-1

This a problem from Mathematics for Computer Science. I'm wondering why this proof of AM-GM is a bogus proof.

Proof: for all nonnegative real numbers $a$ and $b$, show that $$ \frac{a+b}{2} \geq \sqrt{ab} $$

bogus proof. $$ \begin{align*} \frac{a+b}{2} &\overset{?}{\geq} \sqrt{ab} \\ \implies a+b &\overset{?}{\geq} 2 \sqrt{ab} \\ \implies a^2 + 2ab + b^2 &\overset{?}{\geq} 4ab \\ \implies a^2 - 2ab + b^2 &\overset{?}{\geq} 0 \\ \implies (a-b)^2 &\overset{?}{\geq} 0 \end{align*} $$ and the last statement is true because the square of a real number is never negative. This proves the claim.

Thomas Andrews
  • 186,215
  • 8
    It is a valid proof if you just change $\implies$ to $\impliedby$ throughout. – Kavi Rama Murthy Jan 31 '25 at 09:39
  • the problem in the proof you wrote, it's that you suppose the result already true... so it can't work... – Surb Jan 31 '25 at 09:44
  • I would call this an "upside-down proof". Correct proofs start with what you know to be true and work towards what you want to prove. – Henry Jan 31 '25 at 09:46
  • Related: https://math.stackexchange.com/questions/557653 – Henry Jan 31 '25 at 09:47
  • Why do you say it is bogus? It doesn't generalize to more variables tan $2$ easily, but it works fine for $2.$ This,is how it is,often first proved. – Thomas Andrews Jan 31 '25 at 09:52
  • 1
    You do have to note that $a,b\geq0.$ And the proper proof doesn't use the "question mark" inequality language, and reverses the order of the inequalities: $$(a-b)^2\geq0\a^2-2ab+b^2\geq0\\vdots$$ – Thomas Andrews Jan 31 '25 at 09:56

1 Answers1

1

The proof appears convincing at first glance, but it is bogus because of a flaw in the reasoning.

The correct proof of AM-GM should start with the fact that $ (a - b)^2 \geq 0$ and then deduce $ \frac{a+b}{2} \geq \sqrt{ab} $.
Instead, this proof assumes the inequality to be true (by attempting to "prove" it via algebraic manipulations) and then derives a true statement.
This is circular reasoning—it does not actually prove the original inequality but rather confirms that if the inequality holds, then a true statement follows.

A valid proof of AM-GM follows from starting with $ (a - b)^2 \geq 0 $:

$$ a^2 - 2ab + b^2 \geq 0 $$

Rearrange:

$$ a^2 + b^2 \geq 2ab $$

Adding $2ab $ to both sides:

$$ a^2 + 2ab + b^2 \geq 4ab $$

Taking the square root:

$$ a + b \geq 2\sqrt{ab} $$

Dividing by 2:

$$ \frac{a+b}{2} \geq \sqrt{ab} $$

This approach proves the inequality without assuming it.

Second proof

To prove the AM-GM inequality using contradiction, assume that the inequality is false. That is, suppose there exist nonnegative real numbers $a$ and $b$ such that:

$$ \frac{a+b}{2} < \sqrt{ab}. $$

Multiplying both sides by 2 gives:

$$ a + b < 2\sqrt{ab}. $$

Squaring both sides (which preserves the inequality because both sides are nonnegative) results in:

$$ (a + b)^2 < 4ab. $$

Expanding the left-hand side:

$$ a^2 + 2ab + b^2 < 4ab. $$

Subtracting $2ab$ from both sides:

$$ a^2 - 2ab + b^2 < 0. $$

But the left-hand side is $ (a - b)^2$, so we get:

$$ (a - b)^2 < 0. $$

This is a contradiction because the square of any real number is always nonnegative, i.e., $(a - b)^2 \geq 0$.
Thus, our assumption that $ \frac{a+b}{2} < \sqrt{ab}$ must be false.
Therefore, we conclude that:

$$ \frac{a+b}{2} \geq \sqrt{ab}. $$

and we are done. This completes the proof by contradiction.

Below I provide a relevant example from Liebeck's A Concise Introduction to Pure Mathematics which make help you understand better.

enter image description here