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

IMC / 2019 / Problems / Day 2, P9

IMC 2019 · Day 2 · P9

hard

Determine all positive integers nn for which there exist n×nn \times n real invertible matrices AA and BB that satisfy ABBA=B2AAB - BA = B^2 A.

Proposed by Karen Keryan, Yerevan State University & American University of Armenia, Yerevan

Solution (official)

We prove that there exist such matrices AA and BB inf and only if nn is even.

I. Assume that nn is odd and some invertible n×nn \times n matrices A,BA, B satisfy ABBA=B2AAB - BA = B^2 A. Hence B=A1(B2+B)AB = A^{-1} (B^2 + B) A, so the matrices BB and B2+BB^2 + B are similar and therefore have the same eigenvalues. Since nn is odd, the matrix BB has a real eigenvalue, denote it by λ1\lambda_1. Therefore λ2:=λ12+λ1\lambda_2 := \lambda_1^2 + \lambda_1 is an eigenvalue of B2+BB^2 + B, hence an eigenvalue of BB. Similarly, λ3:=λ22+λ2\lambda_3 := \lambda_2^2 + \lambda_2 is an eigenvalue of B2+BB^2 + B, hence an eigenvalue of BB. Repeating this process and taking into account that the number of eigenvalues of BB is finite we will get there exist numbers klk \le l so that λl+1=λk\lambda_{l+1} = \lambda_k. Hence λk+1=λk2+λkλk+2=λk+12+λk+1λl=λl12+λl1λk=λl2+λl.\begin{gather*} \lambda_{k+1} = \lambda_k^2 + \lambda_k \\ \lambda_{k+2} = \lambda_{k+1}^2 + \lambda_{k+1} \\ \dots\dots\dots \\ \lambda_l = \lambda_{l-1}^2 + \lambda_{l-1} \\ \lambda_k = \lambda_l^2 + \lambda_l. \end{gather*} Adding this equations we get λk2+λk+12++λl2=0\lambda_k^2 + \lambda_{k+1}^2 + \dots + \lambda_l^2 = 0. Taking into account that all λi\lambda_i's are real (as λ1\lambda_1 is real), we have λk==λl=0\lambda_k = \dots = \lambda_l = 0, which implies that BB is not invertible, contradiction.

II. Now we construct such matrices A,BA, B for even nn. Let A2=[0110]A_2 = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix} and B2=[1111]B_2 = \begin{bmatrix} -1 & 1 \\ -1 & -1 \end{bmatrix}. It is easy to check that the matrices A2,B2A_2, B_2 are invertible and satisfy the condition. For n=2kn = 2k the n×nn \times n block matrices A=[A2000A2000A2],B=[B2000B2000B2]A = \begin{bmatrix} A_2 & 0 & \dots & 0 \\ 0 & A_2 & \dots & 0 \\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \dots & A_2 \end{bmatrix}, \qquad B = \begin{bmatrix} B_2 & 0 & \dots & 0 \\ 0 & B_2 & \dots & 0 \\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \dots & B_2 \end{bmatrix} are also invertible and satisfy the condition.

How the field did

contestants scored
360
average (of 10)
2.68
solved (≥ 80%)
20.0%
near-0 (≤ 10%)
61.7%
discrimination
0.60

Score distribution (field cohort)

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

Similar problems

IMC 2007 · Day 1 · P3hardavg 3.5/10 · solved 20% · near-0 42% · disc 0.57
IMC 2018 · Day 1 · P3hardavg 2.8/10 · solved 19% · near-0 52% · disc 0.47
IMC 2000 · Day 2 · P12hardavg 2.5/10 · solved 18% · near-0 68% · disc 0.52
IMC 2005 · Day 2 · P9hardavg 3.9/10 · solved 22% · near-0 37% · disc 0.61