Prove that, for any positive integer n: $(a + b)^{n} \leq 2^{n-1}(a^{n}+b^{n}) $ I tried induction theorem, when $n = 1$ it is obviously right. But, say $n=k$, It does not make sense since I cannot expand the $2^{k-1}$($a^{k}$+$b^{k}$).

And I also looked through this similar question but no help.

Prove: $(a + b)^{n} \geq a^{n} + b^{n}$

  • $\begingroup$ An identical question was asked earlier. $\endgroup$ – Anant Mar 19 '14 at 5:38

The simpliest proof:

as $x\to x^n$ is convex, $$ \left(\frac {x+y}2\right)^n \le \frac {x^n+y^n}2 $$ Now multiply by $2^n$ and you are done.

  • $\begingroup$ Can you explain a little more? Our course do not cover things like the "convex". Do you mean increasing? $\endgroup$ – SonicFancy Mar 5 '14 at 23:40
  • $\begingroup$ convex mean, in a context of a real valued differential function, mean that the derivative function is increasing. If this is the case, a theorem yields $a_i > 0 \ \&\ \sum a_i = 1 \Rightarrow f(\sum a_i x_i) \le \sum a_i f(x_i)$. For exemple, the absolute value is convex. $\endgroup$ – mookid Mar 5 '14 at 23:46
  • $\begingroup$ @SonicFancy: why did you unaccept? $\endgroup$ – mookid Oct 26 '14 at 22:10

First of all, this is only true for positive $a$, $b$.

For the induction step: suppose $(a+b)^n \leq 2^{n-1}(a^n+b^n)$ for all $a,b>0$. Multiplying both sides of the inductive hypothesis by $a+b$ (here we use $a+b>0$), we find $(a+b)^{n+1} \leq 2^{n}(a^{n} + b^{n})(a+b)$. It remains to show that $(a^n+b^n)(a+b) \leq 2(a^{n+1}+b^{n+1})$. This is equivalent to $(a^n-b^n)(a-b) \geq 0$, which is true as $x \mapsto x^n$ is an increasing function.

This is the generalized mean inequality for two variables (comparing the arithmetic mean and the $n$-th power mean).

  • $\begingroup$ After multiplying a+b to the supposed part, is that should still be $2^{n-1}$? $\endgroup$ – SonicFancy Mar 5 '14 at 23:44
  • $\begingroup$ And should I consider that the basecase is n=1? $\endgroup$ – SonicFancy Mar 5 '14 at 23:49

$$(a+b)^n\leq 2^{n-1}(a^n+b^n)\Longleftrightarrow\left(\frac{a+b}2\right)^n\leq\frac{a^n+b^n}2\Longleftrightarrow\frac{a+b}2\leq\sqrt[n]\frac{a^n+b^n}2$$ Which is a special case of the generalized mean inequality if $a,b\geq0$ and $n\geq1$ (can even be real).

The full inequality gives you a nice generalized bound $$\left(\sum_{i=1}^k a_i\right)^n\leq k^{n-1}\sum_{i=1}^k a_i^n$$

  • $\begingroup$ This is a little bit hard for me to understand but still thanks a lot. $\endgroup$ – SonicFancy Mar 5 '14 at 23:46

Also, the TL method helps.

Indeed, since our inequality is homogeneous, we can assume that $a+b=2$ and we need to prove that $$a^n+b^n\geq2$$ or $$(a^n-na+n-1)+(b^n-nb+n-1)\geq0,$$ which is true by AM-GM: $$\sum_{cyc}(a^n-nb+n-1)\geq\sum_{cyc}\left(n\sqrt[n]{a^n\cdot1^{n-1}}-na\right)=0$$ and we are done!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.