5

Prove that $${ a }^{ 2 }+2ab+{ b }^{ 2 }\ge 0,\quad\text{for all }a,b\in \mathbb R $$ without using $(a+b)^{2}$.

My teacher challenged me to solve this question from any where. He said you can't solve it. I hope you can help me to solve it.

  • 3
    Use $AM>=GM$ inequality – Archis Welankar Jan 25 '16 at 15:25
  • Take partial derivatives, alternatively. – MathematicsStudent1122 Jan 25 '16 at 15:30
  • 3
    Can you use $(a-b)^2 \geq 0$ ? – lisyarus Jan 25 '16 at 15:31
  • 1
    @ArchisWelankar AM-GM inequality applies to nonnegative numbers, no? – leonbloy Jan 25 '16 at 15:31
  • he said you can't – Science Addict Jan 25 '16 at 15:32
  • @MathematicsStudent1122: If you're going to use calculus, you don't really need partial derivatives. For any $b$ you can differentiate with respect to $a$ to find that the expression has a minimum of $0$ at $a=-b$. – Brian M. Scott Jan 25 '16 at 15:35
  • Is inspecting $a(a+b)+b(a+b)$ for all possible combinations of $a$ and $b$ either positive or negative allowed? Perhaps a bit lengthy, but also really trivial :) – Eric S. Jan 25 '16 at 15:35
  • 1
    @leonbloy WLOG suppose $a>0,$ $b<0$ and apply AM-GM to $a$ and $-b.$ – Justpassingby Jan 25 '16 at 15:42
  • @BrianM.Scott If I'm not mistaken, when considering functions of two variables, fixing one variable and differentiating with respect to the other is partial differentiation. – MathematicsStudent1122 Jan 25 '16 at 16:05
  • The polynomial is homogeneous, so assuming $a \not= 0$ (in which case inequality is trivially true), divide through by $a^2 > 0$ to get $1+2\frac{b}{a}+\left(\frac{b}{a}\right)^2 \geq 0$, and now you only have one variable, and the expression achieves a minimum at $\frac{b}{a} = -1$. – Brian Tung Jan 25 '16 at 17:27
  • So, basically, this is more-or-less equivalent to proving $x^2-2x+1\ge0$, or $x^2+1\ge2x$. – Akiva Weinberger Jan 25 '16 at 21:09
  • @MathematicsStudent1122: The mechanism is the same, but the ideas are logically distinct. With partial differentiation you’re treating it as a function of two variables. My approach treats it as a function of one variable, $a$, and simply proves that no matter what $b$ is, the minimum occurs at $a=-b$ and is $0$. In other words, instead of treating $b$ as a second variable, it treats it as an unspecified constant. – Brian M. Scott Jan 26 '16 at 00:12

8 Answers8

11

Alternatively, you can study the familly of functions

$$f_b(x) = x^2 +2bx + b^2$$

It's easy to see that the minimum of this function is attained at $x = -b$, so the minimum of $f_b$ is $0$ for all values of parameters $b$, hence $\forall a,b, \quad a^2+2ab+b^2 \geq 0$

Tryss
  • 14,310
6

Define a function $$f(a,b)=a^2+2ab+b^2$$

And try to find its extrema:

$$\begin{cases} f'_a=0\\ f'_b=0 \end{cases}$$

$$\begin{cases} 2a+2b=0\\ 2b+2a=0 \end{cases}$$

$$\begin{cases} a+b=0 \end{cases}$$

Since $f''_{ab}{}^2-f''_{aa} f''_{bb}=4-4=0$ there may be an extremum or not. Try to compute $f(a,-a-1)$, $f(a,-a)$ and $f(a,-a+1)$. You will see that

$$f(a,-a-1)=f(a,-a+1)=1\quad\text{but}\quad f(a,-a)=0$$

It means that this function has infinitely many weak minima that lay on the line $a+b=0$. It follows that

$$f(a,b)\geqslant 0$$

since $f$ is continuous (and smooth) as a polynomial.

Kamil Jarosz
  • 4,984
5

This is more of an intuitive explanation:

The statement is clear when $a$ and $b$ have the same sign, because the left hand side is strictly positive. This reduces to showing

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

for $a,b > 0$.

Think in terms of money. Assume you have an $a$-valued coin, and a $b$-valued coin. Assume further that $a \geq b$. Say you are allowed to take $x > 0$ of one coin and $y > 0 $ of another coin. How would you maximize your profit? Of course you take most of the coin with the largest value, and least of the one with lowest value. This would be greater than taking the least of the coin with the greatest value and the most of the coin with the lowest value. Now take $x = a$ and $y = b$.

Improve
  • 1,788
4

Case 1. If both $a,b$ are positive , then each term $a^2, 2ab, b^2$, all are positive. Hence $a^2 + 2ab +b^2>0$.

Case 2. If both $a,b$ are negative, then also each term $a^2, 2ab, b^2$, all are positive. Hence $a^2 + 2ab +b^2>0$.

Case 3. If one of $a,b$ is positive and other is negative, then in this case by changing one of its sign sufficient to show $a^2 -2ab+b^2 > 0$, where $a,b$ both are positive. So, without loss of generality assume $a>b$. Here is a pictorial proof of that fact.

pictorial proof

Timon
  • 2,401
3

Without loss of generality suppose $0<-b<a$ and write $c=-b.$

Multiplying a positive number by the positive scale factor $r=\frac a c>1$ has more effect on large positive numbers than on small positive numbers. Thus

$$ra-a>rc-c$$

in other words

$$a^2/c-a>a-c$$

multiplying by $c$ yields

$$a^2-ac>ac-c^2$$

or

$$a^2+2ab+b^2>0$$

(the inequality is strict because we ruled out the trivial case $c=a$ up front)

Justpassingby
  • 10,029
3

Express the pair $(a,b) \in \mathbb{R}^2$ using polar coordinates

$$a=r \cos \theta,b=r \sin \theta .$$

We have

$$a^2+2ab+b^2=r^2+2r^2 \cos \theta \sin \theta =r^2(1+\sin 2 \theta) \geq 0.$$

user1337
  • 24,381
1

Using the rearrangement inequality on two instances of the sequence $\{a,b\}$

$ ab + ba \le a^2 + b^2$

Intuitive reasoning, if you're multiplying two sequences of numbers pairwise, the maximal value is when the sequences are ordered the same way. Simplifying further, assume both sequences are positive numbers, take one sequence fixed and pick multipliers from the second sequence as weights. You'll achieve the maximum value if you multiple the largest number with a the largest weight and so on.

karakfa
  • 2,675
1

If $a,b$ have the same sign then all summands are non-negative. So, let $a>0$ and $b<0$ (without loss) and write $$a^2+2ab+b^2=a^2-2a|b|+b^2=a^2-a|b|+|b|^2-a|b|=a(a-|b|)+|b|(|b|-a)$$ Now, take two more cases

  1. if $a\ge |b|$, then $a(a-|b|)+|b|(|b|-a)\ge a(a-|b|)+a(|b|-a)=0$ and similarly
  2. if $a<|b|$, then $a(a-|b|)+|b|(|b|-a)>|b|(a-|b|)+|b|(|b|-a)=0$.
Jimmy R.
  • 35,868