3

There is already proof of polynomial long division: Proof of the polynomial division algorithm but alas, it was written in context of question related to abstract algebra.

Maybe this proof can be adapted for pre-calculus level of mathematical knowledge

  • 1
    Are you familiar with mathematical induction? The proof proceeds by induction on the degree of the polynomial. – While I Am Jan 30 '22 at 22:21
  • @William Yes, I'm familiar with mathematical induction. No, I still don't understand the proof that I linked – KarmaPeasant Jan 31 '22 at 07:11
  • It is just taking the largest quotient it can, and then dividing the remainder; rinse and repeat until you are done. In this sense it is exactly like the algorithm for the division of real numbers – FShrike Feb 01 '22 at 18:27
  • Before the line "Rephrased in the language of Gaussian elimination" there is nothing in the proof that involved abstract algebra. Try reading it again. – fleablood Feb 01 '22 at 18:27
  • What's wrong with this: https://en.wikipedia.org/wiki/Polynomial_long_division ? or with these https://www.google.com/search?client=firefox-b-1-d&q=polynomial+division – fleablood Feb 01 '22 at 18:29
  • @fleablood I already tried to re-read it. Maybe repharsing it (instead of adapting) could help. – KarmaPeasant Feb 01 '22 at 18:40
  • @fleablood Wikipedia just shows how polynomial long division is done. Also, I googled and wasn't able to find a relevant proof that I could understand (sometimes because presentation of the proof itself was bad. Other time because I spotted an error in beginning of given proof. Other time proof required knowledge of number theory) – KarmaPeasant Feb 01 '22 at 18:46
  • The proof is that you do it! That you can do it is the proof! – fleablood Feb 01 '22 at 19:21
  • I redid my proof. It should be easier to read now. – fleablood Feb 02 '22 at 02:46

1 Answers1

3

We'll do the proof by mathematical induction on the difference between the degrees of the polynomial.

Base Case: To show: If $p(x)$ is a non-zero polynomial of degree $n$ and $q(x)$ is a non-zero polynomial of degree $m = n$ then there exists a polynomial $d(x)$ of degree $n-m =0$ (in other words a real constant) and a polynomial, $r(x)$ (maybe zero) of degree less than $m$ so that $p(x) = d(x)q(x) + r(x)$

Proof of base case. Let $p(x) = \sum_{k=0}^n a_kx^k; a_n\ne 0$ and $q(x) = \sum_{k=0}^n b_kx^k; b_n\ne 0$. Let $d(x) = \frac {a_n}{b_n}$ (which is a polynomial of degree $0$) and let $r(x)=\sum_{k=0}^{n-1}(a_k -\frac {a_n}{b_n}b_k)x^k$ (which is a polynomial of degree less than $m=n$; Note: it's possible for the leading terms $(a_{n-1} -\frac {a_n}{b_n}b_{n-1})$ or for any terms, $(a_{k}-\frac {a_n}{b_n}b_k)$ to be $0$. so it need not be of degree $n-1$-- it could be less; indeed the entire polynomial may be $0$. It is only required that it is degree less than $n$). And let $r(x)=\sum_{k=0}^{n-1}(a_k -\frac {a_n}{b_n}b_k)x^k=$ .

Then $d(x)q(x)+r(x)=\sum_{k=0}^n\frac {a_n}{b_n}b_k x^k +\sum_{k=0}^{n-1}(a_k -\frac {a_n}{b_n}b_k)x^k=$
$\sum_{k=0}^{n-1}(\frac {a_n}{b_n}b_k + a_k-\frac {a_n}{b_n}b_k)x^k + \frac {a_n}{b_n}b_nx^n=$
$\sum_{k=0}^{n-1}a_kx^k + a_nx^n = \sum_{k=0}^na_kx^k=p(x)$

Induction:

Induction Hypothesis: Assume we know that for any non-zero polynomial, $p(x)$ of degree $n$ and any non-zero polynomial, $q(x)$ of degree $m$ with $m \le n$ and $n-m \le D$ (for a specific non-negative integer $D$) then there will exist a non-zero polynomial, $d(x)$, of degree $n-m$, and a (possibly zero) polynomial $r(x)$ of degree less than $m$ so that $p(x)=d(x)q(x) + r(x)$.

Induction step to prove. If $p(x)$ and $q(x)$ are non-zero polynomials of degrees $n,m$ respectively and $n-m = D+1$ then we need to prove there exist $d(x),r(x)$ where degree of $d(x) = n-m$ and degree of $r(x) < m$.

Proof: Let $p(x) = \sum_{k=0}^n a_kx^k;a_n \ne 0$.

Let $w(x) = \frac {p(x)}x-a_0 = \sum_{h=0}^{n-1}a_{h+1}x^h$ is a polynomial of degree less than $n$. As $a_{n}\ne 0$ it is a polynomial of degree $n-1$.

Now as $(n-1)-m = (n-m)-1 = (D+1) -1 = D$ we know that there exist polynomials $\delta(x)$ of degree $(n-1) -m$ and $\rho(x)$ of degree less than $m$ so that $w(x) = \delta(x)q(x) + \rho(x)$.

And therefore $p(x) = x[w(x)] + a_0= x[\delta(x)q(x) +\rho(x)]+a_0=(x\delta(x))q(x) + (x\rho(x) + a_0)$.

Now $x\delta(x)$ is a polynomial of degree $(n-1)-m +1 = n-m$ and $x\rho(x)+a_0$ is a polynomial of less than $m+1$ so the degree of $x\rho(x)+a_0$ is less than or equal to $m$.

If the degree of $x\rho(x)$ is less than $m$ we are done: Let $d(x) = x\delta(x)$ and let $r(x) = x\rho(x) + a_0$ and we have $p(x) =d(x)q(x) + r(x)$.

If the degree of $x\rho(x) + a_0=m$ then by the induction process we have a polynomial of degree zero (in other words a constant), $v$ and a polynomial $s(x)$ of degree less than $m$ so that $x\rho(x) + a_0 = vq(x) + s(x)$.

In this case we are done: $p(x) = x\delta(x)q(x) + x\rho(x) + a_0 =x\delta(x)q(x) + vq(x) + s(x)= (x\delta(x) + v)q(x) + s(x)$. We just need to define $d(x) =x\delta(x) + v$ and $r(x) = s(x)$.

And.... that's it.

By induction we have shown that for any non-zero polynomials of degrees $n,m; n\ge m$ that so $n-m$ is an non-negative integer (which is always) then there exist a polynomial, $d(x)$ of degree $n-m$ and a polynomial, $r(x)$, of degree less than $m$ so that $p(x) =d(x)q(x) + r(x)$.

That's all.

fleablood
  • 124,253
  • I don't think it is exactly the same as the division algorithm for real numbers at all. For $a,b\in\mathbb{R},b\ne 0$ the algorithm goes: $a=\frac{a}{b}b+0$ and stops. It seems to me to be exactly like the division algorithm for integers. – ancient mathematician Feb 01 '22 at 19:27
  • @ancientmathematician Yeah...you're right.... but "poetic license". – fleablood Feb 01 '22 at 19:35
  • "Pf" means "proof". – fleablood Feb 02 '22 at 20:43
  • In the base case why is $d(x) = \frac {a_n}{b_n}$?
  • Why do you say "not it's possible for the leading $(a_k -\frac {a_n}{b_n}b_k)$ to be $0$", while at same time claiming that entire polynomial may be zero?
  • – KarmaPeasant Feb 03 '22 at 03:53
  • Typing "not" for "note" is a typo I frequently make. It drives me nuts. I'll fix it. – fleablood Feb 03 '22 at 06:05
  • Good to hear, but what about $d(x) = \frac {a_n}{b_n}$? Why did you make d(x) equal to this? – KarmaPeasant Feb 03 '22 at 07:50
  • I also don't understand why $r(x)=\sum_{k=0}^{n-1}(a_k -\frac {a_n}{b_n}b_k)x^k$ – KarmaPeasant Feb 03 '22 at 08:09
  • Please, answer. I can't proceed. – KarmaPeasant Feb 04 '22 at 17:26
  • 1
    " d(x)=anbn? Why did you make d(x) equal to this?" This is the base case. We are assuming $p(x)$ and $q(x)$ are the same degree. The leading term of $q(x)$ is $b_nx^n$. The leading term of $p(x)$ is $a_nx^n$. SO I need to multiply $p(x)$ by something that will turn $b_nx^n\to a_nx^n$. What can I multiply $b_nx^n$ by to make it into $a_nx^n$. The only answer is $\frac {a_n}{b_n}$. – fleablood Feb 04 '22 at 20:36
  • 1
    "I also don't understand why r(x)=∑n−1k=0(ak−anbnbk)xk" Well, $p(x) = \sum b_k x^k$. I just multiplied everything by $\frac {a_n}{b_n}$ to get $\frac {a_n}{b_n}p(x)=\sum \frac {a_n}{b_n}b_k x^k$. That makes the leading term equal to $\frac {a_n}{b_n}b_nx^n = a_nx^n$. But it makes all the other terms equal to $\frac {a_n}{b_n}b_k x^k$. We need to add something to it to make $\frac{a_n}{b_n}b_k x^k \to a_kx^k$. What can I add to $\frac{a_n}{b_n}b_kx^k$ to make the result $a_kx^k$. Then only answer is $(a_n - \frac {a_n}{b_n}b_k)x^k$. – fleablood Feb 04 '22 at 20:40