An awkward sum
I’m not sure where I got this question from. It asks for the value of the infinite sum:
Have a go if you’d like; spoilers are below the line.
—
The first thing I would do is work out the general term. The numerators are going up by two each time, so I reckon it’s
How do we get that? Partial fractions look like they might be our friend.
It turns out that it’s a bit less messy to work out
[*]
Aside: let’s solve this two ways
First way: simultaneous equations. I don’t much care for this way, but it seems that some do.
Matching coefficients:
:
This gives us
Alternatively, we can substitute any numbers we like into [*] and get:
: , so , so , as before.
Back to the question at hand
So,
Let’s write out a few terms of that:
k | |||||
1 | |||||
2 | . | ||||
3 | . | . |
It’s hopefully quite clear that the terms with a denominator of 3 sum to zero, and so will the terms with a denominator of 4, 5 and so on.
All we’re left with are the denominators of 1 (which sum to 1) and 2 (which sum to
Alternatively
We could use generating functions. It’s massively overkill, but it’s a good excuse to use the technique. Let’s imagine that our sum is in fact a function of
We can reuse our partial fractions and rewrite that as
And what we have there are three logarithmic series. Let’s examin them in turn:
-
. -
.
This one’s a bit trickier. I would start with
- or more nicely,
So, altogether,
That screams “regroup!” at me. I’m going to turn the whole thing into a single fraction:
We’re interested in the value of this when
Fortunately, the factor in front of the logarithm also evaluater to zero, and
So, in the limit, we’re left with
—
I’m absolutely not saying that the generating functions way is easier, but the technique is an interesting and wide-ranging one. Try applying it to your favourite difficult sequences today!
Footnotes:
1. Prove this, if you’d like to.