22

How many functions from set $\{1,2,3,\ldots,n\}$ to $\{A,B,C\}$ are surjections? $n \geq 3$

Attempt

I was hoping to count the number of surjections by treating $A,B,C$ like bins, and counting the number of ways to fill them with $1,2,3,\ldots,n$ such that no bins are empty. This is equivalent to putting two "separator" bars in between the $n$ numbers or $n-1 \choose 2$.

However, I think I am missing a step here: prior to putting in the separator bars, I should permutate the $n$ elements. So, multiply by $n!$.

The order of elements in the same bins do not matter. So I should be divided by the number of permutations of the elements in each bins. This is where I am stuck.

Additional Info

I made the question up myself. Feel free to modify the question into a more solvable form if it is too difficult/complicated.

Legendre
  • 2,845

6 Answers6

28

You’ve actually made a pretty good start: your first idea doesn’t work, but it’s a reasonable one to try, and you’ve spotted the difficulty with it. What you want now is an inclusion-exclusion argument. There are $3^n$ functions altogether. $2^n$ of them are functions from $\{1,\dots,n\}$ to $\{A,B\}$, missing $C$ altogether, so we need to throw them out. There are also $2^n$ functions that miss $B$ and $2^n$ that miss $A$, so our improved approximation to the number of surjections is $3^n-3\cdot2^n$.

Unfortunately, we’ve now gone overboard. The function that sends every $k\in\{1,\dots,n\}$ to $A$, for instance, has been thrown out twice, once for not hitting $B$ and once for not hitting $C$. The same goes for the other two constant functions. We need to throw each of them back in once, so that on net we’ve thrown each of them out just once, not twice. The revised number of surjections is then

$$3^n-3\cdot2^n+3=3\left(3^{n-1}-2^n+1\right)\;.\tag{1}$$

A little thought should convince you that no further adjustments are required and that $(1)$ is therefore the desired number.

Brian M. Scott
  • 616,228
  • Doh, the key insight was to count everything and subtract the non-surjections like in probability exercises (birthday paradox). Can't believe I didn't think of that. +1 and accept. Thanks for your help! – Legendre Oct 19 '12 at 05:08
  • @Legendre: You’re welcome! – Brian M. Scott Oct 19 '12 at 05:09
  • 1
    (+1, 9 years later) In this question, it is specified that $n\ge3$, but I thought I'd mention that I just wrote an answer that uses this formula. In that answer, I included one more term to get $3^n-3\cdot2^n+3-0^n$, where $0^0=1$ is used. This extends the validity to $n\ge0$. – robjohn Nov 19 '21 at 17:41
18

Ignoring surjectivity, there are $3^n$ maps $\{1,2,\ldots,n\}\to \{A,B,C\}$. We should subtract the $2^n$ maps that ar actually just maps $\{1,2,\ldots,n\}\to \{A,B\}$, also the $2^n$ maps $\ldots\to \{A,C\}$ and the $2^n$ maps $\ldots\to \{B,C\}$. But now we have subtracted too much: The three constant maps $f(x)=A$, $f(x)=B$ and $f(x)=C$ where subtracted twice (e.g. $f(x)=A$ as a map $\to\{A,B\}$ and also as a map $\{A,C\}$). In summary there are therefore $$ 3^n-3\cdot 2^n+3$$ surjective maps $\{1,2,\ldots,n\}\to\{A,B,C\}$.

9

Using PIE one can find a more generalized form of the number of surjections. The number of surjections of a set containing $m$ elements onto a set containing $n$ elements ($m>n$) is

$$\sum\limits_{i=0}^{n}(-1)^{i}\dbinom{n}{i}(n-i)^{m}$$

In the given question $n=3$ and $m=n$.

  • (+1) This actually works for $m=0$, whereas the answers that don't include the $i=n$ term don't. However, the overloading of the variable $n$ is confusing. – robjohn Nov 18 '21 at 21:29
2

To determine the number of surjective functions from set A={1,2,...,n} to a set B={A,B,C}, you will need to use Sterling's Numbers of the Second Kind, written S(n,k). n would be the size of set A and k would be the size of set B, which is 3.

Sterling Number of The Second Kind

S(n,k) represents the number of ways to partition a set of n labeled objects into k non-empty unlabelled subsets. So it is basically counting the number of different ways the elements of set A can map to elements of B such that each element of B has something from A mapping to it - thus this mapping would represent a surjection. You would want to multiply the number you get from calculating S(n,k) by k! to take care of the labeling. In conclusion, the number of surjections from a set with n elements to a set with k elements is: k!S(n,k)

Ricky
  • 31
0

This in my language (Portuguese) is "combinção completa" and is equivalent to solve \begin{equation} x + y + z = n \end{equation} where $x,y,z$ are integer $\ge 0$ with $x,y,z \ge 1$ (to be surjectve) because you can represents the solutions by a scheme of balls and traces. Thus, doing $x=1 +a, y=1+b$ and $z = 1+c$, we have $a+b+c =n+3$. Therefore, the solution is $CR^{n+3}_{3}= \binom{(n+3)+(3)-1}{n+3}= \binom{n+5}{n+3}$. ($CR^{n+3}_{3}$ is notation in my book in Portuguese.)

user29999
  • 5,211
-2

formula for finding the total number of surjections (onto functions) formula for finding the total number of surjections (onto functions)