Интегральная ошибка формула

Error function
Plot of the error function
General information
General definition
Fields of application	Probability, thermodynamics
Domain, Codomain and Image
Domain
Image
Basic features
Parity	Odd
Specific features
Root	0
Derivative
Antiderivative
Series definition
Taylor series

In mathematics, the error function (also called the Gauss error function), often denoted by erf, is a complex function of a complex variable defined as:^[1]

This integral is a special (non-elementary) sigmoid function that occurs often in probability, statistics, and partial differential equations. In many of these applications, the function argument is a real number. If the function argument is real, then the function value is also real.

In statistics, for non-negative values of x, the error function has the following interpretation: for a random variable Y that is normally distributed with mean 0 and standard deviation 1/√2, erf x is the probability that Y falls in the range [−x, x].

Two closely related functions are the complementary error function (erfc) defined as

and the imaginary error function (erfi) defined as

where i is the imaginary unit

Name[edit]

The name «error function» and its abbreviation erf were proposed by J. W. L. Glaisher in 1871 on account of its connection with «the theory of Probability, and notably the theory of Errors.»^[2] The error function complement was also discussed by Glaisher in a separate publication in the same year.^[3]
For the «law of facility» of errors whose density is given by

(the normal distribution), Glaisher calculates the probability of an error lying between p and q as:

Plot of the error function Erf(z) in the complex plane from -2-2i to 2+2i with colors created with Mathematica 13.1 function ComplexPlot3D

Applications[edit]

When the results of a series of measurements are described by a normal distribution with standard deviation σ and expected value 0, then erf (a/σ √2) is the probability that the error of a single measurement lies between −a and +a, for positive a. This is useful, for example, in determining the bit error rate of a digital communication system.

The error and complementary error functions occur, for example, in solutions of the heat equation when boundary conditions are given by the Heaviside step function.

The error function and its approximations can be used to estimate results that hold with high probability or with low probability. Given a random variable X ~ Norm[μ,σ] (a normal distribution with mean μ and standard deviation σ) and a constant L < μ:

where A and B are certain numeric constants. If L is sufficiently far from the mean, specifically μ − L ≥ σ√ln k, then:

so the probability goes to 0 as k → ∞.

The probability for X being in the interval [L_a, L_b] can be derived as

Properties[edit]

The property erf (−z) = −erf z means that the error function is an odd function. This directly results from the fact that the integrand e^−t2 is an even function (the antiderivative of an even function which is zero at the origin is an odd function and vice versa).

Since the error function is an entire function which takes real numbers to real numbers, for any complex number z:

where z is the complex conjugate of z.

The integrand f = exp(−z²) and f = erf z are shown in the complex z-plane in the figures at right with domain coloring.

The error function at +∞ is exactly 1 (see Gaussian integral). At the real axis, erf z approaches unity at z → +∞ and −1 at z → −∞. At the imaginary axis, it tends to ±i∞.

Taylor series[edit]

The error function is an entire function; it has no singularities (except that at infinity) and its Taylor expansion always converges, but is famously known «[…] for its bad convergence if x > 1.»^[4]

The defining integral cannot be evaluated in closed form in terms of elementary functions, but by expanding the integrand e^−z2 into its Maclaurin series and integrating term by term, one obtains the error function’s Maclaurin series as:

which holds for every complex number z. The denominator terms are sequence A007680 in the OEIS.

For iterative calculation of the above series, the following alternative formulation may be useful:

because −(2k − 1)z²/k(2k + 1) expresses the multiplier to turn the kth term into the (k + 1)th term (considering z as the first term).

The imaginary error function has a very similar Maclaurin series, which is:

which holds for every complex number z.

Derivative and integral[edit]

The derivative of the error function follows immediately from its definition:

From this, the derivative of the imaginary error function is also immediate:

An antiderivative of the error function, obtainable by integration by parts, is

An antiderivative of the imaginary error function, also obtainable by integration by parts, is

Higher order derivatives are given by

where H are the physicists’ Hermite polynomials.^[5]

Bürmann series[edit]

An expansion,^[6] which converges more rapidly for all real values of x than a Taylor expansion, is obtained by using Hans Heinrich Bürmann’s theorem:^[7]

where sgn is the sign function. By keeping only the first two coefficients and choosing c₁ = 31/200 and c₂ = −341/8000, the resulting approximation shows its largest relative error at x = ±1.3796, where it is less than 0.0036127:

Inverse functions[edit]

Given a complex number z, there is not a unique complex number w satisfying erf w = z, so a true inverse function would be multivalued. However, for −1 < x < 1, there is a unique real number denoted erf⁻¹ x satisfying

The inverse error function is usually defined with domain (−1,1), and it is restricted to this domain in many computer algebra systems. However, it can be extended to the disk |z| < 1 of the complex plane, using the Maclaurin series

where c₀ = 1 and

So we have the series expansion (common factors have been canceled from numerators and denominators):

(After cancellation the numerator/denominator fractions are entries OEIS: A092676/OEIS: A092677 in the OEIS; without cancellation the numerator terms are given in entry OEIS: A002067.) The error function’s value at ±∞ is equal to ±1.

For |z| < 1, we have erf(erf⁻¹ z) = z.

The inverse complementary error function is defined as

For real x, there is a unique real number erfi⁻¹ x satisfying erfi(erfi⁻¹ x) = x. The inverse imaginary error function is defined as erfi⁻¹ x.^[8]

For any real x, Newton’s method can be used to compute erfi⁻¹ x, and for −1 ≤ x ≤ 1, the following Maclaurin series converges:

where c_k is defined as above.

Asymptotic expansion[edit]

A useful asymptotic expansion of the complementary error function (and therefore also of the error function) for large real x is

where (2n − 1)!! is the double factorial of (2n − 1), which is the product of all odd numbers up to (2n − 1). This series diverges for every finite x, and its meaning as asymptotic expansion is that for any integer N ≥ 1 one has

where the remainder, in Landau notation, is

as x → ∞.

Indeed, the exact value of the remainder is

which follows easily by induction, writing

and integrating by parts.

For large enough values of x, only the first few terms of this asymptotic expansion are needed to obtain a good approximation of erfc x (while for not too large values of x, the above Taylor expansion at 0 provides a very fast convergence).

Continued fraction expansion[edit]

A continued fraction expansion of the complementary error function is:^[9]

Integral of error function with Gaussian density function[edit]

which appears related to Ng and Geller, formula 13 in section 4.3^[10] with a change of variables.

Factorial series[edit]

The inverse factorial series:

converges for Re(z²) > 0. Here

zⁿ denotes the rising factorial, and s(n,k) denotes a signed Stirling number of the first kind.^[11][12]
There also exists a representation by an infinite sum containing the double factorial:

Numerical approximations[edit]

Approximation with elementary functions[edit]

• Abramowitz and Stegun give several approximations of varying accuracy (equations 7.1.25–28). This allows one to choose the fastest approximation suitable for a given application. In order of increasing accuracy, they are:

  

  (maximum error: 5×10⁻⁴)

  where a₁ = 0.278393, a₂ = 0.230389, a₃ = 0.000972, a₄ = 0.078108

  

  (maximum error: 2.5×10⁻⁵)

  where p = 0.47047, a₁ = 0.3480242, a₂ = −0.0958798, a₃ = 0.7478556

  

  (maximum error: 3×10⁻⁷)

  where a₁ = 0.0705230784, a₂ = 0.0422820123, a₃ = 0.0092705272, a₄ = 0.0001520143, a₅ = 0.0002765672, a₆ = 0.0000430638

  

  (maximum error: 1.5×10⁻⁷)

  where p = 0.3275911, a₁ = 0.254829592, a₂ = −0.284496736, a₃ = 1.421413741, a₄ = −1.453152027, a₅ = 1.061405429

  All of these approximations are valid for x ≥ 0. To use these approximations for negative x, use the fact that erf x is an odd function, so erf x = −erf(−x).

• Exponential bounds and a pure exponential approximation for the complementary error function are given by^[13]

  

• The above have been generalized to sums of N exponentials^[14] with increasing accuracy in terms of N so that erfc x can be accurately approximated or bounded by 2Q̃(√2x), where

  

  In particular, there is a systematic methodology to solve the numerical coefficients {(a_n,b_n)}^N
  _{n = 1} that yield a minimax approximation or bound for the closely related Q-function: Q(x) ≈ Q̃(x), Q(x) ≤ Q̃(x), or Q(x) ≥ Q̃(x) for x ≥ 0. The coefficients {(a_n,b_n)}^N
  _{n = 1} for many variations of the exponential approximations and bounds up to N = 25 have been released to open access as a comprehensive dataset.^[15]

• A tight approximation of the complementary error function for x ∈ [0,∞) is given by Karagiannidis & Lioumpas (2007)^[16] who showed for the appropriate choice of parameters {A,B} that

  

  They determined {A,B} = {1.98,1.135}, which gave a good approximation for all x ≥ 0. Alternative coefficients are also available for tailoring accuracy for a specific application or transforming the expression into a tight bound.^[17]

• A single-term lower bound is^[18]

  

  where the parameter β can be picked to minimize error on the desired interval of approximation.

• Another approximation is given by Sergei Winitzki using his «global Padé approximations»:^{[19][20]: 2–3}

  

  where

  

  This is designed to be very accurate in a neighborhood of 0 and a neighborhood of infinity, and the relative error is less than 0.00035 for all real x. Using the alternate value a ≈ 0.147 reduces the maximum relative error to about 0.00013.^[21]

  This approximation can be inverted to obtain an approximation for the inverse error function:

  

• An approximation with a maximal error of 1.2×10⁻⁷ for any real argument is:^[22]

  

  with

  

  and

  

Table of values[edit]

[edit]

Complementary error function[edit]

The complementary error function, denoted erfc, is defined as

Plot of the complementary error function Erfc(z) in the complex plane from -2-2i to 2+2i with colors created with Mathematica 13.1 function ComplexPlot3D

which also defines erfcx, the scaled complementary error function^[23] (which can be used instead of erfc to avoid arithmetic underflow^[23][24]). Another form of erfc x for x ≥ 0 is known as Craig’s formula, after its discoverer:^[25]

This expression is valid only for positive values of x, but it can be used in conjunction with erfc x = 2 − erfc(−x) to obtain erfc(x) for negative values. This form is advantageous in that the range of integration is fixed and finite. An extension of this expression for the erfc of the sum of two non-negative variables is as follows:^[26]

Imaginary error function[edit]

The imaginary error function, denoted erfi, is defined as

where D(x) is the Dawson function (which can be used instead of erfi to avoid arithmetic overflow^[23]).

Despite the name «imaginary error function», erfi x is real when x is real.

When the error function is evaluated for arbitrary complex arguments z, the resulting complex error function is usually discussed in scaled form as the Faddeeva function:

Cumulative distribution function[edit]

The error function is essentially identical to the standard normal cumulative distribution function, denoted Φ, also named norm(x) by some software languages^{[citation needed]}, as they differ only by scaling and translation. Indeed,

the normal cumulative distribution function plotted in the complex plane

or rearranged for erf and erfc:

Consequently, the error function is also closely related to the Q-function, which is the tail probability of the standard normal distribution. The Q-function can be expressed in terms of the error function as

The inverse of Φ is known as the normal quantile function, or probit function and may be expressed in terms of the inverse error function as

The standard normal cdf is used more often in probability and statistics, and the error function is used more often in other branches of mathematics.

The error function is a special case of the Mittag-Leffler function, and can also be expressed as a confluent hypergeometric function (Kummer’s function):

It has a simple expression in terms of the Fresnel integral.^{[further explanation needed]}

In terms of the regularized gamma function P and the incomplete gamma function,

sgn x is the sign function.

Generalized error functions[edit]

Some authors discuss the more general functions:^{[citation needed]}

Notable cases are:

• E₀(x) is a straight line through the origin: E₀(x) = x/e√π
• E₂(x) is the error function, erf x.

After division by n!, all the E_n for odd n look similar (but not identical) to each other. Similarly, the E_n for even n look similar (but not identical) to each other after a simple division by n!. All generalised error functions for n > 0 look similar on the positive x side of the graph.

These generalised functions can equivalently be expressed for x > 0 using the gamma function and incomplete gamma function:

Therefore, we can define the error function in terms of the incomplete gamma function:

Iterated integrals of the complementary error function[edit]

The iterated integrals of the complementary error function are defined by^[27]

The general recurrence formula is

They have the power series

from which follow the symmetry properties

and

Implementations[edit]

As real function of a real argument[edit]

• In Posix-compliant operating systems, the header math.h shall declare and the mathematical library libm shall provide the functions erf and erfc (double precision) as well as their single precision and extended precision counterparts erff, erfl and erfcf, erfcl.^[28]
• The GNU Scientific Library provides erf, erfc, log(erf), and scaled error functions.^[29]

As complex function of a complex argument[edit]

• libcerf, numeric C library for complex error functions, provides the complex functions cerf, cerfc, cerfcx and the real functions erfi, erfcx with approximately 13–14 digits precision, based on the Faddeeva function as implemented in the MIT Faddeeva Package

See also[edit]

[edit]

• Gaussian integral, over the whole real line
• Gaussian function, derivative
• Dawson function, renormalized imaginary error function
• Goodwin–Staton integral

In probability[edit]

• Normal distribution
• Normal cumulative distribution function, a scaled and shifted form of error function
• Probit, the inverse or quantile function of the normal CDF
• Q-function, the tail probability of the normal distribution

References[edit]

Further reading[edit]

• Abramowitz, Milton; Stegun, Irene Ann, eds. (1983) [June 1964]. «Chapter 7». Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. Applied Mathematics Series. Vol. 55 (Ninth reprint with additional corrections of tenth original printing with corrections (December 1972); first ed.). Washington D.C.; New York: United States Department of Commerce, National Bureau of Standards; Dover Publications. p. 297. ISBN 978-0-486-61272-0. LCCN 64-60036. MR 0167642. LCCN 65-12253.
• Press, William H.; Teukolsky, Saul A.; Vetterling, William T.; Flannery, Brian P. (2007), «Section 6.2. Incomplete Gamma Function and Error Function», Numerical Recipes: The Art of Scientific Computing (3rd ed.), New York: Cambridge University Press, ISBN 978-0-521-88068-8
• Temme, Nico M. (2010), «Error Functions, Dawson’s and Fresnel Integrals», in Olver, Frank W. J.; Lozier, Daniel M.; Boisvert, Ronald F.; Clark, Charles W. (eds.), NIST Handbook of Mathematical Functions, Cambridge University Press, ISBN 978-0-521-19225-5, MR 2723248

External links[edit]

• A Table of Integrals of the Error Functions

Error function
Plot of the error function
General information
General definition
Fields of application	Probability, thermodynamics
Domain, Codomain and Image
Domain
Image
Basic features
Parity	Odd
Specific features
Root	0
Derivative
Antiderivative
Series definition
Taylor series

In mathematics, the error function (also called the Gauss error function), often denoted by erf, is a complex function of a complex variable defined as:^[1]

This integral is a special (non-elementary) sigmoid function that occurs often in probability, statistics, and partial differential equations. In many of these applications, the function argument is a real number. If the function argument is real, then the function value is also real.

In statistics, for non-negative values of x, the error function has the following interpretation: for a random variable Y that is normally distributed with mean 0 and standard deviation 1/√2, erf x is the probability that Y falls in the range [−x, x].

Two closely related functions are the complementary error function (erfc) defined as

and the imaginary error function (erfi) defined as

where i is the imaginary unit

Name[edit]

The name «error function» and its abbreviation erf were proposed by J. W. L. Glaisher in 1871 on account of its connection with «the theory of Probability, and notably the theory of Errors.»^[2] The error function complement was also discussed by Glaisher in a separate publication in the same year.^[3]
For the «law of facility» of errors whose density is given by

(the normal distribution), Glaisher calculates the probability of an error lying between p and q as:

Plot of the error function Erf(z) in the complex plane from -2-2i to 2+2i with colors created with Mathematica 13.1 function ComplexPlot3D

Applications[edit]

When the results of a series of measurements are described by a normal distribution with standard deviation σ and expected value 0, then erf (a/σ √2) is the probability that the error of a single measurement lies between −a and +a, for positive a. This is useful, for example, in determining the bit error rate of a digital communication system.

The error and complementary error functions occur, for example, in solutions of the heat equation when boundary conditions are given by the Heaviside step function.

The error function and its approximations can be used to estimate results that hold with high probability or with low probability. Given a random variable X ~ Norm[μ,σ] (a normal distribution with mean μ and standard deviation σ) and a constant L < μ:

where A and B are certain numeric constants. If L is sufficiently far from the mean, specifically μ − L ≥ σ√ln k, then:

so the probability goes to 0 as k → ∞.

The probability for X being in the interval [L_a, L_b] can be derived as

Properties[edit]

The property erf (−z) = −erf z means that the error function is an odd function. This directly results from the fact that the integrand e^−t2 is an even function (the antiderivative of an even function which is zero at the origin is an odd function and vice versa).

Since the error function is an entire function which takes real numbers to real numbers, for any complex number z:

where z is the complex conjugate of z.

The integrand f = exp(−z²) and f = erf z are shown in the complex z-plane in the figures at right with domain coloring.

The error function at +∞ is exactly 1 (see Gaussian integral). At the real axis, erf z approaches unity at z → +∞ and −1 at z → −∞. At the imaginary axis, it tends to ±i∞.

Taylor series[edit]

The error function is an entire function; it has no singularities (except that at infinity) and its Taylor expansion always converges, but is famously known «[…] for its bad convergence if x > 1.»^[4]

The defining integral cannot be evaluated in closed form in terms of elementary functions, but by expanding the integrand e^−z2 into its Maclaurin series and integrating term by term, one obtains the error function’s Maclaurin series as:

which holds for every complex number z. The denominator terms are sequence A007680 in the OEIS.

For iterative calculation of the above series, the following alternative formulation may be useful:

because −(2k − 1)z²/k(2k + 1) expresses the multiplier to turn the kth term into the (k + 1)th term (considering z as the first term).

The imaginary error function has a very similar Maclaurin series, which is:

which holds for every complex number z.

Derivative and integral[edit]

The derivative of the error function follows immediately from its definition:

From this, the derivative of the imaginary error function is also immediate:

An antiderivative of the error function, obtainable by integration by parts, is

An antiderivative of the imaginary error function, also obtainable by integration by parts, is

Higher order derivatives are given by

where H are the physicists’ Hermite polynomials.^[5]

Bürmann series[edit]

An expansion,^[6] which converges more rapidly for all real values of x than a Taylor expansion, is obtained by using Hans Heinrich Bürmann’s theorem:^[7]

where sgn is the sign function. By keeping only the first two coefficients and choosing c₁ = 31/200 and c₂ = −341/8000, the resulting approximation shows its largest relative error at x = ±1.3796, where it is less than 0.0036127:

Inverse functions[edit]

Given a complex number z, there is not a unique complex number w satisfying erf w = z, so a true inverse function would be multivalued. However, for −1 < x < 1, there is a unique real number denoted erf⁻¹ x satisfying

The inverse error function is usually defined with domain (−1,1), and it is restricted to this domain in many computer algebra systems. However, it can be extended to the disk |z| < 1 of the complex plane, using the Maclaurin series

where c₀ = 1 and

So we have the series expansion (common factors have been canceled from numerators and denominators):

(After cancellation the numerator/denominator fractions are entries OEIS: A092676/OEIS: A092677 in the OEIS; without cancellation the numerator terms are given in entry OEIS: A002067.) The error function’s value at ±∞ is equal to ±1.

For |z| < 1, we have erf(erf⁻¹ z) = z.

The inverse complementary error function is defined as

For real x, there is a unique real number erfi⁻¹ x satisfying erfi(erfi⁻¹ x) = x. The inverse imaginary error function is defined as erfi⁻¹ x.^[8]

For any real x, Newton’s method can be used to compute erfi⁻¹ x, and for −1 ≤ x ≤ 1, the following Maclaurin series converges:

where c_k is defined as above.

Asymptotic expansion[edit]

A useful asymptotic expansion of the complementary error function (and therefore also of the error function) for large real x is

where (2n − 1)!! is the double factorial of (2n − 1), which is the product of all odd numbers up to (2n − 1). This series diverges for every finite x, and its meaning as asymptotic expansion is that for any integer N ≥ 1 one has

where the remainder, in Landau notation, is

as x → ∞.

Indeed, the exact value of the remainder is

which follows easily by induction, writing

and integrating by parts.

For large enough values of x, only the first few terms of this asymptotic expansion are needed to obtain a good approximation of erfc x (while for not too large values of x, the above Taylor expansion at 0 provides a very fast convergence).

Continued fraction expansion[edit]

A continued fraction expansion of the complementary error function is:^[9]

Integral of error function with Gaussian density function[edit]

which appears related to Ng and Geller, formula 13 in section 4.3^[10] with a change of variables.

Factorial series[edit]

The inverse factorial series:

converges for Re(z²) > 0. Here

zⁿ denotes the rising factorial, and s(n,k) denotes a signed Stirling number of the first kind.^[11][12]
There also exists a representation by an infinite sum containing the double factorial:

Numerical approximations[edit]

Approximation with elementary functions[edit]

• Abramowitz and Stegun give several approximations of varying accuracy (equations 7.1.25–28). This allows one to choose the fastest approximation suitable for a given application. In order of increasing accuracy, they are:

  

  (maximum error: 5×10⁻⁴)

  where a₁ = 0.278393, a₂ = 0.230389, a₃ = 0.000972, a₄ = 0.078108

  

  (maximum error: 2.5×10⁻⁵)

  where p = 0.47047, a₁ = 0.3480242, a₂ = −0.0958798, a₃ = 0.7478556

  

  (maximum error: 3×10⁻⁷)

  where a₁ = 0.0705230784, a₂ = 0.0422820123, a₃ = 0.0092705272, a₄ = 0.0001520143, a₅ = 0.0002765672, a₆ = 0.0000430638

  

  (maximum error: 1.5×10⁻⁷)

  where p = 0.3275911, a₁ = 0.254829592, a₂ = −0.284496736, a₃ = 1.421413741, a₄ = −1.453152027, a₅ = 1.061405429

  All of these approximations are valid for x ≥ 0. To use these approximations for negative x, use the fact that erf x is an odd function, so erf x = −erf(−x).

• Exponential bounds and a pure exponential approximation for the complementary error function are given by^[13]

  

• The above have been generalized to sums of N exponentials^[14] with increasing accuracy in terms of N so that erfc x can be accurately approximated or bounded by 2Q̃(√2x), where

  

  In particular, there is a systematic methodology to solve the numerical coefficients {(a_n,b_n)}^N
  _{n = 1} that yield a minimax approximation or bound for the closely related Q-function: Q(x) ≈ Q̃(x), Q(x) ≤ Q̃(x), or Q(x) ≥ Q̃(x) for x ≥ 0. The coefficients {(a_n,b_n)}^N
  _{n = 1} for many variations of the exponential approximations and bounds up to N = 25 have been released to open access as a comprehensive dataset.^[15]

• A tight approximation of the complementary error function for x ∈ [0,∞) is given by Karagiannidis & Lioumpas (2007)^[16] who showed for the appropriate choice of parameters {A,B} that

  

  They determined {A,B} = {1.98,1.135}, which gave a good approximation for all x ≥ 0. Alternative coefficients are also available for tailoring accuracy for a specific application or transforming the expression into a tight bound.^[17]

• A single-term lower bound is^[18]

  

  where the parameter β can be picked to minimize error on the desired interval of approximation.

• Another approximation is given by Sergei Winitzki using his «global Padé approximations»:^{[19][20]: 2–3}

  

  where

  

  This is designed to be very accurate in a neighborhood of 0 and a neighborhood of infinity, and the relative error is less than 0.00035 for all real x. Using the alternate value a ≈ 0.147 reduces the maximum relative error to about 0.00013.^[21]

  This approximation can be inverted to obtain an approximation for the inverse error function:

  

• An approximation with a maximal error of 1.2×10⁻⁷ for any real argument is:^[22]

  

  with

  

  and

  

Table of values[edit]

[edit]

Complementary error function[edit]

The complementary error function, denoted erfc, is defined as

Plot of the complementary error function Erfc(z) in the complex plane from -2-2i to 2+2i with colors created with Mathematica 13.1 function ComplexPlot3D

which also defines erfcx, the scaled complementary error function^[23] (which can be used instead of erfc to avoid arithmetic underflow^[23][24]). Another form of erfc x for x ≥ 0 is known as Craig’s formula, after its discoverer:^[25]

This expression is valid only for positive values of x, but it can be used in conjunction with erfc x = 2 − erfc(−x) to obtain erfc(x) for negative values. This form is advantageous in that the range of integration is fixed and finite. An extension of this expression for the erfc of the sum of two non-negative variables is as follows:^[26]

Imaginary error function[edit]

The imaginary error function, denoted erfi, is defined as

where D(x) is the Dawson function (which can be used instead of erfi to avoid arithmetic overflow^[23]).

Despite the name «imaginary error function», erfi x is real when x is real.

When the error function is evaluated for arbitrary complex arguments z, the resulting complex error function is usually discussed in scaled form as the Faddeeva function:

Cumulative distribution function[edit]

The error function is essentially identical to the standard normal cumulative distribution function, denoted Φ, also named norm(x) by some software languages^{[citation needed]}, as they differ only by scaling and translation. Indeed,

the normal cumulative distribution function plotted in the complex plane

or rearranged for erf and erfc:

Consequently, the error function is also closely related to the Q-function, which is the tail probability of the standard normal distribution. The Q-function can be expressed in terms of the error function as

The inverse of Φ is known as the normal quantile function, or probit function and may be expressed in terms of the inverse error function as

The standard normal cdf is used more often in probability and statistics, and the error function is used more often in other branches of mathematics.

The error function is a special case of the Mittag-Leffler function, and can also be expressed as a confluent hypergeometric function (Kummer’s function):

It has a simple expression in terms of the Fresnel integral.^{[further explanation needed]}

In terms of the regularized gamma function P and the incomplete gamma function,

sgn x is the sign function.

Generalized error functions[edit]

Some authors discuss the more general functions:^{[citation needed]}

Notable cases are:

• E₀(x) is a straight line through the origin: E₀(x) = x/e√π
• E₂(x) is the error function, erf x.

After division by n!, all the E_n for odd n look similar (but not identical) to each other. Similarly, the E_n for even n look similar (but not identical) to each other after a simple division by n!. All generalised error functions for n > 0 look similar on the positive x side of the graph.

These generalised functions can equivalently be expressed for x > 0 using the gamma function and incomplete gamma function:

Therefore, we can define the error function in terms of the incomplete gamma function:

Iterated integrals of the complementary error function[edit]

The iterated integrals of the complementary error function are defined by^[27]

The general recurrence formula is

They have the power series

from which follow the symmetry properties

and

Implementations[edit]

As real function of a real argument[edit]

• In Posix-compliant operating systems, the header math.h shall declare and the mathematical library libm shall provide the functions erf and erfc (double precision) as well as their single precision and extended precision counterparts erff, erfl and erfcf, erfcl.^[28]
• The GNU Scientific Library provides erf, erfc, log(erf), and scaled error functions.^[29]

As complex function of a complex argument[edit]

• libcerf, numeric C library for complex error functions, provides the complex functions cerf, cerfc, cerfcx and the real functions erfi, erfcx with approximately 13–14 digits precision, based on the Faddeeva function as implemented in the MIT Faddeeva Package

See also[edit]

[edit]

• Gaussian integral, over the whole real line
• Gaussian function, derivative
• Dawson function, renormalized imaginary error function
• Goodwin–Staton integral

In probability[edit]

• Normal distribution
• Normal cumulative distribution function, a scaled and shifted form of error function
• Probit, the inverse or quantile function of the normal CDF
• Q-function, the tail probability of the normal distribution

References[edit]

Further reading[edit]

• Abramowitz, Milton; Stegun, Irene Ann, eds. (1983) [June 1964]. «Chapter 7». Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. Applied Mathematics Series. Vol. 55 (Ninth reprint with additional corrections of tenth original printing with corrections (December 1972); first ed.). Washington D.C.; New York: United States Department of Commerce, National Bureau of Standards; Dover Publications. p. 297. ISBN 978-0-486-61272-0. LCCN 64-60036. MR 0167642. LCCN 65-12253.
• Press, William H.; Teukolsky, Saul A.; Vetterling, William T.; Flannery, Brian P. (2007), «Section 6.2. Incomplete Gamma Function and Error Function», Numerical Recipes: The Art of Scientific Computing (3rd ed.), New York: Cambridge University Press, ISBN 978-0-521-88068-8
• Temme, Nico M. (2010), «Error Functions, Dawson’s and Fresnel Integrals», in Olver, Frank W. J.; Lozier, Daniel M.; Boisvert, Ronald F.; Clark, Charles W. (eds.), NIST Handbook of Mathematical Functions, Cambridge University Press, ISBN 978-0-521-19225-5, MR 2723248

External links[edit]

• A Table of Integrals of the Error Functions

График функции

В математике функция ошибок (также называемая Функция ошибок Гаусса ), часто обозначаемая erf, является сложной функцией комплексной определяемой как:

erf ⁡ z = 2 π ∫ 0 ze — t 2 dt. { displaystyle operatorname {erf} z = { frac {2} { sqrt { pi}}} int _ {0} ^ {z} e ^ {- t ^ {2}} , dt.}

Этот интеграл является особой (не элементарной ) и сигмоидной функцией, которая часто встречается в статистике вероятность, и уравнения в частных производных. Во многих из этих приложений аргумент функции является действительным числом. Если аргумент функции является действительным, значение также является действительным.

В статистике для неотрицательных значений x функция имеет интерпретацию: для случайной величины Y, которая нормально распределена с среднее 0 и дисперсия 1/2, erf x — это вероятность того, что Y попадает в диапазон [-x, x].

Две связанные функции: дополнительные функции ошибок (erfc ), определенная как

erfc ⁡ z = 1 — erf ⁡ z, { displaystyle operatorname {erfc} z = 1- operatorname {erf} z,}

и функция мнимой ошибки (erfi ), определяемая как

erfi ⁡ z = — i erf ⁡ (iz), { displaystyle operatorname {erfi} z = -i operatorname {erf} (iz),}

, где i — мнимая единица.

Содержание

• 1 Имя
• 2 Приложения
• 3 Свойства
  • 3.1 Ряд Тейлора
  • 3.2 Производная и интеграл
  • 3.3 Ряд Бюрмана
  • 3.4 Обратные функции
  • 3.5 Асимптотическое разложение
  • 3.6 Разложение на непрерывную дробь
  • 3,7 Интеграл функции ошибок с функцией плотности Гаусса
  • 3.8 Факториальный ряд
• 4 Численные приближения
  • 4.1 Аппроксимация с элементарными функциями
  • 4.2 Полином
  • 4.3 Таблица значений
• 5 Связанные функции
  • 5.1 функция дополнительных ошибок
  • 5.2 Функция мнимой ошибки
  • 5.3 Кумулятивная функци я распределения на
  • 5.4 Обобщенные функции ошибок
  • 5.5 Итерированные интегралы дополнительных функций ошибок
• 6 Реализации
  • 6.1 Как действующая функция действительного аргумента
  • 6.2 Как комплексная функция комплексного аргумента
• 7 См. Также
  • 7.1 Связанные функции
  • 7.2 Вероятность
• 8 Ссылки
• 9 Дополнительная литература
• 10 Внешние ссылки

Имя

Название «функция ошибки» и его аббревиатура erf были предложены Дж. В. Л. Глейшер в 1871 г. по причине его связи с «теорией вероятности, и особенно теорией ошибок ». Дополнение функции ошибок также обсуждалось Глейшером в отдельной публикации в том же году. Для «закона удобства» ошибок плотность задана как

f (x) = (c π) 1 2 e — cx 2 { displaystyle f (x) = left ({ frac {c } { pi}} right) ^ { tfrac {1} {2}} e ^ {- cx ^ {2}}}

(нормальное распределение ), Глейшер вычисляет вероятность ошибки, лежащей между p { displaystyle p}и q { displaystyle q}как:

(c π) 1 2 ∫ pqe — cx 2 dx = 1 2 (erf ⁡ (qc) — erf ⁡ (pc)). { displaystyle left ({ frac {c} { pi}} right) ^ { tfrac {1} {2}} int _ {p} ^ {q} e ^ {- cx ^ {2} } dx = { tfrac {1} {2}} left ( operatorname {erf} (q { sqrt {c}}) — operatorname {erf} (p { sqrt {c}}) right).}

Приложения

Когда результаты серии измерений описываются нормальным распределением со стандартным отклонением σ { displaystyle sigma}и ожидаемое значение 0, затем erf ⁡ (a σ 2) { displaystyle textstyle operatorname {erf} left ({ frac {a} { sigma { sqrt {2}) }}} right)}— это вероятность того, что ошибка единичного измерения находится между −a и + a, для положительного a. Это полезно, например, при определении коэффициента битовых ошибок цифровой системы связи.

Функции и дополнительные функции ошибок возникают, например, в решениях уравнения теплопроводности, когда граничные ошибки задаются ступенчатой ​​функцией Хевисайда.

Функция ошибок и ее приближения Программу присвоили себе преподавателей, которые получили с высокой вероятностью или с низкой вероятностью. Дана случайная величина X ∼ Norm ⁡ [μ, σ] { displaystyle X sim operatorname {Norm} [ mu, sigma]}и константа L < μ {displaystyle L<mu }:

Pr [X ≤ L ] = 1 2 + 1 2 erf ⁡ (L — μ 2 σ) ≈ A ехр (- B (L — μ σ) 2) { Displaystyle Pr [X Leq L] = { frac {1} {2 }} + { frac {1} {2}} operatorname {erf} left ({ frac {L- mu} {{ sqrt {2}} sigma}} right) приблизительно A exp left (-B left ({ frac {L- mu} { sigma}} right) ^ {2} right)}

где A и B — верх числовые константы. Если L достаточно далеко от среднего, то есть μ — L ≥ σ ln ⁡ k { displaystyle mu -L geq sigma { sqrt { ln {k}}}}, то:

Pr [X ≤ L] ≤ A exp ⁡ (- B ln ⁡ k) = A К B { displaystyle Pr [X leq L] leq A exp (-B ln {k}) = { frac {A} {k ^ {B}}}}

, поэтому становится вероятность 0 при k → ∞ { displaystyle k to infty}.

Свойства

Графики на комплексной плоскости Интегрируем exp (-z) erf (z)

Свойство erf ⁡ (- z) = — erf ⁡ (z) { displaystyle operatorname {erf} (-z) = — operatorname {erf} (z)}означает, что функция является ошибкой нечетной функции. Это связано с тем, что подынтегральное выражение e — t 2 { displaystyle e ^ {- t ^ {2}}}является четной функцией.

Для любого комплексное число z:

erf ⁡ (z ¯) = erf ⁡ (z) ¯ { displaystyle operatorname {erf} ({ overline {z}}) = { overline { operatorname {erf} (z)}}}

где z ¯ { displaystyle { overline {z}}}— комплексное сопряжение число z.

Подынтегральное выражение f = exp (−z) и f = erf (z) показано в комплексной плоскости z на рисунках 2 и 3. Уровень Im (f) = 0 показан жирным зеленым цветом. линия. Отрицательные целые значения Im (f) показаны жирными красными линиями. Положительные целые значения Im (f) показаны толстыми синими линиями. Промежуточные уровни Im (f) = проявляются тонкими зелеными линиями. Промежуточные уровни Re (f) = показаны тонкими красными линиями для отрицательных значений и тонкими синими линиями для положительных значений.

Функция ошибок при + ∞ равна 1 (см. интеграл Гаусса ). На действительной оси erf (z) стремится к единице при z → + ∞ и к −1 при z → −∞. На мнимой оси он стремится к ± i∞.

Серия Тейлора

Функция ошибок — это целая функция ; у него нет сингулярностей (кроме бесконечности), и его разложение Тейлора всегда сходится, но, как известно, «[…] его плохая сходимость, если x>1».

определяющий интеграл нельзя вычислить в закрытой форме в терминах элементарных функций, но путем расширения подынтегрального выражения e в его ряд Маклорена и интегрирована почленно, можно получить ряд Маклорена функции ошибок как:

erf ⁡ (z) = 2 π ∑ n = 0 ∞ (- 1) nz 2 n + 1 n! (2 n + 1) знак равно 2 π (z — z 3 3 + z 5 10 — z 7 42 + z 9 216 — ⋯) { displaystyle operatorname {erf} (z) = { frac {2} { sqrt { pi}}} sum _ {n = 0} ^ { infty} { frac {(-1) ^ {n} z ^ {2n + 1}} {n! (2n + 1)}} = { frac {2} { sqrt { pi}}} left (z — { frac {z ^ {3}} {3}} + { frac {z ^ { 5}} {10}} — { frac {z ^ {7}} {42}} + { frac {z ^ {9}} {216}} — cdots right)}

, которое выполняется для каждого комплексного числа г. Члены знаменателя представляют собой последовательность A007680 в OEIS.

Для итеративного вычисления нового ряда может быть полезна следующая альтернативная формулировка:

erf ⁡ (z) = 2 π ∑ n = 0 ∞ (z ∏ К знак равно 1 N — (2 К — 1) Z 2 К (2 К + 1)) знак равно 2 π ∑ N = 0 ∞ Z 2 N + 1 ∏ К = 1 N — Z 2 К { Displaystyle OperatorName { erf} (z) = { frac {2} { sqrt { pi}}} sum _ {n = 0} ^ { infty} left (z prod _ {k = 1} ^ {n} { frac {- (2k-1) z ^ {2}} {k (2k + 1)}} right) = { frac {2} { sqrt { pi}}} sum _ {n = 0} ^ { infty} { frac {z} {2n + 1}} prod _ {k = 1} ^ {n} { frac {-z ^ {2}} {k}}}

потому что что — (2 k — 1) z 2 k (2 k + 1) { displaystyle { frac {- (2k-1) z ^ {2}} {k (2k + 1))}} }выражает множитель для превращения члена k в член (k + 1) (рассматривая z как первый член).

Функция мнимой ошибки имеет очень похожий ряд Маклорена:

erfi ⁡ (z) = 2 π ∑ n = 0 ∞ z 2 n + 1 n! (2 n + 1) знак равно 2 π (z + z 3 3 + z 5 10 + z 7 42 + z 9 216 + ⋯) { displaystyle operatorname {erfi} (z) = { frac {2} { sqrt { pi}}} sum _ {n = 0} ^ { infty} { frac {z ^ {2n + 1}} {n! (2n + 1)}} = { frac {2} { sqrt { pi}}} left (z + { frac {z ^ {3}} {3}} + { frac {z ^ { 5}} {10}} + { frac {z ^ {7}} {42}} + { frac {z ^ {9}} {216}} + cdots right)}

, которое выполняется для любого комплексного числа z.

Производная и интеграл

Производная функция ошибок сразу следует из ее определения:

ddz erf ⁡ (z) = 2 π e — z 2. { displaystyle { frac {d} {dz}} operatorname {erf} (z) = { frac {2} { sqrt { pi}}} e ^ {- z ^ {2}}.}

Отсюда немедленно вычисляется производная функция мнимой ошибки :

ddz erfi ⁡ (z) = 2 π ez 2. { displaystyle { frac {d} {dz}} operatorname {erfi} (z) = { frac {2} { sqrt { pi }}} e ^ {z ^ {2}}.}

первообразная функции ошибок, которые можно получить посредством интегрирования по частям, составляет

z erf ⁡ (z) + е — z 2 π. { displaystyle z operatorname {erf} (z) + { frac {e ^ {- z ^ {2}}} { sqrt { pi}}}.}

Первообразная мнимой функции ошибок, также можно получить интегрированием по частям:

z erfi ⁡ (z) — ez 2 π. { displaystyle z operatorname {erfi} (z) — { frac {e ^ {z ^ {2}}} { sqrt { pi}}}.}

Производные высшего порядка задаются как

erf (k) ⁡ (z) = 2 (- 1) k — 1 π H k — 1 (z) e — z 2 = 2 π dk — 1 dzk — 1 (e — z 2), k = 1, 2, … { Displaystyle operatorname {erf} ^ {(k)} (z) = { frac {2 (-1) ^ {k-1}} { sqrt { pi}}} { mathit {H} } _ {k-1} (z) e ^ {- z ^ {2}} = { frac {2} { sqrt { pi}}} { frac {d ^ {k-1}} {dz ^ {k-1}}} left (e ^ {- z ^ {2}} right), qquad k = 1,2, dots}

где H { displaystyle { mathit {H}}}— физики многочлены Эрмита.

ряд Бюрмана

Расширение, которое сходится быстрее для всех реальных значений x { displaystyle x}, чем разложение Тейлора, получается с помощью теоремы Ганса Генриха Бюрмана :

erf ⁡ (x) = 2 π sgn ⁡ (x) 1 — e — x 2 (1 — 1 12 ( 1 — e — x 2) — 7 480 (1 — e — x 2) 2 — 5 896 (1 — e — x 2) 3 — 787 276480 (1 — e — x 2)) 4 — ⋯) знак равно 2 π знак ⁡ (x) 1 — e — x 2 (π 2 + ∑ k = 1 ∞ cke — kx 2). { displaystyle { begin {align} operatorname {erf} (x) = { frac {2} { sqrt { pi}}} operatorname {sgn} (x) { sqrt {1-e ^ {-x ^ {2}}}} left (1 — { frac {1} {12}} left (1-e ^ {- x ^ {2}} right) — { frac {7} {480}} left (1-e ^ {- x ^ {2}} right) ^ {2} — { frac {5} {896}} left (1-e ^ {- x ^ {2 }} right) ^ {3} — { frac {787} {276480}} left (1-e ^ {- x ^ {2}} right) ^ {4} — cdots right) \ [10pt] = { frac {2} { sqrt { pi}}} operatorname {sgn} (x) { sqrt {1-e ^ {- x ^ {2}}}} left ({ frac { sqrt { pi}} {2}} + sum _ {k = 1} ^ { infty} c_ {k} e ^ {- kx ^ {2}} right). end {выровнено}}

Сохраняя только первые два коэффициента и выбирая c 1 = 31 200 { displaystyle c_ {1} = { frac {31} {200}}}и c 2 = — 341 8000, { displaystyle c_ {2} = — { frac {341} {8000}},}результирующая аппроксимация дает наибольшую относительную ошибку при x = ± 1,3796, { displaystyle x = pm 1,3796,}, где оно меньше 3,6127 ⋅ 10 — 3 { displaystyle 3.6127 cdot 10 ^ {- 3}}:

erf ⁡ (x) ≈ 2 π sign ⁡ (x) 1 — e — x 2 (π 2 + 31 200 e — x 2 — 341 8000 e — 2 х 2). { displaystyle operatorname {erf} (x) приблизительно { frac {2} { sqrt { pi}}} operatorname {sgn} (x) { sqrt {1-e ^ {- x ^ {2 }}}} left ({ frac { sqrt { pi}} {2}} + { frac {31} {200}} e ^ {- x ^ {2}} — { frac {341} {8000}} e ^ {- 2x ^ {2}} right).}

Обратные функции

Обратная функция

Учитывая комплексное число z, не существует уникального комплексного числа w, удовлетворяющего erf ⁡ (w) = z { displaystyle operatorname {erf} (w) = z}, поэтому истинная обратная функция будет многозначной. Однако для −1 < x < 1, there is a unique real number denoted erf — 1 ⁡ (x) { displaystyle operatorname {erf} ^ {- 1} (x)}, удовлетворяющего

erf ⁡ (erf — 1 ⁡ ( х)) = х. { displaystyle operatorname {erf} left ( operatorname {erf} ^ {- 1} (x) right) = x.}

Обратная функция ошибок обычно определяется с помощью домена (- 1,1), и он ограничен этой областью многих систем компьютерной алгебры. Однако его можно продолжить и на диск | z | < 1 of the complex plane, using the Maclaurin series

erf — 1 ⁡ (z) знак равно ∑ К знак равно 0 ∞ ck 2 k + 1 (π 2 z) 2 k + 1, { displaystyle operatorname {erf} ^ {- 1} (z) = sum _ {k = 0} ^ { infty} { frac {c_ {k}} {2k + 1}} left ({ frac { sqrt { pi}} {2}} z right) ^ {2k + 1},}

где c 0 = 1 и

ck = ∑ m = 0 k — 1 cmck — 1 — m (m + 1) (2 m + 1) = {1, 1, 7 6, 127 90, 4369 2520, 34807 16200,…}. { displaystyle c_ {k} = sum _ {m = 0} ^ {k-1} { frac {c_ {m} c_ {k-1-m}} {(m + 1) (2m + 1) }} = left {1,1, { frac {7} {6}}, { frac {127} {90}}, { frac {4369} {2520}}, { frac {34807} {16200}}, ldots right }.}

Итак, у нас есть разложение в ряд (общие множители были удалены из числителей и знаменателей):

erf — 1 ⁡ (z) = 1 2 π ( z + π 12 z 3 + 7 π 2 480 z 5 + 127 π 3 40320 z 7 + 4369 π 4 5806080 z 9 + 34807 π 5 182476800 z 11 + ⋯). { displaystyle operatorname {erf} ^ {- 1} (z) = { tfrac {1} {2}} { sqrt { pi}} left (z + { frac { pi} {12} } z ^ {3} + { frac {7 pi ^ {2}} {480}} z ^ {5} + { frac {127 pi ^ {3}} {40320}} z ^ {7} + { frac {4369 pi ^ {4}} {5806080}} z ^ {9} + { frac {34807 pi ^ {5}} {182476800}} z ^ {11} + cdots right). }

(После отмены дроби числителя / знаменателя характерми OEIS : A092676 / OEIS : A092677 в OEIS ; без отмены членов числителя в записи OEIS : A002067.) Значение функции ошибок при ± ∞ равно ± 1.

Для | z | < 1, we have erf ⁡ (erf — 1 ⁡ (z)) = z { displaystyle operatorname {erf} left ( operatorname {erf} ^ {- 1} (z) right) = z}.

обратная дополнительная функция ошибок определяется как

erfc — 1 ⁡ (1 — z) = erf — 1 ⁡ (z). { displaystyle operatorname {erfc} ^ {- 1} (1-z) = operatorname {erf} ^ {- 1} (z).}

Для действительного x существует уникальное действительное число erfi — 1 ⁡ (x) { displaystyle operatorname {erfi} ^ {- 1} (x)}удовлетворяет erfi ⁡ (erfi — 1 ⁡ (x)) = x { displaystyle operatorname { erfi} left ( operatorname {erfi} ^ {- 1} (x) right) = x}. функция обратной мнимой ошибки определяется как erfi — 1 ⁡ (x) { displaystyle operatorname {erfi} ^ {- 1} (x)}.

Для любого действительного x, Метод Ньютона можно использовать для вычислений erfi — 1 ⁡ (x) { displaystyle operatorname {erfi} ^ {- 1} (x)}, а для — 1 ≤ x ≤ 1 { displaystyle -1 leq x leq 1}, сходится следующий ряд Маклорена:

erfi — 1 ⁡ (z) = ∑ k = 0 ∞ (- 1) ККК 2 К + 1 (π 2 Z) 2 К + 1, { Displaystyle OperatorName {erfi} ^ {- 1} (г) = сумма _ {к = 0} ^ { infty} { гидроразрыва {(-1) ^ {k} c_ {k}} {2k + 1}} left ({ frac { sqrt { pi}} {2}} z right) ^ {2k + 1},}

, где c k определено, как указано выше.

Асимптотическое разложение

Полезным асимптотическим разложением дополнительные функции (и, следовательно, также и функции ошибок) для больших вещественных x

erfc ⁡ (x) = e — x 2 x π [1 + ∑ n = 1 ∞ (- 1) n 1 ⋅ 3 ⋅ 5 ⋯ (2 n — 1) (2 x 2) n] = e — x 2 x π ∑ n = 0 ∞ (- 1) п (2 п — 1)! ! (2 х 2) n, { displaystyle operatorname {erfc} (x) = { frac {e ^ {- x ^ {2}}} {x { sqrt { pi}}}} left [1 + sum _ {n = 1} ^ { infty} (- 1) ^ {n} { frac {1 cdot 3 cdot 5 cdots (2n-1)} {(2x ^ {2}) ^ {n}}} right] = { frac {e ^ {- x ^ {2}}} {x { sqrt { pi}}}} sum _ {n = 0} ^ { infty} ( -1) ^ {n} { frac {(2n-1) !!} {(2x ^ {2}) ^ {n}}},}

где (2n — 1) !! — это двойной факториал числа (2n — 1), которое является произведением всех нечетных чисел до (2n — 1). Этот ряд расходуется для любого конечного x, и его значение как асимптотического разложения состоит в том, что для любого N ∈ N { displaystyle N in mathbb {N}}имеется

erfc ⁡ (Икс) знак равно е — Икс 2 Икс π ∑ N знак равно 0 N — 1 (- 1) N (2 N — 1)! ! (2 х 2) n + RN (x) { displaystyle operatorname {erfc} (x) = { frac {e ^ {- x ^ {2}}} {x { sqrt { pi}}}} sum _ {n = 0} ^ {N-1} (- 1) ^ {n} { frac {(2n-1) !!} {(2x ^ {2}) ^ {n}}} + R_ {N} (x)}

где остаток в нотации Ландау равен

RN (x) = O (x 1 — 2 N e — x 2) { displaystyle R_ {N} ( x) = O left (x ^ {1-2N} e ^ {- x ^ {2}} right)}

при x → ∞. { displaystyle x to infty.}

Действительно, точное значение остатка равно

R N (x): = (- 1) N π 2 1 — 2 N (2 N)! N! ∫ Икс ∞ T — 2 N e — T 2 dt, { Displaystyle R_ {N} (x): = { frac {(-1) ^ {N}} { sqrt { pi}}} 2 ^ { 1-2N} { frac {(2N)!} {N!}} Int _ {x} ^ { infty} t ^ {- 2N} e ^ {- t ^ {2}} , dt,}

который легко следует по индукции, записывая

e — t 2 = — (2 t) — 1 (e — t 2) ′ { displaystyle e ^ {- t ^ {2}} = — (2t) ^ {- 1} left (e ^ {- t ^ {2}} right) ‘}

и интегрирование по частям.

Для достаточно больших значений x, только первые несколько этих асимптотических разностей необходимы, чтобы получить хорошее приближение erfc (x) (в то время как для не слишком больших значений x приведенное выше разложение Тейлора при 0 обеспечивает очень быструю сходимость).

Расширение непрерывной дроби

A Разложение непрерывной дроби дополнительные функции ошибок:

erfc ⁡ (z) = z π e — z 2 1 z 2 + a 1 1 + a 2 z 2 + a 3 1 + ⋯ am = м 2. { displaystyle operatorname {erfc} (z) = { frac {z} { sqrt { pi}}} e ^ {- z ^ {2}} { cfrac {1} {z ^ {2} + { cfrac {a_ {1}} {1 + { cfrac {a_ {2}} {z ^ {2} + { cfrac {a_ {3}} {1+) dotsb}}}}}}}} qquad a_ {m} = { frac {m} {2}}.}

Интеграл функции ошибок с функцией плотности Гаусса

∫ — ∞ ∞ erf ⁡ (ax + б) 1 2 π σ 2 е — (Икс — μ) 2 2 σ 2 dx знак равно erf ⁡ [a μ + b 1 + 2 a 2 σ 2], a, b, μ, σ ∈ R { displaystyle int _ {- infty} ^ { infty} operatorname {erf} left (ax + b right) { frac {1} { sqrt {2 pi sigma ^ {2}}}} e ^ {- { frac {(x- mu) ^ {2}} {2 sigma ^ {2}}}} , dx = operatorname {erf} left [{ frac {a mu + b } { sqrt {1 + 2a ^ {2} sigma ^ {2}}} right], qquad a, b, mu, sigma in mathbb {R}}

Факториальный ряд

• Обратное:

erfc ⁡ z = e — z 2 π z ∑ n = 0 ∞ (- 1) n Q n (z 2 + 1) n ¯ = e — z 2 π z (1 — 1 2 1 (z 2 + 1) + 1 4 1 (z 2 + 1) (z 2 + 2) — ⋯) { displaystyle { begin {align} operatorname {erfc} z = { frac {e ^ {- z ^ {2}}} {{ sqrt { pi}} , z}} sum _ {n = 0} ^ { infty} { frac {(-1) ^ {n} Q_ {n}} {{(z ^ {2} + 1)} ^ { ba r {n}}}} \ = { frac {e ^ {- z ^ {2}}} {{ sqrt { pi}} , z}} left ( 1 — { frac {1} {2}} { frac {1} {(z ^ {2} +1)}} + { frac {1} {4}} { frac {1} {(z ^ {2} +1) (z ^ {2} +2)}} — cdots right) end {align}}}

сходится для Re ⁡ (z 2)>0. { displaystyle operatorname {Re} (z ^ {2})>0.}Здесь

Q n = def 1 Γ (1/2) ∫ 0 ∞ τ (τ — 1) ⋯ ( τ — n + 1) τ — 1/2 е — τ d τ знак равно ∑ К знак равно 0 N (1 2) к ¯ s (n, k), { displaystyle Q_ {n} { stackrel { text {def}} {=}} { frac {1} { Gamma (1/2)}} int _ {0} ^ { infty} tau ( tau -1) cdots ( tau -n + 1) tau ^ {-1/2} e ^ {- tau} d tau = sum _ {k = 0} ^ {n} left ({ frac {1} {2}} right) ^ { bar {k}} s (n, k),}

zn ¯ { displaystyle z ^ { bar {n}}}обозначает возрастающий факториал, а s (n, k) { displaystyle s (n, k)}обозначает знаковое число Стирлинга первого рода.

• Представление бесконечной суммой, составляющей двойной факториал :

ERF ⁡ (Z) знак равно 2 π ∑ N знак равно 0 ∞ (- 2) N (2 N — 1)! (2 N + 1)! Z 2 N + 1 { Displaystyle OperatorName {ERF} (г) = { frac {2} { sqrt { pi}}} sum _ {n = 0} ^ { infty} { frac {( -2) ^ {n} (2n-1) !!} {(2n + 1)!}} Z ^ {2n + 1}}

Численные приближения

Приближение элементов сарными функциями

• Абрамовиц и Стегун дают несколько приближений с точностью (уравнения 7.1.25–28). Это позволяет выбрать наиболее быстрое приближение, подходящее для данного приложения. В порядке увеличения точности они следующие:

erf ⁡ (x) ≈ 1 — 1 (1 + a 1 x + a 2 x 2 + a 3 x 3 + a 4 x 4) 4, x ≥ 0 { displaystyle имя оператора {erf} (x) приблизительно 1 — { frac {1} {(1 + a_ {1} x + a_ {2} x ^ {2} + a_ {3} x ^ {3} + a_ { 4} x ^ {4}) ^ {4}}}, qquad x geq 0}

(максимальная ошибка: 5 × 10)

, где a 1 = 0,278393, a 2 = 0,230389, a 3 = 0,000972, a 4 = 0,078108

erf ⁡ (x) ≈ 1 — (a 1 t + a 2 t 2 + a 3 t 3) e — x 2, t = 1 1 + px, x ≥ 0 { displaystyle operatorname {erf} (x) приблизительно 1- (a_ {1} t + a_ {2} t ^ {2} + a_ {3} t ^ {3}) e ^ {- x ^ {2}}, quad t = { frac {1} {1 + px}}, qquad x geq 0}(максимальная ошибка: 2,5 × 10)

где p = 0,47047, a 1 = 0,3480242, a 2 = -0,0958798, a 3 = 0,7478556

erf ⁡ (x) ≈ 1 — 1 (1 + a 1 x + a 2 x 2 + ⋯ + a 6 x 6) 16, x ≥ 0 { displaystyle operatorname {erf} (x) приблизительно 1 — { frac {1} {(1 + a_ {1} x + a _ {2} x ^ {2} + cdots + a_ {6} x ^ {6}) ^ {16}}}, qquad x geq 0}(максимальная ошибка: 3 × 10)

, где a 1 = 0,0705230784, a 2 = 0,0422820123, a 3 = 0,0092705272, a 4 = 0,0001520143, a 5 = 0,0002765672, a 6 = 0,0000430638

erf ⁡ (x) ≈ 1 — (a 1 t + a 2 t 2 + ⋯ + a 5 t 5) e — x 2, t = 1 1 + px { displaystyle operatorname {erf} (x) приблизительно 1- (a_ {1} t + a_ {2} t ^ {2} + cdots + a_ {5} t ^ {5}) e ^ {- x ^ {2}}, quad t = { frac {1} {1 + px}}}(максимальная ошибка: 1,5 × 10)

, где p = 0,3275911, a 1 = 0,254829592, a 2 = −0,284496736, a 3 = 1,421413741, a 4 = −1,453152027, a 5 = 1,061405429

Все эти приближения действительны для x ≥ 0 Чтобы использовать эти приближения для отрицательного x, викорируйте тот факт, что erf (x) — нечетная функция, поэтому erf (x) = −erf (−x).

• Экспоненциальные границы и чисто экспоненциальное приближение для дополнительных функций задаются как

erfc ⁡ (x) ≤ 1 2 e — 2 x 2 + 1 2 e — x 2 ≤ e — x 2, x>0 erfc ⁡ ( х) ≈ 1 6 е — х 2 + 1 2 е — 4 3 х 2, х>0. { displaystyle { begin {align} operatorname {erfc} (x) leq { frac {1} {2}} e ^ {- 2x ^ {2}} + { frac {1} {2} } e ^ {- x ^ {2}} leq e ^ {- x ^ {2}}, qquad x>0 \ имя оператора {erfc} (x) приблизительно { frac {1} { 6}} e ^ {- x ^ {2}} + { frac {1} {2}} e ^ {- { frac {4} {3}} x ^ {2}}, qquad x>0. end {align}}}

erfc ⁡ (x) ≈ (1 — e — A x) e — x 2 B π х. { displaystyle operatorname {erfc} left (x right) приблизительно { frac { left (1-e ^ {- Ax} right) e ^ {- x ^ {2}}} {B { sqrt { pi}} x}}.}

Они определили {A, B} = {1.98, 1.135}, { displaystyle {A, B } = {1.98,1.135 },}, что дает хорошее приближение для всех x ≥ 0. { displaystyle x geq 0.}

• Одноканальная нижняя граница:

erfc ⁡ (x) ≥ 2 e π β — 1 β е — β Икс 2, Икс ≥ 0, β>1, { Displaystyle OperatorName {erfc} (x) geq { sqrt { frac {2e} { pi}}} { frac { sqrt { beta -1}} { beta}} e ^ {- beta x ^ {2}}, qquad x geq 0, beta>1,}

где параметр β может быть выбран, чтобы минимизировать ошибку на желаемом интервале приближения.