Adieu 2018

Algebra Level 4

Consider the 2018 × 2018 2018\times 2018 matrix A = [ a i j ] A=[a_{ij}] with a i j = i j a_{ij}=ij if i j i\neq j and a i i = 1 + i 2 a_{ii}=1+i^2 . What remainder is left after you divide det A \det A by 2018 2018 ? (No calculator required)


The answer is 1010.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

3 solutions

Mark Hennings
Dec 28, 2018

Consider the n × n n \times n matrix A A , where A i j = δ i j + i j 1 i , j n A_{ij} \; = \; \delta_{ij} + ij \hspace{2cm} 1 \le i,j \le n Then we note that A = I + x x T A \; = \; I + \mathbf{x}\mathbf{x}^T where x = ( 1 2 n ) \mathbf{x} \; = \; \left(\begin{array}{c} 1 \\ 2 \\ \vdots \\ n \end{array} \right) Thus we deduce that A x = ( x 2 + 1 ) x A\mathbf{x} = \big(\Vert\mathbf{x}\Vert^2 + 1\big)\mathbf{x} while A y = y x y = 0 A\mathbf{y} \; = \; \mathbf{y} \hspace{2cm} \mathbf{x}\cdot\mathbf{y} = 0 If we choose an orthogonal matrix P P whose first column is a multiple of x \mathbf{x} , we deduce that A P = P ( x 2 + 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 ) AP \; = \; P\left(\begin{array}{cccccc}\Vert\mathbf{x}\Vert^2+1 & 0 & 0 & 0 & \ldots & 0 \\ 0 & 1 & 0 & 0 & \ldots & 0 \\ 0 & 0 & 1 & 0 & \ldots & 0 \\ 0 & 0 & 0 & 1 & \ldots & 0 \\ \vdots & \vdots & \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & 0 & 0 & \ldots & 1 \end{array} \right) and hence d e t A = d e t ( P T A P ) = x 2 + 1 = 1 6 n ( n + 1 ) ( 2 n + 1 ) + 1 \mathrm{det}\,A \; = \; \mathrm{det}\,(P^TAP) = \Vert\mathbf{x}\Vert^2 + 1 \; = \; \tfrac16n(n+1)(2n+1) + 1 For this problem we have n = 2018 = 2 × 1009 n = 2018 = 2 \times 1009 where 1009 1009 is prime. It is clear that d e t A 1 ( m o d 1009 ) \mathrm{det}\,A \equiv 1 \pmod{1009} , and we also see that d e t A 0 ( m o d 2 ) \mathrm{det}\,A \equiv 0 \pmod{2} . Thus we deduce that d e t A 1010 ( m o d 2018 ) \mathrm{det}\,A \equiv \boxed{1010} \pmod{2018} .

Nice solid solution! Thank you, Mark!

Otto Bretscher - 2 years, 5 months ago
Otto Bretscher
Dec 28, 2018

The symmetric matrix B = [ b i j ] B=[b_{ij}] with b i j = i j b_{ij}=ij has rank 1, with eigenvalues 0 (with muliplicity n 1 n-1 ) and tr B = i = 1 n i 2 \text{tr}B =\sum_{i=1}^ni^2 . Thus the matrix A = B + I n A=B+I_n has the eigenvalues 1 (with multiplicity n 1 n-1 ) and 1 + i = 1 n i 2 1+\sum_{i=1}^ni^2 so det A = 1 + i = 1 n i 2 \det A=1+\sum_{i=1}^ni^2 . (We know that det A \det A is the product of the eigenvalues and tr A \text{tr}A is their sum.)

In the case of n = 2018 n=2018 we find det A = 1 6 × 2018 × 2019 × 4037 + 1 = 1009 × 673 × 4037 + 1 \det A= \frac{1}{6}\times 2018\times 2019 \times 4037+1=1009 \times 673 \times 4037+1 , which leaves a remainder of 1010 \boxed{1010} .

Wow, this is beautiful. I discovered a parttern, and I was trying to prove that det ( A n ) = 1 + ( 1 2 + 2 2 + + n 2 ) \det(A_n) = 1 + (1^2 + 2^2 +\cdots + n^2) via induction. And I then read up on block matrix identities. In particular, det [ A B C D ] = ( D + 1 ) det ( A ) det ( A + B C ) \det \begin{bmatrix}{A} && {B} \\ {C} && {D}\end{bmatrix} = (D+1) \det(A) - \det(A+BC) . I even tried to see if Woodbury matrix identity can be applicable here.

Pi Han Goh - 2 years, 5 months ago
Huan Bui
Dec 28, 2018

I'm sure there is a much more elegant solution to this one. But here's my approach:

Observe that, det ( A n ) = tr ( A n ) ( n 1 ) = 1 n + i = 1 n ( 1 + i 2 ) = 1 + i = 1 n i 2 = 1 + n ( n + 1 ) ( 2 n + 1 ) 6 \det(A_n) = \text{tr}(A_n) - (n-1) = 1-n + \sum_{i=1}^{n}(1+i^2)= 1 + \sum_{i=1}^{n}i^2 = 1 + \frac{n(n+1)(2n+1)}{6} .

For n = 2018 n = 2018 , det ( A n ) mod 2018 = 1 + 2018 ( 2018 + 1 ) ( 2 2018 + 1 ) 6 mod 2018 = 1 + 1009 = 1010 \det(A_n)\text{ }\text{mod}\text{ }2018 = 1 + \frac{2018(2018+1)(2\cdot 2018+1)}{6}\text{ mod }2018 = 1 + 1009 = \boxed{1010} , which is the answer we seek.

Very nicely done, Comrade! Just to clarify for the less experienced readers: How do we know that det ( A n ) = tr ( A n ) ( n 1 ) \det(A_n)=\text{tr}(A_n)-(n-1) ?

You know, I came across this problem today when I needed to prove that the scaling factor of a graph in many variables is what it is.

Otto Bretscher - 2 years, 5 months ago

Log in to reply

Thank you, Otto! It's an observation I made, but @Mark Hennings has written a detailed explanation which also shows why it is true.

Huan Bui - 2 years, 5 months ago

Log in to reply

I posted an explanation as well.

Otto Bretscher - 2 years, 5 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...