11

I'm trying to understand why the number of the monomials of degree $d$ in $n+1$ variables is $C_{n+d,n}$. If someone could help me to remember how to solve this, I would be glad.

Thanks.

user42912
  • 23,582
  • Are you sure that your formula is correct? If $n=0$ (a single variable), you formula gives $0$, but surely there is one monomial of degree $d$! – Jyrki Lahtonen Nov 04 '13 at 07:08
  • 1
    Don't you mean $C_{n+d-1,n-1}$ ? If so, look at:http://en.wikipedia.org/wiki/Monomial , in the part about "number – user99680 Nov 04 '13 at 07:08
  • "Basically, you count (and add together), the number of singletons, pairs, triples, using combinations. – user99680 Nov 04 '13 at 07:10
  • Look at the Wikipedia article on Stars and Bars. We are distributing $d$ identical candies among $n+1$ kids, or less interestingly finding the number of solutions of $a_1+a_2+\cdots+a_{n+1}=d$ in non-negative integers. – André Nicolas Nov 04 '13 at 07:10
  • 3
    This problem is same as distributing $d$ sweets among $n+1$ kids, or solving $x_{1} + \ldots + x_{n+1} = d$ with non-negative integers. The second formulation is quite popular. – Vishal Gupta Nov 04 '13 at 07:15
  • @AndréNicolas Apparently I have this so deeply rooted in my head as "balls in urns" that my first reaction to a different metaphor is to think that it must be wrong somehow. There are worse prejudices to have, I guess... still, it is strange how badly I want to say that it's wasteful, or that the kids will object to many of the arrangements... – Andrew Dudzik Nov 04 '13 at 07:17
  • @user99680 fixed – user42912 Nov 04 '13 at 07:25
  • @JyrkiLahtonen fixed – user42912 Nov 04 '13 at 07:26

2 Answers2

23

The stars and bars argument in combinatorics shows that the number of ways to place $d$ indistinguishable objects (stars) in $n + 1$ distinguishable boxes is to lay the objects in a line and place barriers (bars) to establish boundaries between the boxes.

Here's an example of the bijection with $d = 3$ and $n = 2$.

$$ \begin{align} x_1^3 & \longleftrightarrow \star \star \star \;| \;| \\ x_1^2 x_2 & \longleftrightarrow \star \star \,| \star | \\ x_1^2 x_3 & \longleftrightarrow \star \star \,| \;| \star \\ x_1 x_2^2 & \longleftrightarrow \star \;| \star \star \;| \\ x_1 x_2 x_3 & \longleftrightarrow \star \;| \star \,| \star \\ x_1 x_3^2 & \longleftrightarrow \star \;| \;| \star \star \\ x_2^3 & \longleftrightarrow \,| \star \star \star | \\ x_2^2 x_3 & \longleftrightarrow \,| \star \star \,| \star \\ x_2 x_3^2 & \longleftrightarrow \,| \star \,| \star \star \\ x_3^3 & \longleftrightarrow \,| \;| \star \star \star \end{align} $$

You can see the $\binom{2 + 3}{2} = 10$ monomials of degree $3$ in $2 + 1 = 3$ variables.

Sammy Black
  • 25,273
8

Your question confusingly lets $n$ be one less than the number of variables, a trickery to make the resulting formula more pleasant. But I don't like this kind of fiddling, because it makes me forget just which shift was necessary to get a nice answer (one more or one less?). So put $m=n+1$.

The number you are looking for, the number of monomials of degree$~d$ in the power series $\prod_{i=1}^m(1+X_i+X_i^2+X_i^3+\cdots)=\prod_{i=1}^m\frac1{1-X_i}$, is equal to the coefficient of $X^d$ in the power series $(1-X)^{-m}$ (just substitute $X_i:=X$ for all$~i$), which coefficient is by the binomial formula $(-1)^d\binom{-m}d=\binom{d+m-1}d$. Now you can again replace $m$ by $n+1$ to obtain $\binom{d+n}d=C_{n+d}^n$.

This answer assumes that you know how to "negate the upper index" in a binomial coefficient. The formula, which is well worth learning by heart, is $$ \binom nk=(-1)^k\binom{k-n-1}k \qquad\text{or equivalently}\qquad \binom{-n}k=(-1)^k\binom{k+n-1}k. $$ The formula is very general, the only condition is that $k\in\Bbb N$ (and this is already assumed for the definition $\binom xk=\frac{x(x-1)\ldots(x-k+1)}{k!}$), but $n$ could be anything. As you can see the upper index is not purely negated, but a negative upper index is surely made non-negative. In fact the symmetry is around the interval of values $n$ for which $\binom nk=0$, namely $0\leq n<k$ (those binomial coefficients in column $k$ that stick out above Pascal's triangle, but not as far as into the negative row indexes); this symmetry is $n\mapsto k-1-n$.

  • I was kind of excited to see a proof more alike to my perception of what I would call "rigorous". However, this makes use of the proposition that the given substitution actually constitutes a homomorphism... So now, not having the time to work through the pertaining passages in Zariskis "Comm. Algebra II", I'm kind of sad to seemingly be stuck with weird combinatorial arguments again... :( – polynomial_donut Feb 08 '16 at 18:28
  • 2
    @polynomial_donut So are you asking to prove there exists a ring homomorphism $\Bbb Z[[X_1,\ldots,X_n]]\to\Bbb Z[[X]]$ that sends the monomial $X^\alpha$ (with $\alpha\in\Bbb N^m$ a multi-exponent) to $X^{|\alpha|}$? By the way there is absolutely nothing wrong with combinatorial arguments; in the end nearly all the algebra we use (like the binomial formula) is based on combinatorial arguments. – Marc van Leeuwen Feb 09 '16 at 13:32
  • No, I'm not asking you to prove it... I was just pointing out that this is implicitly used here. I'm not really looking for combinatorial arguments, that's pretty much all. Everybody has their own philosophy about how to go about mathematics, what to regard as a rigorous enough proof, what not, and about the concept of "understanding" something etc. . I didn't tell you what to regard as wrong or right, did I? – polynomial_donut Feb 09 '16 at 18:09