Maths Olympiad Prep

Track / Stage 7 / 220 of 300 #1620 of 1964

Problem 1620

National olympiad second round; IMO P1/P4
Algebra Difficulty 7.5 Prove it

Let AA be a real n×nn\times n matrix satisfying
A+AT=I,A+A^{\text T}=I,where ATA^{\text T} denotes the transpose of AA and II the n×nn\times n identity matrix. Show that detA>0\det A>0.

This one wants a proof. Work it on paper, then read the official solution and mark yourself. Be honest about it: the record is only any use to you if it is.

Official solution

1. Given the matrix equation A+AT=I A + A^{\text{T}} = I , we start by analyzing the properties of A A . Let A=[aij] A = [a_{ij}] . From the given equation, we have:
A+AT=I A + A^{\text{T}} = I
This implies that for the diagonal elements aii a_{ii} :
aii+aii=1    2aii=1    aii=12 a_{ii} + a_{ii} = 1 \implies 2a_{ii} = 1 \implies a_{ii} = \frac{1}{2}
For the off-diagonal elements aij a_{ij} where ij i \neq j :
aij+aji=0    aij=aji a_{ij} + a_{ji} = 0 \implies a_{ij} = -a_{ji}
Therefore, A A can be written as:
A=12I+B A = \frac{1}{2}I + B
where B B is a real skew-symmetric matrix (i.e., BT=B B^{\text{T}} = -B ).

2. Next, we consider the characteristic polynomial of A A . Let pM(x)=det(xIM) p_M(x) = \det(xI - M) be the characteristic polynomial of a matrix M M . For A A , we have:
pA(x)=det(xIA)=det(xI(12I+B))=det((x12)IB)=pB(x12) p_A(x) = \det(xI - A) = \det\left(xI - \left(\frac{1}{2}I + B\right)\right) = \det\left(\left(x - \frac{1}{2}\right)I - B\right) = p_B\left(x - \frac{1}{2}\right)
This relationship shows that the eigenvalues of A A are shifted by 12\frac{1}{2} from the eigenvalues of B B .

3. Since B B is a real skew-symmetric matrix, its eigenvalues are either zero or purely imaginary numbers. Specifically, the eigenvalues of B B are of the form ci ci for some real number c c . If c0 c \neq 0 , the eigenvalues appear in conjugate pairs ci ci and ci -ci .

4. Therefore, the eigenvalues of A A are either:
12(if the eigenvalue of B is 0) \frac{1}{2} \quad \text{(if the eigenvalue of } B \text{ is } 0\text{)}
or
12+ciand12ci(if the eigenvalues of B are ci and ci) \frac{1}{2} + ci \quad \text{and} \quad \frac{1}{2} - ci \quad \text{(if the eigenvalues of } B \text{ are } ci \text{ and } -ci\text{)}

5. To find the determinant of A A , we multiply all its eigenvalues. Let k k be the number of eigenvalues equal to 12\frac{1}{2} and let m m be the number of pairs of purely imaginary eigenvalues. Then:
detA=(12)kj=1m(12+cji)(12cji) \det A = \left(\frac{1}{2}\right)^k \prod_{j=1}^m \left(\frac{1}{2} + c_ji\right)\left(\frac{1}{2} - c_ji\right)
Simplifying the product of conjugate pairs:
(12+cji)(12cji)=(12)2(cji)2=14+cj2 \left(\frac{1}{2} + c_ji\right)\left(\frac{1}{2} - c_ji\right) = \left(\frac{1}{2}\right)^2 - (c_ji)^2 = \frac{1}{4} + c_j^2
Therefore:
detA=(12)kj=1m(14+cj2) \det A = \left(\frac{1}{2}\right)^k \prod_{j=1}^m \left(\frac{1}{4} + c_j^2\right)
Since 14+cj2>0\frac{1}{4} + c_j^2 > 0 for all real cj c_j , and (12)k>0\left(\frac{1}{2}\right)^k > 0, it follows that:
detA>0 \det A > 0

The final answer is detA>0 \boxed{ \det A > 0 }

Source: NuminaMath-1.5, licensed Apache-2.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.