Exponential Integrals

AUTHORS:

  • Benjamin Jones (2011-06-12)

This module provides easy access to many exponential integral special functions. It utilizes Maxima’s special functions package and the mpmath library.

REFERENCES:

[AS](1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13) ‘Handbook of Mathematical Functions’, Milton Abramowitz and Irene A. Stegun, National Bureau of Standards Applied Mathematics Series, 55. See also http://www.math.sfu.ca/~cbm/aands/.

AUTHORS:

  • Benjamin Jones

    Implementations of the classes Function_exp_integral_*.

  • David Joyner and William Stein

    Authors of the code which was moved from special.py and trans.py. Implementation of exp_int() (from sage/functions/special.py). Implementation of exponential_integral_1() (from sage/functions/transcendental.py).

class sage.functions.exp_integral.Function_cos_integral

Bases: sage.symbolic.function.BuiltinFunction

The trigonometric integral \(\operatorname{Ci}(z)\) defined by

\[\operatorname{Ci}(z) = \gamma + \log(z) + \int_0^z \frac{\cos(t)-1}{t}\; dt,\]

where \(\gamma\) is the Euler gamma constant (euler_gamma in Sage), see [AS] 5.2.1.

EXAMPLES:

Numerical evaluation for real and complex arguments is handled using mpmath:

sage: cos_integral(3.0)
0.119629786008000

The alias \(Ci\) can be used instead of \(cos_integral\):

sage: Ci(3.0)
0.119629786008000

Compare cos_integral(3.0) to the definition of the value using numerical integration:

sage: N(euler_gamma + log(3.0) + integrate((cos(x)-1)/x, x, 0, 3.0) - cos_integral(3.0)) < 1e-14
True

Arbitrary precision and complex arguments are handled:

sage: N(cos_integral(3), digits=30)
0.119629786008000327626472281177
sage: cos_integral(ComplexField(100)(3+I))
0.078134230477495714401983633057 - 0.37814733904787920181190368789*I

The limit \(\operatorname{Ci}(z)\) as \(z \to \infty\) is zero:

sage: N(cos_integral(1e23))
-3.24053937643003e-24

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = cos_integral(x)
sage: f.diff(x)
cos(x)/x

sage: f.integrate(x)
x*cos_integral(x) - sin(x)

The Nielsen spiral is the parametric plot of (Si(t), Ci(t)):

sage: t=var('t')
sage: f(t) = sin_integral(t)
sage: g(t) = cos_integral(t)
sage: P = parametric_plot([f, g], (t, 0.5 ,20))
sage: show(P, frame=True, axes=False)

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

REFERENCES:

class sage.functions.exp_integral.Function_cosh_integral

Bases: sage.symbolic.function.BuiltinFunction

The trigonometric integral \(\operatorname{Chi}(z)\) defined by

\[\operatorname{Chi}(z) = \gamma + \log(z) + \int_0^z \frac{\cosh(t)-1}{t}\; dt,\]

see [AS] 5.2.4.

EXAMPLES:

Numerical evaluation for real and complex arguments is handled using mpmath:

sage: cosh_integral(1.0)
0.837866940980208

The alias \(Chi\) can be used instead of \(cosh_integral\):

sage: Chi(1.0)
0.837866940980208

Here is an example from the mpmath documentation:

sage: f(x) = cosh_integral(x)
sage: find_root(f, 0.1, 1.0)
0.5238225713894826

Compare cosh_integral(3.0) to the definition of the value using numerical integration:

sage: N(euler_gamma + log(3.0) + integrate((cosh(x)-1)/x, x, 0, 3.0) -
...     cosh_integral(3.0)) < 1e-14
True

Arbitrary precision and complex arguments are handled:

sage: N(cosh_integral(3), digits=30)
4.96039209476560976029791763669
sage: cosh_integral(ComplexField(100)(3+I))
3.9096723099686417127843516794 + 3.0547519627014217273323873274*I

The limit of \(\operatorname{Chi}(z)\) as \(z \to \infty\) is \(\infty\):

sage: N(cosh_integral(Infinity))
+infinity

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = cosh_integral(x)
sage: f.diff(x)
cosh(x)/x

sage: f.integrate(x)
x*cosh_integral(x) - sinh(x)

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

REFERENCES:

class sage.functions.exp_integral.Function_exp_integral

Bases: sage.symbolic.function.BuiltinFunction

The generalized complex exponential integral Ei(z) defined by

\[\operatorname{Ei}(x) = \int_{-\infty}^x \frac{e^t}{t}\; dt\]

for x > 0 and for complex arguments by analytic continuation, see [AS] 5.1.2.

EXAMPLES:

sage: Ei(10)
Ei(10)
sage: Ei(I)
Ei(I)
sage: Ei(3+I)
Ei(I + 3)
sage: Ei(1.3)
2.72139888023202

The branch cut for this function is along the negative real axis:

sage: Ei(-3 + 0.1*I)
-0.0129379427181693 + 3.13993830250942*I
sage: Ei(-3 - 0.1*I)
-0.0129379427181693 - 3.13993830250942*I

ALGORITHM: Uses mpmath.

TESTS:

Show that the evaluation and limit issue in trac ticket #13271 is fixed:

sage: var('Z')
Z
sage: (Ei(-Z)).limit(Z=oo)
0
sage: (Ei(-Z)).limit(Z=1000)
Ei(-1000)
sage: (Ei(-Z)).limit(Z=1000).n()
-5.07089306023517e-438
class sage.functions.exp_integral.Function_exp_integral_e

Bases: sage.symbolic.function.BuiltinFunction

The generalized complex exponential integral \(E_n(z)\) defined by

\[\operatorname{E_n}(z) = \int_1^{\infty} \frac{e^{-z t}}{t^n} \; dt\]

for complex numbers \(n\) and \(z\), see [AS] 5.1.4.

The special case where \(n = 1\) is denoted in Sage by exp_integral_e1.

EXAMPLES:

Numerical evaluation is handled using mpmath:

sage: N(exp_integral_e(1,1))
0.219383934395520
sage: exp_integral_e(1, RealField(100)(1))
0.21938393439552027367716377546

We can compare this to PARI’s evaluation of exponential_integral_1():

sage: N(exponential_integral_1(1))
0.219383934395520

We can verify one case of [AS] 5.1.45, i.e. \(E_n(z) = z^{n-1}\Gamma(1-n,z)\):

sage: N(exp_integral_e(2, 3+I))
0.00354575823814662 - 0.00973200528288687*I
sage: N((3+I)*gamma(-1, 3+I))
0.00354575823814662 - 0.00973200528288687*I

Maxima returns the following improper integral as a multiple of exp_integral_e(1,1):

sage: uu = integral(e^(-x)*log(x+1),x,0,oo)
sage: uu
e*exp_integral_e(1, 1)
sage: uu.n(digits=30)
0.596347362323194074341078499369

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = exp_integral_e(2,x)
sage: f.diff(x)
-exp_integral_e(1, x)

sage: f.integrate(x)
-exp_integral_e(3, x)

sage: f = exp_integral_e(-1,x)
sage: f.integrate(x)
Ei(-x) - gamma(-1, x)

Some special values of exp_integral_e can be simplified. [AS] 5.1.23:

sage: exp_integral_e(0,x)
e^(-x)/x

[AS] 5.1.24:

sage: exp_integral_e(6,0)
1/5
sage: nn = var('nn')
sage: assume(nn > 1)
sage: f = exp_integral_e(nn,0)
sage: f.simplify()
1/(nn - 1)

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

class sage.functions.exp_integral.Function_exp_integral_e1

Bases: sage.symbolic.function.BuiltinFunction

The generalized complex exponential integral \(E_1(z)\) defined by

\[\operatorname{E_1}(z) = \int_z^\infty \frac{e^{-t}}{t}\; dt\]

see [AS] 5.1.4.

EXAMPLES:

Numerical evaluation is handled using mpmath:

sage: N(exp_integral_e1(1))
0.219383934395520
sage: exp_integral_e1(RealField(100)(1))
0.21938393439552027367716377546

We can compare this to PARI’s evaluation of exponential_integral_1():

sage: N(exp_integral_e1(2.0))
0.0489005107080611
sage: N(exponential_integral_1(2.0))
0.0489005107080611

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = exp_integral_e1(x)
sage: f.diff(x)
-e^(-x)/x

sage: f.integrate(x)
-exp_integral_e(2, x)

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

class sage.functions.exp_integral.Function_log_integral

Bases: sage.symbolic.function.BuiltinFunction

The logarithmic integral \(\operatorname{li}(z)\) defined by

\[\operatorname{li}(x) = \int_0^z \frac{dt}{\ln(t)} = \operatorname{Ei}(\ln(x))\]

for x > 1 and by analytic continuation for complex arguments z (see [AS] 5.1.3).

EXAMPLES:

Numerical evaluation for real and complex arguments is handled using mpmath:

sage: N(log_integral(3))
2.16358859466719
sage: N(log_integral(3), digits=30)
2.16358859466719197287692236735
sage: log_integral(ComplexField(100)(3+I))
2.2879892769816826157078450911 + 0.87232935488528370139883806779*I
sage: log_integral(0)
0

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = log_integral(x)
sage: f.diff(x)
1/log(x)

sage: f.integrate(x)
x*log_integral(x) - Ei(2*log(x))

Here is a test from the mpmath documentation. There are 1,925,320,391,606,803,968,923 many prime numbers less than 1e23. The value of log_integral(1e23) is very close to this:

sage: log_integral(1e23)
1.92532039161405e21

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

REFERENCES:

class sage.functions.exp_integral.Function_log_integral_offset

Bases: sage.symbolic.function.BuiltinFunction

The offset logarithmic integral, or Eulerian logarithmic integral, \(\operatorname{Li}(x)\) is defined by

\[\operatorname{Li}(x) = \int_2^x \frac{dt}{ln(t)} = \operatorname{li}(x)-\operatorname{li}(2)\]

for \(x \ge 2\).

The offset logarithmic integral should also not be confused with the polylogarithm (also denoted by \(\operatorname{Li}(x)\) ), which is implemented as sage.functions.log.Function_polylog.

\(\operatorname{Li}(x)\) is identical to \(\operatorname{li}(x)\) except that the lower limit of integration is \(2\) rather than \(0\) to avoid the singularity at \(x = 1\) of

\[\frac{1}{ln(t)}\]

See Function_log_integral for details of \(\operatorname{li}(x)\). Thus \(\operatorname{Li}(x)\) can also be represented by

\[\operatorname{Li}(x) = \operatorname{li}(x)-\operatorname{li}(2)\]

So we have:

sage: li(4.5)-li(2.0)-Li(4.5)
0.000000000000000

\(\operatorname{Li}(x)\) is extended to complex arguments \(z\) by analytic continuation (see [AS] 5.1.3):

sage: Li(6.6+5.4*I)
3.97032201503632 + 2.62311237593572*I

The function \(\operatorname{Li}\) is an approximation for the number of primes up to \(x\). In fact, the famous Riemann Hypothesis is

\[|\pi(x) - \operatorname{Li}(x)| \leq \sqrt{x} \log(x).\]

For “small” \(x\), \(\operatorname{Li}(x)\) is always slightly bigger than \(\pi(x)\). However it is a theorem that there are very large values of \(x\) (e.g., around \(10^{316}\)), such that \(\exists x: \pi(x) > \operatorname{Li}(x)\). See “A new bound for the smallest x with \(\pi(x) > \operatorname{li}(x)\)”, Bays and Hudson, Mathematics of Computation, 69 (2000) 1285-1296.

Here is a test from the mpmath documentation. There are 1,925,320,391,606,803,968,923 prime numbers less than 1e23. The value of log_integral_offset(1e23) is very close to this:

sage: log_integral_offset(1e23)
1.92532039161405e21

EXAMPLES:

Numerical evaluation for real and complex arguments is handled using mpmath:

sage: N(log_integral_offset(3))
1.11842481454970
sage: N(log_integral_offset(3), digits=30)
1.11842481454969918803233347815
sage: log_integral_offset(ComplexField(100)(3+I))
1.2428254968641898308632562019 + 0.87232935488528370139883806779*I
sage: log_integral_offset(2)
0
sage: for n in range(1,7):
...    print '%-10s%-10s%-20s'%(10^n, prime_pi(10^n), N(Li(10^n)))
10        4         5.12043572466980
100       25        29.0809778039621
1000      168       176.564494210035
10000     1229      1245.09205211927
100000    9592      9628.76383727068
1000000   78498     78626.5039956821

Symbolic derivatives are handled by Sage and integration by Maxima:

sage: x = var('x')
sage: f = log_integral_offset(x)
sage: f.diff(x)
1/log(x)

sage: f.integrate(x)
-x*log_integral(2) + x*log_integral(x) - Ei(2*log(x))

sage: Li(x).integrate(x,2.0,4.5)
-2.5*log_integral(2) + 5.79932114741

sage: N(f.integrate(x,2.0,3.0))
0.601621785860587
Note: Definite integration returns a part symbolic and part
numerical result. This is because when Li(x) is evaluated it is passed as li(x)-li(2).

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

REFERENCES:

class sage.functions.exp_integral.Function_sin_integral

Bases: sage.symbolic.function.BuiltinFunction

The trigonometric integral \(\operatorname{Si}(z)\) defined by

\[\operatorname{Si}(z) = \int_0^z \frac{\sin(t)}{t}\; dt,\]

see [AS] 5.2.1.

EXAMPLES:

Numerical evaluation for real and complex arguments is handled using mpmath:

sage: sin_integral(0)
0
sage: sin_integral(0.0)
0.000000000000000
sage: sin_integral(3.0)
1.84865252799947
sage: N(sin_integral(3), digits=30)
1.84865252799946825639773025111
sage: sin_integral(ComplexField(100)(3+I))
2.0277151656451253616038525998 + 0.015210926166954211913653130271*I

The alias \(Si\) can be used instead of \(sin_integral\):

sage: Si(3.0)
1.84865252799947

The limit of \(\operatorname{Si}(z)\) as \(z \to \infty\) is \(\pi/2\):

sage: N(sin_integral(1e23))
1.57079632679490
sage: N(pi/2)
1.57079632679490

At 200 bits of precision \(\operatorname{Si}(10^{23})\) agrees with \(\pi/2\) up to \(10^{-24}\):

sage: sin_integral(RealField(200)(1e23))
1.5707963267948966192313288218697837425815368604836679189519
sage: N(pi/2, prec=200)
1.5707963267948966192313216916397514420985846996875529104875

The exponential sine integral is analytic everywhere:

sage: sin_integral(-1.0)
-0.946083070367183
sage: sin_integral(-2.0)
-1.60541297680269
sage: sin_integral(-1e23)
-1.57079632679490

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = sin_integral(x)
sage: f.diff(x)
sin(x)/x

sage: f.integrate(x)
x*sin_integral(x) + cos(x)

sage: integrate(sin(x)/x, x)
-1/2*I*Ei(I*x) + 1/2*I*Ei(-I*x)

Compare values of the functions \(\operatorname{Si}(x)\) and \(f(x) = (1/2)i \cdot \operatorname{Ei}(-ix) - (1/2)i \cdot \operatorname{Ei}(ix) - \pi/2\), which are both anti-derivatives of \(\sin(x)/x\), at some random positive real numbers:

sage: f(x) = 1/2*I*Ei(-I*x) - 1/2*I*Ei(I*x) - pi/2
sage: g(x) = sin_integral(x)
sage: R = [ abs(RDF.random_element()) for i in range(100) ]
sage: all(abs(f(x) - g(x)) < 1e-10 for x in R)
True

The Nielsen spiral is the parametric plot of (Si(t), Ci(t)):

sage: x=var('x')
sage: f(x) = sin_integral(x)
sage: g(x) = cos_integral(x)
sage: P = parametric_plot([f, g], (x, 0.5 ,20))
sage: show(P, frame=True, axes=False)

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

REFERENCES:

class sage.functions.exp_integral.Function_sinh_integral

Bases: sage.symbolic.function.BuiltinFunction

The trigonometric integral \(\operatorname{Shi}(z)\) defined by

\[\operatorname{Shi}(z) = \int_0^z \frac{\sinh(t)}{t}\; dt,\]

see [AS] 5.2.3.

EXAMPLES:

Numerical evaluation for real and complex arguments is handled using mpmath:

sage: sinh_integral(3.0)
4.97344047585981
sage: sinh_integral(1.0)
1.05725087537573
sage: sinh_integral(-1.0)
-1.05725087537573

The alias \(Shi\) can be used instead of \(sinh_integral\):

sage: Shi(3.0)
4.97344047585981

Compare sinh_integral(3.0) to the definition of the value using numerical integration:

sage: N(integrate((sinh(x))/x, x, 0, 3.0) - sinh_integral(3.0)) < 1e-14
True

Arbitrary precision and complex arguments are handled:

sage: N(sinh_integral(3), digits=30)
4.97344047585980679771041838252
sage: sinh_integral(ComplexField(100)(3+I))
3.9134623660329374406788354078 + 3.0427678212908839256360163759*I

The limit \(\operatorname{Shi}(z)\) as \(z \to \infty\) is \(\infty\):

sage: N(sinh_integral(Infinity))
+infinity

Symbolic derivatives and integrals are handled by Sage and Maxima:

sage: x = var('x')
sage: f = sinh_integral(x)
sage: f.diff(x)
sinh(x)/x

sage: f.integrate(x)
x*sinh_integral(x) - cosh(x)

Note that due to some problems with the way Maxima handles these expressions, definite integrals can sometimes give unexpected results (typically when using inexact endpoints) due to inconsistent branching:

sage: integrate(sinh_integral(x), x, 0, 1/2)
-cosh(1/2) + 1/2*sinh_integral(1/2) + 1
sage: integrate(sinh_integral(x), x, 0, 1/2).n() # correct
0.125872409703453
sage: integrate(sinh_integral(x), x, 0, 0.5).n() # fixed in maxima 5.29.1
0.125872409703453

ALGORITHM:

Numerical evaluation is handled using mpmath, but symbolics are handled by Sage and Maxima.

REFERENCES:

sage.functions.exp_integral.exponential_integral_1(x, n=0)

Returns the exponential integral \(E_1(x)\). If the optional argument \(n\) is given, computes list of the first \(n\) values of the exponential integral \(E_1(x m)\).

The exponential integral \(E_1(x)\) is

\[E_1(x) = \int_{x}^{\infty} e^{-t}/t dt\]

INPUT:

  • x – a positive real number
  • n – (default: 0) a nonnegative integer; if nonzero, then return a list of values E_1(x*m) for m = 1,2,3,...,n. This is useful, e.g., when computing derivatives of L-functions.

OUTPUT:

A real number if n is 0 (the default) or a list of reals if n > 0. The precision is the same as the input, with a default of 53 bits in case the input is exact.

EXAMPLES:

sage: exponential_integral_1(2)
0.0489005107080611
sage: exponential_integral_1(2,4)  # abs tol 1e-18
[0.0489005107080611, 0.00377935240984891, 0.000360082452162659, 0.0000376656228439245]
sage: exponential_integral_1(40,5)
[1.03677326145166e-19, 2.22854325868847e-37, 6.33732515501151e-55, 2.02336191509997e-72, 6.88522610630764e-90]
sage: exponential_integral_1(0)
+Infinity
sage: r = exponential_integral_1(RealField(150)(1))
sage: r
0.21938393439552027367716377546012164903104729
sage: parent(r)
Real Field with 150 bits of precision
sage: exponential_integral_1(RealField(150)(100))
3.6835977616820321802351926205081189876552201e-46

TESTS:

The relative error for a single value should be less than 1 ulp:

sage: for prec in [20..1000]:  # long time (22s on sage.math, 2013)
....:     R = RealField(prec)
....:     S = RealField(prec+64)
....:     for t in range(8):  # Try 8 values for each precision
....:         a = R.random_element(-15,10).exp()
....:         x = exponential_integral_1(a)
....:         y = exponential_integral_1(S(a))
....:         e = float(abs(S(x) - y)/x.ulp())
....:         if e >= 1.0:
....:             print "exponential_integral_1(%s) with precision %s has error of %s ulp"%(a, prec, e)

The absolute error for a vector should be less than \(c 2^{-p}\), where \(p\) is the precision in bits of \(x\) and \(c = 2 max(1, exponential_integral_1(x))\):

sage: for prec in [20..128]:  # long time (15s on sage.math, 2013)
....:     R = RealField(prec)
....:     S = RealField(prec+64)
....:     a = R.random_element(-15,10).exp()
....:     n = 2^ZZ.random_element(14)
....:     x = exponential_integral_1(a, n)
....:     y = exponential_integral_1(S(a), n)
....:     c = RDF(2 * max(1.0, y[0]))
....:     for i in range(n):
....:         e = float(abs(S(x[i]) - y[i]) << prec)
....:         if e >= c:
....:             print "exponential_integral_1(%s, %s)[%s] with precision %s has error of %s >= %s"%(a, n, i, prec, e, c)

ALGORITHM: use the PARI C-library function eint1.

REFERENCE:

  • See Proposition 5.6.12 of Cohen’s book “A Course in Computational Algebraic Number Theory”.

Previous topic

Bessel Functions

Next topic

Wigner, Clebsch-Gordan, Racah, and Gaunt coefficients

This Page