Mathematical induction question

pc2-brazil
Messages
198
Reaction score
3

Homework Statement


Suppose a and b are real numbers with 0 < b < a. Show that, if n is a positive integer, then
a^n - b^n \leq na^{n-1}(a-b)

Homework Equations



The Attempt at a Solution


I'm trying to show this by induction.
Let P(n) be the proposition that a^n - b^n \leq na^{n-1}(a-b)
I've already verified that P(1) is true, which completes the basis step.
Inductive step:
I must show that, if P(k), then P(k+1).
So, I first assume that this is true for an arbitrary k: a^k - b^k \leq ka^{k-1}(a-b)
Then, I must show that, if P(k) is true, it follows that a^{k+1} - b^{k+1} \leq (k+1)a^k(a-b).
This is where I'm having trouble.
I'm trying to find that a^{k+1} - b^{k+1} is less than or equal to an expression involving a^k - b^k, so that I can use the expression for P(k) to derive an inequality for a^{k+1} - b^{k+1}.
I've tried several ways, like trying to rewrite a^{k+1} - b^{k+1} as aa^k - bb^{k} and then writing that aa^k - bb^k \geq aa^k - ab^k = a(a^k - b^k) (since a > b), but this doesn't help, because I'm looking for something that a^{k+1} - b^{k+1} is less than or equal to, not greater than or equal to.

Thank you in advance.
 
Physics news on Phys.org
EDIT: nevermind. i though i had it but i did not
 
Last edited:
i had one thought, which may or may not be helpful.

if b < a, then a = b + c for some positive number c.

so then your problem becomes:

(b+c)^n - b^n \leq n(b+c)^{n-1}c

now expand the powers of b+c with the binomial formula.
 
It can be instructive to see how the inequality works for the first few values of n. This isn't good enough for a proof, but it may give you some ideas for the proof.

For n=2:
\displaystyle a^2-b^2\ \ ?\,\leq\,? \ \ 2a^1(a-b)

Factor the left side.

\displaystyle (a-b)(a+b)\ \ ?\,\leq\,? \ \ 2a(a-b)

Of course, (a+b) < 2a because a < b .

So, this case doesn't appear to need induction.​

For n=3:
We have a somewhat similar result with the difference of cubes giving a3-b3 = (a-b)(a2+ab+b2). Then the inequality holds if a2+ab+b2 ≤ 3a2, and this is true for a < b. Again, this doesn't require induction.​

For general n:
Did you know that \displaystyle a^n-b^n=(a-b)(a^{n-1}+a^{n-2}b+a^{n-3}b^{2}+\dots+a^{((n-1)-k)}\,b^{k}+\dots +a^{2}b^{n+3}+a\,b^{n-2}+b^{n-1})\,?

Each of the n terms in the last factor is less than or equal to an-1.
For instance: \displaystyle a^{12}-b^{12}=(a-b)(a^{11}+a^{10}b+a^{9}b^{2}+\dots+a^{5}\,b^{6}+ \dots +a^{2}b^{9}+a\,b^{10}+b^{11})​
 
Building on Sammy's post, you might find it easier to put x = b/a, where 0<x<1

Use polynomial long division to see that for positive integral n,

(1-x^n) = (1-x)(1+x+x^2+...x^{n-1})

Alternatively you can just multiply the terms on the RHS, cancel out lots of adjacent terms and end up with the LHS. This is one of the polynomial identities which it's really good to know.

Now,

1+x+x^2+...x^{n-1} has n terms, each of which is less than or equal to 1. Hence the whole sum is less than or equal to n.

So,

a^n - b^n = a^n(1 - x^n) = a^n(1-x)(1+x+x^2+...+x^{n-1}) = a(1-x).a^{n-1}.(1+x+x^2+...+x^{n-1}) = (a-b).a^{n-1}.(1+x+x^2+...+x^{n-1}) \leq (a-b).n.a^{n-1}

and you're done.
 
Last edited:
SammyS said:
It can be instructive to see how the inequality works for the first few values of n. This isn't good enough for a proof, but it may give you some ideas for the proof.

For n=2:
\displaystyle a^2-b^2\ \ ?\,\leq\,? \ \ 2a^1(a-b)

Factor the left side.

\displaystyle (a-b)(a+b)\ \ ?\,\leq\,? \ \ 2a(a-b)

Of course, (a+b) < 2a because a < b .

So, this case doesn't appear to need induction.​

For n=3:
We have a somewhat similar result with the difference of cubes giving a3-b3 = (a-b)(a2+ab+b2). Then the inequality holds if a2+ab+b2 ≤ 3a2, and this is true for a < b. Again, this doesn't require induction.​

For general n:
Did you know that \displaystyle a^n-b^n=(a-b)(a^{n-1}+a^{n-2}b+a^{n-3}b^{2}+\dots+a^{((n-1)-k)}\,b^{k}+\dots +a^{2}b^{n+3}+a\,b^{n-2}+b^{n-1})\,?

Each of the n terms in the last factor is less than or equal to an-1.
For instance: \displaystyle a^{12}-b^{12}=(a-b)(a^{11}+a^{10}b+a^{9}b^{2}+\dots+a^{5}\,b^{6}+ \dots +a^{2}b^{9}+a\,b^{10}+b^{11})​
Very interesting idea.
I knew that result for an - bn, but I didn't think of using it.
This leads directly to the desired result, not requiring induction.
 
There are two things I don't understand about this problem. First, when finding the nth root of a number, there should in theory be n solutions. However, the formula produces n+1 roots. Here is how. The first root is simply ##\left(r\right)^{\left(\frac{1}{n}\right)}##. Then you multiply this first root by n additional expressions given by the formula, as you go through k=0,1,...n-1. So you end up with n+1 roots, which cannot be correct. Let me illustrate what I mean. For this...
Back
Top