6 - Updated - Problems and Solutions
6 - Updated - Problems and Solutions
⎡ ⎤
2AD 2 AD 2 + BD 2 − AB 2 AD 2 + CD 2 − AC 2
1 ⎣
det AD + BD 2 − AB 2
2
2BD 2 BD 2 + CD 2 − BC 2 ⎦ .
2 AD 2 + CD 2 − AC 2 BD + CD 2 − BC 2
2
2CD 2
The determinant here is the Cayley–Menger determinant for the tetrahedron ABCD and
its value is 2882 , where is the volume of ABCD. Hence (∗) is equal to 1442 , which
is clearly nonnegative. This yields the desired inequality, and equality holds if and only if
= 0, in other words A, B, C, and D are coplanar.
Editorial comment. The Cayley–Menger determinant generalizes Heron’s formula for the
area of a triangle to simplices of higher dimension.
Also solved by M. Bataille (France), R. Chapman (UK), G. Fera & G. Tescaro (Italy), D. Fleischman,
E. A. Herman, W. Janous (Austria), M. Kaplan & M. Goldenberg, B. Karaivanov (USA) & T. S. Vassilev
(Canada), K. T. L. Koo (China), A. Stadler (Switzerland), R. Stong, T. Wiandt, and the proposer.
A Matrix Equation
12173 [2020, 275]. Proposed by Florin Stanescu, Serban Cioculescu School, Gaesti,
Romania. Suppose that X and Y are n-by-n complex matrices such that 2Y 2 = XY − Y X
and the rank of X − Y is 1. Prove Y 3 = Y XY .
Solution by Roger A. Horn, Tampa, FL. Let z and w be nonzero complex n-vectors such
that X − Y = zw ∗ . It suffices to show that if
2Y 2 = zw ∗ Y − Y zw ∗ , (1)
then Y zw ∗ Y = 0. Jacobson’s lemma (see page 126 of R. Horn and C. Johnson (2018),
Matrix Analysis, 2nd ed., New York: Cambridge University Press) states that if BC −
CB commutes with C, then BC − CB is nilpotent. Consequently, Y 2 (and hence Y ) is
nilpotent. The rank of Y 2 is at most 2, since it is the sum of two matrices whose ranks are
at most 1. Therefore, the Jordan canonical form of Y is a direct sum of nilpotent Jordan
blocks that are not larger than 4-by-4. There are three cases.
Case (a): Y 2 = 0 (no block larger than 2-by-2). If Y 2 = 0, then Y 2 z = 0 and
0 = 2Y 3 = Y 2Y 2 = Y zw ∗ Y − Y 2 zw ∗ = Y zw ∗ Y. (2)
0 = 2Y 4 = Y 2 2Y 2 = Y 2 zw ∗ Y − Y 3 zw ∗ = (Y 2 z)(w ∗ Y ).
Either w ∗ Y = 0 and we are done, or w ∗ Y
= 0 and Y 2 z = 0. In the latter case, (2) also
holds, and it ensures that Y zw∗ Y = 0.
Case (c): Y 3
= 0 and Y 4 = 0 (the largest block is 4-by-4). Let v be a complex n-vector
such that Y 3 v
= 0. Suppose Y z
= 0. We compute
0 = 2Y 5 v = 2Y 2 Y 3 v = zw ∗ Y 4 v − Y zw ∗ Y 3 v = −(w ∗ Y 3 v)Y z,
so w ∗ Y 3 v = 0. We also have
0 = 2Y 4 v = 2Y 2 Y 2 v = zw ∗ Y 3 v − Y zw ∗ Y 2 v = −(w ∗ Y 2 v)Y z,
950
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 128