Академический Документы
Профессиональный Документы
Культура Документы
Contents
Page
Mathematical Properties . . . . . . . . . . . . . . . . . . . . 297
7.1. Error Function. . . . . . . . . . . . . . . . . . . . . . 297
7.2. Repeated Integrals of the Error Function . . . . . . . . . . 299
7.3. Fresnel Integrals . . . . . . . . . . . . . . . . . . . . . 300
7.4. Definite and Indefinite Integrals . . . . . . . . . . . . . . 302
Numerical Methods . . . . . . . . . . . . . . . . . . . . . . 304
7.5. Use and Extension of the Tables . . . . . . . . . . . . . . 304
References . . . . . . . . . . . . . . . . . . . . . . . . . . 308
Table 7.1. Error Function and its Derivative (0 5 ~ 5 2 ) . . . . . . 310
(2/&)e-', 2=2(.01)10, 8s
Table 7.3. Complementary Error Function ( 2 5 2 5 m ) . . . . . . . 316
e - ~ ~ e ~ 2 d2=0(.02)
t, 2, IOD
ze-I1Jet2dt, ~-~=.25(-.005)0, 9D
0
1 Guest worker, National Bureau of Standards, from The American University. (Presently
Purdue University.)
296
296 ERROR FUNCTION AND FRESNEL INTEGRALS
Page
Table 7.6. (3/I'(l/3))J2e-t3dt ( 0 5 ~ 5 2 . 3 ) . . . . . . . . . . . . . 320
0
Z= 0 (.02)1.7(.M)
2.3, 7D
Table 7.7. Fresnel Integrals (02255). . . . . . . . . . . . . . . 321
C ( Z ) = cos
~ (it2) dt, S ( z ) = l s i n c t2)dt, ~=0(.02)5, 7D
Table 7.8. Auxiliary Functions (0225a) . . . . . . . . . . . . . 323
Z= 0 (.02)1 , ~ '=
- 1 (- .02)0, 15D
Table 7.9. Error Function for Complex Argumentg (05z23.9,05y<3) . 325
w(z)=e-e2 erfc (-iz), z=z+iy, ~=0(.1)3.9,y=0(.1)3, 6D
Table 7.10. Complex Zeros of the Error Function (l<n510) . . . . 329
zn,erf z,=O,n=1(1)10, 8D
Table 7.11. Complex Zeros of Fresnel Integrals (O<n<5). . . . . 329
z,, z:, C(z,)=O, 8(~:)=O,n=0(1)5, 4D
Table 7.12. Maxima and Minima of Fresnel Integrals ( O s n s 5 ) . . . 329
Definitions
2
7.1.5 erf z=-x
-
(-l)nz2n+l
r G+i)
In 7.1.2 the path of integration is subject to the
restriction arg t+a with la/<T as t+m along the For In-+(z),
see chapter 10.
4
path. (a=? is permissible if WPremains bounded Symmetry Relations
4
to the left.) 7.1.9 erf (-z)=-erf z
- -
Integral Representation 7.1.10 erf z=erf z
7.1.4
7.1.11 w(--2)=2e-z2-w(z)
w (z)=-
i (D -=-J
e-"dt 2iz o e-"dt
-
r S-m 2-i! a z2-t2 (yz>o) 7.1.12 w(7)=w (- 2)
FIQUBE
7 . 1y=ezsLm
. e-"&. FIQUBB
7.2. y=e-"~e"&.
298 ERROR FUNCTION AND FRESNEL INTEGRALS
7.1.15
ss_*
1 e-'2&_------...
OD 1 1/2 1 3/2 2
2--t 2-2-2-2-2-
1 n Hp)
--
-6 lim C-
n+m k P 1 z-xp)
(9-2# 0)
F(z)==e-$ eg2dt l
7.1.17 F(.92413 88730. . . )=.54104 42246. . .
7.1.18 F(1.50197 52682 . . . )=.42768 66160 . . .
Derivatives
7.1.19
dR+l 2
erfz=(-1)" -H,,(z)e-2 (n=O, 1,2, . . .)
dz"+l fi
7.1.20
(2) +22W'"+"
w(*+2) (2) +2 (n+ 1)w'R' (2) =o
(n=O, 1,2, . . .)
FIQUBE7.3. Altitude Chart of la(~). w(0)(2) =w (2) , w' (2) = -2 2 w(2) +-2i
fi
(For the Hermite polynomials HJz) see chapter
22.)
Inequalities [7.11], [7.17]
7.1.13 Relation to Confluent Hypergeometric Function (see
chapter 13)
1 1 7.1.21
x+pT2<eT
e-"dt 5
X+dZ 1
(22.0)
erf z = e M
+ (i, %,-z2)=$ e-"M 1
( 9 :
-9 z2
)
The Normal Distribution Function With Mean m and
(For other inequalities see [7.2].) Standard Deviation u (see chapter 26)
(t-m)2
1
7.1.22 - L e -- 202 d t = a ( l + e r f ( s ) )
Continued Fractions adg S-m
7.1.14 Aeymptotic Expansion
R"(2)= (- 1)
1 . 3 . . . (2n-1)
0,
erf (x+iy)=erf x+agz [(I-cos 2xy)+i sin 2xy]
(222)"
-n-+
0'Ce-l (l+$) dt (la- zl<;)
where
jn(x,y) =2x-2x cosh ny cos 2xy4-n sinh ny sin 2xy
For x real, Rn(x) is less in absolute value than gn(x, y)=2x cosh ny sin 2xy+n sinh ny cos 2xy
the first neglected term and of the same sign. Ie(x, y)I =10-161erf (x+iy)l
7.2. Repeated Integrals of the Error Function
Rational Approximations * (0 z
<< m)
7.1.25 Definition
1 7.2.1
erf x=1-(alt+azt2+a#) e-Z'+e(x), t=-
1 +PX inerfcz= i"-'erfct dt (n=0,1,2, . . . )
L
O D
l(X)112.5XlO--~
2
p = .47047 al= .34802 42 az= -.09587 98 i-1 erfc z=- e-z2, io erfc z=erfc z
a3=.74785 56 &
7.1.26 Differential Equation
erf x= 1- (alt+a2t2+a3ts+a4t4+a6t6~e-~'+e(x)
,
7.2.2
(t-2)"
7.2.3 i" erfc
7.1.27
1 Power Seriea a
erf x=1-
1
[I +a1x+az2+a~23+a~x
4 4 + 4 4
((X)115x10-4
al= .278393 ~=.230389
as= .000972 a*=.078108 Recurrence Relations
7.1.28 7.2.5
1 Z 1
erf x= 1 - in erfc z=-- n in-1erfc z+- 2n in-2 erfc z
[l+alz+az22+ . . +a6z6I ,6+44
l(X) I 5 3 x lo-' (n=1, 2,3, . . . )
al= .07052 30784 az= .04228 20123 7.2.6
a3=.00927 05272 a4=.00015 20143
2(n+l)(n+2)i"+2erfc z
a6= .00027 65672 a(= .00004 30638 1
= (2n+l +2zz)i"erfc z - ~in-2erfc z
2 Approximations1.1.25-7.1.28 are from C. Hastings, Jr., (n=l, 2,3, . . .>
Approximations for digital computers. Princeton Univ.
Press, Princeton, N. J., 1955 (with permission). * The terms in this series corresponding to k=n+2,
n+4, n+6, . . . are understood to be zero.
300 ERROR FUNCTION AND FRESNEL INTEGRALS
[ l
- Zn+l M6+1,;,z2)]
2n-117 ( T )
Asymptotic Expansion
7.2.14
in erfc Z--
2 e-$ 2 (-1)m(2m+n)!
+ (2z)n+'m-0
~
n!m!(2z)h
=s,' 6
I
G x2) 7.3.14
7.3.8 S(x)=S1
(x&)=S2
S(z)=-cos 6 22) 51 3 .
n-0
(-l)n,p+1
. . (4n+3)
z4n+3
6 6
*
1
7.3.9 C(z)=2+f(z) sin z')-g(z)
(">
cos 3 22 +sin 22) 5 1 3(- 1)
n=o *
n7r2n
. . . (4n+l)
z4n+l
1
7.3.10 S(z)=--j(z)
2
cos 6 z2)-g(z) sin ("z 2')
+J&)
793.15 c2(z>=Jl/z(z)+J5,2(~)
7-3-16 S~(Z)=J~,~(Z)+J,/~(Z)SJ~~,
...
+...
- '1
7'3*11 c(z)=z
(- "(7r/2)
(2n)!(4n+l)
2n 24n+ 1
Symmetry Relations
7.3.17 C(-Z) =-C(Z) , S(-Z)=-S(Z)
7.3.12 7.3.18 C(iz)=iC(~), S(iz)=-iS(~)
C(2) =cos 6 22) 51
n-0 *
(-l)n?P
3 . . . (4~~4-1)
z4n+l 7.3.19 C(z>=C?Zs; S ( z > = S 3
+sin G 9s (- 1)n p + 1
1 . 3 . . . (4n+3)
24n+a
7.3.20
1
C(X)+y
Value at Infinity
S(z)-'i
1
@-+a)
dj(4
Derivatives
-1
dgo=nxj(x)
V
7.3.21 dz --rxg(z), dx
7.3.23 g(x)
=L% F w [$ ,l+ilx]}
7.3.24 f ( x ) = 4 { q w [g (l+i)z]}
v 3; >
-t -t
2
=zea M l,~,-i-zz
FIQURB
7.5. Fresnel Integrals.
v-CGr), v-s(z)
1-3-26 C~(Z)
-2 -y J-+(t)dt,S2(z)=;J' 0 J+(t)dt
302 ERROR FUNCTION AND FRESNEL INTEGRALS
7.4.3
7.3.28
1 - 3 . . . (4mfl)
rzg(2)- 2 (-1)m
m -0
7.4.4
Lm . . ( 2 n - l ) d-
tae-a'dt= 1 . 3 .2n+lan
If R$f)(z),RAg)(z) are theremaindersaftern terms
in 7.3.27, 7.3.28, respectively, then =-r(n+3) (9u>O; n=o, 1,2, . . .)
2U"++
7.3.29 7.4.5
1 3 . . . (4n-1)
B y (2) = (- 1)
*
( T Z y
e('),
7.4.6
1-
7.4.7
e-&* sin (2zt)dt=J1
U
e-r2/afk eWt
7.3.33
.L 1
g(x)=2+4.142z+3.492x2+6.67023+
Ic (z)I52 x 10-3
(For more accurate approximations see [7.1].)
7.4. Definite and Indefinite Integrals
For a more extensive list of integrals see [7.5],
17 81, [7.15].
eval2dt r
7.4.1 7.4.12 -=;I ea[1-(erf 4iV] (a>O)
7.4.18
1 OD
7.4.19
7.4.20
7.4.21
7.4.23
304 ERROR FUNCTION AND FRESNEL INTEGRALS
7.4.33 7.4.38
Jcos (az2+2bx+c)dx
b
+w (---+iax)]+const. (a#o>
7.4.39
1
7.4.35 Serf xdx=x erf x+- e-zz+const. ssin (a2+2bx+c)dx
fi
7.4.36
=& {cos S [-;(ux+b)]
(F)
7.4.37
+const. (azo) -sin ( e c[g
)
(az+b)]}+const.
+const. (a#O)
7.4.41 SS(x)dx=xS(x)+A cos 7r 6 x2)+const.
Numerical Methods
7.5. Use and Extension of the Tables erf .745=.70467 80779+ (.5)(.00652 58247)X
[l -(.005)(.74)+ (.OOOOO 83333)(.0952)]
Example 1. Compute erf .745 and e-(*746)2
using Taylor's series. =.70792 8920
With the aid of Taylor's theorem and 7.1.19
it can be shown that e-(.746)1=-?! 4- (.65258 24665) [1-.0074
2
erf (xo+ph)=erf xo + (.000025)(.0952)+ (.OOOOO 00833)(.74)(1.9048)]
+s2 e-dph [ 1 - p h ~ ~ +1~ p ~ h ~ ( 2 g - 1 ) ] + ~ = .57405 7910.
As a check the computation was repeated with
q=.75, p= -.5.
Example 2. Compute erfc x to 5 s for x=4.8.
We have 1/22=.0434028. With Table 7.2 and
linear interpolation in Table 7.3, we obtain
u &(U)
6.20310 68 .432sooB . W 9 42
6.31808 80 ,41673 97 - - . O W80 ,42732 63
6.08938 01 .45083 88 .1M)61 99 691 63 ,42718 63
With x=1.72, m=15 we obtain ~ 4 3 4 3 270 . 3 m M - . i ~ 270 750 go 8 52 .an7 71
4.97881 11 .48860 94 .26308 89 674 79 9 39 61 .an7 67
&(5.24)= .427177
Example 9. Compute S2(5.24) using Taylors
17 0 12 (3) 2.1011 7 (n 2.6879 2 (11) 1.2820 series and Table 7.8.
16 1 11 (4) 1.3831 6 (8) 1.6669 1
:% Using 7.3.21 we can write Taylorsseries forfz (u)
(4:) (4:)
16 3.44 10 9:8006 6 (8) 8.9787 0
14 (1) 4.3834 9 6) 6.4143 4 (9) 4.9170 -1 13) 1.0087
13 (2) 26399 8 6) 4 1686 3 (10) 28031
=f and g2(a)=g in the form
306 ERROR FUNClION AND FRESNEL INTEGRALS
(k=O, 1, 2, . . . .).
Finally,
t-3I2
1
dt=7.33XlO-?.
Example 10. Compute &(2) using 7.3.16. The answer correct to 8D is .39527 290 (Table
11.2).
Generating the values of J,,++(2) as deacribed
in chapter 10 we fmd Example 12. Compute w(.44+.67i) using bi-
variate linear interpolation.
=J&) +J7/2(2) +Jlli2(2) +J,,/2(2) + * - By linear interpolation in Table 7.9 along the
=.49129+.06852+.00297+.00006=.56284. xdirection at y= .6 and y= .7
Example 11. Computelm dt by numerical w(.44+ .Si) SJ .6(.522246+ .167880i) + .4(.498591
integration using Tables 9.1 and 7.8. [Yo(t) is the +.2026666)=.512784+.181794i
Bessel function of thesecond kinddehedin9.1.16.]
w(.44+ .7i) IJ .6(.487556+ .147975i)+ .4(.467521
We decompose the integral into three parts +.179123i)= .479542+.160434i.
-
lZo--x1l
--Pl+PZ,
I zo-rll
_ =-
lz --x I
I lzl -P1-
20-
P2 Example 17. Compute erf (2+i).
From 7.1.3, 7.1.12 we have
{
= 1.003606-.0112590~.
where z=zo+.l(pl+ip2). Thus, with &=.4+.6;,
ll=.45+.65i, 12=.45+.55i, p1=.4, pz=.l, we get Example 18. Compute S, ((a+i)fl)-
from Table 7.9 From 7.3.22, 7.3.8, 7.3.18 we have
&?w({~)=%(.522246+ .498591+.487556+.467521)
=.493979
= :(.522246 +.498591+.561252+ .533157)
S,(z)=Z--
1 1 4
4
e@w
r (l+i) -
&I
9 ~ ( { 2 )
= .528812 --
l+i e
4
-f2Sw [(& l);]
Bw(z)
x (1 - (.4+ .1)]{ [ l - (.4- .1)].522246
+ (.4-.l).528812}+ (.4+ .l)X Setting z =(i+i),E! and making use of 7.1.11,
[1-(.4-.1)] .493979+(.4-.1).498591)=.509789. 7.1.12, and Table 7.9
The correct answer is .509756. Straightforward
bivariate interpolation gives .509460. 8 1 ((i+i) 42)=
Example 14. Compute Yw1.39+.61i) to 6D ----
using Taylorsseries. -i - 3 sin 2) w (2+Z
e-* (cos z-i 1 3 i)
2 4 2
Let z=.39+.61i, zo=.4+.6i. From 7.1.20,
and using Table 7.9, we have +yH+i e2 (cos 2
>
sin 2 w a+:( i)
~(&)=.522246+.1678803 -_ .990734-.6816193.
W(&)=-.21634 + .36738;, 2 - 2 0 ~ (-1 +i)lO-
&U(&) = -.215-. 1854 (2- q)2=-2iX lo- Example 19. Compute
using Table 7.9.
s.-e-c14)ta-3rcos (2t)dt
Yw(z)=
.167880- .0021634-.0036738
4-.0000430=. 162086. Setting b=y+ix, c=O in 7.4.2 and using 7.1.3,
7.1.12 we find
Example 15. Compute 24.4- 1.3i).
From 7.1.11, 7.1.12
1
e -0ta-2vt cos (2xt)dt=51 4-
;aw?*)
W( .4-1.3;) =W(-.4-1.39=2e- (.4-1.3:)
(a>O, 2, y real).
-w(4+1.3i). Hence from Table 7.9
using Tables 7.9, 4.4 and 4.6
w(.4- 1.3i)=4.33342+8.042013. Jm e-(ll4)+at cos (2tjdt= fiaw(2+3i) = .231761.
308 ERROR FUNCTION AND FRESNEL INTEGRALS
References
TeXts [7.22] G. N. Watson, A treatise on the theory of Bessel
[7.1] J. Boemma, Computation of Fresnel integrals, functions, 2d ed. (Cambridge Univ. Press, London,
Math. Comp. 14, 380 (1960). England, 1958).
[7.2] A. V. Boyd, Inequalities for Mills' ratio, Rep. Tables
Statist. Appl. Res. Un. Jap. Sci. Engrs. 6, 44-46
(1959).
[7.3] 0. Emersleben, Numerische Werte des Fehler-
[7.23] M. Abramowitz, Table of the integral
s,' e-y'
Math. Phys. 30, 162-163 (1951). z=0(.01)2.5,
du, J.
integrals fiir G,Z. Angew. Math. Mech. 31, 8D.
393-394 (1951). [7.24] P. C. Clemmow and Cara M. Munford, A table of
[7.4] A. Erd6lyi et al., Higher transcendental functions,
vol. 2 (McGraw-Hill Book Co., Inc., New York, d(+r)e+'*p2J e-+i'x2d~ for complex values of p,
N.Y., Toronto, Canada, London, England, 1953). Philos. Trans. Roy. SOC.London { A ) , 245, 189-
[7.5] A. Erd6lyi et al., Tables of integral transforms, 211 (1952). IpI =0(.01).8, arg p=0(10)450, 4D.
vol. 1 (McGraw-Hill Book Co., Inc., New York, 17.251 R. B. Dingle, Doreen Arndt and S. K. Roy, The
N.Y., Toronto, Canada, London, England, 1954). integrals
[7.6] W. Gautschi, Note on bivariate linear interpolation
for analytic functions, Math. Tables Aids Comp. ~,(z)= (p!)-LS,"' s,(s*+z*)-'e-vic
13, 91-96 (1959). and
[7.7 W. Gautschi, Recursive computation of the re-
peated integrals of the error function, Math. ~ , ( z=
) (pI)-lJm cV(g+z2)-*e-*da
Comp. 15, 227-232 (1961). and their tabulation, Appl. Sci. Res. B 6, 155-164
[7.8] W. Grobner and N. Hofreiter, Integraltafel (1956). C(z), S(z), 2===0(1)20, 12D.
(Springer-Verlag, Wien and Innsbruck, Austria, [7.261 V. N. Faddeeva and N. M. Terent'ev, Tables of
1949-50). 2i
[7.9] D. R. Hartree, Some properties and applications of values of the function w(z) =e+ ( l + z t e P d t )
the repeated integrals of the error function, Mem. for complex argument. Translated from the
Proc. Manchester Lit. Philos. SOC.80, 85-102 Russian by D. G. Fry (Pergamon Press, New
(1936). York, N.Y., 1961). w(z),z=z+iy; z,y=0(.02)3;
[7.10] C. Hastings, Jr., Approximations for digital com- ~=3(.1)5,y=0(.1)3; 2=0(.1)5, y=3(.1)5; 6D.
puters (Princeton Univ. Press, Princeton, N.J., [7.27] B. D. Fried and S. D. Conte, The plasma dispersion
1955). function (Academic Press, New York, N.Y. and
[7.11] Y. Komatu, Elementary inequalities for Mills' London, England, 1961). i&w(z), id&'(z),
ratio, Rep. Statist. Appl. Res. Un. Jap. Sci. a=z+iy; z=O(.1)9.9, y= -9.1(.1)10; z=var.
Engrs. 4, 69-70 (1955-57).
[7.12] E. Kreyszig, On the zeros of the Fresnel integrals,
Canad. J. Math. 9, 118-131 (1957).
[7.13] Th. Laible, Hohenkarte des Fehlerintegrals, Z.
(.1)9.9, y= -lO(.l)-9.2; 6s.
[7.28] K. A. Karpov, Tablitsy funktaii w(z) =.-*at
kompleksnoi oblasti (Izdat. Akad. Nauk SSSR.,
ez%h v