Unofficial archive — problems, solutions & results © IMC, reproduced with permission.

IMC / 2020 / Problems / Day 1, P2

IMC 2020 · Day 1 · P2

medium

Let AA and BB be n×nn \times n real matrices such that rk(ABBA+I)=1\operatorname{rk}(AB - BA + I) = 1 where II is the n×nn \times n identity matrix.

Prove that trace(ABAB)trace(A2B2)=12n(n1).\operatorname{trace}(ABAB) - \operatorname{trace}(A^2 B^2) = \frac12 n (n - 1). (rk(M)\operatorname{rk}(M) denotes the rank of matrix MM, i.e., the maximum number of linearly independent columns in MM. trace(M)\operatorname{trace}(M) denotes the trace of MM, that is the sum of diagonal elements in MM.)

Rustam Turdibaev, V. I. Romanovskiy Institute of Mathematics

Solution (official)

Let X=ABBAX = AB - BA. The first important observation is that trace(X2)=trace(ABABABBABAAB+BABA)=2trace(ABAB)2trace(A2B2)\operatorname{trace}(X^2) = \operatorname{trace}(ABAB - ABBA - BAAB + BABA) = 2 \operatorname{trace}(ABAB) - 2 \operatorname{trace}(A^2 B^2) using that the trace is cyclic. So we need to prove that trace(X2)=n(n1)\operatorname{trace}(X^2) = n(n-1).

By assumption, X+IX + I has rank one, so we can write X+I=vtwX + I = v^t w for two vectors v,wv, w. So X2=(vtwI)2=I2vtw+vtwvtw=I+(wvt2)vtw.X^2 = (v^t w - I)^2 = I - 2 v^t w + v^t w v^t w = I + (w v^t - 2) v^t w. Now by definition of XX we have trace(X)=0\operatorname{trace}(X) = 0 and hence wvt=trace(wvt)=trace(vtw)=nw v^t = \operatorname{trace}(w v^t) = \operatorname{trace}(v^t w) = n so that indeed trace(X2)=n+(n2)n=n(n1).\operatorname{trace}(X^2) = n + (n - 2) n = n (n - 1). An alternative way to use the rank one condition is via eigenvalues: Since X+IX + I has rank one, it has eigenvalue 0 with multiplicity n1n - 1. So XX has eigenvalue 1-1 with multiplicity n1n - 1. Since trace(X)=0\operatorname{trace}(X) = 0 the remaining eigenvalue of XX must be n1n - 1. Hence trace(X2)=(n1)2+(n1)12=n(n1).\operatorname{trace}(X^2) = (n-1)^2 + (n-1) \cdot 1^2 = n (n-1).

How the field did

contestants scored
453
average (of 10)
5.12
solved (≥ 80%)
44.8%
near-0 (≤ 10%)
38.4%
discrimination
0.37

Score distribution (field cohort)

Computed on contestants with a meaningful total (field cohort); discrimination is the corrected item–total correlation.

Similar problems

IMC 2003 · Day 1 · P3mediumavg 5.8/10 · solved 45% · near-0 25% · disc 0.57
IMC 2000 · Day 1 · P3mediumavg 5.4/10 · solved 45% · near-0 33% · disc 0.58
IMC 2019 · Day 1 · P2mediumavg 6.1/10 · solved 44% · near-0 13% · disc 0.38
IMC 2023 · Day 2 · P6mediumavg 5.3/10 · solved 47% · near-0 36% · disc 0.48