matrix theorems with expo of matrix
DOCX · 20.5 KB
Open DOCX file
Short note by Phil dated 11.20.16 from his matrix binder. It proves the exponential product theorem by first treating scalars, using a double-sum reordering theorem to match the series, then showing the binomial expansion needs A and B to commute. A corollary shows exp(i(n+dn)J) differs from the product of exponentials because the angular momentum generators have nonzero commutators.
AI-written summary; may contain errors.
Extracted text (machine-read; may contain errors)
Matrix theorems with expo of matrix PhL 11.20.16
I guess I don't have this stuff written down anywhere, so do it here today.
1. Theorem 1: eAeB = eBeA = eA+B provided A and B commute.
Proof: If A and B commute, then clearly An and Bm commute, so we have
eAeB = (ΣnAn/n!) (ΣmBm/m!) = (ΣmBm/m!)(ΣnAn/n!) = eBeA
Next, consider this simpler problem of showing for non-matrix x and y
exey = ex+y
We have to show then that
(Σnxn/n!) (Σmym/m!) = Σn(x+y)n/n! .
First write the left side as
LHS = (Σnxn/n!) (Σmym/m!) = ΣnΣm xnym / (n!m!)
On the right side we know that
(x+y)n = Σk=0n (n,k) xn-kyk = Σk=0n xn-kyk (**)
So then the right side is this
RHS = Σn(x+y)n/n! = Σn Σk=0n xn-kyk /n! = Σn=0∞ Σk=0n xn-kyk
So now we want to show that
LHS = RHS
Σn=0∞Σm=0∞ xnym / (n!m!) = Σn=0∞ Σk=0n xn-kyk (*)
Now forget (*) for a moment and recall this general summation reordering Theorem 3 from integrals and series/ double sum... doc:
Σm=0∞ Σn=0∞ fn,m = Σk=0∞ Σm=0k fk-m,m m +n = k Theorem 3 (1)
Now swap order on the left, then rename dummies on the right so that k→n and m→k
Σn=0∞ Σm=0∞ fn,m = Σn=0∞ Σk=0n fn-k,k m +n = k Theorem 3 (2)
Now let
fn,m = xnym / (n!m!) fn-k,k = xn-kyk / ((n-k)!k!)
The theorem in form (2) then says
Σn=0∞ Σm=0∞ xnym / (n!m!) = Σn=0∞ Σk=0n xn-kyk / ((n-k)!k!)
But this is exactly equation (*) so we have thus proven that (*) is correct and therefore that exey = ex+y
Now how is this different if x → A and y→ B are matrices? Look at step (**)
(x+y)n = Σk=0n (n,k) xn-kyk
With matrices this would be for example
(A+B)2 =(A+B)(A+B) = A2 + BA + AB + B2
only if A and B commute can you write this as
A2 + 2AB + B2 = Σk=02 (2,k) An-kBk
So to make the above proof work for matrices, they must commute. Then we have
eAeB = eA+B iff A and B commute QED
Corollary: exp(i [ n + dn]J) ≠ exp(i dnJ) exp(i nJ) Ji = generators
Write
A = in J = iniJi
B = i dnJ = i dniJi
Then
[A,B] = [ iniJi,i dnjJj] = - nidnj [Ji,Jj] = - nidnj i εijkJk ≠ 0
[Ji, Vj ] = iijkVk