4

I have to prove that $\frac{1}{5}n^5+\frac{1}{3}n^3+\frac{7}{15}n$ is an integer for any $n$. I solved this by brute-force, exhausting all the possibilities methods. I was wondering if there was a way to solve this at-a-glance with some sort of theory? Because

Although my answer is correct, I was apparently supposed to use some theory to answer this question.

I started by putting everything on common denominator and factoring out $n$:

$$\frac{1}{5}n^5+\frac{1}{3}n^3+\frac{7}{15}n = \frac{n(3n^4+5n^2+7)}{15}$$

then I proceeded to plug in $\pm1 ,\pm2, \pm3 ,..., \pm7$ into $3n^4+5n^2+7 \pmod {3,5,\text{ or }15}$

$$3(\pm 1)^4 +5(\pm 1)^2 +7 \equiv 15 \equiv 0 \pmod{15}$$

$$3(\pm 2)^4 +5(\pm 2)^2 +7 \equiv 75 \equiv 5\cdot 15 \equiv 0 \pmod{15}$$

$$3(\pm 3)^4 +5(\pm 3)^2 +7 \equiv 295 \equiv 0 \pmod{5} \text{ while } n \equiv 0 \pmod 3$$

$$3(\pm 4)^4 +5(\pm 4)^2 +7 \equiv 855 \equiv 57\cdot 15\equiv 0 \pmod{15}$$

$$3(\pm 5)^4 +5(\pm 5)^2 +7 \equiv 2007 \equiv 0 \pmod{3}\text{ while } n \equiv 0 \pmod 5$$

$$3(\pm 6)^4 +5(\pm 6)^2 +7 \equiv 4075 \equiv 0 \pmod{5}\text{ while } n \equiv 0 \pmod 3$$

$$3(\pm 7)^4 +5(\pm 7)^2 +7 \equiv 7455 \equiv 497 \cdot 15\equiv 0 \pmod{15}$$

to conclude that if $n\equiv 0 \pmod {15}$ then $\frac{n}{15}$ is an integer from which $\frac{1}{5}n^5+\frac{1}{3}n^3+\frac{7}{15}n$ is an integer,

and if $n \not\equiv 0 \pmod {15}$ then either $3n^4+5n^2+7 \equiv 0 \pmod{15}$ or $n \equiv 0 \pmod 3$ and $ 3n^4+5n^2+7 \equiv 0 \pmod{5}$ or $n \equiv 0 \pmod 5$ and $ 3n^4+5n^2+7 \equiv 0 \pmod{3}$ from which the statement is clearly true.

Is there a less brute-force-ish way of concluding this? Is there some theory I should be using that would cause me to not be excessively lengthy in calculation if I were given different, larger numbers than $15$?

Sil
  • 16,612
  • 5
    It's still a bit computation heavy, but assuming the result is correct then you should be able to write $\frac{1}{5}n^5 + \frac{1}{3}n^3 + \frac{7}{15}n = c_5 \binom{n}{5} + c_4 \binom{n}{4} + \cdots + c_1 n + c_0$ for some constant integers $c_5, \ldots, c_0$. – Daniel Schepler Mar 04 '21 at 21:21
  • With $d=\prod_j p_j^{e_j}$ and $f\in \Bbb{Z}[x]$ then $\forall n,\frac{f(n)}{d}\in \Bbb{Z}$ iff $\forall n,j, f(n)\equiv 0\bmod p_j^{e_j}$.

    When $e_j=1$ this happens iff $\forall j,f(x)=0 \in \Bbb{F}_{p_j}[x]/(x^{p_j}-x)$. The case $e_j> 1$ is a bit more complicated.

    – reuns Mar 04 '21 at 22:12

5 Answers5

3

The question asks how to prove a polynomial $\,f(n)\,$ takes only integer values for any integer $\,n\,.$ If the polynomial is of degree $0$, then it is a constant and if that constant is an integer we are done. If the polynomial is of degree $1$, then if any two consecutive values, $\,f(k), f(k+1)\,$ are integers, then it is an arithmetic progression and we are done. In general, for any degree $\,d\,$ polynomial $\,f(n),\,$ it is sufficient to verify that $$f(k),f(k+1), \dots,f(k+d-1) $$ are all integers for some integer $\,k,\,$ which implies that the polynomial takes on only integer values for all integer $\,n.\,$

A proof of this can be done by using a difference table to find an explicit expression for the polynomial as a sum of binomial coefficients. For example, using the particular polynomial in the question, $$ f(n) := \frac{1}{5}n^5+\frac{1}{3}n^3+\frac{7}{15}n $$ the forward difference table is:

$$ \begin{matrix} \Delta^6f(n) &&&&&&& 0 \\ \Delta^5f(n) &&&&&& 24 && 24 \\ \Delta^4f(n) &&&&& 48 && 72 && 96 \\ \Delta^3f(n) &&&& 32 && 80 && 152 && 248 \\ \Delta^2f(n) &&& 8 && 40 && 120 && 272 && 520\\ \Delta f(n) && 1 && 9 && 49 && 169 && 441 && 961 \\ f(n) & 0 && 1 && 10 && 59 && 228 && 669 && 1630 \\ n & 0 && 1 && 2 && 3 && 4 && 5 && 6 \end{matrix} $$ where the 5th differences are constant as expected for a 5th degree polynomial. Thus, $$ f(n) = 0{n \choose 0} +1{n \choose 1} + 8{n \choose 2} + 32{n \choose 3} + 48{n \choose 4} + 24{n \choose 5}$$ where the coefficient of $\,{n\choose k}=\Delta^kf(0),\,$ the $k$th difference of $\,f\,$ at zero. The Wikipedia article finite difference explains some of this theory. The forward differences of $\,f\,$ are $$ \Delta f(n) := f(n+1)-f(n), \:\: \Delta^2 f(n) := \Delta f(n+1) - \Delta f(n), \:\: \dots. $$

In this particular case, the 1st backward difference is the OEIS sequence A058031 $$ a_n := (n^2-n+1)^2 = \nabla f(n) := f(n)-f(n-1).$$ Thus the polynomial is the partial sums of an integer sequence and therefore is an integer sequence itself.

Somos
  • 35,251
  • 3
  • 30
  • 76
3

To show that $15$ divides $n(3n^4+5n^2+7)$, show that $3$ and $5$ do.

$3$ does because either $3$ divides $n$ or $n^2\equiv1\pmod3$ by Fermat's little theorem,

in which case $3n^4+5n^2+7\equiv5n^2+7\equiv5+7=12\equiv0\pmod3.$

$5$ does because either $5$ divides $n$ or $n^4\equiv1\pmod5$ by Fermat's little theorem,

in which case $3n^4+5n^2+7\equiv3n^4+7\equiv3+7=10\equiv0\pmod5$.

J. W. Tanner
  • 60,406
2

$$ \begin{align}f(n+1)-f(n)&=\frac{(n+1)^5-n^5}5+\frac{(n+1)^3-n^3}3+\frac7{15} \\&=\frac{5n^4+10n^3+10n^2+5n+1}5+\frac{3n^2+3n+1}3+\frac7{15}\\&=\text{integer}+\frac15+\frac13+\frac 7{15}\end{align}$$ is an integer for all integers $n$ and so is $f(0)$. The claim follows by induction.


Here's a generalization for you: Suppose $p_1,p_2,\ldots,p_m$ are primes, $a_1,a_2,\ldots,a_m$ are integers and $c$ is real. Let $$ f(n)=\frac{a_1n^{p_1}}{p_1}+\frac{a_2n^{p_2}}{p_2}+\cdots +\frac{a_mn^{p_m}}{p_m}+cn.$$ If $f(1)$ is an integer, then $f(n)$ is an integer for all integers $n$.

1

You can cut some work by just looking at the remainder mod $3$.

Let $n=3q+r$. Easy calculation gives $$\frac15(3q+r)^5+\frac13(3q+r)^3+\frac{7}{15}(3q+r) = \text{some integer} + \frac{243q^5+7q}{5} + \frac{r^5}5+\frac{r^3}3+\frac{7r}{15}.$$

By Little Fermat we have $q^5 \equiv q \pmod{5}$ so $$243q^5+7q \equiv 250q \equiv 0\pmod{5}.$$ The second term $$\frac{r^5}5+\frac{r^3}3+\frac{7r}{15}$$ is the same as the original expression but you only have to check $r=0,1,2$ which is almost trivial.

mechanodroid
  • 46,490
1

Fermat's Little Theorem

In particular, this polynomial admits a lot of simplification. $$ \begin{align} 3n^5+5n^3+7n &\equiv2n^3+n&\pmod3\tag{1a}\\ &\equiv2n+n&\pmod3\tag{1b}\\ &\equiv0&\pmod3\tag{1c} \end{align} $$ Explanation:
$\text{(1a)}$: reduce coefficients $\bmod3$
$\text{(1b)}$: apply Fermat's Little Theorem
$\text{(1c)}$: $3n\equiv0\pmod3$

$$ \begin{align} 3n^5+5n^3+7n &\equiv3n^5+2n&\pmod5\tag{2a}\\ &\equiv3n+2n&\pmod5\tag{2b}\\ &\equiv0&\pmod5\tag{2c} \end{align} $$ Explanation:
$\text{(2a)}$: reduce coefficients $\bmod5$
$\text{(2b)}$: apply Fermat's Little Theorem
$\text{(2c)}$: $5n\equiv0\pmod5$

Thus, $3n^5+5n^3+7n\equiv0\pmod{15}$.


Binomial Polynomials

In general, any polynomial that takes only integer values can be written as an integer combination of binomial polynomials: $$ \frac{3n^5+5n^3+7n}{15}=24\binom{n}{5}+48\binom{n}{4}+32\binom{n}{3}+8\binom{n}{2}+\binom{n}{1}\tag3 $$


The Computatation of the Coefficients

For a degree $d$ polynomial $P$, $$ P(n)=\sum_{k=0}^dc_k\binom{n}{k}\tag4 $$ We can compute $c_n$ inductively, using $$ P(n)-\overbrace{\sum_{k=0}^{n-1}c_k\binom{n}{k}}^{\substack{\text{using the $c_k$}\\\text{computed}\\\text{previously}}}=c_n\overset{\substack{1\\[3pt]\downarrow\\[3pt]\,}}{\binom{n}{n}}+\overbrace{\sum_{k=n+1}^dc_k\binom{n}{k}}^{\substack{\text{the binomial}\\\text{coefficients}\\\text{are $0$}}}\tag5 $$

$c_0=P(0):$ $$ c_0=\frac{3\cdot0^5+5\cdot0^3+7\cdot0}{15}=0 $$ $c_1=P(1)-c_0\binom{1}{0}:$ $$ c_1=\frac{3\cdot1^5+5\cdot1^3+7\cdot1}{15}-0\binom{1}{0}=1 $$ $c_2=P(2)-c_0\binom{2}{0}-c_1\binom{2}{1}:$ $$ c_2=\frac{3\cdot2^5+5\cdot2^3+7\cdot2}{15}-0\binom{2}{0}-1\binom{2}{1}=8 $$ $c_3=P(3)-c_0\binom{3}{0}-c_1\binom{3}{1}-c_2\binom{3}{2}:$ $$ c_3=\frac{3\cdot3^5+5\cdot3^3+7\cdot3}{15}-0\binom{3}{0}-1\binom{3}{1}-8\binom{3}{2}=32 $$ $c_4=P(4)-c_0\binom{4}{0}-c_1\binom{4}{1}-c_2\binom{4}{2}-c_3\binom{4}{3}:$ $$ c_4=\frac{3\cdot4^5+5\cdot4^3+7\cdot4}{15}-0\binom{4}{0}-1\binom{4}{1}-8\binom{4}{2}-32\binom{4}{3}=48 $$ $c_5=P(5)-c_0\binom{5}{0}-c_1\binom{5}{1}-c_2\binom{5}{2}-c_3\binom{5}{3}-c_4\binom{5}{4}:$ $$ c_5=\frac{3\cdot5^5+5\cdot5^3+7\cdot5}{15}-0\binom{5}{0}-1\binom{5}{1}-8\binom{5}{2}-32\binom{5}{3}-48\binom{5}{4}=24 $$ Thus, the converted polynomial is $$ \frac{3n^5+5n^3+7n}{15}=0\binom{n}{0}+1\binom{n}{1}+8\binom{n}{2}+32\binom{n}{3}+48\binom{n}{4}+24\binom{n}{5} $$

robjohn
  • 345,667
  • how does it go wrong if you can’t write it as an integer linear combination of binomial polynomials? – no lemon no melon Mar 06 '21 at 00:51
  • If you read the link provided above, it shows that if a polynomial maps $\mathbb{Z}\to\mathbb{Z}$, then it can be written that way. If it can’t be written that way, it will map some integer to a non-integer. – robjohn Mar 06 '21 at 01:04
  • yup, but in calculating the integer coefficients it was assumed that $\frac{3n^5 + 5n^3 + 7n}{15}$ mapped integers to integers first and then it ended up working out, I'm wondering if you started naively and it wasn't meant to work out as above, how would you know that a contradiction arose? – no lemon no melon Mar 06 '21 at 01:11
  • Nothing was assumed. If the coefficients don’t come out integers, you know it doesn’t map $\mathbb{Z}\to\mathbb{Z}$ – robjohn Mar 06 '21 at 01:41
  • I was referring to the first comment on my question, but thanks, it makes a lot of sense now :) – no lemon no melon Mar 06 '21 at 03:06
  • Ah, I hadn't noticed that Daniel Schepler had suggested the binomial polynomial approach in comments. In any case, I'm glad that it all makes sense! – robjohn Mar 06 '21 at 03:12