Solution (#1752) If we set  

1 t t2 /2! · · · tr−1 /(r − 1)!

 .. .. 

 0 1 t . . 

λt  .. 
M(t) = e  0 0 1 . t2 /2! ,
 
 .... .. .. 
 . . . . t 
0 ··· 0 0 1
then M(0) = I and M ′ (t) equals
   
1 t t2 /2! · · · tr−1 /(r − 1)! 0 1 t · · · tr−2 /(r − 2)!
 .. ..   .. .. 
 0 1
 t . . 

 0 0
 1 . . 

λt  . ..  λt  . .. 
λe  0 0 1 t2 /2! +e  0 0 0 t ,
   
 . . .. ..   . .. .. .. 
 .. .. . . t   .
. . . . 1 
0 ··· 0 0 1 0 ··· 0 0 0
 
λ λt + 1 λt2 /2! + t · · · λtr−1 /(r − 1)! + tr−2 /(r − 2)!
 .. .. 
 0
 λ λt + 1 . . 

 .. 
= eλt  0 0 λ . 2
λt /2! + t 
 
 . . . . 
 .. .. .. .. λt + 1 
0 ··· 0 0 λ1
  2 
λ 1 0 ··· 0 1 t t /2! · · · tr−1 /(r − 1)!
 .. .  .. .. 
 0 λ
 1 . ..  
 0 1 t . . 

 ..   .. 
= eλt  0 0 λ . 0  0 0 1 . t 2
/2!  = J(λ, r)M (t)
  
 . . . .  . . . . 
 .. .. .. . . 1   .. .. .. .. t 
0 ··· 0 0 λ 0 ··· 0 0 1
as required. Hence eJ(λ,r)t = M (t). Note then that we have
r
det eJ(λ,r) = eλ = eλr = etraceJ(λ,r) .
Now by #1079 every matrix A is similar to one in Jordan normal form. So there exists invertible P such that
P −1 AP = diag (J(λ1 , r1 ), J(λ2 , r2 ), . . . , J(λk , rk )) .
Then
det eA = det(P −1 eA P ) [by the product rule for determinants]
P −1 AP
= det e [by #1746]
= det diag eJ(λ1 ,r1 ) , eJ(λ2 ,r2 ) , . . . , eJ(λk ,rk ) [by #1744]
J(λ1 ,r1 ) J(λ1 ,r1 ) J(λ1 ,r1 )
= det e × det e × · · · × det e [by #820]
= etraceJ(λ1 ,r1 ) × etraceJ(λ2 ,r2 ) × · · · × etraceJ(λk ,rk )
= exp [traceJ(λ1 , r1 ) + traceJ(λ2 , r2 ) + · · · + traceJ(λk , rk )]
= exp trace P −1 AP
= exp traceA [by the trace product rule].