Fact-checked by Grok 2 weeks ago

Sums of powers

Sums of powers, also known as power sums, refer to the finite sum S_p(n) = \sum_{k=1}^n k^p, where p is a non-negative and n is a positive representing the upper of . These sums express the total of the p-th powers of the first n positive integers and result in a in n of degree p+1. They are fundamental in mathematics, appearing in for studying arithmetic progressions and figurate numbers, in through connections to integrals and series, and in as moments of discrete distributions. The investigation of sums of powers has ancient origins. The Pythagoreans in the 6th century BCE recognized the sum for p=1, S_1(n) = \frac{n(n+1)}{2}, as the formula for triangular numbers. In the BCE, derived the formula for p=2, S_2(n) = \frac{n(n+1)(2n+1)}{6}, using geometric methods in his works On Conoids and Spheroids and On Spirals to compute areas and volumes. Early medieval contributions include Nicomachus's work on cubes in the 1st century CE and formulas for p=3 and p=4 by Islamic mathematicians such as Al-Karaji around 1000 CE and Alhazen () in the 11th century. Significant advancements occurred in the with European mathematicians. and developed recursive methods using binomial coefficients, while Johann Faulhaber published explicit formulas for sums up to p=17 (and mentioned up to p=100) in his 1631 treatise Academia Algebræ, expressing them in terms of binomial coefficients without a unified general form. A breakthrough came posthumously in 1713 when Jakob introduced the Bernoulli numbers in , enabling a compact general expression for all p. This led to in its modern form: S_p(n) = \frac{1}{p+1} \sum_{j=0}^p (-1)^j \binom{p+1}{j} B_j n^{p+1-j}, where B_j are the Bernoulli numbers (with B_1 = -\frac{1}{2} in the convention used here). provided a rigorous proof of this formula for all positive integers p in 1834. Beyond closed forms, sums of powers connect to broader mathematical structures, such as the Euler-Maclaurin formula for approximating integrals and Nicomachus's theorem stating S_3(n) = \left( S_1(n) \right)^2. They also relate to the via S_p(n) = \zeta(-p) - \zeta(-p, n+1) for analytic continuations, highlighting their role in .

Fundamentals

Definition

In , the p-th power , often denoted S_p(n), is defined as the finite S_p(n) = \sum_{m=1}^n m^p, where n and p are positive integers. This expression aggregates the p-th powers of the first n natural numbers, providing a foundational object in and discrete analysis. Power sums differ from , which are infinite expansions of the form \sum_{k=0}^\infty a_k x^k for a x, or from convergent infinite series such as \sum_{k=1}^\infty \frac{1}{k^2} = \frac{\pi^2}{6}. Instead, they emphasize finite over the initial segment of natural numbers, avoiding issues associated with infinite terms. These sums hold relevance in assessing behaviors, such as and leading coefficients in their closed forms, and act as discrete analogs to integrals, approximating quantities like \int_0^n x^p \, dx through interpretations scaled appropriately. Bernoulli numbers serve as key tools for deriving general expressions for power sums. For illustration, when p=1 and n=3, the computation yields S_1(3) = 1^1 + 2^1 + 3^1 = 6.

Notation and basic examples

In , the of the p-th powers of the first n positive integers, \sum_{k=1}^n k^p, is commonly denoted as S_p(n), where the subscript p indicates the exponent and n specifies the number of terms. Alternative notations include p_k(n) to emphasize the k-th power, with the subscript or superscript varying by context to highlight the power . Uppercase letters such as S typically represent the summation, while lowercase p or k denotes the power, reflecting conventions in and . To illustrate, consider basic computations for small values of n and p. For p=1 and n=3, the sum is $1^1 + 2^1 + 3^1 = 6, showing linear accumulation. For p=2 and n=4, it yields $1^2 + 2^2 + 3^2 + 4^2 = 1 + 4 + 9 + 16 = 30, where the increasing contributions from larger bases highlight quadratic growth. These manual summations for modest n and p reveal emergent patterns, such as accelerating increments with higher powers, without relying on closed forms. Power sums also arise descriptively in generating functions, where they serve as coefficients in formal power series expansions, particularly for encoding sequences in combinatorial identities and symmetric polynomial bases. The following table lists computed values for p=1,2,3 and n=1 to $5, offering a concise reference for these foundational cases:
n \ p123
1111
2359
361436
41030100
51555225
While efficient methods like enable direct evaluation for larger parameters (covered later), such examples foster intuition for the sums' nature.

Historical development

Early contributions

The study of sums of powers traces its origins to . The Pythagoreans in the 6th century BCE recognized the sum for p=1, S_1(n) = \frac{n(n+1)}{2}, as the formula for triangular numbers. In the 3rd century BCE, derived the formula for p=2, S_2(n) = \frac{n(n+1)(2n+1)}{6}, using geometric methods in his works On Conoids and Spheroids and The Quadrature of the Parabola. in the 1st century CE contributed to sums of cubes (p=3). Subsequent advancements occurred in ancient , where (c. 476–550 CE) provided some of the earliest known formulas for the sums of the first n squares and cubes of natural numbers in his treatise (499 CE). These contributions represented a significant advancement in understanding arithmetic series raised to powers, focusing on explicit expressions for low exponents to aid astronomical and geometric calculations. In the , scholars built upon such ideas, with Abu Bakr al-Karaji (d. 1019) making notable progress in on related problems, including a geometric proof of the identity linking the sum of the first n cubes to the square of the sum of the first n natural numbers. Al-Karaji's work extended to pyramidal numbers, which involve cumulative sums akin to higher-order power sums, demonstrating iterative methods for computing these aggregates without a general algebraic framework. His approaches emphasized practical computation for engineering and algebraic texts. Overall, pre-17th-century work on power sums centered on specific low exponents, driven by practical needs and constrained by the absence of a comprehensive theoretical structure.

Key advancements by Bernoulli and Euler

In 1631, Johann Faulhaber published empirical formulas for the sums of powers of the first n positive integers up to the 17th power in his book Academia Algebrae, marking a significant empirical advancement in the study of power sums. These formulas, expressed using a notation involving "cossic" terms for powers, extended earlier particular cases and included expressions up to the 23rd power in encoded form, though presented explicitly only for higher exponents like the 13th power as a in n divided by 105. Jakob Bernoulli advanced this work in his posthumously published Ars Conjectandi in 1713, where he derived symbolic formulas for sums of powers up to the 10th power and conjectured a general pattern involving a sequence of coefficients now known as numbers. linked these sums to finite differences and s, using notation \int for and observing that the general \sum_{k=1}^n k^p could be expressed as \frac{1}{p+1} n^{p+1} + \frac{1}{2} n^p + \sum terms with numbers B_j, specifically conjecturing the form \sum_{m=1}^n m^p = \frac{1}{p+1} \sum_{j=0}^p (-1)^j \binom{p+1}{j} B_j n^{p+1-j}. In the , Leonhard Euler provided rigorous proofs of 's conjectures around 1730 through his development of a general formula, confirming the pattern for arbitrary powers and establishing its theoretical foundation. Euler further refined the approach by connecting numbers to exponential generating functions in the early 1730s, defining them as coefficients in the expansion \frac{x}{e^x - 1} = \sum_{k=0}^\infty B_k \frac{x^k}{k!}, which facilitated broader applications in .

General closed-form expressions

Faulhaber's formula

gives an explicit for the sum of the m-th powers of the first n positive integers as \sum_{k=1}^n k^m = \frac{1}{m+1} \sum_{j=0}^m (-1)^j \binom{m+1}{j} B_j n^{m+1-j}, where B_j denotes the j-th (with the convention B_1 = -\frac{1}{2}). This representation reveals that the power sum is a in n of degree exactly m+1, with leading term \frac{n^{m+1}}{m+1}. The components of the formula include binomial coefficients \binom{m+1}{j}, which determine the scaling of each term, alternating signs (-1)^j that alternate the contributions, and powers n^{m+1-j} that decrease from m+1 to 1, ensuring the polynomial structure. The numbers B_j serve as coefficients that encapsulate the combinatorial of the . Faulhaber first published specific formulas for up to the 17th power in his 1631 treatise Academia Algebrae, expressing them as explicit polynomials without reference to numbers, which were developed later by Jakob . A modern general form equivalent to Faulhaber's approach, avoiding numbers, rewrites the using Stirling numbers of the second kind S(m,j) and falling factorials: \sum_{k=1}^n k^m = \sum_{j=0}^m S(m,j) \frac{(n+1)^{\underline{j+1}}}{j+1}, where x^{\underline{j}} = x(x-1)\cdots(x-j+1). To verify the formula, consider m=1: \sum_{k=1}^n k = \frac{1}{2} \sum_{j=0}^1 (-1)^j \binom{2}{j} B_j n^{2-j}. For j=0: (-1)^0 \binom{2}{0} B_0 n^2 = 1 \cdot 1 \cdot n^2 = n^2. For j=1: (-1)^1 \binom{2}{1} B_1 n = -2 \cdot \left(-\frac{1}{2}\right) n = n. Thus, \frac{1}{2} (n^2 + n) = \frac{n(n+1)}{2}, matching the known formula.

Role of Bernoulli numbers

Bernoulli numbers B_k are a of rational numbers defined by the exponential \frac{x}{e^x - 1} = \sum_{k=0}^\infty B_k \frac{x^k}{k!}, which holds for |x| < 2\pi. The first few values are B_0 = 1 and B_1 = -\frac{1}{2}, with subsequent terms determined recursively. A key property is that odd-indexed Bernoulli numbers vanish beyond the first: B_{2n+1} = 0 for all integers n \geq 1. This vanishing simplifies the expressions in power sum formulas, as terms involving odd powers greater than 1 drop out, leading to more compact polynomials for even exponents and adjustments for odd ones. In the context of sums of powers, the Bernoulli numbers serve as the coefficients in the polynomial expansion provided by , weighting the powers of n to yield the exact sum \sum_{k=1}^n k^p. They arise naturally from the theory of finite differences because the associated B_m(x) satisfy the difference equation B_m(x+1) - B_m(x) = m x^{m-1}, allowing the indefinite sum of powers to be expressed as a telescoping series involving these polynomials. The Bernoulli numbers are computed using the recursive relation \sum_{j=0}^m \binom{m+1}{j} B_j = 0 for m > 1, with the initial condition B_0 = 1. This relation enables sequential determination of higher terms. The first ten Bernoulli numbers are given in the following table:
nB_n
01
1-\frac{1}{2}
2\frac{1}{6}
30
4-\frac{1}{30}
50
6\frac{1}{42}
70
8-\frac{1}{30}
90
10\frac{5}{66}
These values illustrate the alternation in sign for even indices and the zeros for odd indices greater than 1.

Explicit formulas for small exponents

Sum of first powers

The sum of the first n positive integers is given by the formula \sum_{k=1}^n k = \frac{n(n+1)}{2}. This result can be derived by pairing terms in the sum: the first term $1 pairs with the last n to give n+1, the second $2 with n-1 to give n+1, and so on, yielding n/2 such pairs each summing to n+1, for a total of \frac{n(n+1)}{2}. Geometrically, this sum represents the nth T_n, formed by arranging n unit disks or dots into an with n dots along each side. The formula has been known since antiquity, dating back to the Pythagoreans in the BCE, who studied as part of theory. A popular but apocryphal attributes its rediscovery to the young , who reportedly summed the integers from 1 to 100 using the pairing method as a schoolboy in the late , though the story first appeared in print over a century later. For example, when n=10, the is \frac{10 \times 11}{2} = 55. The formula can also be verified by : the base case n=1 holds as $1 = \frac{1 \times 2}{2}; assuming it holds for n=m, then for n=m+1, the is \frac{m(m+1)}{2} + (m+1) = \frac{(m+1)(m+2)}{2}, completing the proof. As a special case of for power sums, this expression arises directly when the exponent p=1.

Sum of squares

The sum of the squares of the first n natural numbers is given by the formula \sum_{k=1}^n k^2 = \frac{n(n+1)(2n+1)}{6}. This captures the quadratic nature of the sum, as it is a cubic in n, reflecting the accumulation of squared terms. One standard derivation uses a based on the difference of cubes. Start with the identity (k+1)^3 - k^3 = 3k^2 + 3k + 1. Summing both sides from k=1 to n yields \sum_{k=1}^n \left[ (k+1)^3 - k^3 \right] = 3 \sum_{k=1}^n k^2 + 3 \sum_{k=1}^n k + \sum_{k=1}^n 1. The left side telescopes to (n+1)^3 - 1^3 = (n+1)^3 - 1. Substituting the known sums \sum_{k=1}^n k = \frac{n(n+1)}{2} and \sum_{k=1}^n 1 = n allows solving for \sum_{k=1}^n k^2, confirming the formula after algebraic simplification. An equivalent expanded form is \sum_{k=1}^n k^2 = \frac{2n^3 + 3n^2 + n}{6}, which highlights the leading cubic term dominating for large n. For example, when n=5, the direct sum is $1^2 + 2^2 + 3^2 + 4^2 + 5^2 = 1 + 4 + 9 + 16 + 25 = 55, matching \frac{5 \cdot 6 \cdot 11}{6} = 55. In statistics, this sum relates to the variance of a discrete uniform distribution over \{1, 2, \dots, n\}, where the second moment E[X^2] = \frac{1}{n} \sum_{k=1}^n k^2 = \frac{(n+1)(2n+1)}{6}, and the variance is E[X^2] - (E[X])^2 = \frac{n^2 - 1}{12}. This connection underscores its role in measuring dispersion for equally likely integer outcomes. The formula also emerges as a special case of using Bernoulli numbers, where the general expression for p=2 simplifies directly to this result.

Sum of cubes

The sum of the cubes of the first n positive integers is given by the formula \sum_{k=1}^n k^3 = \left( \frac{n(n+1)}{2} \right)^2, where \frac{n(n+1)}{2} is the nth . This identity, known as Nicomachus's theorem, highlights the elegant connection between cubes and squared triangular numbers, demonstrating that the total volume of stacked unit cubes up to n^3 rearranges perfectly into a square layer. The result was first noted by the Greek mathematician around 100 AD in his work Introduction to Arithmetic, where he explored patterns in odd numbers summing to cubes, leading to this broader observation. A formal algebraic proof was provided by Claude Bachet de Méziriac in , using expansion and verification for small n. One standard proof uses . Assume the formula holds for n = m, so \sum_{k=1}^m k^3 = \left( \frac{m(m+1)}{2} \right)^2. For n = m+1, \sum_{k=1}^{m+1} k^3 = \left( \frac{m(m+1)}{2} \right)^2 + (m+1)^3 = \frac{(m+1)^2 m^2}{4} + (m+1)^3 = (m+1)^2 \left[ \frac{m^2}{4} + (m+1) \right]. The term in brackets simplifies to \frac{m^2 + 4m + 4}{4} = \frac{(m+2)^2}{4}, so the expression becomes (m+1)^2 \cdot \frac{(m+2)^2}{4} = \left( \frac{(m+1)(m+2)}{2} \right)^2, confirming the formula. The base case n=1 holds as $1^3 = 1^2. Alternatively, Nicomachus's visual proof involves dissecting and rearranging the cubes into gnomons around a central square, forming the overall square of side \frac{n(n+1)}{2}. For example, when n=4, $1^3 + 2^3 + 3^3 + 4^3 = 1 + 8 + 27 + 64 = 100 = 10^2, and the 4th is \frac{4 \cdot 5}{2} = 10. Although the formula expands to the \frac{n^4 + 2n^3 + n^2}{4}, the compact squared form is preferred for its insight into geometric structure. This expression arises as a special case of when the exponent is 3.

Sum of fourth powers

The sum of the fourth powers of the first n positive integers is given by the formula \sum_{k=1}^n k^4 = \frac{n(n+1)(2n+1)(3n^2 + 3n - 1)}{30}. This expression can be derived using Faulhaber's method, which expresses power sums in terms of binomial coefficients and Bernoulli numbers, or alternatively by telescoping the difference of fifth powers, where \sum_{k=1}^n (k^5 - (k-1)^5) simplifies to yield the fourth-power sum after accounting for lower-order terms. An equivalent expanded form as a is \sum_{k=1}^n k^4 = \frac{1}{5} n^5 + \frac{1}{2} n^4 + \frac{1}{3} n^3 - \frac{1}{30} n. As an even power sum, \sum k^4 exhibits symmetry properties inherent to even exponents and appears in the computation of statistical moments, such as the fourth central moment in discrete uniform distributions over \{1, \dots, n\}. The polynomial form simplifies through the involvement of Bernoulli numbers, specifically B_4 = -\frac{1}{30}, which contributes the linear term coefficient in Faulhaber's general expansion for p=4. For example, when n=3, the direct computation is $1^4 + 2^4 + 3^4 = 1 + 16 + 81 = 98, which matches the formula: \frac{3 \cdot 4 \cdot 7 \cdot (27 + 9 - 1)}{30} = \frac{3 \cdot 4 \cdot 7 \cdot 35}{30} = 98. This formula for fourth powers illustrates the increasing complexity for higher even exponents, where additional terms introduce more fractional coefficients beyond the cubic case.

Properties and identities

Symmetry and generating functions

The power sums S_p(n) = \sum_{k=1}^n k^p possess a fundamental recursive defined by the relation S_p(n) - S_p(n-1) = n^p for n \geq 1, with the S_p(0) = 0. This underscores the additive structure of the sums, allowing n^p to be isolated as the incremental contribution to the total sum. Such symmetry facilitates inductive constructions and is central to algorithms for computing power sums efficiently, as each term builds directly on the previous sum. Generating functions offer an elegant framework for analyzing the sequence \{S_p(n)\}_{n=1}^\infty. The ordinary takes the form \sum_{n=1}^\infty S_p(n) x^n = \frac{x A_p(x)}{(1-x)^{p+2}}, where A_p(x) is the p-th Eulerian polynomial, a of p-1 with nonnegative coefficients given by A_p(x) = \sum_{j=0}^{p-1} \langle p \atop j \rangle x^j, and \langle p \atop j \rangle denote the Eulerian numbers. This is obtained by interchanging the order of summation in the double defining S_p(n), yielding \sum_{n=1}^\infty S_p(n) x^n = \frac{1}{1-x} \sum_{k=1}^\infty k^p x^k, with the inner \sum_{k=1}^\infty k^p x^k = \frac{x A_p(x)}{(1-x)^{p+1}}. The Eulerian polynomials thus encode the combinatorial structure underlying the coefficients of the power in this series expansion. Bernoulli numbers aid in determining these coefficients through related expansions, such as in integrated into the context. As an illustrative case, for p=1, A_1(x) = 1, so the simplifies to \frac{x}{(1-x)^3}, which generates the triangular numbers S_1(n) = \frac{n(n+1)}{2}. This highlights how the at x=1 of order p+2 reflects the asymptotic growth of S_p(n) \sim \frac{n^{p+1}}{p+1}. emerges as a direct consequence of symmetric polynomials within the framework of finite s. Specifically, expressing the monomial n^p in the basis of falling factorials (n)_k = n(n-1)\cdots(n-k+1), which are symmetric under permutations of their arguments in the difference table, allows the indefinite sum (antidifference) to yield the closed form. The coefficients in this expansion, involving of the second kind, lead to the expression for S_p(n) = \frac{1}{p+1} \sum_{j=0}^p (-1)^j \binom{p+1}{j} B_j n^{p+1-j}, where B_j are numbers; the symmetry ensures the formula's consistency across the discrete derivative structure.

Relations to other sums

Power sums are closely related to harmonic numbers, particularly in the context of generalized summations. For the zeroth power, the sum \sum_{k=1}^n k^0 = n, which aligns with the degenerate case of the H_n^{(0)} = n, though the focus on positive powers distinguishes them from the standard harmonic series \sum_{k=1}^n \frac{1}{k} = H_n. More broadly, generalized harmonic numbers H_n^{(s)} = \sum_{k=1}^n \frac{1}{k^s} for s > 0 involve negative exponents, contrasting with the positive power sums \sum_{k=1}^n k^p for p > 0, yet both share connections through and in their closed-form expressions. A key interconnection arises via the binomial theorem and Stirling numbers of the second kind, which express powers in terms of binomial coefficients. Specifically, any power k^p can be rewritten as k^p = \sum_{m=0}^p S(p,m) \cdot k^{\underline{m}}, where S(p,m) are the Stirling numbers of the second kind and k^{\underline{m}} = k(k-1)\cdots(k-m+1) is the falling factorial, equivalently k^{\underline{m}} = m! \binom{k}{m}. This representation allows power sums to be transformed into sums over falling factorials or binomial terms, facilitating relations to combinatorial identities. Faulhaber's formula briefly leverages these Stirling numbers for general expressions, though the emphasis here is on the structural ties. The relation to falling factorial sums is direct through these Stirling conversions, enabling the summation \sum_{k=1}^n k^p = \sum_{m=0}^p S(p,m) \sum_{k=1}^n k^{\underline{m}}. The inner \sum_{k=1}^n k^{\underline{m}} simplifies to \frac{(n+1)^{\underline{m+1}}}{m+1}, linking power sums to higher-order falling factorials and underscoring their combinatorial underpinnings. An illustrative connecting quadratic terms to sums is \sum_{k=1}^n k(k-1) = 2 \sum_{k=1}^n \binom{k}{2} = \frac{n(n-1)(n+1)}{3}, which reduces the second-degree falling factorial sum to a cubic expression and demonstrates how power sums of degree 2 relate to lower-degree or combinatorial quantities.

Asymptotic approximations

Euler-Maclaurin formula application

The Euler-Maclaurin formula provides a powerful method for approximating sums by integrals with corrective terms involving numbers, enabling the derivation of asymptotic expansions for various functions, including power sums. The formula states that for a smooth function f on the interval [a, b], \sum_{k=a}^{b} f(k) \approx \int_{a}^{b} f(x) \, dx + \frac{f(a) + f(b)}{2} + \sum_{k=1}^{m} \frac{B_{2k}}{(2k)!} \left( f^{(2k-1)}(b) - f^{(2k-1)}(a) \right) + R_m, where B_{2k} are the even-indexed numbers, f^{(2k-1)} denotes the (2k-1)-th , and R_m is the term. This expansion is particularly useful for when b = n is large and a = 1 is fixed, as the contributions from the lower become negligible compared to those at the upper . When applied to power sums S_p(n) = \sum_{k=1}^n k^p with f(x) = x^p (assuming p > -1 for of the integral), the integral term evaluates to \int_1^n x^p \, dx = \frac{n^{p+1}}{p+1} - \frac{1}{p+1}, which for large n is asymptotically \frac{n^{p+1}}{p+1}. The endpoint correction simplifies to approximately \frac{n^p}{2}, and the higher-order terms involve the derivatives of x^p, yielding the S_p(n) \sim \frac{n^{p+1}}{p+1} + \frac{n^p}{2} + \sum_{k=1}^{\lfloor p/2 \rfloor} \frac{B_{2k}}{2k} \binom{p}{2k-1} n^{p+1-2k}. Here, the sum is finite for fixed integer p since higher derivatives vanish, and Bernoulli numbers B_{2k} (with B_2 = 1/6, B_4 = -1/30, etc.) provide the coefficients for the subleading powers. This expansion captures the behavior of S_p(n) up to lower-order terms, with the dominant contribution from \frac{n^{p+1}}{p+1}. The leading terms establish the scale: for large n, S_p(n) = \frac{n^{p+1}}{p+1} + O(n^p), reflecting that the sum grows like the of x^p. Error bounds depend on the R_m, which can be estimated by the next unevaluated term or via the integral form of the ; for polynomials like x^p, truncating after the terms where derivatives vanish gives an representation up to constants, but in the asymptotic regime, the error from ignoring the lower contributions is O(1). For practical verification, consider p=2: the sum is \sum_{k=1}^n k^2 = \frac{n(n+1)(2n+1)}{6}, while the asymptotic up to the B_2 term is \frac{n^3}{3} + \frac{n^2}{2} + \frac{n}{6}. For n=100, the value is 338,350, and the approximation yields 338,350, matching exactly modulo O(1) constant terms from the lower .

Stirling's series connection

Stirling's series offers an for the n! as n \to \infty, n! \sim \sqrt{2\pi n} \left( \frac{n}{e} \right)^n \left( 1 + \frac{1}{12n} + \frac{1}{288n^2} - \frac{139}{51840 n^3} + \cdots \right), where the coefficients in the series are determined by the even-indexed Bernoulli numbers B_{2k}. Equivalently, the logarithmic form is \ln n! \sim n \ln n - n + \frac{1}{2} \ln (2\pi n) + \sum_{k=1}^\infty \frac{B_{2k}}{2k(2k-1) n^{2k-1}}, again featuring Bernoulli numbers in the correction terms. This expansion arises from approximating \ln n! = \sum_{k=1}^n \ln k via its \int_1^n \ln x \, dx = n \ln n - n + 1, with higher-order corrections supplied by the numbers through the underlying Euler-Maclaurin technique. The sum \sum \ln k serves as a non-polynomial analog to the power sums S_p(n) = \sum_{k=1}^n k^p, where both share the structure of an leading term plus -driven refinements; for the , \Gamma(z+1) \sim \sqrt{2\pi / z} (z/e)^z \exp\left( \sum_{k=1}^\infty \frac{B_{2k}}{2k(2k-1) z^{2k-1}} \right) as z \to \infty, linking the factorial directly to these asymptotic . Power sums connect to this framework through their own exact representation using Bernoulli polynomials, which generalize the Bernoulli numbers appearing in Stirling's series: S_p(n) = \frac{B_{p+1}(n+1) - B_{p+1}}{p+1}, for fixed positive integer p and integer n \geq 1, where B_m(x) denotes the m-th and B_m = B_m(0). For large n, this yields the asymptotic expansion S_p(n) \sim \frac{n^{p+1}}{p+1} + \frac{n^p}{2} + \sum_{j=2}^p \frac{B_j}{j} \binom{p}{j-1} n^{p+1-j}, with the Bernoulli numbers providing the subleading terms beyond the approximation \int_1^n x^p \, dx = \frac{n^{p+1} - 1}{p+1}. A concrete illustration occurs for p=1, where S_1(n) = \sum_{k=1}^n k = \frac{n(n+1)}{2} = \frac{n^2 + n}{2}, asymptotically \frac{n^2}{2} + \frac{n}{2} for large n, matching the integral \int_1^n x \, dx = \frac{n^2 - 1}{2} plus a linear correction from B_2( n+1 ) - B_2 = (n+1)^2 - (n+1) = n^2 + n , divided by 2 to give the exact sum. This mirrors the harmonic number approximation H_n = \sum_{k=1}^n \frac{1}{k} \approx \ln n + \gamma + \frac{1}{2n} - \sum_{k=1}^\infty \frac{B_{2k}}{2k n^{2k}}, where the \frac{1}{2n} endpoint correction parallels the \frac{n}{2} term in the power sum, both rooted in the same Bernoulli structure that refines Stirling's series for \ln n! \approx n H_n - \frac{n^2}{2} + \cdots.

Applications

In calculus and analysis

In calculus, sums of powers provide a discrete counterpart to continuous integration. The finite sum \sum_{k=1}^n k^p approximates the \int_1^n x^p \, dx = \frac{n^{p+1} - 1}{p+1}, particularly for large n, where the dominant term \frac{n^{p+1}}{p+1} captures the asymptotic growth. This analogy arises because the sum can be viewed as a partitioned accumulation akin to the area under the of x^p. The accuracy of this approximation is refined through error analysis via the Euler-Maclaurin formula, which expands the difference between the sum and the integral as \sum_{k=1}^n f(k) = \int_1^n f(x) \, dx + \frac{f(1) + f(n)}{2} + \sum_{m=1}^M \frac{B_{2m}}{(2m)!} \left( f^{(2m-1)}(n) - f^{(2m-1)}(1) \right) + R_M, where f(x) = x^p, B_{2m} are Bernoulli numbers, and R_M is the remainder term involving the (2M+1)-th . For f(x) = x^p with p a positive , the leading correction is \frac{1}{2} n^p, with subsequent terms of order n^{p-2}, n^{p-4}, and so on, until the expansion terminates for functions. This systematic error decomposition enables precise bounds and higher-order approximations in analytical contexts. Power sums further connect to integration through Riemann sums. Rewriting the sum as \sum_{k=1}^n k^p = n^{p+1} \sum_{k=1}^n \left( \frac{k}{n} \right)^p \frac{1}{n}, reveals it as a right Riemann sum for \int_0^1 x^p \, dx = \frac{1}{p+1} over [0, 1] partitioned into n subintervals of width $1/n. As n \to \infty, the Riemann sum converges to the integral, confirming the asymptotic relation \sum_{k=1}^n k^p \sim \frac{n^{p+1}}{p+1}. This formulation underscores how discrete power sums motivate the definition of the definite integral and facilitate the derivation of integration rules for power functions. For negative exponents, finite power sums \sum_{k=1}^n k^{-p} with p > 1 represent truncations of the infinite series defining the , \zeta(p) = \sum_{k=1}^\infty k^{-p}, which converges absolutely for \operatorname{Re}(p) > 1. The truncation error is bounded by the tail \sum_{k=n+1}^\infty k^{-p} \leq \int_n^\infty x^{-p} \, dx = \frac{n^{1-p}}{p-1}, allowing finite sums to approximate \zeta(p) with controlled precision in computational analysis. In numerical quadrature, expresses \sum_{k=1}^n k^p exactly as a of p+1 involving numbers, \sum_{k=1}^n k^p = \frac{1}{p+1} \sum_{j=0}^p (-1)^j \binom{p+1}{j} B_j n^{p+1-j}, enabling precise evaluation when such sums approximate integrals of polynomials. This exactness supports the verification of rules, such as ensuring they integrate monomials correctly up to the rule's degree of precision.

In discrete mathematics and number theory

In discrete mathematics, sums of powers play a key role in combinatorial identities, particularly in the enumeration of lattice points within polytopes via Ehrhart polynomials. The Ehrhart polynomial L_P(t) of an integral polytope P counts the number of lattice points in the t-dilate tP, and for lattice-face polytopes, its coefficients can be expressed using power sums P_k(x) = \sum_{i=1}^x i^k combined with Bernoulli polynomials to relate lattice point counts to volumes of projections of P. Specifically, the formula i(P, m) = \sum_{k=0}^d \mathrm{Vol}_k(\pi^{d-k}(P)) m^k links the Ehrhart polynomial to lower-dimensional volumes, where power sums appear in the explicit computation through the Euler-Maclaurin summation involving Bernoulli numbers. This connection facilitates combinatorial proofs and identities for counting problems, such as the number of integer solutions to inequalities defining the polytope. In number theory, explicit formulas exist for the sum of m-th powers modulo an odd prime p. If p-1 does not divide m, then \sum_{k=1}^{p-1} k^m \equiv 0 \pmod{p}; otherwise, \sum_{k=1}^{p-1} k^m \equiv -1 \pmod{p}. More refined enumerations consider the distribution of power sums over subsets: the function N(p, m, \alpha) counts subsets S \subseteq \{1, \dots, p-1\} such that \sum_{x \in S} x^m \equiv \alpha \pmod{p}, and under the condition that $2 or -2 is an m-th power residue modulo p, N(p, m, 0) = (2^{p-1} + p - 1)/p and N(p, m, \alpha) = (2^{p-1} - 1)/p for \alpha \not\equiv 0 \pmod{p}. These results extend to bounds like |N(p, m, \alpha) - p^{-1} 2^{p-1}| < \exp(c m p^{-1/2} \log p) for some constant c > 0. Connections to Wieferich primes arise in higher-order congruences for sums of powers modulo p^2; generalizations of Wieferich primes to higher orders involve conditions where a^{p-1} \equiv 1 \pmod{p^{k+1}} for base a and order k, impacting the valuation of power sums like \sum k^{p-1} modulo prime powers. Discrete calculus employs forward differences to analyze sums of powers, where the forward difference operator \Delta f(n) = f(n+1) - f(n) applied to the power sum S_p(n) = \sum_{k=1}^n k^p yields \Delta S_p(n) = (n+1)^p. This relation mirrors the , enabling the of powers as a : integrating k^p discretely recovers S_p(n) up to a constant, with higher-order differences \Delta^{p+1} S_p(n) = 0 confirming S_p(n) as a of degree p+1. Such tools underpin identities in finite differences, like Newton's divided difference for power sums. Power sums also feature in the generating functions for integer partitions within combinatorial contexts, particularly through the power sum symmetric functions p_\lambda = \prod p_i^{\lambda_i}, where p_i = \sum x_j^i over variables, forming a basis for the ring of symmetric functions indexed by partitions \lambda. In P-partition theory, weighted P-partitions generalize to quasisymmetric functions, where combinatorial power sums encode class values of characters and facilitate expansions of s for restricted partitions, such as those with bounded part sizes. For instance, the for partitions into distinct parts involves products over power sums, aiding under power constraints like maximum exponent in parts.

Generalizations

Sums over arithmetic progressions

The sum of the p-th powers of the first n terms of an arithmetic progression with first term a and common difference d is given by \sum_{k=0}^{n-1} (a + k d)^p. This generalizes the standard power sum S_p(n) = \sum_{k=1}^n k^p (with a=1, d=1) to arbitrary starting points and steps in the sequence. Expanding each term using the binomial theorem yields (a + k d)^p = \sum_{j=0}^p \binom{p}{j} a^{p-j} (k d)^j, so the full sum becomes \sum_{k=0}^{n-1} (a + k d)^p = \sum_{j=0}^p \binom{p}{j} a^{p-j} d^j \sum_{k=0}^{n-1} k^j. For j > 0, the inner sum \sum_{k=0}^{n-1} k^j = \sum_{k=1}^{n-1} k^j, which reduces to standard power sums up to degree p, while the j=0 term is simply n a^p. This approach expresses the arithmetic progression sum as a linear combination of the basic power sums S_j(n) for j = 0 to p. A concrete example arises for sums of even powers, where a = 2 and d = 2, giving \sum_{k=0}^{n-1} (2 + 2k)^p = 2^p \sum_{k=1}^{n} k^p = 2^p S_p(n). This scaling illustrates how the common difference directly factors into the standard power sum. For the linear case p=1, the sum simplifies to the arithmetic series formula \sum_{k=0}^{n-1} (a + k d) = \frac{n}{2} \left(2a + (n-1)d\right), which follows directly from pairing terms or the with p=1.

Multidimensional and infinite sums

Multidimensional power sums extend the concept of one-dimensional sums to higher dimensions, considering grids in \mathbb{R}^d. A typical form is the sum over integer coordinates k = (k_1, \dots, k_d) with $1 \leq k_i \leq n_i for each i, of the p-th power of the components: \sum_{k_1=1}^{n_1} \cdots \sum_{k_d=1}^{n_d} (k_1^p + \cdots + k_d^p). This expression is separable due to the linearity of the and independence of the variables; it factors into a sum over each dimension separately. Specifically, \sum_{i=1}^d \left( \prod_{j \neq i} n_j \right) \sum_{k_i=1}^{n_i} k_i^p, where each inner sum is a standard one-dimensional power sum. For equal limits n_i = n in all dimensions, this simplifies to d n^{d-1} \sum_{k=1}^n k^p. For example, in two dimensions with p=2 and equal limits n, the sum is $2n \sum_{k=1}^n k^2 \approx 2n \cdot \frac{n^3}{3} = \frac{2n^4}{3}, which approximates the double integral \iint_{[0,n]^2} (x^2 + y^2) \, dx \, dy = \frac{2n^4}{3} via the Euler-Maclaurin formula, relating discrete lattice sums to continuous volumes or areas in the large-n limit. Infinite power sums \sum_{k=1}^\infty k^p diverge for p \geq -1, as the terms do not decay sufficiently fast (e.g., the series for p=-1). For p < -1, the series converges absolutely, equaling the Riemann function \zeta(-p) where \operatorname{Re}(-p) > 1. Through , \zeta(s) extends to negative arguments, yielding finite values even where the series diverges formally. For negative powers p = -m with m a positive , \sum_{k=1}^\infty k^{-m} = \zeta(m), and explicit values at negative s are given by \zeta(-n) = -\frac{B_{n+1}}{n+1} for nonnegative s n, where B_k are numbers (e.g., \zeta(-1) = -1/12). In physics, infinite power sums via the zeta function appear in Bose-Einstein for ideal quantum gases. The particle number in excited states is N_e = g \frac{V}{\lambda^3} \zeta(3/2) in three dimensions, where g is the spin degeneracy, V the volume, \lambda the thermal wavelength, and \zeta(3/2) \approx 2.612 determines the critical temperature for .

References

  1. [1]
    Power Sum -- from Wolfram MathWorld
    S_p(n)=sum_(k=1)^nk^p. (2). General power sums arise commonly in statistics. For example, k-statistics are most commonly defined in terms of ...
  2. [2]
    [PDF] Sums of numerical powers in discrete mathematics: Archimedes ...
    + n: They are called 'discrete'sums because the items being added together have individual, separated, effect on the result, unlike what happens.
  3. [3]
    Faulhaber's Formula -- from Wolfram MathWorld
    In a rare 1631 work entitled Academiae Algebrae, J. Faulhaber published a number of formulae for power sums of the first n positive integers.
  4. [4]
    Bernoulli Number -- from Wolfram MathWorld
    These numbers arise in the series expansions of trigonometric functions, and are extremely important in number theory and analysis.
  5. [5]
    [PDF] sums of powers by l'hopital's rule - Williams College
    Aug 5, 2024 · Abstract. For a positive integer d, let pd(n) := 0d + 1d + 2d + ··· + nd; i.e., pd(n) is the sum of the first dth-powers up to n.
  6. [6]
    [PDF] generatingfunctionology - Penn Math
    May 21, 1992 · This book is about generating functions and some of their uses in discrete mathematics. The subject is so vast that I have not attempted to give ...
  7. [7]
    Sums of Powers of Positive Integers - Aryabhata (b. 476), northern ...
    Sums of Powers of Positive Integers - Aryabhata (b. 476), northern India · =(n(n+1)2)2. · =(1+2+3)2, · =(1+2+3+⋯+n)2 · =(n(n+1)2)2. · =n(n+1)(2n+1)6 · =n(n+1)(2n+1)6.
  8. [8]
    Sums of Powers of Positive Integers - Abu Bakr al-Karaji (d. 1019 ...
    The mathematician Abu Bakr Al-Karaji worked in Baghdad in what is now Iraq for at least part, if not all, of his life. He wrote books on engineering and ...Missing: Tartaglia Billy
  9. [9]
    Sums of Powers of Positive Integers - Introduction
    In this article, we first recount some of the early history of formulas for sums of integer powers. ... Sums of Powers of Positive Integers - Aryabhata (b. 476), ...Missing: Tartaglia Billy
  10. [10]
    [PDF] Elementary Number Theory in Nine Chapters, Second Edition
    Elementary Number Theory in Nine Chapters is primarily intended for a one-semester course for upper-level students of mathematics, in particular,.
  11. [11]
    Sums of Powers of Positive Integers - Johann Faulhaber (1580-1635 ...
    In his 1631 Academia Algebrae, Faulhaber presented formulas for sums of powers of the first n positive integers from the 13 th to the 17 th powers.
  12. [12]
    Sums of Powers of Positive Integers - Jakob Bernoulli (1654-1705 ...
    Bernoulli derived symbolic formulas for the sums of positive integer powers using the method conjectured above for Fermat, then noted a pattern that would make ...
  13. [13]
    [PDF] Sums of numerical powers in discrete mathematics: Archimedes ...
    You will carry out and contrast various ways of proving statements, from using geometry to algebra to proof by generalizable example or by mathematical ...
  14. [14]
    Bernoulli numbers and generating functions - TracingCurves
    Nov 13, 2023 · According to [1], in the early 1730's Euler posed the following definition of the Bernoulli numbers as the coefficients of the exponential ...
  15. [15]
    A remark on an explicit formula for the sums of powers of integers
    Mar 5, 2025 · In this short note, we show that Samsonadze's formula corresponds to a well-known formula for S_k(n) involving the Stirling numbers of the second kind.
  16. [16]
    DLMF: §24.2 Definitions and Generating Functions ‣ Properties ...
    Bernoulli numbers, definition, generating function; Referenced by: §2.10(i) ... Euler numbers, Euler polynomials, definition, generating function; Notes ...
  17. [17]
    DLMF: §24.4 Basic Properties ‣ Properties ‣ Chapter 24 Bernoulli ...
    §24.4(i) Difference Equations, §24.4(ii) Symmetry, §24.4(iii) Sums of Powers, §24.4(iv) Finite Expansions, §24.4(v) Multiplication Formulas, Raabe'Missing: recursive | Show results with:recursive
  18. [18]
    Sum of First n Natural Numbers - UCSD CSE
    We prove the formula 1+ 2+ ... + n = n(n+1) / 2, for n a natural number. There is a simple applet showing the essence of the ...
  19. [19]
    Gauss's Day of Reckoning | American Scientist
    There is yet another way of thinking about the summation process: n(n + 1)/2 has been known since antiquity as the formula for the triangular numbers, those in ...
  20. [20]
    [PDF] Figurate Numbers: A Historical Survey of an Ancient Mathematics
    Figurate numbers comprise one of the oldest areas of mathematics, dating back to the Pythagoreans of the 6𝑡𝑡ℎ century BCE and capturing the attention of ...
  21. [21]
    Series Sums and Gauss's Formula | CK-12 Foundation
    The story is likely apocryphal (a legend) ... When Gauss was asked to add up the first 100 integers, he found the sum very quickly, by pairing the numbers:.
  22. [22]
    [PDF] Mathematical Induction - Stanford University
    Proof: By induction. Let P(n) be “the sum of the first n positive natural numbers is n(n + 1) / 2.” We show that P(n) is true for all n ∈ ℕ.
  23. [23]
    Sum of n, n², or n³ | Brilliant Math & Science Wiki
    This technique generalizes to a computation of any particular power sum one might wish to compute. Sum of the Squares of the First n n n Positive Integers.Missing: notation | Show results with:notation
  24. [24]
    Sum of Squares of n Natural Numbers - Formula, Even and Odd, Proof
    The sum of squares of n natural numbers can be calculated using the formula [n(n+1)(2n+1)] / 6. The sum of squares of n odd and n even numbers are ...
  25. [25]
    Discrete Uniform Distributions - Milefoot
    The sum of the first N integers is given by N∑x=1x=N(N+1)2, and the sum of the first N squares is given by N∑x=1x2=N(N+1)(2N+1)6. Now the expected value formula ...
  26. [26]
    Nicomachus's Theorem -- from Wolfram MathWorld
    Nicomachus noted that when the odd numbers are grouped in blocks of length, 1, 2, 3, ..., i.e., 1; 3+5; 7+9+11; 13+15+17+19; ..., the n th cubic number n^3 ...
  27. [27]
    Nicomachus' theorem - PlanetMath.org
    Mar 22, 2013 · Nicomachus' theorem. Theorem (Nicomachus). The sum of the cubes of the first n n integers is equal to the square of the n n th triangular ...
  28. [28]
    Sums of powers of integers – how Fermat may have found them - jstor
    comparable expressions for the sum of cubes, the sum of fourth powers, the ... + n3 from Bachet [6]. But this, Ferm?t says, is where Bachet stopped. Ferm?t ...
  29. [29]
    Graham, Knuth, and Patashnik: Concrete Mathematics - CS Stanford
    Based on the course Concrete Mathematics taught by Knuth at Stanford University from 1970--1989. The second edition includes important new material about the ...
  30. [30]
    Sums of powers of integers and generalized Stirling numbers of the ...
    Nov 21, 2022 · We derive a couple of infinite families of explicit formulas for S_k(n). One of the families involves the r-Stirling numbers of the second kind.
  31. [31]
    [PDF] Bernoulli numbers and the Euler-Maclaurin summation formula
    In many applications, the Euler-Maclaurin summation formula provides an asymptotic expansion, in which case the divergent nature of the series in eq. (14) ...
  32. [32]
    The Euler-Maclaurin formula, Bernoulli numbers, the zeta function ...
    Apr 10, 2010 · Here is interesting result which one can use to come up with Euler -Maclaurin formula.I found out that 1/(1+X) is convergent for X>1 if we ...
  33. [33]
    5.11 Asymptotic Expansions ‣ Properties ‣ Chapter 5 Gamma ...
    The scaled gamma function Γ ∗ ⁡ ( z ) is defined in (5.11.3) and its main property is Γ ∗ ⁡ ( z ) ∼ 1 as z → ∞ in the sector | ph ⁡ z | ≤ π − δ.Missing: Faulhaber | Show results with:Faulhaber
  34. [34]
    [PDF] The Bernoulli numbers, power sums, and zeta values
    XVIII of André Weil's Number Theory: An Approach. Through History for more on Euler and the zeta function.
  35. [35]
    [PDF] Euler-Maclaurin Formula 1 Introduction - People | MIT CSAIL
    Euler-Maclaurin summation formula is an important tool of numerical analysis. Simply put, it gives us an estimation of the sum P n i=0 f(i) through the ...
  36. [36]
  37. [37]
    [PDF] Contributions to the Theory of Ehrhart Polynomials - DSpace@MIT
    ... Ehrhart polynomial of an integral cyclic polytope, we define a more general ... power sums, so we will first discuss some properties of power sums and ...
  38. [38]
    Sum of powers mod p - Math Stack Exchange
    Jun 10, 2018 · It's a standard sort of trick for showing that various sums vanish (modp). The point is that multiplying by g permutes the non-zero residues ...
  39. [39]
    [PDF] Enumeration of power sums modulo a prime - MIT Mathematics
    We consider, for odd primes p, the function N(p, m, e) which equals the number of subsets SC{1 ,...,p - 1) with the property that &sxm = OL (modp).
  40. [40]
    Bits of 3n in binary, Wieferich primes and a conjecture of Erdős
    We prove our results by proving a nonexistence theorem for “higher Wieferich primes”. ... In [Erd79] Erdős conjectured that there are only finitely many powers of ...
  41. [41]
    [PDF] Discrete Calculus
    We define the discrete derivative of a function f(n), denoted ∆nf(n), to be f(n + 1) − f(n). This operator has some interesting properties. 2.1 Properties. • ...Missing: ΔS_p( | Show results with:ΔS_p(
  42. [42]
    [PDF] Discrete Calculus and its Applications
    Mar 9, 2014 · It turns out that the forward difference is very useful in computing sums. Theorem 3.3 (The Fundamental Theorem of Discrete Calculus). For ...<|control11|><|separator|>
  43. [43]
    [2112.06457] P-partition power sums - arXiv
    Dec 13, 2021 · We develop the theory of weighted P-partitions, which generalises the theory of P-partitions from labelled posets to weighted labelled posets.
  44. [44]
    [PDF] European Journal of Combinatorics P-partition power sums✩
    Power sum symmetric functions are of especial interest because they encode the class values of the characters of the symmetric group under the Frobenius.
  45. [45]
    [PDF] On Sums of Powers of Arithmetic Progressions, and Generalized ...
    Jul 13, 2017 · [9] R. L. Graham, D. E. Knuth and O. Patashnik, Concrete Mathematics, 2nd edition, 1994, Addison-Weseley,. Reading, Massachusetts, 1991. [10] ...