Suppose f ( x ) is a polynomial with complex coefficients of degree 2013 with no multiple roots. What are the last three digits of the smallest possible number of distinct complex roots of f ( f ( x ) ) ?
Details and assumptions
Each root of f ( f ( x ) ) is counted once, regardless of multiplicity.
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.
How do you get the fact that the number of distinct complex roots of a polynomial p is equal to deg { p } − deg { g cd ( p ′ , p ) } ? Thanks.
If the distinct roots of f ( x ) = 0 are λ 1 , … , λ 2 0 1 3 , then the set of roots of f ( f ( x ) ) = 0 is the set R = j = 1 ⋃ 2 0 1 3 R j = j = 1 ⋃ 2 0 1 3 { x ∈ C ∣ ∣ f ( x ) = λ j } Certainly R i ∩ R j = ∅ for i = j , and so ∣ R ∣ = ∑ j = 1 2 0 1 3 ∣ R j ∣ .
If μ ∈ R j is a repeated root of the equation f ( x ) = λ j , then f ( x ) − λ j = ( x − μ ) a g ( x ) for some a ≥ 2 and some polynomial g ( x ) . But then f ′ ( x ) = ( x − μ ) a − 1 h ( x ) for some polynomial h ( x ) , and hence μ is an ( a − 1 ) -fold repeated root of the polynomial f ′ ( x ) , which has degree 2 0 1 2 . The effect of μ is to reduce the size of ∣ R j ∣ by a − 1 . Thus ∣ R ∣ is equal to 2 0 1 3 × 2 0 1 3 minus a reduction due to the effects of repeated roots in the sets R j . Each such repeated root contributes to this reduction by an amount equal to its multiplicity as a zero of f ′ ( x ) . Thus the greatest possible theoretical reduction to ∣ R ∣ that is due to repeated roots is 2 0 1 2 , the degree of f ′ ( x ) . That would make the least theoretically possible size of ∣ R ∣ to be 2 0 1 3 × 2 0 1 3 − 2 0 1 2 = 4 0 5 0 1 5 7 It remains to show that this least theoretically possible size is achievable. If we define λ = e 2 0 1 2 i π , so that λ 2 0 1 2 = − 1 , consider f ( x ) = x 2 0 1 3 + λ . Certainly the zeros of f ( x ) are the 2 0 1 3 distinct 2 0 1 3 th roots of − λ . Since λ 2 0 1 3 + λ = λ ( λ 2 0 1 2 + 1 ) = 0 , λ is one of these. But 0 is a 2 0 1 3 -fold repeated root of f ( x ) = λ .
In other words, it is possible to achieve this maximum possible reduction of 2 0 1 2 to the size of ∣ R ∣ , and so the minimum possible size of ∣ R ∣ is indeed 4 0 5 0 1 5 7 .
If S a is the solution set of f ( x ) = a , then the number of distinct (complex) roots of f ( f ( x ) ) is ∑ a ∈ S 0 ∣ S a ∣ . (The S a sets are all distinct, since f ( x ) cannot have two different values for the same x .)
But if x is a double [triple, quadruple, etc.] root of f ( x ) = a , then it is a single [respectively double, triple, etc.] root of f ′ ( x ) = 0 . That is to say,
∑ a ∈ S 0 ( 2 0 1 3 − ∣ S a ∣ ) ≤ d e g ( f ′ ) = 2 0 1 2
so ∑ a ∈ S 0 ∣ S a ∣ ≥ 2 0 1 3 2 − 2 0 1 2 .
On the other hand, f ( x ) = ( x + 2 ) 2 0 1 3 − 1 has no multiple roots, and it yields
∑ a ∈ S 0 ∣ S a ∣ = 2 0 1 3 ∗ 2 0 1 2 + ∣ S − 1 ∣ = 2 0 1 3 ∗ 2 0 1 2 + 1
which is the same as 2 0 1 3 2 − 2 0 1 2 = 4 0 5 0 1 5 7 , and the last three digits are 1 5 7 .
Suppose f ( x ) = a ( x − a 1 ) ⋅ . . . ⋅ ( x − a 2 0 1 3 ) . Then the roots of f ( f ( x ) ) are the roots of f ( x ) − a i , for i = 1 , 2 , . . . , 2 0 1 3 . Since for i = j a i = a j , so for a fixed x f ( x ) cannot equal a i and a j simultaneously. Thus, none of the roots of f ( x ) − a i is a root of f ( x ) − a j . The following lemma is useful for more than just this problem.
Lemma. If x 0 is a root of a polynomial g , then its multiplicity in g is exactly one bigger than its multiplicity in g ′ , the derivative of g . (if x 0 is not a root of g ′ , then we say that its multiplicity in g ′ is zero).
Proof. If the multiplicity of x 0 in g is k ≥ 1 , then g ( x ) = ( x − x 0 ) k ⋅ h ( x ) , where h ( x 0 ) = 0 . So g ′ ( x ) = k ( x − x 0 ) k − 1 h ( x ) + ( x − x 0 ) k h ′ ( x ) = ( x − x 0 ) k − 1 ( k h ( x ) + ( x − x 0 ) h ′ ( x ) ) Note that k h ( x 0 ) + ( x 0 − x 0 ) h ′ ( x 0 ) = k h ( x 0 ) = 0 , so the lemma is proven.
For each i , the total number of roots of f ( x ) − a i , if counted with multiplicity, is 2 0 1 3 , for a total of 2 0 1 3 2 roots of f ( f ( x ) ) . Since we count the roots once, we must subtract k − 1 for each root x 0 of some f ( x ) − a i of multiplicity k . By the above lemma, applied to g ( x ) = f ( x ) − a i , this number k − 1 equals the multiplicity of x 0 in g ′ ( x ) . Note that for every i g ′ ( x ) = f ′ ( x ) , so the sum of the numbers ( k − 1 ) that we need to subtract, over all roots of f ( x ) − a i , is at most the degree of f ′ ( x ) , which equals 2 0 1 2 . Therefore, the total number of roots is at least 2 0 1 3 2 − 2 0 1 2 . We will now construct an example of f ( x ) that achieves this bound.
Consider f ( x ) = x 2 0 1 3 + c , where c is some complex number, such that c 2 0 1 2 = − 1 . Clearly, all roots of f are simple (by the Lemma above, or directly). Then f ( f ( x ) ) = ( x 2 0 1 3 + c ) 2 0 1 3 + c = ( x 2 0 1 3 2 + . . . + 2 0 1 3 x 2 0 1 3 c 2 0 1 2 + c 2 0 1 3 ) + c Note that c 2 0 1 3 + c = 0 , so 0 is a root of f ( f ( x ) ) with multiplicity 2 0 1 3 . Therefore, the total number of roots of f ( f ( x ) ) , counted without multiplicity, is at most 2 0 1 3 2 − 2 0 1 2 .
Finally, the last three digits of 2 0 1 3 2 − 2 0 1 2 are the same as 1 3 2 − 1 2 = 1 5 7 .
Problem Loading...
Note Loading...
Set Loading...
The key fact is that the number of distinct complex roots of a polynomial p is equal to the degree of p minus the degree of gcd ( p ′ , p ) . (This is because a root that appears with multiplicity k in p appears with multiplicity k − 1 in gcd ( p ′ , p ) .)
So if we are to minimize the number of distinct roots of f ( f ( x ) ) , we want to maximize the degree of the gcd of it and its derivative. Its derivative is f ′ ( f ( x ) ) f ′ ( x ) , and note that since f has no repeated roots, f ′ and f are relatively prime. So f ′ ( f ( x ) and f ( f ( x ) ) are relatively prime. So the gcd of f ( f ( x ) ) and its derivative reduces to gcd ( f ( f ( x ) ) , f ′ ( x ) ) .
This gcd has degree at most deg f ′ ( x ) = 2 0 1 2 , so the number of distinct zeroes is at least 2 0 1 3 2 − 2 0 1 2 . To show that this lower bound is attained, we need to find an f such that this gcd has degree 2 0 1 2 , that is, a polynomial f such that f ′ ( x ) divides f ( f ( x ) ) .
Take a to be a complex number satisfying a 2 0 1 2 = − 1 . Then let f ( x ) = x 2 0 1 3 + a . Then an easy computation shows that f ( f ( x ) ) is divisible by x 2 0 1 2 , which is a constant multiple of f ′ ( x ) . Hence we are done.
Finally, 2 0 1 3 2 − 2 0 1 2 ≡ 1 3 2 − 1 2 ≡ 1 5 7 mod 1 0 0 0 .