How to calculate #A^n#,with #A=((1,2,3),(0,1,2),(0,0,1))#?

2 Answers
May 28, 2017

#A^n = ((1, 2n, 2n^2+n),(0, 1, 2n),(0, 0, 1))#

Explanation:

Let us look at the first few powers of #A# to find the pattern:

#A^0 = ((1, 0, 0),(0, 1, 0),(0, 0, 1))#

#A^1 = ((1, 2, 3),(0, 1, 2),(0, 0, 1))#

#A^2 = ((1, 2, 3),(0, 1, 2),(0, 0, 1))((1, 2, 3),(0, 1, 2),(0, 0, 1)) = ((1, 4, 10),(0, 1, 4),(0, 0, 1))#

#A^3 = ((1, 2, 3),(0, 1, 2),(0, 0, 1))((1, 4, 10),(0, 1, 4),(0, 0, 1)) = ((1, 6, 21),(0, 1, 6),(0, 0, 1))#

#A^4 = ((1, 2, 3),(0, 1, 2),(0, 0, 1))((1, 6, 21),(0, 1, 6),(0, 0, 1)) = ((1, 8, 36),(0, 1, 8),(0, 0, 1))#

It looks like the form of #A^n# is something like:

#A^n = ((1, 2n, a_n),(0, 1, 2n),(0, 0, 1))#

with the #a_n#'s forming the sequence:

#color(blue)(3), 10, 21, 36,..." "# for #n=1,2,3,4,...#

Writing down the sequence of differences between consecutive terms, we get:

#color(blue)(7), 11, 15#

Writing down the sequence of differences between those differences, we get:

#color(blue)(4), 4#

Having reached a constant sequence, we can use the first term of each sequence to give a direct formula for #a_n#:

#a_n = color(blue)(3)/(0!)+color(blue)(7)/(1!)(n-1)+color(blue)(4)/(2!)(n-1)(n-2)#

#color(white)(a_n) = 3+7n-7+2n^2-6n+4#

#color(white)(a_n) = 2n^2+n#

So it looks like:

#A^n = ((1, 2n, 2n^2+n),(0, 1, 2n),(0, 0, 1))#

We can prove this by induction:

Case #n=0#

#((1, 2(color(blue)(0)), 2(color(blue)(0))^2+(color(blue)(0))),(0, 1, 2(color(blue)(0))),(0, 0, 1)) = ((1, 0, 0),(0, 1, 0),(0, 0, 1)) = A^0#

Induction step

Suppose:

#A^n = ((1, 2n, 2n^2+n),(0, 1, 2n),(0, 0, 1))#

Then:

#((1, 2(color(blue)(n+1)), 2(color(blue)(n+1))^2+(color(blue)(n+1))),(0, 1, 2(color(blue)(n+1))),(0, 0, 1)) = ((1, 2n+2, 2n^2+5n+3),(0, 1, 2n+2),(0, 0, 1))#

#color(white)(((1, 2(n+1), 2(n+1)^2+(n+1)),(0, 1, 2(n+1)),(0, 0, 1))) = ((1, 2, 3),(0, 1, 2),(0, 0, 1))((1, 2n, 2n^2+n),(0, 1, 2n),(0, 0, 1)) = A A^n = A^(color(blue)(n+1))#

#color(white)()#
Remarks

Given that we now have a formula for #A^n#, what about non-integer values of #n#?

Using our formula:

#A^(1/2) = ((1, 1, 1),(0,1,1),(0,0,1))#

We find:

#((1, 1, 1),(0, 1, 1),(0, 0, 1))((1, 1, 1),(0, 1, 1),(0, 0, 1)) = ((1, 2, 3), (0, 1, 2), (0, 0, 1))#

as expected.

This would be quite interesting to generalise to other forms of matrix. Does it give a generic way of finding a square root of a matrix?

May 28, 2017

See below.

Explanation:

Matrix #A# characteristic polynomial is

#p(lambda)=1 - 3 lambda + 3 lambda^2 - lambda^3 =- (lambda-1)^3#

and #A# obeys this characteristic equation so

#A^3=3A^2-3A+I#

There is a difference equation associated to this relationship

#a_(n+3)=3a_(n+2)-3a_(n+1)+a_n#

The difference equation solution is

#a_n=C_0+nC_1+n^2C_2# so finally

#A^n = C_0+nC_1+n^2C_2#

Here

#C_0 = ((1,0,0),(0,1,0),(0,0,1))#
#C_1=((0, 2, 1),(0, 0, 2),(0, 0, 0))#
#C_2=((0,0,2),(0,0,0),(0,0,0))#