PreprintPDF Available

On a recurrence relation for the sums of powers of integers

Authors:
  • Independent Researcher
Preprints and early-stage research may not have been peer reviewed yet.

Abstract

Recently, Thomas and Namboothiri [1] derived a recurrence identity expressing an exponential power sum with negative powers in terms of another exponential power sum with positive powers. From this result, the authors obtained a corresponding recurrence relation for the ordinary power sums $S_k(n) = 1^k + 2^k + \cdots + n^k$. In this short note, we provide an alternative simple proof of the latter recurrence. Our proof is based on the following two ingredients: (i) an expression for $S_k(n+m)$, and (ii) the symmetry property of the power sum polynomials $S_k(n)$.
On a recurrence relation for the sums
of powers of integers
*
Jos´e Luis Cereceda
Collado Villalba, 28400 (Madrid), Spain
jl.cereceda@movistar.es
Abstract
Recently, Thomas and Namboothiri [1] derived a recurrence identity expressing an exponen-
tial power sum with negative powers in terms of another exponential power sum with positive
powers. From this result, the authors obtained a corresponding recurrence relation for the or-
dinary power sums Sk(n)=1k+ 2k+· · · +nk. In this short note, we provide an alternative
simple proof of the latter recurrence. Our proof is based on the following two ingredients: (i) an
expression for Sk(n+m), and (ii) the symmetry property of the power sum polynomials Sk(n).
1 Introduction
In a remarkable paper [1], Thomas and Namboothiri derived a recurrence identity expressing the
exponential power sum Pk1
s=1 spe
2πims
kin terms of another exponential power sum with positive
powers (see Proposition 2.1 of [1]). As a consequence of this result, the authors obtained a corre-
sponding recurrence relation for the ordinary power sums Sk(n)=1k+ 2k+· · · +nk. Specifically,
they showed that (see, equation (8) of [1])
k1
X
s=1
sp=kp(k1) +
p1
X
a=0
(1)pap
aka
k1
X
s=1
spa.
It is easy to see that the above equation is equivalent to the recurrence relation given in Propo-
sition 1.1 below. In Section 2, we provide a simple novel proof of the recurrence in question. In
Section 3, we derive a variant of Proposition 1.1. We conclude in Section 4 with some final remarks.
Proposition 1.1. For integers k, n 1, let Sk(n)denote the power sum 1k+ 2k+· · · +nk. Then,
k1
X
j=0
(1)jk
jnkjSj(n1) = 2Sk(n1),if k is odd;
0,if k is even,(1)
where S0(n) = n, and where it is assumed that Sj(0) = 0 for all j.
2 Proof of Proposition 1.1
Let us first observe that, as shown in [2], Sk(n+m) can be expressed as
Sk(n+m) = Sk(n) + Sk(m) +
k1
X
j=0 k
jnkjSj(m),(2)
*
This is a slightly corrected version of the paper published in the Far East Journal of Mathematical Education
24, 2023, pp. 45-50. The two corrections made here are: the inclusion of the factor (1)kin the right-hand side of
equation (3), and the quotation of Theorem 10 of [3], instead of Theorem 3 of [3].
1
because
Sk(n+m)=1k+ 2k+· · · +nk+ (n+ 1)k+· · · + (n+m)k
=Sk(n) +
k
X
j=0 k
jnkj1j+
k
X
j=0 k
jnkj2j
+· · · +
k
X
j=0 k
jnkjmj
=Sk(n) +
k
X
j=0 k
jnkj1j+ 2j+· · · +mj.
Therefore, making the transformations n nand mn1, it follows that
Sk(1) = Sk(n) + Sk(n1) + (1)k
k1
X
j=0
(1)jk
jnkjSj(n1).(3)
On the other hand, by invoking the symmetry property of the power sum polynomials Sk(n) (cf.,
e.g., Theorem 10 of [3])
Sk(n1) = δk,0+ (1)k+1Sk(n),(4)
we immediately deduce that, for all k1,
Sk(n) = (1)k+1Sk(n1).(5)
In particular, from (5) we find that Sk(1) = 0 provided k1. Hence, substituting (5) into (3)
and taking into account that Sk(1) = 0 for k1 yields
1+(1)k+1 Sk(n1) =
k1
X
j=0
(1)jk
jnkjSj(n1), k 1,
which is just the recurrence relation (1).
Remark 2.1. By letting n= 2 in (1), we obtain the following identity for k1:
k1
X
j=0 1
2jk
j=1/2k1,if k is odd;
0,if k is even.
3 A variant of Proposition 1.1
It is worthwhile to point out that, by making the transformation n nand using the symmetry
property (4), we can express (1) in the alternate form
nk+
k1
X
j=0
(1)jk
jnkjSj(n) = 2Sk(n),if kis odd ;
0,if kis even.(6)
2
Let us note, incidentally, that letting n= 1 in the preceding equation leads to the well-known
binomial identity
k
X
j=0
(1)jk
j= 0,for all k1.
On the other hand, putting n=1 in (2) and recalling that Sk(1) = 0 for all k1, we obtain
mk= (1)k1
k1
X
j=0
(1)jk
jSj(m), k 1.
Thus, renaming mas nin the last equation and substituting the resulting expression for nkinto
(6) gives rise to the following variant of (1):
k1
X
j=0
(1)jk
jnkj(1)kSj(n) = 2Sk(n),if kis odd ;
0,if kis even.(7)
For the case where k1 is odd, from (7) we have
Sk(n) = 1
2
k1
X
j=0
(1)jk
jnkj+ 1Sj(n),odd k1,
which should be compared with the formula for Sk(n) that is obtained from (1) when k1 is odd,
namely
Sk(n) = 1
2
k1
X
j=0
(1)jk
j(n+ 1)kjSj(n),odd k1.(8)
Likewise, for the case in which k2 is even, from (7) we obtain the identity
k1
X
j=0
(1)jk
jnkj1Sj(n)=0,even k2.(9)
Notice that (9) allows one to obtain the odd-indexed power sum Sk1(n) in terms of S0(n),
S1(n), . . . , Sk2(n).
4 Concluding remarks
We end this note with the following remarks.
Remark 4.1. The recurrence relation (1)was already proved by El-Mikkawy and Atlan in Theorem
3.1 (viii) of [4] by manipulating the exponential generating function of Sk(n).
Remark 4.2. In Proposition 8.6.1 of [5], Treeby obtained the following recursive formula
Sk(n) = 1
k1
k1
X
j=1
(1)jk
j1(n+ 1)kjSj(n),(10)
which holds for any odd integer k3.
3
Remark 4.3. By combining (8)and (10), we obtain in turn the recursive formula
Sk(n) = 1
k+ 1
k1
X
j=0
(1)jk+ 1
j(n+ 1)kjSj(n),odd k1,(11)
which applies to any odd integer k1. Note that, for each j= 0,1, . . . , k 1, the product
(n+ 1)kjSj(n)is a polynomial in nof degree k+ 1.
Example 4.4. When k= 3, from (11)it follows that
S3(n) = 1
4(n+ 1)3S0(n)4(n+ 1)2S1(n) + 6(n+ 1)S2(n)
=1
4n(n+ 1)32n(n+ 1)3+n(n+ 1)2(2n+ 1),
which, after simplifying, gives us the well-known result
S3(n) = 1
4n2(n+ 1)2=S1(n)2.
A beautiful proof without words of the identity S3(n)=(S1(n))2can be found in the recent
paper [6]. Finally, it should be noted that the formula (11) can also be obtained directly starting
from (9). We leave it as an easy exercise for the interested reader to check this fact.
References
[1] N. E. Thomas and K. V. Namboothiri, On an exponential power sum, preprint (2023), available
at https://arxiv.org/abs/2303.10853v1
[2] D. R. Snow, Some identities for sums of powers of integers, Proceedings of Utah Academy of
Sciences, Arts, and Letters, 52 Part 2, pp. 29–31 (1975).
[3] N. J. Newsome, M. S. Nogin, and A. H. Sabuwala, A proof of symmetry of the power sum
polynomials using a novel Bernoulli number identity, Journal of Integer Sequences 20, Article
17.6.6 (2017).
[4] M. El-Mikkawy and F. Atlan, Notes on the power sum 1k+ 2k+. . . +nk,Annals of Pure and
Applied Mathematics 9(2), 215–232 (2015).
[5] D. Treeby, Optimal block stacking and combinatorial identities via Archimedes’ method, PhD
thesis, Monash University, 2018.
[6] F. Mart´ınez de la Rosa, Proof without words: sums of cubes in puzzle form, Far East Journal
of Mathematical Education 23, 7 (2022).
4
ResearchGate has not been able to resolve any citations for this publication.
Thesis
Full-text available
What is the maximum overhang that can be obtained when a set of blocks of variable width are stacked so that there is one block at each level? We determine the configuration of blocks that maximises the overhang when these blocks are stacked in a prescribed order. We also describe infinite sets of blocks of variable width that can be stacked to yield towers with arbitrarily large overhang. This investigation is a generalisation of a classic problem, which asks how n identical blocks should be stacked to create a tower of maximal overhang. The solution to this problem is well-known: one can achieve an overhang that is proportional to the nth harmonic number. Since the harmonic numbers diverge, one can theoretically construct a balanced tower of any conceivable overhang, given a sufficient quantity of blocks. When considering blocks of variable width, we show that one can obtain arbitrarily large overhang if and only if a particular series diverges. The series in question has not been studied to date. Using this result, we find characterisations of sets of blocks that can be stacked to yield towers with arbitrarily large overhang. We will obtain a counterpart to the Abel-Dini Theorem, which, informally, states that there is no slowest diverging series of positive terms. The block stacking counterpart asserts that there is no stack whose overhang diverges more slowly than all others. We then allow the blocks to be permuted, obtaining a partial description of how a finite set of blocks should be stacked, and in what order, to maximise the stack’s overhang. For small sets of blocks, this partial description will allow us to determine the maximal overhang in exponential time. In an additional line of inquiry, we will consider various symmetric configurations of blocks and point masses in the plane. By considering the centres of mass of such configurations we establish various well-known summation formulas, and discover others that are new and appealing. The method that we employ can be regarded as a discrete counterpart to the physical method conceived by Archimedes to determine the areas and volumes of various geometric figures. For instance, Archimedes used his Law of the Lever to establish the area of a parabolic segment, while we use similar arguments to establish a formula for the sum of consecutive squares. We extend the investigation by finding essentially physical arguments to find formulas for other sums of powers, combinatorial identities and Fibonacci summation formulas.
Article
Full-text available
Abstract. In this paper we present some efficient computational methods to compute the power sum. Some new identities are given.
On an exponential power sum
  • N E Thomas
  • K V Namboothiri
N. E. Thomas and K. V. Namboothiri, On an exponential power sum, preprint (2023), available at https://arxiv.org/abs/2303.10853v1
Some identities for sums of powers of integers
  • D R Snow
D. R. Snow, Some identities for sums of powers of integers, Proceedings of Utah Academy of Sciences, Arts, and Letters, 52 Part 2, pp. 29-31 (1975).
  • N J Newsome
  • M S Nogin
  • A H Sabuwala
N. J. Newsome, M. S. Nogin, and A. H. Sabuwala, A proof of symmetry of the power sum polynomials using a novel Bernoulli number identity, Journal of Integer Sequences 20, Article 17.6.6 (2017).