Linear-Algebra-Done-Right Β· Unit 26 Β· Video 5 Β· Interactive Practice

Why ST βˆ’ TS Can Never Be the Identity

IKey Formulas

FormulaNameHypotheses
tr⁑T=⟨Te1,e1⟩+β‹―+⟨Ten,en⟩\operatorname{tr} T = \langle Te_1, e_1\rangle + \cdots + \langle Te_n, e_n\rangleTrace on an inner product spaceVV a finite-dimensional inner product space, e1,…,ene_1, \dots, e_n an orthonormal basis, T∈L(V)T \in \mathcal{L}(V)
tr⁑(λT)=λtr⁑T\operatorname{tr}(\lambda T) = \lambda \operatorname{tr} T,  tr⁑(S+T)=tr⁑S+tr⁑T\ \operatorname{tr}(S + T) = \operatorname{tr} S + \operatorname{tr} Ttr⁑\operatorname{tr} is a linear functional on L(V)\mathcal{L}(V)S,T∈L(V)S, T \in \mathcal{L}(V), λ∈F\lambda \in \mathbf{F}
tr⁑(ST)=tr⁑(TS)\operatorname{tr}(ST) = \operatorname{tr}(TS)Order does not matter inside the traceS,T∈L(V)S, T \in \mathcal{L}(V)
tr⁑(STβˆ’TS)=0β‰ dim⁑V=tr⁑I\operatorname{tr}(ST - TS) = 0 \neq \dim V = \operatorname{tr} ISTβˆ’TS=IST - TS = I is impossibleVV finite-dimensional and nonzero

Key Insight: Every commutator STβˆ’TSST - TS has trace 00, while the identity has trace dim⁑V\dim V; in infinite dimensions there is no trace, and the equation can hold.

IIEvery Commutator Has Trace 00

Whatever SS and TT are, the diagonal of STβˆ’TSST - TS adds up to 00, so it never matches tr⁑I=2\operatorname{tr} I = 2.

IIIDiagonal Entries ⟨Tek,ek⟩\langle Te_k, e_k\rangle

Turning the orthonormal basis changes each ⟨Tek,ek⟩\langle Te_k, e_k\rangle but not their sum, tr⁑T=4\operatorname{tr} T = 4 for T(x,y)=(x+2y, 3y)T(x, y) = (x + 2y,\ 3y).

IVQuiz Questions

Problem 1 Β· Linearity of the Trace

Given: S,T∈L(V)S, T \in \mathcal{L}(V) with tr⁑S=4\operatorname{tr} S = 4 and tr⁑T=βˆ’1\operatorname{tr} T = -1 β€” find tr⁑(3Sβˆ’2T+STβˆ’TS)\operatorname{tr}(3S - 2T + ST - TS).

βœ… Correct! 3β‹…4βˆ’2β‹…(βˆ’1)+(tr⁑(ST)βˆ’tr⁑(TS))=12+2+0=143 \cdot 4 - 2 \cdot (-1) + \bigl(\operatorname{tr}(ST) - \operatorname{tr}(TS)\bigr) = 12 + 2 + 0 = 14.
❌ Check the sign. βˆ’2tr⁑T=βˆ’2β‹…(βˆ’1)=+2-2 \operatorname{tr} T = -2 \cdot (-1) = +2.
❌ The commutator contributes nothing. tr⁑(STβˆ’TS)=tr⁑(ST)βˆ’tr⁑(TS)=0\operatorname{tr}(ST - TS) = \operatorname{tr}(ST) - \operatorname{tr}(TS) = 0 for all S,TS, T.
❌ Keep the coefficients. Linearity gives tr⁑(3S)=3tr⁑S=12\operatorname{tr}(3S) = 3 \operatorname{tr} S = 12 and tr⁑(βˆ’2T)=2\operatorname{tr}(-2T) = 2.
Show solution

The trace is linear and tr⁑(ST)=tr⁑(TS)\operatorname{tr}(ST) = \operatorname{tr}(TS):

tr⁑(3Sβˆ’2T+STβˆ’TS)=3tr⁑Sβˆ’2tr⁑T+tr⁑(ST)βˆ’tr⁑(TS)=12+2+0=14.\operatorname{tr}(3S - 2T + ST - TS) = 3\operatorname{tr} S - 2\operatorname{tr} T + \operatorname{tr}(ST) - \operatorname{tr}(TS) = 12 + 2 + 0 = 14.

Problem 2 Β· Which Is Not a Commutator?

Given: four 2Γ—22 \times 2 matrices β€” find the one that is NOT of the form M(STβˆ’TS)\mathcal{M}(ST - TS) for any S,T∈L(C2)S, T \in \mathcal{L}(\mathbf{C}^2) (matrices with respect to the standard basis).

βœ… Correct! Its trace is 11, but every STβˆ’TSST - TS has trace 00.
❌ That one is a commutator. With S=(1000)S = \begin{pmatrix} 1 & 0 \\ 0 & 0 \end{pmatrix} and T=(0100)T = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix}, STβˆ’TS=Tβˆ’0=TST - TS = T - 0 = T.
❌ That one is a commutator. With S=(0100)S = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix} and T=(0010)T = \begin{pmatrix} 0 & 0 \\ 1 & 0 \end{pmatrix}, STβˆ’TS=(1000)βˆ’(0001)ST - TS = \begin{pmatrix} 1 & 0 \\ 0 & 0 \end{pmatrix} - \begin{pmatrix} 0 & 0 \\ 0 & 1 \end{pmatrix}.
❌ That one is a commutator. Take S=TS = T: then STβˆ’TS=0ST - TS = 0.
Show solution

Since tr⁑(STβˆ’TS)=tr⁑(ST)βˆ’tr⁑(TS)=0\operatorname{tr}(ST - TS) = \operatorname{tr}(ST) - \operatorname{tr}(TS) = 0, a matrix with nonzero trace is never a commutator. Option (d) has trace 11.

The other three have trace 00 and are commutators: (a) =[1000][0100]βˆ’[0100][1000]= \bigl[\begin{smallmatrix} 1 & 0 \\ 0 & 0 \end{smallmatrix}\bigr]\bigl[\begin{smallmatrix} 0 & 1 \\ 0 & 0 \end{smallmatrix}\bigr] - \bigl[\begin{smallmatrix} 0 & 1 \\ 0 & 0 \end{smallmatrix}\bigr]\bigl[\begin{smallmatrix} 1 & 0 \\ 0 & 0 \end{smallmatrix}\bigr]; (b) =[0100][0010]βˆ’[0010][0100]= \bigl[\begin{smallmatrix} 0 & 1 \\ 0 & 0 \end{smallmatrix}\bigr]\bigl[\begin{smallmatrix} 0 & 0 \\ 1 & 0 \end{smallmatrix}\bigr] - \bigl[\begin{smallmatrix} 0 & 0 \\ 1 & 0 \end{smallmatrix}\bigr]\bigl[\begin{smallmatrix} 0 & 1 \\ 0 & 0 \end{smallmatrix}\bigr]; (c) =SSβˆ’SS= SS - SS for any SS.

Problem 3 Β· Diagonal Entries in an Orthonormal Basis

Given: T∈L(R2)T \in \mathcal{L}(\mathbf{R}^2) defined by T(x,y)=(2x+y,Β βˆ’x+4y)T(x, y) = (2x + y,\ -x + 4y), the Euclidean inner product, and the orthonormal basis e1=15(3,4)e_1 = \tfrac15(3, 4), e2=15(βˆ’4,3)e_2 = \tfrac15(-4, 3) β€” find ⟨Te1,e1⟩\langle Te_1, e_1\rangle and ⟨Te2,e2⟩\langle Te_2, e_2\rangle.

⟨Te1,e1⟩\langle Te_1, e_1\rangle

⟨Te2,e2⟩\langle Te_2, e_2\rangle

βœ… Correct! 8225+6825=6=2+4=tr⁑T\tfrac{82}{25} + \tfrac{68}{25} = 6 = 2 + 4 = \operatorname{tr} T.
❌ e1e_1 has length 11 only with the factor 15\tfrac15. 82=⟨T(3,4),(3,4)⟩82 = \langle T(3, 4), (3, 4)\rangle must be divided by 525^2.
❌ 22 is the entry for the standard basis vector (1,0)(1, 0). Diagonal entries depend on the orthonormal basis; only their sum does not.
❌ 44 is the entry for (0,1)(0, 1). Compute ⟨Te2,e2⟩\langle Te_2, e_2\rangle for e2=15(βˆ’4,3)e_2 = \tfrac15(-4, 3), or subtract from tr⁑T=6\operatorname{tr} T = 6.
❌ The two entries need not be equal. They must add up to tr⁑T=6\operatorname{tr} T = 6.
❌ Not quite. Te1=15(10,13)Te_1 = \tfrac15(10, 13); take its inner product with 15(3,4)\tfrac15(3, 4).
❌ Not quite. The two diagonal entries add up to tr⁑T=2+4=6\operatorname{tr} T = 2 + 4 = 6.
Show solution
Te1=15(6+4,Β βˆ’3+16)=15(10,13),⟨Te1,e1⟩=125(30+52)=8225.Te_1 = \tfrac15(6 + 4,\ -3 + 16) = \tfrac15(10, 13), \qquad \langle Te_1, e_1\rangle = \tfrac1{25}(30 + 52) = \tfrac{82}{25}. Te2=15(βˆ’8+3,Β 4+12)=15(βˆ’5,16),⟨Te2,e2⟩=125(20+48)=6825.Te_2 = \tfrac15(-8 + 3,\ 4 + 12) = \tfrac15(-5, 16), \qquad \langle Te_2, e_2\rangle = \tfrac1{25}(20 + 48) = \tfrac{68}{25}.

The sum is 15025=6\tfrac{150}{25} = 6, which is tr⁑T=2+4\operatorname{tr} T = 2 + 4 computed from the standard basis.

Problem 4 Β· A Scaled Identity as a Commutator

Given: VV with dim⁑V=5\dim V = 5, and λ∈F\lambda \in \mathbf{F} such that STβˆ’TS=Ξ»IST - TS = \lambda I for some S,T∈L(V)S, T \in \mathcal{L}(V) β€” find which values of Ξ»\lambda are possible.

βœ… Correct! Taking traces gives 0=5Ξ»0 = 5\lambda, so Ξ»=0\lambda = 0, and Ξ»=0\lambda = 0 does occur (take S=TS = T).
❌ Take the trace of both sides. tr⁑(STβˆ’TS)=0\operatorname{tr}(ST - TS) = 0 while tr⁑(Ξ»I)=5Ξ»\operatorname{tr}(\lambda I) = 5\lambda.
❌ 55 is tr⁑I\operatorname{tr} I, not a possible λ\lambda. The trace equation is 0=5λ0 = 5\lambda.
❌ Ξ»=0\lambda = 0 is possible. Any operator commutes with itself, so SSβˆ’SS=0=0β‹…ISS - SS = 0 = 0 \cdot I.
Show solution

Apply the trace, which is linear and satisfies tr⁑(ST)=tr⁑(TS)\operatorname{tr}(ST) = \operatorname{tr}(TS):

0=tr⁑(ST)βˆ’tr⁑(TS)=tr⁑(STβˆ’TS)=tr⁑(Ξ»I)=Ξ»dim⁑V=5Ξ».0 = \operatorname{tr}(ST) - \operatorname{tr}(TS) = \operatorname{tr}(ST - TS) = \operatorname{tr}(\lambda I) = \lambda \dim V = 5\lambda.

So Ξ»=0\lambda = 0. Conversely Ξ»=0\lambda = 0 is attained, for instance by S=T=IS = T = I. The case Ξ»=1\lambda = 1 is the video's statement that STβˆ’TS=IST - TS = I is impossible.

Solved: 0 / 4