Find the sum of the values of the parameter a such that for some non-constant polynomials with real coefficients f ( x ) and g ( x ) ( g ( x ) ) 2 = ( f ( x ) ) 3 − 1 2 ( f ( x ) ) 2 + 4 5 f ( x ) − a
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.
Very nice explanation! Totally different from my thinking and I learned something! Anyway I thought the way that I solved this problem would be very similar to how a lot of you have. Well since I can't go back to submit a solution, I have to take up some of your valuable replying space for my solution:
In contrast to your analysis on degrees, I will focus on roots.
Since we have a polynomial equal to some other polynomial squared, this means that the roots of the first polynomial(namely the RHS) have multiplicity of an even number.
Let r 1 , r 2 , r 3 be the roots of x 3 − 1 2 x 2 + 4 5 x − a . Consider the three equations f ( x ) = r 1 , r 2 , r 3 , the solutions of these equations will give us all the roots of RHS (because if the degree of f ( x ) is n , then the three equations give us 3 n roots and the degree of f ( x ) 3 is 3 n ).
If r 1 , r 2 , r 3 are all distinct, then the three solution sets from the three equations have no common elements because we can't have some x 1 such that f ( x 1 ) = r 1 = r 2 . Since we are aiming for roots with a multiplicity of an even number, these three solution sets must each individually have multiplicity of an even number. In other words, f ( x ) − r 1 , r 2 , r 3 must all be equal to a polynomial squared. Of course no such f ( x ) could satisfy that because subtracting r i only changes the constant term. This means that r 1 , r 2 , r 3 cannot all be distinct.
Next, we investigate the case that WLOG r 1 , r 2 = r 3 are the roots. Then since we get the same two solution sets for f ( x ) = r 2 = r 3 , we have a set of roots with multiplicity of an even number. On top of that, f ( x ) = r 1 must have roots with a multiplicity of an even number and, surely, such f ( x ) can be constructed given any r 1 . So r 1 , r 2 = r 3 as roots for x 3 − 1 2 x 2 + 4 5 x − a can give us f ( x ) which makes the RHS=polynomial square. Now we just have to find all the a that makes: x 3 − 1 2 x 2 + 4 5 x − a = ( x − r 1 ) ( x − r 2 ) 2 = x 3 − ( 2 r 2 + r 1 ) x 2 + ( r 2 2 + 2 r 1 r 2 ) x − r 1 r 2 2 By comparing coefficients we have equations: 2 r 2 + r 1 = 1 2 , r 2 2 + 2 r 1 r 2 = 4 5 Solving this gives solutions ( r 1 , r 2 ) = ( 6 , 3 ) = ( 2 , 5 ) , which makes a = 5 0 , 5 4
Finally, we have the case r 1 = r 2 = r 3 . Like the case above, we can certain construct a function f such that the solution set for f ( x ) = r 1 = r 2 = r 3 has a multiplicity of an even number. So let's see if there's an a to make this case possible: x 3 − 1 2 x 2 + 4 5 x − a = ( x − r 1 ) 3 = x 3 − 3 r 1 x 2 + 3 r 1 2 x − r 1 3 which gives us a no solution, so no a for this case.
Since these are all the cases, and a = 5 0 , 5 4 is all we have, this gives answer 5 0 + 5 4 = 1 0 4 □
Side note: It seems like the phrase [multiplicity of an even number] took up some space in my solution, I don't have a shorter word to describe this at the moment so sorry!
Log in to reply
Nice solution! And definitely more "elementary" than my posted one :)
Sometime ago, I read about the following lemma:
Lemma : Suppose we have some distinct complex numbers (because for us to apply to this question, we cannot assume that they are reals), let's denote them as p 1 = p 2 = … = p n . Let us also select arbitrary polynomials f , g over C . (not to be confused with those in the problem at the moment; we will see the application later) Then we have that the following statement ( i ) implies statement ( i i ) :
( i ) f ( x ) 2 = ∏ i = 1 n ( g ( z ) − p i )
( i i ) f ( z ) , g ( z ) are constant.
I think at this moment, one can see where I'm getting; the question defines the polynomials as
non-constant
So if we can prove the lemma, we would have shown that in the question, (still retaining the notation of P in my solution), we need P ( x ) to have multiple roots. In essence, this solution is equivalent to yours, and the lemma is essentially just a generalisation of your "multiplicity of an even number" argument.
And oh, by the way, defining P and taking derivatives is quite a neat trick. Because of double roots, (in other words, the lemma) P ′ ( x ) = 3 ( x − 3 ) ( x − 5 ) = 0 needs to hold. Therefore, this is also a triple streamlined version of my posted solution.
So look no further! We will prove this crucial lemma.
Proof : Well, we are going to employ proof by contradiction. Basically, we are going to apply your trick of the multiplicity and note first that C is a UFD , this is because C is a field, and all fields are automatically UFDs.
So, suppose for the sake of contradiction that such f , g exist. (Again, sorry for the clash in notation, we are referring to the lemma) By definition, p i are distinct so clearly g ( x ) − p i for i = 1 , … , n are disjoint from each other. So, since the roots on the LHS of ( i ) come in pairs, the roots of g ( x ) − p i come in pairs too. To derive a contradiction, we first simplify the observations that we have made and represent them in an easier-to-visualise form. For instance, we can factor f ( x ) as follows:
f ( x ) = ∏ i = 1 n f i ( x ) where we define f i ( x ) 2 = g ( x ) − p i for i = 1 , … , n
What happens if we choose two distinct indices r , s ? Squares should remind one of the difference of squares , which is a neat representation. So, consider:
( f r ( x ) + f s ( x ) ) ( f r ( x ) − f s ( x ) ) = ( g ( x ) − p r ) − ( g ( x ) − p s ) = p r − p s = 0 (remember that p i are distinct by assumption)
which is a constant. Clearly this implies that both f ( x ) and g ( x ) are constant, contradiction. □
I found a simpler solution, building up on the idea that if non-constant polynomials f , g exist, then we must have that P ( x ) has multiple roots .
Although this is not immediately obvious, it is hinted by my solution above. As above, we remark that for some n ∈ Z + , we have de g f = 2 n and de g g = 3 n .
So to deal with multiple roots, like I mentioned, it is always a good idea to take derivatives . Phrasing our goal another way, we need only prove that P is not coprime with P ′ . So retaining the notations of my solution posted, we have:
P ′ ( f ( x ) ) f ′ ( x ) = 2 g ( x ) g ′ ( x ) ( ∗ ) .
Now notice the following:
If g cd ( P , P ′ ) = 1 then there are polynomials t , v such that P ( y ) t ( y ) + P ′ ( y ) v ( y ) = 1 . Subbing in f ( x ) for y , we get:
P ( f ( x ) ) t ( f ( x ) ) + P ′ ( f ( x ) ) v ( f ( x ) ) = 1
So as a result, we have g cd ( P ( f ( x ) ) , P ′ ( f ( x ) ) ) = 1 . Recall that ( g ( x ) ) 2 = P ( f ( x ) ) , so this implies that g cd ( g ( x ) , P ′ ( x ) ) = 1 so by ( ∗ ) , we in fact have g ( x ) ∣ f ′ ( x ) . However, with the beginning observation of degrees, de g g = 3 u > 2 u − 1 = de g f ′ which is impossible.
We can derive similar conclusions as the solution posted.
You may be curious to know that the motivation for the original problem came from the theory of elliptic curves. When a degree three polynomial p ( x ) has three distinct (complex) roots, the equation y 2 = p ( x ) defines an algebraic curve, known as elliptic curve. When two or all three roots are the same, this curve "degenerates" to a rational curve, i.e., admits a parametrization by rational functions. The Wikipedia page on elliptic curves is very informative, though not easy to read.
Good job! :-)
Also a minor typo: it should be de g f ′ < d − 1 , sorry about that.
Let's say f ( x ) 3 − 1 2 f ( x ) 2 + 4 5 f ( x ) − a = ( f ( x ) − c 1 ) ( f ( x ) − c 2 ) ( f ( x ) − c 3 ) Then divide into the following cases :
i) All the roots are the same ( c 1 = c 2 = c 3 ) , from which we will get 3 c 1 = 1 2 ⇒ c 1 = 4 But then the right hand side will be ( f ( x ) − 4 ) 3 = f ( x ) 3 − 1 2 f ( x ) 2 + 4 8 f ( x ) − 6 4 = f ( x ) 3 − 1 2 f ( x ) 2 + 4 5 f ( x ) − a for any a . Hence, no solution from this case.
ii) Exactly two of the roots are the same (WLOG, c 1 = c 2 ) , from which we will get 2 c 1 + c 3 = 1 2 2 c 1 c 3 + c 1 2 = 4 5 From the 1st equation, substitute c 3 = 1 2 − 2 c 1 to the 2nd equation, and we will get 3 c 1 2 − 2 4 c 1 + 4 5 = 0 and thus the following pairs ( c 1 , c 3 ) = ( 3 , 6 ) , ( 5 , 2 ) . One can easily get a = 5 4 , 5 0 , respectively. And any of this a will work since now f ( x ) 3 − 1 2 f ( x ) 2 + 4 5 f ( x ) − a = ( f ( x ) − c 1 ) ( f ( x ) − c 1 ) ( f ( x ) − c 3 ) = ( f ( x ) − c 1 ) 2 ( f ( x ) − c 3 ) and we can choose f ( x ) = h ( x ) 2 + c 3 for some polynomial h so that the above's form is a square of a polynomial.
(iii) All the roots are different. In this case, let's say p is a complex root of g ( x ) , then we must have even power of ( x − p ) in g ( x ) 2 and so in ( f ( x ) − c 1 ) ( f ( x ) − c 2 ) ( f ( x ) − c 3 ) too. But then, p is the root of exactly one of f ( x ) − c 1 , f ( x ) − c 2 , f ( x ) − c 3 since f ( p ) has exactly one value (taken from c 1 , c 2 , c 3 ) and therefore one of them has even power of ( x − p ) as its factors. Moreover, this works for all roots of g ( x ) , and so each of f ( x ) − c 1 , f ( x ) − c 2 , f ( x ) − c 3 is a square of polynomial, let's call it f 1 , f 2 , f 3 respectively. Now we have f 1 2 − f 2 2 = ( f ( x ) − c 1 ) − ( f ( x ) − c 2 ) = c 2 − c 1 = c = 0 ( f 1 − f 2 ) ( f 1 + f 2 ) = c which forces f 1 , f 2 to be constant functions, and so thus f . Hence, no solution from this case, too. Therefore the only possible values of a are 5 4 , 5 0 and the sum is then 1 0 4 .
This is a very detailed solution which analyzes all the issues well.
Mistakes in other solutions included the following.
1) It is not obvious that when two non-constant polynomials differ by a constant, they cannot be both complete squares. The trick is to apply the difference of squares formula, as done here in Case iii.
2) Just proving that a can only be 50 or 54 is not enough. One has to show that for these values of a the required polynomials actually exist.
Let N=x^3-12x^2+45x-a We will have N(f(x))=[g(x)]^2 If \deg g(x)=d then \deg f(x)=\frac{2d}{3} Differentating both side of the equation yields N'(f(x))f'(x)=2g'(x)g(x) If n is the root (real or complex) of g(x) then N'(f(n))=0 or f'(n)=0 Since \deg f(x)< \deg g(x) there is a number p such that g(p)=0 and f(p) \ne 0 So there must be one root n of g(x) which also the root of N(f'(x)) N'(x)=0 \Rightarrow x \in {3; 5} f(n) \in {3; 5} \Rightarrow a \in {50; 54} So our solution is 104
Let T ( x ) = x 3 − 1 2 x 2 + 4 5 x − a then T ( f ( x ) ) = [ g ( x ) ] 2 ( 1 ) . Therefore if g ( x ) has degree d, f ( x ) has degree 3 2 d .
Differentiating (1) we also get $$T'(f(x))f'(x)=2g(x)g'(x)\ \ \ (*)$$ Thus, if r is a real or complex root of g ( x ) then either T ′ ( f ( r ) ) = 0 or f ′ ( r ) = 0 .
Since de g f ( x ) < de g g ( x ) , there exists an either real or complex number α satisfying g ( α ) = 0 and f ( α ) = 0 . In other words, from (*) we infer that, there must be one root r of g ( x ) which is also the root of T ( f ′ ( x ) .
Also, note that T ′ ( x ) = 3 x 2 − 2 4 x + 4 5 = 0 ⇔ x ∈ { 3 ; 5 } .Therefore, f ( r ) ∈ { 3 ; 5 } ⇒ P ( 3 ) = 0 ∨ P ( 5 ) = 0 ⇔ 5 4 − a = 0 ∨ 5 0 − 1 = 0
which is why a ∈ { 5 0 ; 5 4 } and so the answer is 104.
Note first that the RHS can be rewritten as [ ( f ( x ) ) 3 − 1 2 ( f ( x ) ) 2 + 4 8 f ( x ) − 6 4 ] − 3 f ( x ) − ( a − 6 4 ) which can then be simplified to ( f ( x ) − 4 ) 3 − 3 f ( x ) − ( a − 6 4 ) . Let h ( x ) = f ( x ) − 4 . Then the equation reduces to ( g ( x ) ) 2 = ( h ( x ) ) 3 − 3 ( h ( x ) + 4 ) − ( a − 6 4 ) ( g ( x ) ) 2 = ( h ( x ) ) 3 − 3 h ( x ) + ( a − 5 2 ) . There are several things we can notice about the fact that the RHS must be a perfect square polynomial. First, note that g ( x ) can not be an irreducible poly-polynomial in terms of h ( x ) (for example, ( h ( x ) ) 2 + h ( x ) + 1 ). To prove this, note that the degree of the RHS is 3 , which is an odd number. If g ( x ) were irreducible, then ( g ( x ) ) 2 would have an even degree, contradiction. Therefore ( g ( x ) ) 2 must be broken down into linear factors.
Because of this, let ( h ( x ) ) 3 − 3 h ( x ) + ( a − 5 2 ) = ( h ( x ) − a 1 ) ( h ( x ) − a 2 ) ( h ( x ) − a 3 ) for some real a 1 , a 2 , a 3 . We must next note that it is impossible for all three of these factors to be perfect square polynomials. To prove this, suppose the opposite, that a ( x ) and b ( x ) are both perfect square polynomials that differ by a constant. Then the constant terms in a ( x ) and b ( x ) would differ. But note that due to the distributive property, the constant terms of each of these polynomials affect other coefficients as well, which is a direct contradiction of the fact that they only differ in their constant term. Therefore only one of the three linear factors can be a perfect square polynomial. It follows that two of the a i must be equal.
Because of this, rewrite our original equation as ( h ( x ) ) 3 − 3 h ( x ) + ( a − 5 2 ) = ( h ( x ) − a 1 ) 2 ( h ( x ) − a 2 ) . It is our job now to determine the possible values for a 1 and a 2 . Consider this as a polynomial in terms of h ( x ) , with roots a 1 , a 1 , a 2 . By Vieta's formulas, we must have a 1 + a 1 + a 2 = 0 ⟹ a 2 = − 2 a 1 . Then we can use Vieta's one more time to get that a 1 a 1 + a 1 a 2 + a 1 a 2 = − 3 a 1 2 − 2 a 1 2 − 2 a 1 2 = − 3 − 3 a 1 2 = 3 a 1 = ± 1
By substituting to find a 2 , we see that the only possible factorizations of ( h ( x ) ) 3 − 3 h ( x ) + ( a − 5 2 ) that result in perfect square trinomials are ( h ( x ) + 1 ) 2 ( h ( x ) − 2 ) and ( h ( x ) − 1 ) 2 ( h ( x ) + 2 ) . We can check to see that suitable h ( x ) indeed produce valid g ( x ) . (For example, h ( x ) = x 2 + 2 x + 3 satisfies the first case, and h ( x ) = x 2 + 2 x − 1 satisfies the second case.) The constant terms in these polynomials are − 2 and 2 , respectively. Letting a − 5 2 = − 2 gives a = 5 4 , and letting a − 5 2 = 2 gives a = 5 0 . Hence the only two possible values of a are 5 4 and 5 0 , and the sum of these two values is 1 0 4 .
Lemma: For α 1 , α 2 , … , α n ∈ R such that α 1 = α 2 = … = α n and p ( z ) , q ( z ) any two polynomials over R : ( p ( z ) ) 2 = ∏ i = 1 n ( q ( z ) − α i ) ⇒ p ( z ) , q ( z ) are constant (Proof) Assume the contrary, suppose there are two non-constant polynomials p ( z ) , q ( z ) such that ( p ( z ) ) 2 = ∏ i = 1 n ( q ( z ) − α i ) . Since α 1 = α 2 = … = α n , the set of roots for q ( z ) − α i , i = 1 , … , n are disjoint from each other. Since the roots of the above equation come in pairs and R is a UFD, the roots of each q ( z ) − α i also come in pairs. This means we can factor p ( z ) into n pieces such that: p ( z ) = ∏ i = 1 n p i ( z ) and ( p i ( z ) ) 2 = q ( z ) − α i , i = 1 , 2 , … , n Notice that for 1 ≤ i = j ≤ n , ( p i ( z ) + p j ( z ) ) ( p i ( z ) − p j ( z ) ) = ( q ( z ) − α i ) − ( q ( z ) − α j ) = α j − α i = 0 are constants. Thiis implies all p i ( z ) are constants. As a result, p ( z ) = ∏ i = 1 1 0 p i ( z ) and q ( z ) = ( p i ( z ) ) 2 + α i are also constants, a contradiction!
Using the lemma to the problem, ( g ( x ) ) 2 = P ( f ( x ) ) = ( f ( x ) ) 3 − 1 2 ( f ( x ) ) 2 + 4 5 f ( x ) − a , for it to have non-constant solutions, a necessary condition is that the cubic polynomial P ( z ) = z 3 − 1 2 z 2 + 4 5 z − a has repeated roots. Let c be any repeated root, ( z − c ) 2 will be a factor of P ( z ) and hence P ′ ( c ) = 3 c 2 − 2 4 c + 4 5 = 3 ( c − 3 ) ( c − 5 ) = 0 ⇒ c = 3 or c = 5 . P ( 3 ) = 0 gives a = 5 4 and P ( 5 ) = 0 gives a = 5 0 . Therefore, the answer is 5 0 + 5 4 = 1 0 4 .
This is clearly a transcript of a solution posted by Achille Hui on the stackexchange.com on March 7th. And it is almost identical to the solution by boytoitot@yahoo.com (id# 22392).
http://math.stackexchange.com/questions/322812/sum-of-the-values-of-a-for-which-g2-f3-12f245f-a-has-non-constant-soluti
Claim: For a 1 , a 2 , … , a n ∈ R such that a 1 = a 2 = … = a n and p ( z ) , q ( z ) any two polynomials over R : ( p ( z ) ) 2 = ∏ i = 1 n ( q ( z ) − a i ) ⇒ p ( z ) , q ( z ) are constant Proof: Assume the contrary, suppose there are two non-constant polynomials p ( z ) , q ( z ) such that ( p ( z ) ) 2 = ∏ i = 1 n ( q ( z ) − a i ) . Since a 1 = a 2 = … = a n , the set of roots for q ( z ) − a i , i = 1 , … , n are disjoint from each other. Since the roots of the above equation come in pairs and R is a Unique Factorization Domain, the roots of each q ( z ) − a i also come in pairs. This means we can factor p ( z ) into n pieces such that: p ( z ) = ∏ i = 1 n p i ( z ) and ( p i ( z ) ) 2 = q ( z ) − a i , i = 1 , 2 , … , n Notice that for 1 ≤ i = j ≤ n , ( p i ( z ) + p j ( z ) ) ( p i ( z ) − p j ( z ) ) = ( q ( z ) − a i ) − ( q ( z ) − a j ) = a j − a i = 0 are constants. Thiis implies all p i ( z ) are constants. As a result, p ( z ) = ∏ i = 1 1 0 p i ( z ) and q ( z ) = ( p i ( z ) ) 2 + a i are also constants, a contradiction!
Applying the claim to the original problem, ( g ( x ) ) 2 = P ( f ( x ) ) = ( f ( x ) ) 3 − 1 2 ( f ( x ) ) 2 + 4 5 f ( x ) − a , for it to have non-constant solutions, a necessary condition is that the cubic polynomial P ( z ) = z 3 − 1 2 z 2 + 4 5 z − a has repeated roots. Let c be any repeated root, ( z − c ) 2 will be a factor of P ( z ) and hence P ′ ( c ) = 3 c 2 − 2 4 c + 4 5 = 3 ( c − 3 ) ( c − 5 ) = 0 ⇒ c = 3 or c = 5 . P ( 3 ) = 0 gives a = 5 4 and P ( 5 ) = 0 gives a = 5 0 . Hence, the answer is 5 0 + 5 4 = 1 0 4 .
This is clearly a transcript of a solution posted by Achille Hui on the stackexchange.com on March 7th. And it is almost identical to the solution by rm_cosico@yahoo.com (id# 20947)
http://math.stackexchange.com/questions/322812/sum-of-the-values-of-a-for-which-g2-f3-12f245f-a-has-non-constant-soluti
Let P ( x ) = x 3 − 1 2 x 2 + 4 5 x , we have P ’ ( x ) = 0 when x = 3 or x = 5 . Note that P ( 3 ) = 5 4 and P ( 5 ) = 5 0 . Looking at the behaviour of P ( x ) , we note that the cubic equation P ( x ) = a has 1 real root and 2 distinct complex roots for a < 5 0 , a > 5 4 , 3 distinct real roots for 5 0 < a < 5 4 , and 2 real roots (with 1 repeated) for a = 5 0 , 5 4 . Assume that a = 5 0 , 5 4 . We will show that the cubic equation P ( x ) = a has at least 1 real root c and no repeated complex roots. Factorise g ( x ) 2 = P ( f ( x ) ) − a = ( f ( x ) − c ) ( f ( x ) 2 + p f ( x ) + q ) , where p , q ∈ R . We show that ( f ( x ) − c ) and ( f ( x ) 2 + p f ( x ) + q ) have no common complex root. Assume that they have a common root α , then f ( α ) is a common complex root of x − c and x 2 + p x + q , so f ( α ) is a repeated complex root of P ( x ) = a , a contradiction.
Now each complex root in g ( x ) 2 has even multiplicity, so each complex root in ( f ( x ) 2 + p f ( x ) + q ) has even multiplicity. Since ( f ( x ) 2 + p f ( x ) + q ) ∈ R [ x ] and ( f ( x ) 2 + p f ( x ) + q ( has positive leading coefficient, this implies that f ( x ) 2 + p f ( x ) + q ) = h ( x ) 2 for some h ( x ) ∈ R [ x ] . Finally, observe that since x 2 + p x + q has no repeated roots, we have x 2 + p x + q = ( x + 2 p ) 2 + r , where r = 0 , r ∈ R . Thus h ( x ) 2 = ( f ( x ) 2 + p f ( x ) + q ) = ( f ( x ) + 2 p ) 2 + r , so r = ( h ( x ) − ( f ( x ) + 2 p ) ) ( h ( x ) + ( f ( x ) + 2 p ) (, so both ( h ( x ) − ( f ( x ) + 2 p ) ( and ( h ( x ) + ( f ( x ) + 2 p ) ( are constant, so that f ( x must be constant, a contradiction.
We can check directly with a = 5 0 , a = 5 4 . Therefore, the sum we want to find is 5 0 + 5 4 = 1 0 4 .
Problem Loading...
Note Loading...
Set Loading...
As with all polynomial problems, there are only two major ideas:
(i) Analysis of roots - whether it has double roots,etc. (a really good method would be to differentiate to bring down the highest power)
(ii) Analysis of degree
And not to forget, when doing algebra, we try our best to simplify the polynomial equations we are working with. In particular, the LHS of this equation looks very complicated, but notice that suppose we define a new polynomial P ( x ) = x 3 − 1 2 x 2 + 4 5 x − a , then we can rewrite this equation neatly as:
g ( x ) 2 = P ( f ( x ) ) ( ∗ )
Now, we know nothing of the roots, so we apply idea (ii) and do some analysis on the degree. For instance, if de g g = d , then de g f = 3 2 d < d . We see no apparent use for this, so we leave this for later.
Next, we turn to idea (i), in the sleek form, we can differentiate ( ∗ ) to get:
P ′ ( f ( x ) ) × f ′ ( x ) = 2 g ( x ) g ′ ( x )
Aha, so if q is a root of g ( x ) , then either P ′ ( f ( r ) ) = 0 or f ′ ( r ) = 0 . We will attempt the former first, because it is considerably simpler.
So suppose P ′ ( f ( r ) ) = 0 . Firstly, we evaluate P ′ ( x ) to get: 3 x 2 − 2 4 x + 4 5 , being a quadratic equation, it is easy to see that it in fact can be simplified to the following: 3 ( x − 3 ) ( x − 5 ) . Voila! We have that f ( r ) is either 3 or 5 . In which case, we can simplify P ( 3 ) = 0 or P ( 5 ) = 0 . However, P ( 3 ) = 5 4 − a and P ( 5 ) = 5 0 − a so a = 5 0 , 5 4 .
I basically was too lazy to work out the other part and tried 5 0 + 5 4 = 1 0 4 and it worked.
Now, with this "conjecture" in mind, we deal with the other case f ′ ( r ) = 0 . Notice that other than the fact of its degree, we know virtually nothing about the behaviour, etc. of f ( x ) , which makes things tough. However, due to the "conjecture", we are motivated to try to reduce f ′ ( r ) = 0 to the above case of P ′ ( f ( r ) ) = 0 . So firstly, a list of things that we know:
(i) de g f < d − 1
And... that's all. However, we can compare this with g . For instance, this clearly implies that there definitely exists (at least) a root r that has higher multiplicity as a root of g than as a root for f ′ . One clever way to represent the growth rate/exponents, is to use Landau O-notation . So, we write (for some n and positive m ) f ( x ) = f ( r ) + m ( x − r ) n + O ( ( x − r ) n + 1 ) and g ( x ) = O ( ( x − r ) n ) . Hmm, this looks better! So if we substitute everything back into ( ∗ ) , (remember that we want to reduce to P ′ ( f ( r ) ) = 0 ):
P ( f ( x ) ) = P ′ ( f ( r ) ) m ( x − r ) n + O ( ( x − r ) n + 1 ) = O ( ( x − r ) n + 1 ) = g ( x ) 2
So simplifying, we do indeed get that we need P ′ ( f ( r ) ) = 0 as desired.