ten seconds.

  10. MSQStandard

    A matrix is normal if AA†=A†A. Which statements are true?
    1. Every normal matrix can be diagonalised by a unitary matrix
    2. Eigenvectors of a normal matrix belonging to different eigenvalues are orthogonal
    3. Every normal matrix has real eigenvalues
    4. Hermitian, anti-Hermitian and unitary matrices are all normal
    Answer and solution

    Answer: A, B, D

    The spectral theorem holds exactly for normal matrices: A=UDU†, so the eigenvectors (columns of U) are orthonormal. Each of H†=H, A†=−A, U†=U−1 makes A commute with A†. Eigenvalues are real only for the Hermitian ones; a unitary matrix like iI is normal with eigenvalue i.

    Why people get this wrong: Normality gives orthogonal eigenvectors, not real eigenvalues.

  11. NATStandard

    The difference between the two eigenvalues of the Hermitian matrix (02i−2i3) is ______.
    Answer and solution

    Answer: 5

    Trace =3, determinant =0⋅3−(2i)(−2i)=−4. So λ2−3λ−4=0, λ=4 or −1, difference 5. Quick formula: the difference is (tr)2−4det⁡=9+16=5.

    Why people get this wrong: (2i)(−2i)=−4i2=+4, so the determinant is −4, not +4.

  12. MCQStandard

    Which unitary matrix U makes U†σxU diagonal?
    1. (100−1)
    2. (0110)
    3. 12(111−1)
    4. 12(1ii1)
    Answer and solution

    Answer: C

    The columns of U must be the normalised eigenvectors of σx: (1,1)/2 for +1 and (1,−1)/2 for −1. That is the Hadamard matrix, and U†σxU=σz. The first two options commute with nothing useful: σzσxσz=−σx and σxσxσx=σx, still off-diagonal.

    Why people get this wrong: The last option is unitary, but its columns (1,i) and (i,1) are not eigenvectors of σx.

  13. MCQStandard

    The matrix exp⁡(iπ2σx) equals
    1. σx
    2. iσx
    3. −I
    4. I
    Answer and solution

    Answer: B

    Because σx2=I, the series splits into even and odd parts: eiθσx=Icos⁡θ+iσxsin⁡θ. At θ=π/2: iσx.

    Why people get this wrong: Forgetting the factor i from the odd terms gives σx.

  14. NATStandard

    For A=(0ln⁡2ln⁡20), the value of tr(eA) is ______ (to two decimal places).
    Answer and solution

    Answer: 2.5 (±0.01)

    A has eigenvalues ±ln⁡2, so eA has eigenvalues eln⁡2=2 and e−ln⁡2=12. Trace =2.5.

    Why people get this wrong: Exponentiating entry by entry gives tr=2 (e0+e0), which is wrong.

  15. MSQStandard

    Which of these matrices can be diagonalised (over the complex numbers)?
    1. (1102)
    2. (2102)
    3. (01−10)
    4. (0010)
    Answer and solution

    Answer: A, C

    Distinct eigenvalues always give a full set of eigenvectors: the first has 1 and 2, the third has ±i. The second has eigenvalue 2 twice but A−2I has rank 1, so only one eigenvector. The fourth is nilpotent and non-zero: eigenvalue 0 twice with one eigenvector, so it is not diagonalisable.

    Why people get this wrong: Complex eigenvalues are no obstacle over ℂ; repeated eigenvalues with too few eigenvectors are.

  16. MCQStandard

    A real 2×2 orthogonal matrix has determinant −1. Its eigenvalues are
    1. e±iθ for some θ
    2. 1 and 1
    3. i and −i
    4. +1 and −1
    Answer and solution

    Answer: D

    Eigenvalues have modulus 1 and multiply to −1. A complex pair e±iθ multiplies to +1, so both must be real: +1 and −1. Geometrically it is a reflection: the mirror line is kept (+1), the perpendicular direction is flipped (−1).

    Why people get this wrong: e±iθ belongs to rotations, which have determinant +1.

  17. NATStandard

    H=2P1+5P2 with projectors P1=12(1111) and P2=12(1−1−11). The top-right entry of the positive square root H is ______ (to three decimal places).
    Answer and solution

    Answer: -0.411 (±0.003)

    This is a spectral decomposition (P1P2=0, P1+P2=I), so f(H)=f(2)P1+f(5)P2. Then H=2P1+5P2 and its top-right entry is 12(2−5)=12(1.414−2.236)≈−0.411.

    Why people get this wrong: Taking square roots of the entries of H is not H.

  18. NATStandard

    A 3×3 matrix A has eigenvalues 1, 2 and −1. The value of det⁡(A2+I) is ______.
    Answer and solution

    Answer: 20

    If Ax→=λx→ then (A2+I)x→=(λ2+1)x→, so A2+I has eigenvalues 2,5,2. The determinant is their product, 20.

    Why people get this wrong: det⁡(A2)+det⁡(I)=4+1=5: the determinant is not additive.

  19. MCQStandard

    For A=(1234), the Cayley-Hamilton theorem gives
    1. A2=5A−2I
    2. A2=4A+2I
    3. A2=−5A+2I
    4. A2=5A+2I
    Answer and solution

    Answer: D

    A 2×2 matrix satisfies A2−(trA)A+(det⁡A)I=0. Here trA=5 and det⁡A=4−6=−2, so A2−5A−2I=0, i.e. A2=5A+2I. Check: A2=(7101522)=5A+2I.

    Why people get this wrong: Sign slip with the negative determinant gives 5A−2I.

  20. MSQStandard

    B=P−1AP for an invertible P. Which properties must A and B share?
    1. The same eigenvalues
    2. The same eigenvectors
    3. The same rank
    4. If A is symmetric, so is B
    Answer and solution

    Answer: A, C

    det⁡(B−λI)=det⁡(P−1(A−λI)P)=det⁡(A−λI): same characteristic polynomial, so same eigenvalues (and trace and determinant). Multiplying by invertible matrices does not change rank. But if Ax→=λx→ then B(P−1x→)=λP−1x→: the eigenvectors are transformed by P−1, not shared. Symmetry survives only for orthogonal P (P−1=P𝖳): for example A=diag(1,2) and the shear P=(1101) give P−1AP=(1−102), which is not symmetric.

    Why people get this wrong: Similar matrices describe the same map in different bases: basis-free things (eigenvalues, rank) agree; components (eigenvector entries, symmetry) need not.

  21. NATExam level

    The circulant matrix (123312231) has one real eigenvalue and two complex ones. The product of the two complex eigenvalues is ______.
    Answer and solution

    Answer: 3

    Each row sums to 6, so (1,1,1) is an eigenvector with eigenvalue 6. The determinant is 1(1−6)−2(3−4)+3(9−2)=18, the product of all three, so the complex pair multiplies to 18/6=3. (Directly: λ=1+2ω+3ω2=−32−32i and its conjugate, with |λ|2=94+34=3.)

    Why people get this wrong: Missing the row-sum eigenvector makes this a long cubic.

  22. MCQExam level

    A 3×3 matrix A has characteristic polynomial λ3−6λ2+11λ−6. By the Cayley-Hamilton theorem, A−1 equals
    1. 16(A2−6A+11I)
    2. −16(A2−6A+11I)
    3. 16(A2+6A+11I)
    4. 16(A2−6A−11I)
    Answer and solution

    Answer: A

    A3−6A2+11A−6I=0, so A(A2−6A+11I)=6I and A−1=16(A2−6A+11I). Check with the eigenvalues 1, 2, 3: for λ=2, 16(4−12+11)=12 ✓.

    Why people get this wrong: Moving −6I across the equals sign flips its sign; slipping here gives the second option.

  23. NATExam level

    For A=(1111), the top-left entry of eA is ______ (to two decimal places).
    Answer and solution

    Answer: 4.19 (±0.01)

    A2=2A, so An=2n−1A for n≥1. Then eA=I+∑n≥12n−1n!A=I+e2−12A. Top-left: 1+e2−12=e2+12≈4.19. (Or: eigenvalues 2 and 0 with projectors 12A and I−12A.)

    Why people get this wrong: Entry-wise exponentiation gives e≈2.72.

  24. MSQExam level

    Let H be Hermitian and U=eiH. Which statements are true?
    1. U is unitary
    2. The eigenvalues of U are eiλ, with λ the eigenvalues of H
    3. det⁡U=eitrH
    4. U is Hermitian
    Answer and solution

    Answer: A, B, C

    U†=e−iH†=e−iH, and e−iHeiH=I since H commutes with itself: unitary. In the eigenbasis of H, U is diagonal with entries eiλ. det⁡U=∏eiλ=ei∑λ=eitrH. U is Hermitian only if every eiλ is real (λ a multiple of π).

    Why people get this wrong: This is why time evolution e−iHt/ℏ is unitary: the Hamiltonian is Hermitian.

  25. MCQExam level

    Which matrix commutes with A=(2112)?
    1. (1002)
    2. (0100)
    3. (3−1−13)
    4. (1101)
    Answer and solution

    Answer: C

    A has distinct eigenvalues (3 and 1), so a matrix commutes with it exactly when it shares the eigenvectors (1,1) and (1,−1), that is when it has the form (abba). Only the third option does (=4I−A). For example (1002)A=(2124) but A(1002)=(2214).

    Why people get this wrong: Multiplying out is safe but slow; check whether the candidate shares the eigenvectors.

  26. NATExam level

    Let A=(111222333). The value of det⁡(I+A) is ______.
    Answer and solution

    Answer: 7

    A=u→v→T with u→=(1,2,3)T and v→=(1,1,1)T: rank 1. Its eigenvalues are v→Tu→=6 (eigenvector u→) and 0,0. So I+A has eigenvalues 7,1,1 and determinant 7. This is the matrix determinant lemma det⁡(I+u→v→T)=1+v→Tu→.

    Why people get this wrong: det⁡(I+A)≠1+det⁡A; here det⁡A=0 would give 1.

  27. MCQExam level

    A non-zero 4×4 matrix N satisfies N2=0. The largest possible rank of N is
    1. 1
    2. 2
    3. 3
    4. 4
    Answer and solution

    Answer: B

    N2=0 means everything in the image of N is sent to zero, so image⊆kernel and rank≤nullity=4−rank. Hence rank ≤2, reached by N=(0I200) in 2×2 blocks.

    Why people get this wrong: All eigenvalues are 0, but that alone does not fix the rank.

  28. NATExam level

    The value of trexp⁡[iπ33(σx+σy+σz)] is ______ (to two decimal places).
    Answer and solution

    Answer: 1 (±0.01)

    Write the exponent as iθnˆ⋅σ→ with nˆ=(1,1,1)/3 and θ=π/3. Since (nˆ⋅σ→)2=I, eiθnˆ⋅σ→=Icos⁡θ+inˆ⋅σ→sin⁡θ. The Pauli part is traceless, so the trace is 2cos⁡(π/3)=1.

    Why people get this wrong: Taking the angle as π/(33) forgets that σx+σy+σz=3nˆ⋅σ→; pull out the unit vector first.

  29. MCQExam level

    2×2 matrices A and B have eigenvalues {1,2} and {3,−1}. For the 4×4 Kronecker product A⊗B,
    1. tr=5, det⁡=−6
    2. tr=6, det⁡=−6
    3. tr=5, det⁡=36
    4. tr=6, det⁡=36
    Answer and solution

    Answer: D

    The eigenvalues of A⊗B are all products aibj: 3,−1,6,−2. Sum =6 (also trAtrB=3×2). Product =36 (also (det⁡A)2(det⁡B)2=4×9).

    Why people get this wrong: The eigenvalues of A⊗B multiply, they do not add; adding gives the spectrum of A⊗I+I⊗B.

  30. MSQExam level

    For a Hermitian matrix H, which conditions are each equivalent to H being positive definite?
    1. All eigenvalues of H are positive
    2. v→†Hv→>0 for every non-zero v→
    3. All diagonal entries of H are positive
    4. H=B†B for some invertible B
    Answer and solution

    Answer: A, B, D

    In the eigenbasis v→†Hv→=∑λk|ck|2, which is positive for all v→≠0 exactly when every λk>0. If so, B=H works; conversely v→†B†Bv→=‖Bv→‖2>0. Positive diagonal entries are necessary but not enough: (1221) has eigenvalue −1.

    Why people get this wrong: Positive diagonal entries follow from positive definiteness, not the other way round.

← All topics · Timed mock test →