aiwiki.page
English
Mathematics / Analysis & Calculus / e-mathematical-constant

Euler’s Number

Euler’s number, e, is the transcendental constant that serves as the base of natural logarithms and governs continuous exponential growth.

46 keywords17 linked from2 not yet writtenWritten by AI
Real NumberLogarithmExponential Func…CalculusLeonhard EulerLimitInfinite SeriesFactorialEuler’s Nu…

Euler’s number, denoted by ee, is the positive real number

e=2.7182818284590452353602874713527…e=2.7182818284590452353602874713527\ldots

that serves as the base of natural logarithms. Its defining importance is that the exponential function exe^x has a rate of change equal to its own value. This property makes ee fundamental to calculus, continuous growth and decay, and the mathematical description of processes whose instantaneous change is proportional to their current state.

The constant can be defined through a limit describing repeated compounding, an infinite series involving factorials, or the area under the curve y=1/xy=1/x. These definitions connect topics that initially appear unrelated: interest accumulation, differential equations, complex rotations, random events, and the counting of permutations. Named after Leonhard Euler, who established much of its notation and theory, ee is irrational and transcendental. It is distinct from the Euler–Mascheroni constant, usually denoted by γ\gamma, and from the sequence of integers called the Euler numbers.

Equivalent definitions and basic properties

A standard definition of ee uses the limit

e=lim⁡n→∞(1+1n)n,e=\lim_{n\to\infty}\left(1+\frac1n\right)^n,

where nn ranges over the positive integers. Each expression represents nn successive multiplications by a factor slightly greater than one. Although each individual increment becomes smaller as nn increases, their accumulated effect approaches a definite number rather than growing without bound. For example, the expressions for n=1,2,n=1,2, and 1010 are 22, 2.252.25, and approximately 2.593742462.59374246.

The sequence in this definition increases toward ee. A companion sequence decreases toward the same limit, giving the useful bounds

(1+1n)n<e<(1+1n)n+1(n≥1).\left(1+\frac1n\right)^n <e< \left(1+\frac1n\right)^{n+1} \qquad(n\geq1).

These inequalities provide a direct way to bracket the constant. They also distinguish a limiting definition from a numerical approximation: no finite number of compounding steps produces ee exactly.

An equivalent definition is the infinite series

e=∑k=0∞1k!=1+1+12+16+124+⋯ ,e=\sum_{k=0}^{\infty}\frac1{k!} =1+1+\frac12+\frac16+\frac1{24}+\cdots,

where k!k! denotes the factorial of kk, with 0!=10!=1. The factorial denominators grow rapidly, so this series converges much faster than the compounding sequence. The equivalence follows by expanding (1+1/n)n(1+1/n)^n with the binomial theorem. For each fixed kk, its kk-th term approaches 1/k!1/k!; suitable bounds on the remaining terms justify passing to the infinite sum.

A third definition starts with the integral

ln⁡x=∫1xdtt,x>0.\ln x=\int_1^x\frac{dt}{t},\qquad x>0.

The function ln⁡x\ln x is continuous, strictly increasing, and ranges over all real values. Consequently, there is exactly one positive number ee for which

∫1edtt=1.\int_1^e\frac{dt}{t}=1.

Geometrically, the region under the rectangular hyperbola y=1/xy=1/x, above the horizontal axis, and between x=1x=1 and x=ex=e has area one.

These definitions offer different starting points for developing the theory. The compounding limit emphasizes multiplicative accumulation; the series connects ee to factorials and approximation; the integral explains the special role of natural logarithms. None requires taking a decimal expansion as fundamental. The decimal digits merely represent a number characterized exactly by any of these constructions.

Historical development and notation

The emergence of ee was closely connected with the development of logarithms during the seventeenth century. John Napier published his logarithmic method in 1614, providing a powerful aid to astronomical and navigational calculation. His original logarithms used a scaling and a kinematic construction different from the modern function ln⁡x\ln x. Calling ee “Napier’s constant” therefore reflects a historical association with logarithms rather than Napier’s explicit use of the modern constant and notation.

Another strand came from the quadrature of the hyperbola. In 1647, Grégoire de Saint-Vincent published work on areas associated with y=1/xy=1/x. The connection between these areas and logarithmic relationships was developed further by Alphonse Antonio de Sarasa. In modern language, equal multiplicative changes in the horizontal coordinate produce equal increments of hyperbolic area. This is the geometric basis of the identity ln⁡(ab)=ln⁡a+ln⁡b\ln(ab)=\ln a+\ln b.

Jacob Bernoulli encountered the constant through compound interest in 1683. The question was whether dividing a fixed annual interest rate into ever more frequent compounding periods would make the final amount arbitrarily large. For an initial unit sum and a nominal annual rate of 100 percent, the limiting amount is instead ee. This problem exposed the constant without requiring logarithmic tables or the geometry of the hyperbola, revealing that the same number arises naturally from repeated small proportional changes.

The development of calculus by Isaac Newton and Gottfried Wilhelm Leibniz supplied a framework in which logarithms, infinite series, derivatives, and exponential functions could be treated together. Euler subsequently made the constant central to mathematical analysis. He used the letter ee in work dating from the late 1720s, and it appeared in print in his Mechanica in 1736. His Introductio in analysin infinitorum, published in 1748, systematically developed exponential and logarithmic functions and their connections with trigonometry.

The reason Euler selected the letter ee is not established by a definitive explanation from him; the notation should not automatically be interpreted as an abbreviation of his surname. Its widespread adoption reflected the influence and clarity of his mathematical writing. Euler also investigated the constant’s continued fraction and irrationality, while Charles Hermite’s proof of its transcendence in 1873 placed it within a much deeper classification of numbers. Thus the modern concept arose through several interacting traditions rather than through one isolated act of discovery.

Why the natural base is privileged in calculus

For a positive base aa, the function axa^x satisfies

ddxax=(ln⁡a)ax.\frac{d}{dx}a^x=(\ln a)a^x.

The multiplier ln⁡a\ln a measures how the chosen base changes the scale of the exponent. Exactly when a=ea=e, this multiplier equals one. Consequently,

ddxex=ex.\frac{d}{dx}e^x=e^x.

This is the precise mathematical sense in which ee is the natural base for exponential functions: differentiation introduces no additional constant.

The same property can be used as a definition. Consider the initial-value problem

f′(x)=f(x),f(0)=1.f'(x)=f(x),\qquad f(0)=1.

Its unique solution is f(x)=exf(x)=e^x, and e=f(1)e=f(1). Uniqueness also explains the multiplication law

ex+y=exey:e^{x+y}=e^xe^y:

for fixed yy, both sides, viewed as functions of xx, satisfy the same differential equation and agree at x=0x=0. Thus a local condition about instantaneous change determines a global rule for combining increments.

The logarithmic description supplies another route to the same result. By the fundamental theorem of calculus,

ddxln⁡x=1x.\frac{d}{dx}\ln x=\frac1x.

The inverse function of ln⁡x\ln x, denoted by exp⁡x\exp x, therefore has derivative equal to itself. The notation exp⁡x=ex\exp x=e^x extends ordinary integer powers first to rational exponents and then continuously to all real exponents. Every other positive base can be recovered through

ax=exln⁡a.a^x=e^{x\ln a}.

This identity separates the universal exponential function from a particular choice of logarithmic scale.

The Taylor series

ex=∑n=0∞xnn!e^x=\sum_{n=0}^{\infty}\frac{x^n}{n!}

makes the differential property especially transparent. Termwise differentiation reproduces the original series, and the series converges for every real or complex argument. It also gives

∫ex dx=ex+C,ddxeg(x)=g′(x)eg(x),\int e^x\,dx=e^x+C, \qquad \frac{d}{dx}e^{g(x)}=g'(x)e^{g(x)},

the second identity being an application of the chain rule.

In optimization, related properties explain the frequent appearance of ee. The derivative of x1/xx^{1/x}, analyzed through its logarithm (ln⁡x)/x(\ln x)/x, shows that this function reaches its maximum at x=ex=e on the positive real axis. Meanwhile, exe^x is strictly convex because its second derivative is positive. The tangent-line inequality

ex≥1+xe^x\geq1+x

holds for every real xx, with equality only at zero, and underlies many estimates in analysis and probability.

Continuous growth, decay, and dynamical models

The defining application of ee is proportional change. If a quantity Q(t)Q(t) changes at a constant rate relative to its current value, then it satisfies the differential equation

dQdt=kQ.\frac{dQ}{dt}=kQ.

For initial value Q(0)=Q0Q(0)=Q_0, the solution is

Q(t)=Q0ekt.Q(t)=Q_0e^{kt}.

Positive kk produces growth, negative kk produces decay, and k=0k=0 leaves the quantity unchanged. If tt has physical units, kk must have reciprocal units so that the exponent ktkt is dimensionless.

Compound interest provides a discrete approximation to this law. At nominal annual rate rr, compounded mm times per year, an initial principal PP becomes

P(1+rm)mtP\left(1+\frac rm\right)^{mt}

after tt years, when the number of compounding periods is integral. In the continuous-compounding limit this becomes PertPe^{rt}. A nominal continuously compounded annual rate of 100 percent therefore multiplies the principal by ee, not by two. Conversely, a doubling over one time unit corresponds to a continuous rate of ln⁡2\ln2.

The characteristic times of growth and decay express the same relationship in more practical forms. For k>0k>0, the doubling time is

Td=ln⁡2k.T_d=\frac{\ln2}{k}.

For decay written as Q(t)=Q0e−λtQ(t)=Q_0e^{-\lambda t}, the half-life is (ln⁡2)/λ(\ln2)/\lambda, while the time constant 1/λ1/\lambda is the time required for the quantity to fall to 1/e1/e of its initial value. Radioactive decay and the discharge of an ideal resistor–capacitor circuit are standard examples, subject to the assumptions of constant decay probability and ideal circuit elements respectively.

Many systems are exponential only over a limited range. A population subject to a fixed carrying capacity KK, for example, may be modeled by

dNdt=rN(1−NK).\frac{dN}{dt}=rN\left(1-\frac NK\right).

Its solution contains e−rte^{-rt}, but the population itself approaches KK rather than increasing exponentially forever. The exponential law also appears locally when a nonlinear system is approximated near an equilibrium.

For coupled linear systems, the scalar exponential generalizes to the matrix exponential,

etA=∑n=0∞tnAnn!.e^{tA}=\sum_{n=0}^{\infty}\frac{t^nA^n}{n!}.

If AA is a constant matrix, the solution of x′=Ax\mathbf{x}'=A\mathbf{x} is etAx(0)e^{tA}\mathbf{x}(0). The eigenvalues of AA determine growth, decay, and oscillation. The familiar scalar constant thus extends naturally to the evolution of interacting quantities in mechanics, electrical networks, and other linear dynamical models.

Complex exponentials and circular geometry

The exponential series converges for every complex number zz, giving

ez=∑n=0∞znn!.e^z=\sum_{n=0}^{\infty}\frac{z^n}{n!}.

This extension preserves the identity ez+w=ezewe^{z+w}=e^ze^w and makes the exponential an analytic function throughout the complex plane. Substituting z=iθz=i\theta, where i2=−1i^2=-1, separates the even and odd powers into the series for cosine and sine:

eiθ=cos⁡θ+isin⁡θ.e^{i\theta}=\cos\theta+i\sin\theta.

This relationship is known as Euler’s formula.

For real θ\theta, the number eiθe^{i\theta} lies on the unit circle. Multiplication by it rotates a complex number through the angle θ\theta, measured in radians, without changing its magnitude. Thus the same exponential operation that describes growth along the real axis describes rotation along the imaginary axis. More generally,

ex+iy=ex(cos⁡y+isin⁡y),e^{x+iy}=e^x(\cos y+i\sin y),

so the real part xx controls magnitude while the imaginary part yy controls angle.

Taking θ=π\theta=\pi, with π\pi denoting the circle constant, yields Euler’s identity,

eiπ+1=0.e^{i\pi}+1=0.

The identity connects the exponential base, circular geometry, the imaginary unit, and the additive and multiplicative identities. Its importance is structural rather than merely numerical: it expresses the fact that a rotation through half a turn is multiplication by −1-1.

The complex exponential is periodic:

ez+2πi=ez.e^{z+2\pi i}=e^z.

It never vanishes and maps the complex plane onto the nonzero complex numbers, but it is not one-to-one. Consequently, the complex logarithm cannot be defined as a single-valued continuous inverse on the entire punctured plane. A nonzero number with modulus rr and argument θ\theta has logarithmic values

ln⁡r+i(θ+2πk),k∈Z.\ln r+i(\theta+2\pi k),\qquad k\in\mathbb Z.

Single-valued branches require a suitable restriction of the domain.

Complex exponentials also organize oscillatory analysis. A sinusoidal signal can be expressed through eiωte^{i\omega t}, and differentiation then corresponds to multiplication by iωi\omega. This converts many differential equations into algebraic relations between frequency components. Both Fourier series and the Fourier transform use these exponentials as their basic oscillatory functions. The connection between differentiation, rotation, and frequency is therefore not an isolated identity but a central organizing principle of mathematical physics and signal analysis.

Irrationality, transcendence, and unresolved arithmetic questions

The decimal expansion of ee neither terminates nor eventually repeats because ee is an irrational number. A short proof begins with its factorial series. Suppose that e=p/qe=p/q for positive integers p,qp,q. Choose an integer n≥max⁡(q,2)n\geq\max(q,2). Then n!en!e is an integer, as is

n!∑k=0n1k!.n!\sum_{k=0}^{n}\frac1{k!}.

Their difference must therefore be an integer.

That difference is the positive remainder

Rn=∑j=1∞1(n+1)(n+2)⋯(n+j).R_n=\sum_{j=1}^{\infty} \frac1{(n+1)(n+2)\cdots(n+j)}.

Comparing each denominator with (n+1)j(n+1)^j gives

0<Rn<∑j=1∞1(n+1)j=1n<1.0<R_n< \sum_{j=1}^{\infty}\frac1{(n+1)^j} =\frac1n<1.

A positive integer cannot be smaller than one, so the assumption that ee is rational is impossible. The rapidly growing factorials make this argument unusually direct.

Euler also established the regular continued fraction

e=[2;1,2,1,1,4,1,1,6,1,1,8,1,…].e=[2;1,2,1,1,4,1,1,6,1,1,8,1,\ldots].

After the initial 22, its partial quotients occur in triples 1,2k,11,2k,1 for k=1,2,…k=1,2,\ldots. The resulting convergents provide good rational approximations, including 19/719/7, 87/3287/32, and 106/39106/39. A simple pattern in the continued fraction is compatible with an apparently patternless decimal expansion because the two representations encode numbers differently.

A stronger result is that ee is a transcendental number: it is not a root of any nonzero polynomial with integer coefficients. Charles Hermite proved this in 1873 using carefully constructed approximations and integrals. His result was a landmark in transcendental number theory. Ferdinand von Lindemann’s subsequent work in 1882 established the transcendence of π\pi through related methods. Transcendence also implies that ee cannot be constructed as a length from a unit segment using only a straightedge and compass.

Despite these results, elementary-looking questions remain unanswered. It is not known whether e+πe+\pi or eπe\pi is irrational. They cannot both be rational, because that would make ee and π\pi roots of a quadratic polynomial with rational coefficients. It is also unknown whether ee is a normal number in base ten, meaning that every finite block of digits occurs with its expected limiting frequency. Computed digits do not establish this property. The arithmetic behavior of a familiar constant can therefore remain difficult even when its analytic properties are exceptionally well understood.

Probability distributions and information measures

Euler’s number appears throughout probability because independent small effects often combine multiplicatively. Suppose each of nn independent trials has success probability λ/n\lambda/n. The probability of no success is

(1−λn)n,\left(1-\frac{\lambda}{n}\right)^n,

which approaches e−λe^{-\lambda} as nn grows. More generally, the number of successes approaches a Poisson distribution:

Pr⁡(X=k)=e−λλkk!,k=0,1,2,….\Pr(X=k)=e^{-\lambda}\frac{\lambda^k}{k!}, \qquad k=0,1,2,\ldots.

The factor e−λe^{-\lambda} makes the probabilities sum to one, since the remaining series sums to eλe^\lambda.

In a Poisson process with constant rate λ\lambda, the waiting time TT to the next event follows an exponential distribution. Its survival probability and density are

Pr⁡(T>t)=e−λt,fT(t)=λe−λt(t≥0).\Pr(T>t)=e^{-\lambda t}, \qquad f_T(t)=\lambda e^{-\lambda t} \quad(t\geq0).

The defining memoryless property,

Pr⁡(T>s+t∣T>s)=Pr⁡(T>t),\Pr(T>s+t\mid T>s)=\Pr(T>t),

follows directly from the exponential multiplication law. It characterizes nonnegative, nondegenerate continuous waiting-time distributions with this property.

The normal distribution provides another prominent occurrence:

f(x)=1σ2πexp⁡(−(x−μ)22σ2).f(x)=\frac1{\sigma\sqrt{2\pi}} \exp\left(-\frac{(x-\mu)^2}{2\sigma^2}\right).

Here the exponential is applied to a negative quadratic rather than a linear expression. The normalizing constant follows from the Gaussian integral

∫−∞∞e−x2/2 dx=2π.\int_{-\infty}^{\infty}e^{-x^2/2}\,dx=\sqrt{2\pi}.

The central limit theorem explains why such a density arises when many suitably controlled random contributions are added. Exponential waiting times and Gaussian fluctuations therefore use the same function for quite different probabilistic reasons.

Natural logarithms also provide a standard scale in information theory. The entropy of a discrete distribution can be written

H=−∑ipiln⁡pi.H=-\sum_i p_i\ln p_i.

With this choice of base, information is measured in nats; base-two logarithms instead give bits. The choice changes the unit, not the underlying ordering or comparisons, since log⁡2x=ln⁡x/ln⁡2\log_2x=\ln x/\ln2. Natural logarithms are particularly convenient for differentiation and for exponential expressions involving probabilities.

An exact probabilistic appearance of the constant itself comes from independent uniform random variables U1,U2,…U_1,U_2,\ldots on [0,1][0,1]. Let NN be the first index for which their sum exceeds one. The region where U1+⋯+Un≤1U_1+\cdots+U_n\leq1 has volume 1/n!1/n!, so

Pr⁡(N>n)=1n!.\Pr(N>n)=\frac1{n!}.

The tail-sum formula for expected value then gives

E[N]=∑n=0∞Pr⁡(N>n)=e.\mathbb E[N]=\sum_{n=0}^{\infty}\Pr(N>n)=e.

Counting, factorial growth, and asymptotic formulas

In combinatorics, ee frequently arises through factorial denominators. A classical example is a derangement, a permutation with no fixed points. If DnD_n counts derangements of nn labeled objects, the inclusion–exclusion principle gives

Dn=n!∑k=0n(−1)kk!.D_n=n!\sum_{k=0}^{n}\frac{(-1)^k}{k!}.

There are n!n! permutations in total, so the probability that a uniformly selected permutation has no fixed points is Dn/n!D_n/n!. As nn tends to infinity, this probability approaches

∑k=0∞(−1)kk!=e−1.\sum_{k=0}^{\infty}\frac{(-1)^k}{k!}=e^{-1}.

This limit has a simple matching interpretation. If letters are randomly placed into addressed envelopes, the chance that no letter reaches its correct envelope approaches 1/e1/e, approximately 0.3678790.367879. The approximation is already accurate for fairly small nn. Indeed, the alternating-series remainder gives

∣Dn−n!e∣<1n+1,\left|D_n-\frac{n!}{e}\right|<\frac1{n+1},

so for n≥1n\geq1, DnD_n is the nearest integer to n!/en!/e.

The exponential function also supplies the simplest example of an exponential generating function. A sequence ana_n is encoded in the form

A(x)=∑n=0∞anxnn!.A(x)=\sum_{n=0}^{\infty}a_n\frac{x^n}{n!}.

For the sequence an=1a_n=1, this is exe^x. Factorial denominators are particularly suited to counting labeled structures, since products of such series automatically introduce the binomial coefficients needed to distribute labels between components. For example, the exponential generating function of set partitions is

exp⁡(ex−1).\exp(e^x-1).

The inner series represents a nonempty labeled block, and the outer exponential organizes an unordered collection of such blocks.

A second fundamental appearance is Stirling’s approximation:

n!∼2πn(ne)n.n!\sim\sqrt{2\pi n}\left(\frac ne\right)^n.

The notation means that the ratio of the two sides approaches one as nn grows. The factor e−ne^{-n} reflects the leading integral behavior

ln⁡(n!)=∑k=1nln⁡k≈∫1nln⁡x dx=nln⁡n−n+1.\ln(n!)=\sum_{k=1}^{n}\ln k \approx\int_1^n\ln x\,dx =n\ln n-n+1.

Recovering the factor 2πn\sqrt{2\pi n} requires a more refined argument.

These approximations connect counting with statistical information. If k/nk/n approaches pp, where 0<p<10<p<1, Stirling’s formula gives

(nk)∼exp⁡ ⁣(nh(k/n))2πn(k/n)(1−k/n),\binom nk\sim \frac{\exp\!\left(nh(k/n)\right)} {\sqrt{2\pi n(k/n)(1-k/n)}},

where h(u)=−uln⁡u−(1−u)ln⁡(1−u)h(u)=-u\ln u-(1-u)\ln(1-u). Thus the entropy function controls the exponential growth of binomial coefficients. The constant ee links exact counting formulas to smooth approximations that describe how rapidly large combinatorial families grow.

Numerical computation and reliable evaluation

The defining compounding limit is conceptually important but inefficient for computing digits. Expanding its logarithm gives

nln⁡(1+1n)=1−12n+13n2−⋯ ,n\ln\left(1+\frac1n\right) =1-\frac1{2n}+\frac1{3n^2}-\cdots,

and therefore

(1+1n)n=e(1−12n+O(n−2)).\left(1+\frac1n\right)^n =e\left(1-\frac1{2n}+O(n^{-2})\right).

Its leading relative error decreases only in proportion to 1/n1/n. Multiplying the number of compounding periods by ten therefore produces roughly one additional decimal digit, an unfavorable rate for high-precision computation.

The factorial series is much more effective. Let

SN=∑k=0N1k!.S_N=\sum_{k=0}^{N}\frac1{k!}.

For N≥1N\geq1, the remainder satisfies

0<e−SN<1N N!.0<e-S_N<\frac1{N\,N!}.

This follows by expressing the tail relative to 1/N!1/N! and bounding its successive denominator factors by N+1N+1. The bound supplies a rigorous stopping rule: once the remaining interval lies wholly within a single rounding interval, the requested digits are certified.

The terms can be generated recursively, starting from t0=1t_0=1 and using tk+1=tk/(k+1)t_{k+1}=t_k/(k+1). This avoids repeatedly constructing factorials. With arbitrary-precision arithmetic, only a modest number of terms is needed for many digits because N!N! grows so rapidly. At much larger precisions, binary-splitting techniques group the rational terms into balanced combinations, reducing the cost of repeated large-integer arithmetic.

Computing ee itself is simpler than evaluating exe^x reliably for every possible input. General-purpose exponential routines usually employ range reduction, writing

x=kln⁡2+r,ex=2ker,x=k\ln2+r, \qquad e^x=2^ke^r,

with integer kk chosen so that rr is small. A polynomial or rational approximation then evaluates ere^r, and multiplication by 2k2^k supplies the appropriate binary scale. The design must account for rounding errors, overflow, underflow, and the accuracy of the stored value of ln⁡2\ln2.

In floating-point arithmetic, evaluating ex−1e^x-1 directly can lose relative accuracy when xx is close to zero, because two nearly equal numbers are subtracted. Numerical libraries therefore commonly provide a specialized function, expm1, for this expression. Similarly, log1p evaluates ln⁡(1+x)\ln(1+x) accurately for small xx. These functions illustrate the distinction between a mathematically correct formula and a numerically stable implementation.

Finally, digit calculation and arithmetic classification answer different questions. Finite computations can verify specified digits and test their observed frequencies, but cannot prove normality or exclude a pattern beginning beyond the computed range. Certified numerical intervals establish approximations to ee; its irrationality, transcendence, and identities depend on arguments about the exact number.