Abstract
A partition of a positive integer \(n\) is a non-increasing sequence of positive integers whose sum is \(n\). It may be represented by a Ferrers diagram. These diagrams contain corners which are points of degree two. We define corners of types \((a,b)\), \((a+,b)\) and \((a+,b+)\), and also define the size of a corner. Via a generating function, we count corners of each type and corners of size \(m\). We also find asymptotics for the number of corners as \(n\) tends to infinity.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
A partition of a positive integer \(n\) is a non-increasing sequence of positive integers whose sum is \(n\), and the number of such partitions is denoted by \(P(n)\) (we define \(P(0)=1\)). A standard result is
Many researchers, for example George Andrews, have focused on the number of partitions satisfying certain conditions (for example, see [1–5, 8, 10, 11] and references therein). For instance, if \(Q(n)\) is the number of partitions of \(n\) with distinct parts and \(Q(0):=1\), then we have the generating function
In this paper, we define a new statistic, namely a corner for such partitions. This is related both to descents and the number of occurrences of a part (of fixed size). It is simple to describe corners in terms of the associated Ferrers diagrams.
The Ferrers diagram of an integer partition proves to be a useful tool for visualizing partitions. It is constructed by stacking left-justified rows of cells, where the number of cells in the \(i\)th row corresponds to the size of the \(i\)th part.
A corner of a partition \(\pi \) is a point of degree two in the corresponding Ferrers diagram. We denote the number of corners of \(\pi \) by \(cor(\pi )\). For example, if \(\pi =4422111\), then \(cor(\pi )=6\), see Fig. 1. Moreover, we define \(cor_k(\pi )\) to be the number of corners at line \(y=k\) in the Ferrers diagram of \(\pi \), where the topmost horizontal line of the Ferrers diagram is the line \(y=0\). For example, for the partition illustrated below, \(cor_0(\pi )=2\) (corners \(E\) and \(F\)), \(cor_2(\pi )=1\) (corner \(C\)) and so on. More generally, we are interested in several types of corners. We say that a corner is of type \((a,b)\) if it is at the bottom right of a specific maximal \(a\times b\) rectangle (where \(b\) is its length and \(a\) its height). For such a maximal rectangle, there are no cells below it and no cells to its right. Thus, corner \(B\) is at the bottom right of the 2 by 1 rectangle, with cells marked by X. So that \(B\) is a \((2,1)\) corner. Here, we shall only consider corners which are at the bottom-right extremities of such \(b\) by \(a\) rectangles. So for convenience, we ignore the 3 corners \(D\), \(E\) and \(F\) at levels \(x=0\) and \(y=0\). Thus, the partition in Fig. 1 has corners \(C\) of type \((2,2)\), \(B\) of type \((2,1)\) and \(A\) of type \((3,1)\).
We observe that a corner of type \((a,b)\) in the Ferrers diagram of \(\pi \) corresponds to consecutive parts \(c\,,\,c-b\) in the partition \(\pi \) such that the multiplicity of \(c\) is exactly \(a\) and the last occurrence of \(c\) is followed by a part of size \(c-b\). We also consider corners of type \((a+,b)\); these are corners of type \((j,b)\) for any \(j \ge a\). Similarly, we define corners of type \((a+,b+)\). Finally, we define the size of corners of type \((a,b)\) to be \(a+b\).
We summarize some of the main results obtained in this paper in Table 1.
2 Corners at level \(k\)
Let \(P_m=P_m(x,t_0,t_1,\ldots )\) be the generating function for the number of partitions of \(n\) where the first part is \(m\) and \(t_k\) marks the number of corners at level \(k\) (\(y=k\), measured downwards with \(y=0\) at the top). Each partition \(\pi \) with first part \(m\) can be written as \(\pi =m\pi '\), where \(\pi '\) is either empty or its largest part is less than or equal to \(m\). Therefore we obtain
Define \(P(z,x,t_0,t_1,\ldots ):=\sum _{m\ge 1}P_m(x,t_0,t_1,\ldots )z^m\). Multiplying the above equation by \(z^m\) and summing over \(m\ge 1\), we derive the following result.
Proposition 2.1
The generating function \(P(z,x,t_0,t_1,\ldots )\) satisfies the following functional equation:
Now, we are ready to study the total number of corners in a partition. Let \(P(z,x,t):=P(z,x,t,t,\ldots )\). Proposition 2.1 with \(t_j=t\) for all \(j\ge 0\) gives
which is equivalent to
By iterating this equation for \(z=1,x,x^2,\ldots \), and solving for \(P(1,x,t)\), we obtain the following result.
Theorem 2.2
The generating function for the number of partitions of \(n\) according to the number of corners counted by \(t\) is given by
We shall need the lemma below for the work that follows:
Lemma 2.3
For all \(a,b\ge 1\),
and
Proof
(a) We have
which implies that \(f(b)=\frac{1}{1-x^b}f(b+1)\). We use the fact that \(f(1)=\frac{1}{\prod _{i\ge 1}(1-x^i)}\) to complete the proof of the first identity.
(b) Next we prove the second identity using induction on \(a\). For \(a=1\), we have
Now, assume that the claim holds for all \(1,2,\ldots ,a\), and let us prove it for \(a+1\). Firstly, we have
Thus, by the induction hypothesis, we have
which completes the induction. \(\square \)
We now find the derivative of the generating function \(t^{-4}P(1,x,t)\) with respect to \(t\) and thereafter substitute \(t=1\):
Theorem 2.4
Thus, the generating function for the total number of corners is
This implies that the total number of corners in all partitions of \(n\) is given by
Remark 2.5
For any given partition \(\pi \), the total number of corners is equal to the number of distinct part sizes in \(\pi \) plus 3. Distinct part sizes in partitions have previously been studied in [7, 10].
By using Theorem 2.2 and the \(q\)-binomial theorem, we obtain
Theorem 2.6
The generating function for the number of partitions with exactly \(m+4\) corners is given by
Another application of our general result in Proposition 2.1 is to study the total number of corners at line \(y=2k\) for any \(k\ge 0\). In order to do that, we define
Proposition 2.1 shows
and
Therefore,
Iterating infinitely many times, we obtain
Thus, we may state the following result.
Theorem 2.7
The generating function \(Q(1,x,t)\), where \(t\) marks the number of corners at even levels, is given by
For the total number of such corners, we calculate
which implies
Note that \(\sum _{j\ge 0}\frac{x^jy^j}{\prod _{i=1}^j\big (1-x^i\big )}=\frac{1}{\prod _{j\ge 1}\big (1-yx^j\big )}\). By splitting the respective sums above into those with the largest part odd (and generating function \(\prod _{j \ge 1}\frac{1}{1-x^j}-\prod _{j \ge 1} \frac{1}{1+x^j}\)), or the largest part even (with generating function \(\prod _{\mathrm{j} \ge 1}\frac{1}{1-x^j}+\prod _{\mathrm{j} \ge 1} \frac{1}{1+x^j}\)), we obtain
Using the facts that \(\frac{1}{\prod _{j\ge 0}\big (1-x^j\big )}=\sum _{j\ge 0}P(j)x^j\) and \(\frac{1}{\prod _{j\ge 0}\big (1+x^j\big )}=\sum _{j\ge 0}P'(j)x^j\) (see sequence A081362 in [9]), we find
Theorem 2.8
Let \(n\ge 1\), then the total number of corners at even level \(2k\), where \(k\ge 0\), in all partitions of \(n\) is given by
3 Corners of type \((a,b)\)
In this section, we study partitions according to the number of corners of type \((a,b)\). Let
be the generating function for the number of partitions of \(n\) according to the number of corners of type \((a,b)\) with first part of size \(k\), where \(x\) marks the size of the partition and \(q\) the number of such corners. Then
where \(P_{0;a,b}(x):=1\). This is equivalent to
Define \(P_{a,b}(x,y,q):=\sum _{k\ge 0}P_{k;a,b}(x)y^k\). By multiplying the last recurrence by \(y^k\) and summing over \(k\ge 0\), we get the following result.
Theorem 3.1
The generating function \(P_{a,b}(x,y,q)\) satisfies
Now we will find an explicit formula for the generating function \(P_{1,b}(x,y,q)\). To do that we need the following lemma.
Lemma 3.2
Let \(\mathcal {A}_0=\{1\}\) and \(\mathcal {A}_1=\{a_0\}\). For \(n\ge 2\), we define \(\mathcal {A}_n\) to be the set
where \(B\cdot x=\{\pi \cdot x\mid \pi \in B\}\). Assume that \(\alpha _i=a_i\alpha _{i+1}+b_i\alpha _{i+2}\) for all \(i\ge 0\). Then
for all \(n\ge 0\).
Proof
We proceed by induction on \(n\). For \(n=0\), the statement reduces to \(\alpha _0=a_0\alpha _1+b_0\alpha _2\), which holds. Assume that the claim holds for \(n\), and let us prove it for \(n+1\). By the induction hypothesis, we have
which, by the definition of the sets \(\mathcal {A}_n\), implies
This completes the induction step. \(\square \)
Clearly, the set \(\mathcal {A}_n\) in the statement of the above Lemma 3.2 can be described bijectively as all sequences \(\pi _0\pi _1\cdots \pi _s\) such that \(\pi _0=0\), \(\pi _s\in \{n-1,n-2\}\) and \(\pi _j-\pi _{j-1}=1,2\) for all \(j=1,2,\ldots ,s\). We denote all such sequences by \(\mathcal {B}_n\). For each word \(\pi =\pi _0\pi _1\cdots \pi _j\in \mathcal {B}_n\), we define
Therefore,
By Theorem 3.1 with \(a=1\), the generating function \(P_{a,b}(x,y,q)\) satisfies
Define \(\alpha _i=P_{1,b}\big (x,x^iy,q\big )\), \(a_i=\frac{1}{1-x^{i+1}y}+(q-1)y^bx^{b(i+1)}\) and \(b_i=(1-q)y^bx^{b(i+2)}\); so
where \(\lim _{j\rightarrow \infty }\alpha _j=1\). Hence, by Lemma 3.2 (note that \(b_i\rightarrow 0\) when \(i\rightarrow \infty \)), we have
where \(\mathcal {B}_\infty =\lim _{n\rightarrow \infty }\mathcal {B}_n\) which is the set of all sequences \(\pi _0\pi _1\pi _2\cdots \) such that \(\pi _0=0\) and \(\pi _{i+1}-\pi _i\) is 1 or 2. So we can state the following result.
Theorem 3.3
The generating function \(P_{1,b}(x,y,q)\) is given by
Note that \(J(\pi )=\emptyset \) if and only if \(\pi =0123\cdots \). Thus the above theorem with \(q=1\) becomes
as is well known.
Also, note that \(|J(\pi )|=1\) if and only if there exists \(i\ge 0\) such that
and \(|J(\pi )|=2\) if and only if there exists \(i'>i+1\ge 1\) such that
(Similarly, we can characterize all the infinite sequences with \(J(\pi )=m\)). Then the above theorem shows that
Another way to solve the recurrence \(\alpha _i=a_i\alpha _{i+1}+b_i\alpha _{i+2}\) in Lemma 3.2 is by using continued fractions. Note that \(\frac{\alpha _i}{\alpha _{i+1}}=a_i+\frac{b_i}{\frac{\alpha _{i+1}}{\alpha _{i+2}}}\), for all \(i\ge 0\). Therefore,
So, multiplying over all \(i\ge 0\), assuming that \(\lim _{j\rightarrow \infty }\alpha _j=1\), we obtain
Thus, if we define \(\alpha _i=P_{1,b}(x,x^iy,q)\), \(a_i=\frac{1}{1-x^{i+1}y}+(q-1)y^bx^{b(i+1)}\) and \(b_i=(1-q)y^bx^{(2+i)b}\), we have the following result.
Theorem 3.4
The generating function \(P_{1,b}(x,y,q)\) is given by
Note that by comparing Theorems 3.3 and 3.4, we get the following corollary.
Corollary 3.5
We have
3.1 Total number of corners of type \((a,b)\)
Using Theorem 3.1, we find the generating function for the total number of corners of type \((a,b)\) to be
Using \(P_{a,b}(x,y,1)=\frac{1}{\prod _{j\ge 1}(1-x^jy)}\), this is equivalent to
By iterating this an infinite number of times, and noting that \(\frac{\mathrm{d}}{\mathrm{d}q}P_{a,b}(x,x^s,q){\Big |}_{q=1} \rightarrow 0\) as \(s \rightarrow \infty \), we have
Thus, by Lemma 2.3 and Eq. (3),
This is captured in the following theorem.
Theorem 3.6
The generating function for the total number of corners of type \((a,b)\) in all partitions is given by
Example 3.7
Theorem 3.6 for \(b=1\) leads to
which implies that the total number of corners of type \((a,1)\) in all partitions of \(n\) is given by
Let \(\mathcal {P}_n\) denote the set of all partitions of size \(n\). There is a simple correspondence between the total number of corners of type \((a,b)\) in \(\mathcal {P}_n\) and the number of parts \(\ge b\) of multiplicity \(a+1\) in \(\mathcal {P}_{n+b}\). We illustrate the bijection in Fig. 2.
Note, however, that the distributions of corners of type \((a,b)\) and of parts of multiplicity \(a+1\) are quite different. For example, for the case where \(a=1\), the generating function for partitions with no parts of multiplicity \(2\) is \(\prod _{i=1}^\infty (\frac{1}{1-x^i}-x^{2i})\), whereas by setting \(q=0\) in Theorem 3.4, the generating function for partitions with no corner of type \((1,b)\) is
4 Corners of type \((a+,b)\)
We define a corner of type \((a+,b)\) to be a corner of type \((j,b)\) for any \(j \ge a\). Let
be the generating function for the number of partitions according to the number of corners of type \((a+,b)\) with first part \(k\). Then for \(k\ge b\),
with \(P_{k;a+,b}(x)=\frac{x^k}{(1-x)(1-x^2)\cdots (1-x^k)}\) for \(k=0,1,\ldots ,b-1\). Thus,
Define \(P_{a+,b}(x,y,q):=\sum _{k\ge 0}P_{k;a+,b}(x,q)y^k\). By multiplying the last recurrence by \(y^k\) and summing over \(k\ge 0\), we get
We solve the recursion in the case where \(a=1\).
Theorem 4.1
The generating function \(P_{1+,b}(x,y,q)\) is given by
Moreover, the generating function for the number of partitions without corners of types \((1+,b)\) is given by
Proof
By (4) with \(a=1\), we have
which, by repeated iteration and using \(\lim _{i\rightarrow \infty }P_{1+,b}(x,x^iy,q)=1\), implies
as required. \(\square \)
4.1 Total number of corners of type \((a+,b)\)
Differentiating Eq. (4) with respect to \(q\), then substituting \(q=1\) and using \(P_{a+,b}(x,y,1)=\frac{1}{\prod _{j\ge 1}(1-x^jy)}\), we obtain
This implies
Iterating infinitely many times, we obtain
By Lemma 2.3, this leads to the following result.
Theorem 4.2
The generating function for the total number of corners of type \((a+,b)\) in all partitions of \(n\) is given by
Example 4.3
Theorem 4.2 for \(a=1\) shows that
Thus, the total number of corners of type \((1+,b)\) in all the partitions of \(n\) is given by
We note that a simple bijection between a Ferrers diagram and its transpose implies that the above results also count corners of type \((a,b+)\).
We also note that the number of corners of type \((a+,b)\) in the partitions of \(n\) equals the number \(\ge b\) of parts of size \(a+1\) in the partitions of \(n+b\). We illustrate the bijection in Fig. 3.
First, we add \(b\) cells to the first part of size \(j-b\) and then conjugate the \(a+1\) parts of size \(j\); thereafter we rearrange all parts (in the right-hand side drawing in Fig. 3) in descending order.
Example 4.4
For \(a=2\), the corners of type \((2+,1)\) in the partition \(2211\) maps to the two different partitions: \(322\) and \(331\). Also, the corner of type \((2+,1)\) in the partition \(111111\) (resp. \(21111\), \(3111\), \(411\)) maps to the partition \(31111\) (resp. \(3211\), \(331\), \(43\)).
5 Corners of type \((a+,b+)\)
Let \(P_{k;a+,b+}(x):=P_{k;a+,b+}(x,q)\) be the generating function for the number of partitions of \(n\) according to the number of corners of type \((a+,b+)\) with first part \(k\). Then for \(k\ge b\),
with \(P_{k;a+,b+}(x)=\frac{x^k}{(1-x)(1-x^2)\cdots (1-x^k)}\) for \(k=0,1,\ldots ,b-1\). Thus,
Define \(P_{a+,b+}(x,y,q):=\sum _{k\ge 0}P_{k;a+,b+}(x,q)y^k\). Multiplying the last recurrence by \(y^k\) and summing over \(k\ge 0\), we get
For \(a=1\), we obtain the following result.
Theorem 5.1
The generating function \(P_{1+,b+}(x,y,q)\) is
Note that \(P_{a+,b+}(x,y,1)=\frac{1}{\prod _{j\ge 1}(1-x^jy)}\). By differentiating Eq. (5) with respect to \(q\) and then substituting \(q=1\), we obtain
which implies
Iterating infinitely many times, we obtain
By Lemma 2.3b, this implies the following result.
Theorem 5.2
The generating function for the total number of corners of type \((a+,b+)\) in all partitions is given by
Example 5.3
Theorem 5.2 for \(a=1\) shows that
Thus, the total number of corners of type \((1+,b+)\) in all partitions of \(n\) is given by
6 Corners of size \(m\)
We shall define the size of a corner of type \((a,b)\) to be \(m\) when \(m=a+b\). In this section, we find the generating function that counts the total number of corners of size \(m\) in partitions of \(n\). For a corner of size \(m\), where \(m\) is fixed, there are \(p\) parts of size \(k\) followed by a part of size \(k+p-m\), for \(1 \le p \le m-1\). We denote the generating function by \(P_{k,m}(x):=P_{k,m}(x,q)\), where \(q\) marks such corners, as illustrated in Fig. 4.
Clearly, the generating function satisfies the following equation:
Multiplying by \((1-x^k)\) and simplifying,
We now define \(P_{m}(x,y,q):=\sum _{k \ge 0}P_{k,m}(x,y,q)y^k\). Thus after multiplying (6) by \(y^k\), summing over all \(k\) and interchanging the order of summation, we have
Following the previous procedure for finding the total number of corners, we compute
The generating function for all partitions is \(P_m(x,y,1)=\frac{1}{\prod _{j\ge 1}(1-x^jy)}\). Therefore
After infinitely many iterations and noticing that \(\frac{{\mathrm{d}} P_m(x,x^s,q)}{{\mathrm{d}}q} \rightarrow 0\) as \(s \rightarrow \infty \), we obtain our result
Theorem 6.1
The generating function for the total number of corners of size \(m\) is
7 Asymptotics
To find asymptotic estimates for the various types of corners studied above, we need to study generating functions of the form \(P(x)F(x)\), where \(P(x)\) is the generating function for the number of partitions. In the paper [6], the authors show how such asymptotic expansions can be obtained in a quasi-automatic way from expansions of \(F(x)\) around \(x = 1\). For the convenience of the reader, we state the relevant results from [6] that we need:
Theorem 7.1
Suppose that the function \(F(z)\) satisfies
and \(F(e^{-t}) = a t^b + \mathcal {O}(f(|t|))\) as \(t \rightarrow 0\), \(\mathfrak {R}t > 0\), for real numbers \(a,b\). Then one has
as \(n \rightarrow \infty \) for any \(0 < \epsilon < \frac{1 - \eta }{2}\), where \(I_{\nu }\) denotes a modified Bessel function of the first kind.
It is also shown in [6] that the quotient of modified Bessel functions simplifies, with \(h = |b+3/2|-1/2\) and \(m = \sqrt{\frac{2\pi ^2}{3} \left( n - \frac{1}{24} \right) }\), to
In our applications, we apply (13) together with Theorem 7.1. Thereafter, we use computer algebra to obtain asymptotic formulae in terms of powers of \(n\).
Firstly, in the case of the total number of corners of type \((a,b)\), we have \(\frac{\mathrm{d}}{\mathrm{d}q}P_{a,b}(x,y,q){\Big |}_{q=1}=P(x)F_{a,b}(x)\), where
To apply Theorem 7.1, we need to consider the Taylor expansion of \(F_{a,b}(e^{-t})\):
Now
and
Substituting these into (14) gives
Using this in Theorem 7.1 leads to
Theorem 7.2
The average number of corners of type \((a,b)\) in a random partition of \(n\) is
In the case of the total number of corners of type \((a+,b)\), we have \(\frac{\mathrm{d}}{\mathrm{d}q}P_{a,b}(x,y,q){\Big |}_{q=1}=P(x)F_{a,b}(x)\), where
From this we find
Using this in Theorem 7.1 leads to
Theorem 7.3
The average number of corners of type \((a+,b)\) in a random partition of \(n\) is
Next for the total number of corners of type \((a+,b+)\), we have
This yields
Using this in Theorem 7.1 leads to
Theorem 7.4
The average number of corners of type \((a+,b+)\) in a random partition of \(n\) is
For the total number of corners treated in Theorem 2.4, we use \(F(x)=\frac{2x-1}{1-x}+4\) in Theorem 7.1 to obtain
Theorem 7.5
The average number of corners in a random partition of \(n\) is
For the total number of corners at even level, the contribution of \(\frac{1}{2\prod _{j\ge 1}(1+x^j)}\) to (2) is asymptotically negligible; therefore, we may take \(F(x)=\frac{5-3x^2}{2(1-x^2)}\) in Theorem 7.1 to obtain
Theorem 7.6
The average number of corners at even level in a random partition of \(n\) is
Finally, for the total number of corners of size \(m\) we take
in Theorem 7.1, with the aid of the expansions (17) applied to cases \((a,m-a)\) with \(1\le a\le m-1\), to obtain
Theorem 7.7
The average number of corners of size \(m\) in a random partition of \(n\) is
References
Ahlgren, S., Lovejoy, J.: The arithmetic of partitions into distinct parts. Mathematika 48(1–2), 191–202 (2001)
Alon, N.: Restricted integer partition functions. Integers 13, A16 (2013)
Andrews, G.E.: The Theory of Partitions, Addison-Wesley, Reading 1976; reprinted. Cambridge University Press, Cambridge (1998)
Andrews, G.E.: Two theorems of Euler and a general partition theorem. Proc. Am. Math. Soc. 20, 499–502 (1969)
Glaisher, J.W.L.: A theorem in partitions. Messenger Math. 12, 158–170 (1883)
Grabner, P., Knopfmacher, A., Wagner, S.: A general asymptotic scheme for moments of partition statistics, accepted to special issue of Combinatorics, Probability and Computing dedicated to Philippe Flajolet.
Hirschorn, M.: The number of different parts in the partitions of \(n\), Fibonacci Quarterly, to appear.
Remmel, J.B.: Bijective proofs of some classical partition identities. J. Combin. Theory Ser. A 33(3), 273–286 (1982)
Sloane, N.J.A.: The On-Line Encyclopedia of Integer Sequences, published electronically at http://oeis.org (2010)
Wilf, H.S.: Three problems in combinatorial asymptotics. J. Combin. Theory Ser. A 35, 199–207 (1983)
Wilf, H.S.: Identically Distributed Pairs of Partition Statistics. Seminaire Lotharingien, The European digital Mathematics Library, 44, B44c, p. 3 (2000)
Author information
Authors and Affiliations
Corresponding author
Additional information
Charlotte Brennan and Arnold Knopfmacher were supported by the National Research Foundation under Grant Numbers 86329 and 81021, respectively.
Rights and permissions
About this article
Cite this article
Blecher, A., Brennan, C., Knopfmacher, A. et al. Counting corners in partitions. Ramanujan J 39, 201–224 (2016). https://doi.org/10.1007/s11139-014-9666-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11139-014-9666-4