Keep it real (and distinct)

Calculus Level 3

Let B B be the set of all real numbers b b such that the polynomial x 4 + 4 x 3 + b x 2 + c x + d x^4+4x^3+bx^2+cx+d has four distinct real roots for some values of c c and d d . Find the supremum (i.e., the least upper bound) of B B . Enter 666 if you come to the conclusion that B B has no upper bound.

(from a recent calculus exam)


The answer is 6.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

3 solutions

Otto Bretscher
Dec 5, 2018

If f ( x ) f(x) has four distinct real roots, then f ( x ) f'(x) will have three (as in Pi Han Goh's "Method 2") and f ( x ) = 12 x 2 + 24 x + 2 b f''(x)=12x^2+24x+2b will have two, by the mean value theorem (or Rolle's theorem). The discriminant of f ( x ) f''(x) , namely, 96 ( 6 b ) 96(6-b) , will be positive, so b < 6 b<6 .

It is not hard to see that all b < 6 b<6 are attained. Note that ( x + 1 ) 4 = x 4 + 4 x 3 + 6 x 2 + 4 x + 1 (x+1)^4=x^4+4x^3+6x^2+4x+1 . We can perturb the coefficients a bit (as in Mark Hennings' solution) to create four distinct real roots while keeping the coefficient of x 2 x^2 as close to 6 6 as we wish. Consider ( ( x + 1 ) 2 ϵ ) ( ( x + 1 ) 2 2 ϵ ) = ( x + 1 ) 4 3 ϵ ( x + 1 ) 2 + 2 ϵ 2 \left((x+1)^2-\epsilon\right)\left((x+1)^2-2\epsilon\right)=(x+1)^4-3\epsilon(x+1)^2+2\epsilon^2 for positive ϵ \epsilon . The coefficient of x 2 x^2 will be 6 3 ϵ 6-3\epsilon , and we have four distinct real roots, 1 ± ϵ , 1 ± 2 ϵ -1\pm \sqrt{\epsilon},-1\pm \sqrt{2\epsilon} . Thus the supremum of B B is 6 \boxed{6} .

Pi Han Goh
Dec 4, 2018

DISCLAIMER: All the methods below are wrong because I have only proven that B < 6 B<6 , but I didn't prove that B B has a supremum of 6.


Method 1: Let p > q > r > s p>q>r>s be the real roots to the polynomial x 4 + 4 x 3 + b x 2 + c x + d x^4 + 4x^3 + bx^2 + cx + d . Then by Vieta's Formula , the following equations are true, { p + q + r + s = 4 p q + p r + p s + q r + q s + r s = b . \begin{cases} p + q + r + s = -4 \\ pq + pr + ps + qr + qs + rs = b. \end{cases} Since p > q > r > s p>q>r>s is true, the following inequality holds, ( p q ) 2 + ( p r ) 2 + ( p s ) 2 + ( q r ) 2 + ( q s ) 2 + ( r s ) 2 > 0 (p-q)^2 + (p-r)^2 + (p-s)^2 + (q-r)^2 + (q-s)^2 + (r-s)^2 > 0 Expanding and simplifying it gives 3 ( p 2 + q 2 + r 2 + s 2 ) 2 ( p q + p r + p s + q r + q s + r s ) > 0. 3(p^2 + q^2 + r^2 + s^2) - 2(pq + pr + ps + qr + qs + rs) > 0 . With p 2 + q 2 + r 2 + s 2 = ( p + q + r + s ) 2 2 ( p q + p r + p s + q r + q s + r s ) p^2 + q^2 + r^2 + s^2 = (p+q+r+s)^2 - 2(pq + pr + ps + qr + qs + rs) , we can solve for the range of b b , 3 ( ( 4 ) 2 2 b ) 2 b > 0 b < 6 . 3( (-4)^2 - 2b) - 2b > 0 \qquad \Rightarrow \qquad b < \boxed{6} .


Method 2: Since the quartic polynomial f ( x ) : = x 4 + 4 x 3 + b x 2 + c x + d f(x):=x^4 + 4x^3 + bx^2 + cx + d has 4 distinct real roots, then the cubic polynomial f ( x ) = 4 x 3 + 12 x 2 + 2 b x + c f'(x) = 4x^3 + 12x^2 + 2bx + c has 3 distinct real roots. So the (cubic) discriminant of f ( x ) f'(x) is positive, ( 12 ) 2 ( 2 b ) 2 4 ( 4 ) ( 2 b ) 3 4 ( 12 ) 3 ( c ) 27 ( 4 ) 2 ( c ) 2 + 18 ( 4 ) ( 12 ) ( 2 b ) ( c ) > 0 c 2 ( 27 ) + c ( 432 108 b ) + ( 8 b 3 36 b 2 ) < 0. (12)^2 (2b)^2 - 4(4)(2b)^3 - 4(12)^3 (c) - 27(4)^2 (c)^2 + 18(4)(12)(2b)(c) > 0 \quad \Rightarrow \quad c^2 (27) + c(432 - 108b) + (8b^3 - 36b^2 ) < 0. This quadratic inequality (in c c ) has a positive (quadratic) discriminant , ( 432 108 b ) 2 4 ( 27 ) ( 8 b 3 36 b 2 ) < 0 b < 6 . (432 - 108b)^2 - 4(27)(8b^3 - 36b^2) < 0\quad \Rightarrow \quad b< \boxed6.


Method 3: Since the quartic polynomial f ( x ) : = x 4 + 4 x 3 + b x 2 + c x + d f(x):=x^4 + 4x^3 + bx^2 + cx + d has 4 distinct real roots, then there exists a straight line ( x ) : m x + n \ell(x) : mx + n that is (below and) tangent to f ( x ) f(x) at exactly 2 points. Thus, f ( x ) ( x ) f(x) - \ell(x) has double roots, so for reals j k j\ne k , f ( x ) ( x ) = ( x j ) 2 ( x k ) 2 ( x 4 + 4 x 3 + b x 2 + c x + d ) ( m x + n ) = ( x j ) 2 ( x k ) 2 x 4 + x 3 ( 4 ) + x 2 ( b ) + x ( c m ) + ( d n ) = x 4 + x 3 ( 2 j 2 k ) + x 2 ( j 2 + 4 j k + k 2 ) + ( 2 j 2 k 2 j k 2 ) + ( j 2 k 2 ) \begin{aligned} f(x) - \ell(x) &=& (x-j)^2 (x-k)^2 \\ (x^4 + 4x^3 + bx^2 + cx + d) - (mx + n) &=& (x-j)^2 (x-k)^2 \\ x^4 + x^3(4) + x^2 (b) + x(c-m) + (d-n) &=& x^4 + x^3(-2j - 2k) + x^2(j^2 + 4jk + k^2 )+ (-2j^2k - 2jk^2) + (j^2 k^2) \end{aligned} By comparing the coefficients of the powers of x x , we get { 2 j 2 k = 4 j 2 + 4 j k + k 2 = b { j + k = 2 j k = b 4 2 . \begin{cases} -2j - 2k = 4 \\ j^2 + 4jk + k^2 = b \end{cases} \quad \Rightarrow \quad \begin{cases} j + k = -2 \\ jk = \frac{b-4}2. \end{cases} And finally ( j k ) 2 > 0 j 2 + k 2 2 j k > 0 ( j 2 + k 2 + 4 j k ) 6 j k > 0 b 6 ( b 4 2 ) > 0 b < 6 . (j-k)^2 > 0 \quad \Rightarrow \quad j^2 + k^2 - 2jk > 0\quad \Rightarrow \quad (j^2 + k^2 + 4jk) - 6jk > 0 \quad \Rightarrow \quad b - 6 \left( \frac{b-4}2\right) > 0 \quad \Rightarrow \quad b < \boxed6.

Thank you, Comrade! All of your methods show that b < 6 b<6 , but are you actually showing (explicitly) that the supremum is 6 (and not less)?

Otto Bretscher - 2 years, 6 months ago

Log in to reply

Nope. All of these are wrong. I just don't feel like deleting my solution because I hope to come back with a complete proof in the near future.

Pi Han Goh - 2 years, 6 months ago

Log in to reply

It's not wrong, just incomplete. You need to show that all values b < 6 b<6 are attained, which is the easy part.

Otto Bretscher - 2 years, 6 months ago

Log in to reply

@Otto Bretscher What if you change the polynomial x 4 + 4 x 3 + b x 2 + c x + d x^4 + 4x^3 + bx^2 + cx +d such that the coefficient of x 3 x^3 is no longer 4 (let's call that number G G )? Can you still figure the answer out in terms of (any value of) G G ?

Pi Han Goh - 2 years, 6 months ago

Log in to reply

@Pi Han Goh Yes, my solution will still work, mutatis mutandis. You will get the condition b < 3 G 2 8 b < \frac{3G^2}{8} . The limiting polynomial in my second paragraph will be ( x + G 4 ) 4 \left(x+\frac{G}{4}\right)^4 . You can see why I picked G = 4 G=4 ;)

Otto Bretscher - 2 years, 6 months ago

Log in to reply

@Otto Bretscher Yup, I finally got it. I don't know why I kept messing up my algebra. Thank you.

Pi Han Goh - 2 years, 5 months ago
Mark Hennings
Dec 4, 2018

If f ( x ) = x 4 + 4 x 3 + b x 2 + c x + d f(x) = x^4 + 4x^3 + bx^2 + cx + d has four distinct real zeros, then f ( x 1 ) = x 4 + ( b 6 ) x 2 + ( c 2 b + 8 ) x + ( d c + b 3 ) f(x-1) \; = \; x^4 + (b-6)x^2 + (c-2b+8)x + (d-c+b-3) has four distinct real zeros that add to zero. If these zeros are α , β , γ , δ \alpha,\beta,\gamma,\delta , then b 6 = α β + α γ + α δ + β γ + β δ + γ δ = 1 2 [ ( α + β + γ + δ ) 2 ( α 2 + β 2 + γ 2 + δ 2 ) ] = 1 2 ( α 2 + β 2 + γ 2 + δ 2 ) b-6 \; = \; \alpha\beta + \alpha\gamma + \alpha\delta + \beta\gamma + \beta\delta + \gamma\delta \; = \; \tfrac12\big[(\alpha+\beta+\gamma+\delta)^2 - (\alpha^2 + \beta^2 + \gamma^2 +\delta^2)\big] \; = \; -\tfrac12(\alpha^2 + \beta^2 + \gamma^2 + \delta^2) so that b = 6 1 2 ( α 2 + β 2 + γ 2 + δ 2 ) < 6 b \; = \; 6 - \tfrac12(\alpha^2+\beta^2+\gamma^2+\delta^2) \; < \; 6 By choosing f ( x ) = ( ( x + 1 ) 2 u 2 ) ( ( x + 1 ) 2 v 2 ) = ( x + 1 ) 4 ( u 2 + v 2 ) ( x + 1 ) 2 + u 2 v 2 f(x) \; = \; \big((x+1)^2-u^2\big)\big((x+1)^2 - v^2\big) \; = \; (x+1)^4 - (u^2+v^2)(x+1)^2 + u^2v^2 for distinct positive reals u , v u,v , we see that we can certainly achieve b = 6 u 2 v 2 b = 6 - u^2 - v^2 for any such u , v u,v . Thus B = ( , 6 ) B = (-\infty,6) , so the supremum is 6 \boxed{6} .

Wait, what's wrong with the following method?

Let p , q , r , s p,q,r,s denote the roots of the given polynomial. And define S 1 S_1 and S 2 S_2 as the 1st and 2nd symmetric sum of p , q , r , s p,q,r,s , respectively. Then the sum of squares of these 4 numbers must be positive (as at most one of p , q , r , s p,q,r,s can be 0 only), p 2 + q 2 + r 2 + s 2 = S 1 2 2 S 2 = ( 4 ) 2 2 b > 0 b < 8 . p^2 + q^2 + r^2 + s^2 = S_1^2 - 2S_2 = (-4)^2 - 2b > 0 \Rightarrow b < \boxed8.

Why did I get 8 instead of 6? I fulfilled all the criteria, no?

This makes me skeptical of all my other methods in my solution below.

Pi Han Goh - 2 years, 6 months ago

Log in to reply

It is certainly true that b < 8 b < 8 . It is just that this inequality is not sharp. You are basing your argument for 8 8 on the (true) statement that p 2 + q 2 + r 2 + s 2 > 0 p^2 + q^2 + r^2 + s^2 > 0 . However, you are interested in finding a lower bound for this sum of squares given the restriction that p + q + r + s = 4 p+q+r+s=-4 . The minimum value of p 2 + q 2 + r 2 + s 2 p^2 + q^2 + r^2 + s^2 , given that p + q + r + s = 4 p+q+r+s=-4 , is 4 4 , obtained when p = q = r = s = 1 p=q=r=s=-1 . Since the roots are distinct we have p 2 + q 2 + r 2 + s 2 > 4 p^2 + q^2 + r^2 + s^2 > 4 , given the restriction. Thus 16 2 b > 4 16-2b > 4 , and so b < 6 b < 6 .

Mark Hennings - 2 years, 6 months ago

Log in to reply

BUT OF COURSE!!!!!

Pi Han Goh - 2 years, 6 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...