3

How can I prove that ${2^n-1\choose k}$ and ${2^n-k\choose k}$ always returns odd numbers? It is possible to prove this by congruence?

by the way : $0 \leq k \leq (2^n-1)$

7 Answers7

11

$$\binom{2^n-1}{k} = \frac{(2^n-1)(2^n-2)(2^n-3)\cdots(2^n-k)}{1\cdot 2\cdot 3\cdots k} $$ and $2^n-a$ is divisible by exactly as many factors of $2$ as $a$ is.

This means that every factor of $2$ in the numerator is matched by a factor of $2$ in the denominator. After cancelling those, what's left in the numerator must be an odd number, and you can't get an even integer by dividing an odd integer by any integer.

  • A different take at it, great! But I'm more than a bit worried that veterans of the site don't search for duplicates. I mean, you probably suspected that this question had been covered earlier. Do you feel that searching for a previous incarnation is futile? – Jyrki Lahtonen Jul 26 '16 at 09:21
3

Kummers theorem:

The maximum power of the prime $p$ that divides $\binom{n}{k}$ is equal to the number of carries when adding $n-k$ and $k$ in base $p$.

Clearly if $a$ and $b$ add up to $111\dots 1$ in binary there can't be any carries, since the digit in the position of the first carry is always $0$.

Asinomás
  • 105,651
3

See Lucas's theorem on binomial coefficients. More generally, for any prime $p$, $\displaystyle {{p^n-1} \choose k}$ is never divisible by $p$ when $0 \le k \le p^n-1$.

Robert Israel
  • 448,999
  • 2
    I think you mean not divisible by $p$? – awllower Jul 25 '16 at 17:19
  • 1
    Yes, thanks for catching that. – Robert Israel Jul 25 '16 at 17:41
  • A minor addition. When $p>2$ the similar non-divisibility holds for binomial coefficients $\binom {ap^n-1}k$ for all $k, 0\le k\le ap^n-1$ and all $a, 0<a<p$. It follows easily from Lucas theorem that this list of non-divisible rows of Pascal triangles is complete. – Jyrki Lahtonen Jul 26 '16 at 09:16
  • And, as a veteran of the site you probably suspected that this question had been covered earlier. Do you feel that searching for a previous incarnation is futile? – Jyrki Lahtonen Jul 26 '16 at 09:19
2

In formula $(7)$ of this answer, it is shown that the number of factors of $p$ in $\binom{n}{k}$ for a prime $p$ is $$ \frac{\sigma_p(k)+\sigma_p(n-k)-\sigma_p(n)}{p-1}\tag{1} $$ where $\sigma_p(k)$ is the sum of the digits in the base-$p$ expansion of $k$. This is also the number of carries performed when adding $n-k$ and $k$ in base-$p$. Since all the base-$2$ digits of $2^n-1$ are $1$, $k$ and $2^n-1-k$ are one's complements and so the sum of all their digits will be the sum of the digits of $2^n-1$ (which is $n$).

Therefore, by $(1)$, the number of factors of $2$ in $\binom{2^n-1}{k}$ is $0$. That is, $\binom{2^n-1}{k}$ is odd.

robjohn
  • 345,667
2

For a completely different approach, you can prove it by induction on $n$ if you first prove, also by induction on $n$, that $\binom{2^n}k$ is odd if and only if $k=0$ or $k=2^n$. Assume this result for $n$. By Vandermonde’s identity we have

$$\binom{2^{n+1}}k=\sum_{\ell=0}^{2^n}\binom{2^n}\ell\binom{2^n}{k-\ell}\;,\tag{1}$$

where by symmetry we may assume that $k\le 2^n$. The $\ell$ term in the summation in $(1)$ is odd if and only if $\binom{2^n}\ell$ and $\binom{2^n}{k-\ell}$ are both odd, which by the induction hypothesis is the case if and only if $\ell=k=0$, $\ell=0$ and $k=2^n$, or $\ell=k=2^n$. Thus, $\binom{2^{n+1}}k$ is even if $0<k<2^n$,

$$\binom{2^{n+1}}{2^n}\equiv\binom{2^n}0\binom{2^n}{2^n}+\binom{2^n}{2^n}\binom{2^n}0\equiv 0\pmod2\;,$$

and

$$\binom{2^{n+1}}0\equiv\binom{2^n}0^2+\binom{2^n}0\binom{2^n}{2^n}+\binom{2^n}{2^n}\binom{2^n}0\equiv 1\pmod2\;,$$

as desired.

Now assume the actual desired result for $n$. By Vandermonde’s identity we have

$$\binom{2^{n+1}-1}k=\sum_{\ell=0}^{2^n}\binom{2^n}\ell\binom{2^n-1}{k-\ell}\;,$$

where by symmetry we may assume that $k<2^n$. By the induction hypothesis $\binom{2^n-1}{k-\ell}$ is odd for each $\ell\le k$, so

$$\binom{2^n}\ell\binom{2^n-1}{k-\ell}\equiv\binom{2^n}\ell\pmod2$$

for each $\ell\le k$, and

$$\binom{2^{n+1}-1}k\equiv\sum_{\ell=0}^k\binom{2^n}\ell\pmod2\;.\tag{2}$$

From the preliminary result we know that $\binom{2^n}\ell$ is even for $1\le\ell\le k$ (since $k<2^n$), so the only odd term in the summation in $(2)$ is $\binom{2^n}0=1$, and $\binom{2^{n+1}-1}k$ is odd.

Brian M. Scott
  • 616,228
1

For the first proposition, i.e., "$2^n-1\choose k$ is odd", I came up to a proof by mathematical induction and using some basic properties of binomial coefficient ${r\choose s}$. The other proposition "$2^n-k\choose k$ is odd" is not true (consider $n=k=3$).

The aforementioned properties are:

  1. Chu–Vandermonde identity: $${r\choose s}=\sum_{j=0}^{s}{{m\choose j}}{r-m\choose s-j}\label{eqcv}\tag{1}$$
  2. Pascal's rule: $${r\choose s}={r-1\choose s}+{r-1\choose {s-1}}\label{eqpascal}\tag{2}$$

Here is the proof. Obviously, the proposition is true for $n=1$. Assuming it to be true for a natural number $n$, we have to show it is also true for $n+1$, i.e., ${2^{n+1}-1\choose k}$ is odd for $k=0,1,\dots,2^{n+1}-1$.

To do this, we start from (\ref{eqcv}) and let $r=2^{n+1}-1$, $s=k$, and $m=2^n-1$. Also we first assume $0\leq k\leq 2^n-1$. So: $${2^{n+1}-1\choose k}=\sum_{j=0}^{k}{{2^n-1\choose j}}{2^n\choose k-j}$$ which after taking out the last term (the term with $j=k$) leads to: $${2^{n+1}-1\choose k}={2^n-1\choose k}+\sum_{j=0}^{k-1}{{2^n-1\choose j}}{2^n\choose k-j}\label{eq3}\tag{3}$$

Now, according to the induction hypothesis, ${2^n-1\choose k}$ and ${2^n-1\choose j}$ are odd numbers. Thus, we just need to show ${2^n\choose k-j}$ is even. By using (\ref{eqpascal}) and letting $r=2^n$ and $s=k-j$ we have: $${2^n\choose k-j}={{2^n-1}\choose k-j}+{{2^n-1}\choose k-j-1}$$

which is the sum of two odd numbers (induction hypothesis), and hence, it's even. Notice that the assumption $0\leq k\leq 2^n-1$ together with $j\leq k-1$, results in: $$2^n-1\geq k-j\geq k-j-1\geq0$$ which satisfies the desired limits for using binomial coefficients.

The last step is to show the proposition is also true for $2^{n+1}-1\geq k>2^n-1$. In such cases, we define $k'=2^{n+1}-1-k$, which leads to $0\leq k'<2^n$, or equivalently $0\leq k'\leq 2^n-1$ which is the same as the limit we first considered for $k$. Since ${2^{n+1}-1\choose{k}}={2^{n+1}-1\choose{k'}}$ (due to symmetry), we can repeat the proof, this time with $k$ replaced by $k'$. The proof is now complete.

babakks
  • 1,640
  • is false the following reasoning ? According to the Vandermondes identity is $$\binom{2^{n}-1}{k}= \sum_{\ell=0}^{k}\binom{2^{n}}{\ell}\binom{-1}{k-\ell}$$ and $$\binom{-x}{k} = (-1)^{k}\binom{x+k-1}{k}$$

    Then we have $$\sum_{\ell=0}^{k}\binom{2^{n}}{\ell}\binom{-1}{k-\ell} = \sum_{\ell=0}^{k}\binom{2^{n}}{\ell}(-1)^{k-l}\binom{1+k-\ell-1}{k-\ell} =\sum_{\ell=0}^{k}\binom{2^{n}}{\ell}(-1)^{k-l}\binom{k-\ell}{k-\ell}= (-1)^{k}\sum_{\ell=0}^{k}\binom{2^{n}}{\ell}$$ Where $\sum_{\ell=0}^{k}$$\binom{2^{n}}{\ell}$ is always even :/

    – Darío A. Gutiérrez Jul 26 '16 at 19:43
  • 1
    First note, "Vandermondes" identity assumes parameters to be non-negative. The identity that you used is in fact the "Chu-Vandermondes" identity, which is also true for complex values. Second, $\sum_{\ell=0}^{k}{2^n\choose \ell}$ is not always even. For example, consider $k=0$. In fact only the first term of the sum (i.e., for $\ell=0$) and the last term (for $\ell=2^n$, if possible) are odd. Thus, for $0\leq k\leq 2^n -1$ (as in the original problem) the sum only contains one odd value (the first one), and since $k$ can never reach $2^n$, the other odd value never gets part in summation. – babakks Jul 26 '16 at 20:44
  • 1
    Last note, in your final equation, $(-1)^k$ has to be corrected to $(-1)^{k-\ell}$ and goes within the summation, unless I'm wrong. – babakks Jul 26 '16 at 20:44
0

Everything in row $2^n - 1$ of Pascal's triangle is odd because everything in row $2^n$ is even except for the two ends.