i e], pi1] e],singer/zr1/zr1_2008_09/as_25.pdf · 25. numerical interpolation, differentiation, and...
TRANSCRIPT
25. Numerical Interpolation, Differentiation, and Integration
PHILIP J. DAVIS AND IVAN POLONSKY *
Contente Formulas
25.1. Differences . . . . . . . . . . . . . . . . . . . . . . 25.2. Interpolation . . . . . . . . . . . . . . . . . . . . . 25.3. Differentiation . . . . . . . . . . . . . . . . . . . . 25.4. In tepa tion . . . . . . . . . . . . . . . . . . . . . . 25.5. Ordinary Differential Equations . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table 25.1. +Point Lagrangian Interpolation Coefficients (3 I n I S ) . .
n=3 ,4 ,p=- [“;‘I - (.01) E], Exact
n=5, 6, p=-p$] (.01) E], 10D
n=7,8, p=- pi1] - (.1) E], 1OD
Table 25.2. +Point coefficients for k-th Order Daerentiation ( 1 5k I 5 ) . . . . . . . . . . . . . . . . . . . . . . . . . .
k=1, n=3(1)6, Exact k=2(1)5, n=k+1(1)6, Exact
Table 25.3. n-Point Lagrangian Integration Coe5cients (3 I n 510) . .
Table 25.4. Abscissae and Weight Factors for Gaussian Integration
n=3(1)10, Exact
(2In596) . . . . . . . . . . . . . . . . . . . . . . . . . .
n=2(1)10, 12, 15D n= 16(4)24(8)48(16)96, 21D
Table 25.5. Abscissas for Equal Weight Chebyshev Integration (21n59) . . . . . . . . . . . . . . . . . . . . . . . . . .
n=2(1)7, 9, 10D
Table25.6. Abscissae and Weight Factors for Lobatto Integration (3 5 n I 10). . . . . . . . . . . . . . . . . . . . . . . . . .
Table 25.7. Abscissas and Weight Factors for Gaussian Integration for Integrands with a Logarithmic Singularity (25n54) . . . . .
n=3(1)10, 8-10D
n=2(1)4, 6D
Pege 877 878 882 885 896
898
900
914
915
916
920
920
920
National Bureau of Standards. * National Bureau of Standards. (Preaently, Bell Tel. Labe., Whippeny, N.J.1
876
876 NUMERICAL ANALYSIS
Table 25.8. Abscissae and Weight Factors for Gaussian Integration of Page Momenta (1 SnS8). . . . . . . . . . . . . . . . . . . . . . 921
k=0(1)5, n=1(1)8, 10D
Table 25.9. Abscissae and Weight Factors for Laguerre Integration (2SnS15). . . . . . . . . . . . . . . . . . . . . . . . . . 923
n=2(1)10, 12, 15, 12D or S
Table 25.10. Abscissae and Weight Factors for Hermite Integration (2Sn120) . . . . . . . . . . . . . . . . . . . . . . . . . . 924
n=2(1)10, 12, 16, 20, 13-15D or S
Table 25.11. Coefficienta for Filon’a Quadrature Formula (0 5e-g 1) . , 924 e=o(.oi).i (.1) 1 , 8D
25. Numerical Interpolation, Differentiation, and Integration
Numerical analyste have a tendency to ac- cumulate a multiplicity of tools each designed for highly specialized operations and each requiring special knowledge to use properly. From the vast stock of formulas available we have culled the present selection. We hope that it will be useful. As with all such compendia, the reader may miss his favorites and find others whose utility he thinks is marginal.
We would have liked to give examplea to illuminate the formulas, but this haa not been feasible. Numerical analysis is partially a ecience and partially an art, and short of writing a text- book on the subject it has been imposeible to indicate where and under what circumstances the various formulas are useful or accurate, or to elucidate the numerical difficulties to which one might be led by uncritical use. The formulas are therefore issued together with a caveat against their blind application.
Formulas
mainders.
25.1. Dif€emncea
Forwud DiiTerencea 25.1.1
bk=~$:(.--~) if n and k am of same parity.
Forward Diferencu Central Dif.tncsr
2-1 f-1
25.1.3
25.1.4
Divided Mfllerencxm
877
878 NUMERICAL ANALYSIS
25.1.6 where *n(Z)=(Z-%) (Z-ZI) . . . (Z-zn) and r:(z) is ita derivative:
25.1.7
*:(z&)= (zt-5) . * * (z&-zt-l)(Zt-zt+l) . . . (zt-z.)
Let D be a simply connected domain with a piecewise smooth boundary C and contain the points a,, . . ., zn in its interior. Let f(z) be qalytic in D and continuous in D+C. Then,
25.1.10
25.1.11
Reciprocal DHenncea
25.1.12
Remainder in Lagrmqe Interpolation Formula
25.2.3
25.2.4
25.2.5
The conditions of 25.1.8 are assumed here.
For n odd, (-; (n-l)Skl 3 1 (n-1)). 25.2.7
25.2.8
n even.
n odd.
(s<€<zl)
k has the same range aa in 25.2.6.
Lyuase Two Point Interpolation Formula (Linear Interpolation)
NUMERICAL ANALYSIS 879
w a n g e Three Point Interpolation Formula
25.2.11
f(zo+~h)=A-if-i+Aofo+Alfi+R2
25.2.12
Rib) .065hy(€) = .065Aa (IpI 5 1) Lagrange Four Point Interpolation Formula
25.2.13
f (&+ph)=A-i f-i+Aojo+Ai fi+Ao fa+Ra
-P(P--l) (P--2)f-l+(p’--1)(P--2)fo 6 2 k
25.2.18 R&) .0049hyce)(€) = .0049Ae (O<p<l) .0071hef“’([) =.0071A’ (-l<p<O, l<p<2) .024hy(‘(() k.024Ae (-2<p<-l, 2<p<3)
(z-r<€<zs) Lagrange Seven Point Interpolation Formula
25.2.19 f(zo+ph)=g AJf+Re
25.2.20 i-4
.0025h7f‘f‘7)(() A..0025A7 (Ip(<l)
.0046h’f7’(€) =.0046A7 (l<Jp1<2)
.Ol9h’f7’(€) = .019A7 (2<lp1<3)
(z-a<€<zs)
Lagrange Eight Point Interpolation Formula
25.2.21
25.2.22
Rr (P) =
Let f(zlzo,zl, . . .,zk) denote the unique poly- nomial of k” degree which coincides in value with f(z) at a, . 25.2.23
., zr.
880 NUMERICAL ANALYSIS
Taylor Erplnaion 25.2.24
25.2.25 R,,=J:f(n+l) (t) @=$ dt
(zo<€<G) (For r,, see 25.1.6.)
Newton% Forward Difference Formule
25.2.28
f(~o+Ph)=fo+P&+(G+ . . . +@*;+En
A2 xa . f a
25.2.29
(&<€<zn)
Relation Between Newton and Lagranee Coefficients
25.2.30
Everett'e Formula 25.2.31
P(P-1) (P--2) 6; 3! f (%+ PN = (1 --P)fo+ Pf1-
Relation Between Everett and Lagrange Coefficients
25.2.33
E2=A'_1 E4=A!2 Ea=AEa
$',=A: F4=A! Fa=& Everett's Formula With Throwback
(Modified Central Difference)
25.2.34
25.2.54 m=2n+l
f ( ~ + ph) = (1 - p )jo +pji + E&. o +F&, 1 + E4%, O+F4$. 1 +
25.2.44 8:s 6'- .0131266+ .004368-.001610
25.2.45 6&= 9- .278276'+ .06856'- .01661°
25.2.% R = .00000083lp$,l+ .00000946'
-1's Formula With Throwback 25.2.47
f(2o +ph) = (1 -p>jo+pj1+ Bz (6:. 0 + 6;. 1)
P (P -1 ) (P-)) +B36;+RJ B 2 = p v ~ B3- 6
1 f (x) "2 s+& (ak cos kx+bk s b kz) k-1
(k=OJ1,. . ,Jn)
25.2.55 m=2n
(k=OJIJ. . .,n)
b, is arbitrary.
(k=OJlJ. . .Jn-l)
Subtabulation
Let j(x) be tabulated initially in intervals of width h. It is desired to subtabulate j(z) in intervals of width h/m. Let A and designate differences with respect to the original and the
h a 1 intervals respectively. Thus &=j (z,,+z) -j(;co). Assuming that the original 5* order differences are zero
25.2.56
h
At (1 - m) (1 -2m) (1 - 3m) 24m4 +
From this information we may construct the b a l tabulation by addition. For m= 10,
25.2.57 - &=.1&-.045&+.0285~-.02066~ @=.01&-.009&+.007725a g=.OOl&- .00135G
- - 2=.OOO1g
h e a r Inverse Interpolation
Find p, given jp(=j(ro+ph)). Linear
25.2.58
882 NUMERICAL ANALYSIS
Quadratic Inverse Interpolation 25.2.59
(f1- 2fo+f- l)P8 + (fl -f- l)P + 2 (fo-fp) = 0
Inverse Interpolation by Revemion of !hries OD
25.2.60 Given f(zo+ph) =fp=C akpk
25.2.61
k=O
p=h+C2h8+C3h3+ . . ., h=(fp-%)/$
25.2.62 ca= -a~lai
-a, 3a; 2l& 14at c5=-+>+--- al ai a; ai +- at
Inversion of Newton's Forward Difference Formula
25.2.63 %=fo ale&--+--- Ai @ %+
2 3 4 - . *
G At+ cca=s-4 . . *
a4=a+. % . .
(Used in conjunction with 25.2.62.)
25.2.64 Inversion of Everett'r Formula
%=jo s: s: s: 6' aI=s4---- +-+'+ . . .
aa=8-a - * .
%= 6-- 24 + . a .
3 6 20 30 6: ff+
-s:+s: 4+s:
a4=a+ 4 . . .
-st+%+ a,= ~ 120 . -
(Used in conjunction with 25.2.62.)
Bivariate Interpolation
Three Point Formula (Linear) 25.2.65
+ f(zo+Ph,Yo+ n 4 = (1 -P- n)fo. 0
+PfL o+ do, 1 + OW) Four Point Formula
25.2.66
i- f (%+Ph,YO+ nk) = (1 -PI (1 - n)fo.o+P(1- n v i . o
+ n(1 - Plfo, 1 +Pdl. 1 + W2) Six Point Formula
25.2.67
25.3. Dmerentiation
Lagrange's Formula
k=O 25.3.1
(See 25.2.1.)
NUMERICAL
25.3.3 rn(z) d (*+I) ) (€
t=t (%I (G<€<xn) Rb(x)= (n+ fn+l) 1) ! (€>?rb(z) +(n+l)! &j
Equally Spaced Abscissas
Three Points 25.3.4
f:=Y (zo+ph) 1
-h -- { (p- $If-* - 2p j o + (P+# f l 1 +R;
Four Points 25.3.5
f.=f’(G+ph)=k{ - 6 3P2-6P 4-2 j-l
+3p2-4p-l 3p2-2p-2 jl
3p’- 1
2 2 fo-
++.} +R;
Five Points 25.3.6
1 2p3-3p2-p+l j-2 { 12 - 4p3- 3P2---8P+4 j-l+2P3;5Pjo
j:=f’(%+Ph)=~
6
- 4p3i-3p2-8p-4 6 jl
2p3+3p2-p-1 j2) +R:
For numerical values of differentiation coeffi-
12 +
cients see Table 25.2.
Markoff’s Formulas
(Newton’s Forward Dserence Formula Differentiated)
25.3.7 p(ao+ph)-Ipo+22p-l@ 1
2
3p’-6p+2A?i+ . . . +& d P (n) &j]+R: ‘ 6
25.3.8
R‘,=hy(n+l) (6) d ( p )+hn+l ( p )” jb+l)(t) dp n+l n+l dx
(%<€<%)
1 1 1 25.3.9 hfi=&-,j@+B&-;iG+ . . .
ANALYSIS 883
11 5 12 6 25.3.10 h ~ i 2 ’ = @ - ~ + - ~ - - A ~ + . . .
25.3.11 hSjia)=&-- 3 7 15 ,&+;i&-sG+. . .
25.3.12 17 7
hy44)=&--2&+ 6g--zg+ . . .
25.3.13
h6j i5)=g- , jg+s@-s@+. 5 25 35 . .
Everett’s Formula 25.3.14
hf’(G+&) - - j o+ j ~ - 6 3p2-@+2 8; ; 3Pi-1 6:
+ - * - -[( 2n+1)]@+[(2.+1)1
5p‘-20p3+ 15p2 + 1 Op- 6 *.+5p4- 1 5p2+4 St 120 -
120 p+n--l P+n IS?
25.3.15 1 1 1 1
hji=-jo+jl-, 6:-- 6 S:+ 20 - St+- 30 St
MBFerences in Terms of Derivative8
25.3.16 h3 h4 hs
&-hji+z j ~’+~ j i a’+~ j i 4’+- j ~’’ 5!
25.3.17 7 1 & c5 h2jia) + hyia) + e h’ji4) +z hyf)
3 5 25.3.18 @=hyia’+,j h‘fi”+;i fA5)
25.3.19
25.3.20 g c5 hyi5) & c5 hyi4) + 2hyi6’
Partial Derivatives
25.3.21
+ (j1.0-f -l.o)+o(ha) -- bjo, 0 - l
bx 2h
NUMERICAL ANALYSIS 884 25.3.22 I 25.3.26
-- a;?-& ~f1.1-~f0.I+f-l,l+f1.0-2f0.0+f~~.0
+f1.-1--2fo, -I+f-l.-l,+O(h”>
Cfl.l-~l,-l-f-l.l+f-l. -,,+0(h2) 1 a2.fo.0- 1 axay a 2
-- a&o-a (A 1 -f- 1.1 +A, - 1 -f- 1. - 1) + 0(h2)
25.3.23 25.3.27
1 .- ;;f&-p ~fl,l+f-l,l+fl.-l+f-l.--l
--2f1,,-2f -l,o-~fo,1-~fo,-l+~fo.o~+~~~B)
-3- c f i . o+f-1, o+ fo. 1 +fo. -1 axay 212 a2f0 O 1 ~f1,0-2f0.0+f-1,0~+0~h2~ a2f 0.0- - 1 az2 h2
-2fo.o-f1.1-f-I. -1) +O(h2)
25.3.a 25.3.28
--- -- a;+0- I& (-fa. 04- 1 6f1, o- 30f0, o a;$ ’-$ cfa, 0-4f1, O+ 6f0. 0-4f-I, o+f-2. o) + 0 (h2) + w-~. o-j-2, o) + o(h41
25.3.25 25.3.29
NUMERICAL hNALYSIS
Laplacian 25.3.33
25.3.30
acu acu V4%,0= (2 -+2 az’ay2+5do*o -
=p 1 [ ~ 0 ~ 0 . 0 - ~ ( ~ 1 , 0 + ~ 0 . I s u - l , o+% . -1)
+2(UI,l+Ul, - l+~ - l , l+~ - - I . -1)
+ (%. 2+ u2. o+u-2. O+%. -2) 1 + 0(h2)
885
Modified Trapeeoidal Rule 25.4.4
Jrf(z)dz=h [$’+fl+ . . . +fm-l+$]
h 1 lm 6 (4) ) +E [-f-1+f1+fm-1-fm+1l+720 hf (€
886 NUMERICAL ANALYSIS
Simpeon'e Rule 25.4.5
Extended Simpn's Rule 25.4.6
Euler-Maclaurin Summation Formula
25.4.7
(For &k, Bernoulli numbers, see chapter 23.) If f(a+2)(z) and f("#)(z) do not change sign for
q<z<~ then lRul is less than the first neglected term. If f("+')(z) does not change sign for z,,<z<zm, ]Rut is less than twice the first neglected term.
Lapnnge Formula 25.4.8
f(z)dz=k (L:" (a) --LP (a))ft +Rn 1-0
(See 25.2.1.)
25.4.9
Equally Spaad Abecissae 25.4.11
(See Table 25.3 for A,(m).)
Newton-Cotes F O ~ U ~ M ( C l d Type)
(For Trapezoidal and Simpson's Rules see 25.4.1- 25.4.6.) 25.4.13 (Simpeon'e 3 5 rule)
25.4.14 (Bode% rule)
8 f(" (E) h7 +32js+7jJ - 945
25.4.15
275 f (*) (t)h7 12096 +75f4+19fs)-
25.4.16
25.4.17
25.4.18
25.4.19
*See page XI.
NUMERICAL
25.4.20
~of(2Mz=- 5h { 16067Cf0+f10) 299376
t 1 06300 Cf1 +fo) - 48525 (f2 +fa) +272400 C f 8 +f71
-26O55OCfi+fd +427368fsI
- 1346350 f(12) (€)h13 32691 8592
Newton-Cotes Formulas (Open 'bp) 25.4.21
3h f(*) (€)ha J;f(2)d2=F ( f I + f 2 1 + 4
25.422
28f (4) (4) h5 [f(x)dz=$ (2f,-f2+2f3)+ 90
25.4.23
95f4) (€)he ~ f ( z ) ~ z = ~ (11f1+.f2+f3+llf4)+ 144
25.4.24
c f ( x ) d z = s (1 1fi- 14j2+26fr 14f4 + 1 1fs) 6h
+41f'"(€)h7 140
25.4.25
[f(X)d%=- 7h (6llf1-453f2+562f8+562j. 1440 5257
-453f~+611f , )+~~~')(€)W 25.4.26
J;f(Z)dZ=- 945 (46Of1-954f2+2196f3-2459fr 8h
+2196fs-954f6+460f7)+~f(*)(~)hg 3956
Five Point Rule for Analytic Functions
to+ ih 25.4.27
ANALYSIS 887
J_:Df(z)dz=& {24f(z0) +4[f(z0+h) +f(zo-h)l
-[f(zo+a) +f(zo--ih)] 1 +R
1 4 7 IRI S a 9 If'"(z)l, S designates the square withvertices zo+i'h(k=O, 1,2,3); hcan becomplex.
Chebyahev's Equal Weight Integration Formula
2 n 25.4.28 J",j(x)dx=- c j(st> + E n n 1-1
Abscissas: 2, is the itb zero of the polynomial part of
n . . .] 2s exp [------ -n n 2.322 4-52 6-72'
(See Table 25.5 for z,.)
complex. Remainder :
For n=8 and n210 some of the zeros are
+1 %=+I - (n+l) R~=J-~ (n+ 1) ! f ( ~ 2
2 2.+y(n+I) (€f 1 2 -- n(n+l>! 1-1
where (=€(d satisfies O I E I s and O l E t l s
(i=l, . . . ,n)
Integration Formulas of Gaussian Type
(For Orthogonal Polynomials see chapter 22)
Gaud Formula
1
2504.29 J - l j ( ~ ) d ~ = ~ a= 1 wij(zt)+Rn
Related orthogonal polynomials: Legendre poly- nomials Pn(z) 9 Pn(l)=l
Abscissas: x, is the ith zero of Pn(x)
Weights : w, = 2/( 1 - 8) [Pn ( 2, ,I2 (See Table 25.4 for z1 and w,.)
f(2n)(€) (--1<€<1) 22n+1 (n!) 4
'n=(2n+i) [(24 !13
Gauss' Formula, Arbitrary Interval
b-a n
Y"(9) Z,+(T) b+a
25.4.30 J)(Y)dY=T c 1-1 WJ(Yf) +a
*See page 11.
888 NUMERICAL ANALYSIS
Related orthogonal polynomials: Pn(z), Pn(l)= 1 Abscissas: z( is the i" zero of P,(Z) Weights: wt=2/(1 -$) [P:(zt)I2
Radau's Integration Formula
25.4.31
Related polynomials:
Abscissas: xf is the im zero of
Pn-l(z)+Z'n(d z+1
1 1-zt - 1 1 Weights:
wf=2 [Pn4(zt)]2 1-z: [PL(zt)I2
Remainder:
Lobatto's Integration Formula
25.4.32
Related polynomials: P:-l(z)
Abscissas: zr is th6 (i-lPzero of P:,,(z)
Weights: 2
W{= (X'P f 1) n(n- 1) [Pn- 1 ( 4 I' (See Table 25.6 for z: and wi.9
Remainder:
Related orthogonal polynomials:
(21 =&~ZGRP~~O) (1-22)
(For the Jacobi polynomials P!". O' see chapter 22.)
Abscissas:
zt is the i" zero of q,,(z)
Weights:
&
(See Table 25.8 for zt and wt.)
Remainder :
25.4.34
Related orthogonal polynomials:
Abscissas: zt=l-[j where is the im positive zero of P2n+l(z). Weights: ~,=2[jw:~"+~) where w12"+') are the Gaussian weights of order 2n+l. Remainder :
25.4.35
yt=a+ (b-ulzt
Related orthogonal polynomials:
1 -p2n+1 (G), ~2,+1(1)=1 JCi
Abscissas: x,=l-tj where is the itb positive zero of P2,,+l(z). Weights: ~~=2[:w/*~+l) where ~ i ( " + ~ ) are the Gaussian weights of order 2n+ 1.
NUMERICAL ANALYSIS 889
Related orthogonal polynomials: -
~2~ (Jl-x), P2n(1)=1
Abscissas: xi=l-t: where ti is the i" positive zero of PZn (x) . Weights: wf=2wjZn) wiZn) are the Gaussian weights of order 2n. Remrticder :
25.4.37 ey dy=d c a 2 i-1 wJ(yt) +Rn
yf=a+ (b-ah
Related orthogonal polynomials: -
Pzn (41 - z) ,P2n (1) = 1 Abscissas: xi=l-t: where tf is the it" positive zero of PZn(x).
Weights: w , = ~ w / ~ ~) wjZn' are the Gaussian weights of order 2n.
n 25.4.38 fo d x = x wi j(xf) +R,
-1 JiZ 1-1
Related orthogonal polynomials: Chebyshev Poly- nomials of First Kind
Abscissas: (2i- 1)r 2n Xt=COS
b+a b--a Yi=-+-x* 2 2
Related orthogonal polynomials: 1
Tn (2) t Tn (1) =F
Bbscissas: (2i-1)r 2n xf=cos
Weights: r Wa c - n
25.4.40
Related orthogonal polynomials: Chebyshev Poly- nomials of Second Kind
sin [(n+l) arccos x] * un(x)= sin (arccos x)
Abscissas: *
r i n+ 1 x*=cos -
Weights: r i *
wi=- sinZ - n+l n+l
Remainder:
(2%) !22n+1 f""'(t) (-1<t<1) 2L4.41
b+a b-a y*=-+- 2 2 Xf
un(z)= sin (arccos x)
Related orthogonal polynomials: * sin [ (n+ 1) arccos x]
Abscissas: * z U xi =cos - n+l
Weights : i * U wi=- sin2 - r n+l n+l
Related orthogonal polynomials:
Abscissas: 2i-1 r xi =cos2 - - - 2n+l 2
Weights : 2r
Wf=- Xf
*See pnge 11.
NUMERICAL ANALYSIS 890
Related orthogonal polynomials:
Abscissas: 2i-1 r x: =cos2 - * - 2n+l 2
Weights: 2r w:=2n+l z:
Related orthogonal polynomials: polynomials or- thogonal with respect the weight function --In z Abscissas: See Table 25.7 Weights: See Table 25.7
25.4.45
Related orthogonal polynomials: Laguerre poly- nomials Ln(z). Abscissas: z: is the i* zero of Ln(Z) Weights:
'Zr wi= (~+~)TL+~(s:)I*
*
(See Table 25.9 for xf and wt.) Remainder:
25.4.46
Related orthogonal polynomials: Hermite poly- nomials Hn(z). Abscissas: z: is the it" zero of Hn(z) Weights:
2"-Ln! 6 n2[Hn - 1(zd l2
(See Table 25.10 for zf and toc.)
Remainder:
f""(€> (- <€< 1 n! Jr Rn=- 2"(2n)!
Filon'a Integration Formula a 25.4.47
l+co2 e sin 28 ~(e)=2 ( 8" -- @ )
For small 8 we have
25.4.53 2e3 2e6 28' a=--- 45 315 4725 " *
2 2e2 48' 28' @-' +---+-- . . . -5 15 105 567
* For certain difficulties associated with this formula, see the article by J. W. Tukey, p. 400, "On Numerical Approximation," Ed. R. E. Langer, Madison, 1959.
*See page 11.
25.4.56 S2n-1=gf2f-1 sin (tx2f-l) I= 1
n 25-4-57 C ~ , - ~ = ~ j $ ? . ~ I= 1 COS (t22f-1)
"> m +0(hh"-2
1 1 2Tn nn - J ~(X,Y)G%=- C f(h COS mj h sin - 2uh r 2m 12-1
Circle C: i+ y2< ha. 25.4.61
(Xf, Y i) wi
(020) 112 ~=0(h4 )
(hh,O), (0,fh) 1/8
O r , Y 3 wi
(090) \12 (fh,O) 1/12 ~=0(h4)
(fz'fzJ3) h h 1/12
1 uh2 - JJ c f(x,Y)d;rdy=k i- 1 wJ(z*,yJ+R
(Xi, Y d Wt
(090) 419
( f h ,O) 1/9
( fhl fh) 1/36 R=O(h4)
/n i h \ 1 In
(xi, Yf ) Wt
(OJ 'I 1 /6
(fh,O) 1/24 R=O(h6)
Square' S: l x l l h , J Y ! S h 25.4.62
1 4h >JJf('J ?/)dsdy=& i=1 WJ(xf,Y<) +R
R
(xt, Vi) Wi
(0,O) 1/9
( J 6 - l h ,,%!,Js-Ih sin 2Tk IS+& 10 10 10 i.) 3 6 0
10 10 -) 3 6 0
('"' * ' 'J1o) -
(.,/ ! I ! $h~~~ , $+Ahs inM 10 IS-&
R=O(h'')
4 For regions, such as the square, cube, cylinder, etc., which are the Cartesian products of lower dimensional regions, one may always develop integration rules by "multiplying together'' the lower dimensional rules. Thus if
Sp1j(z)h a 5 wij(Zi) i=l
is a one dimensional rule, then
S,'S,lf(zj v)&dgs iJ-1 & wimif(zilzi) beoomea a two dimensional rule. Such rules are not neceassrily the most "economical".
893
R=o(~*) (-3tfzfi) h h 3/60
(-kt 0) 8/60
($*t 43) 8/60
PJUMERICAL ‘ANALYSIS
(A4 h , - + 4 h) 251324
<or*$ h) 10181
($ h,O) lOl8l
Equilateral Triangle T
R=O(h6)
Radius of Circumscribed Circle=h
(XC,YO 25.4.63
- JJTf(z,Y)dzdy=e Wff(2f,Yf)+R a@ 270/1200 1
5 a h 2 4 1-1
G + 1
-m+1 ((7) h,o) 155-fl
1200 R = O(h6)
G + 1 (( 14 >h’
4 7 ) 4
(Xf, YO Wf ((fly) h,*(+-) m)
( ( - F ) h , O ) I 1554 -G
0 1 1200
R=O(hs) C0,O) 314 (h, 0) 1/12 Regular Hexagon H
(+$) 25.4.64
Radius of Circumscribed Circle=h 1/12
1 - JJj(x,Y)dzdy=k 1-1 WfAXf , Yf)+R ? a h z 2 H
(Zf ,Yf ) Wf
(090) 21/36
h h 5/72 R=O(h4) ( AS’fZ d-3) (fh,O) 5/72
894 NUMERICAL ANALYSIS
(xr,?/r, 23 W#
(f4 A, *& h,O)
I (O ,*&h, fdh) . - I
(*& ho, *& h) 1/15
R=O(V) ( f h, 0,O) (X#,Y#) 201
(0, *h, 0) 1/30 (0,O) 258/1008
(A+-& m) 125/1008 R=OW ((40, fh) (X#,Yf, z,) 'u)f (fh 9 9 0) 125/1008
Surface of Sphere 2: zZ+y2+z2=h* (*& h, fg h, f & h) 27/840
25.4965 1 (*.&hf&h,O)
(0, f & h, f & h)
- S s f ( x t ~ , 4 d ~ = & f r l w.f(x:,vt,Zd+R
(f .& h,O, f & h) 32/840 R=O(h*)
(fh,O, 0)
(0, f h, 0) 40/840
(090, f h)
Sphere S: $+yP+z*<h*
25.4.66
1
(21, %, 23 wi (fh,O,O) 1/6
(o,fh,o) 1/6 :*ha B ~=0(h4) - sssl(z ,Y,Z)dnlyd~=~ wd(wh , z:)+R
(O,O,fh) 1/6
(Zc,Yr,Zr) Wt
(O,O, 0) 2/5
(fh,O,o) 1/10
(0, fh,O) 1/10
(O,O, &I&) 1/10
~ = 0 ( h 4 )
Cube6 C: IzlSh
IYl S h
IzlSn.
cf,=sum of values off at the 6 points midway
cfr=sum of values o f f at the 6 centers of the
cfo=sum of values off at the 8 vertices of C. cfe=SUm of values o f f at the 12 midpoints of
edges of C. cfd=sum of values o f f at the 4 points on the
diagonals of each face at a distance of
i&h from the center of the face. 2
from the center of C to the 6 faces.
faces of c.
(zt, Yt, zo wr
(-+h,O,O) 1/6 R=o(~*)
(0, fh,O) 1/6
(O,O, *h) 1/6
25.4.68
& J$Jf(r,Y,z)ddYdz C
1 360
25.4.69
=- [-496fm+128Cf,+8cf,+5cfo]+O(h~)
1 =- 450 [91Cf1-4OCfe+ 1 6Cfd]+O(h6)
where fm=f(o, 0, 0) .
I see footnote to 26.4.62.
Tetrahedron: F 25.470
'$$Sf (w, z)dzdydz=qj 1 Cfw+z 9 CfY V
+terms of 4th order
=Gfm+% C f o + ~ C f a
+terms of 4" order
9-
32 1 4
where V: Volume of T
zfo: Sum of values of the function at the vertices
cfe: Sum of values of the function at midpoints
CfY: Sum of values of the function at the center
fm: Value of function at center of gravity of T.
of F.
of the edges of F.
of gravity of the faces of 5.
NuadERIcAL ANALYSIB 896
25.5. Ordinary DiEerential Equations'
First Order: y'=j(x, y)
Point Slope Formula
25.5.1
25.5.2
Y .+l 'Y .+ hY: + 0(h2)
Y n+1 =yn--l+ 2hY:+ o(ha)
Trapeddal Formula
h 25.5.3 Y,+l=y.+z (Y:+,+Y:) +o(m
6 The reader is cautioned against possible instabilities especially in formulas 25.5.2 and 25.5.13. See, e.g. (25.1 1 I, [ 25.121.
NUMERICAL ANALYSIS 897
Systems of Differential Equations
First Order: y’=f(r,y,z), z’=g(z,y,z).
Second Order Runge-Kutta 25.5.17
1 yn+l=yn+2 (k1+kz)+O(h3),
1 Zn+l=zn+S (h+&) +O(h3)
Runge-Kutta Method 25.5.20
1 1 Yn+l=Yn+h [y:+g (k,+kz+b> +O(h6)
Runge-Kutta Method
25.5.22 yn+,=yn+h
*See page 11.
898 NUMERICAL ANALYSIS
References
Texts
(For textbooks on numerical analysis, see texts in chapter 3)
[25.1] J. Balbrecht and L. Collats, Zur numerischen Auswertung mehrdimensionaler Integrale, Z. Angew. Math. Mech. 38, 1-15 (1958).
[25.2] Berthod-Zaborowski, Le calcul des integrales de
la forme: f(x) log x dx.
H. Mineur, Techniques de calcul numerique, pp. 555-556 (Librairie Polytechnique Ch. BBranger, Paris, France, 1952).
[25.3] W. G. Bickley, Formulae for numerical integra- tion, Math. Gas. 23, 352 (1939).
[25.4] W. G. Bickley, Formulae for numerical differentia- tion, Math. Gas. 25, 19-27 (1941).
[25.5] W. G. Bickley, Finite difference formulae for the square lattice, Quart. J. Mech. Appl. Math., 1,
[25.6] G. Birkhoff and D. Young, Numerical quadrature of analytic and harmonic functions, J. Math.
[25.7] L. Fox, The use and construction of mathematical tables, Mathematical Tables vol. I, National Physical Laboratory (Her Majesty’s Stationery Office, London, England, 1956).
[25.8] S. Gill, Process for the step-by-step integration of differential equations in an automatic digital computing machine, Proc. Cambridge Philos. SOC. 47, 96-108 (1951).
[25.9] P. C. Hammer and A. H. Stroud, Numerical evaluation of multiple integrals 11, Math. Tables Aids Comp. 12, 272-280 (1958).
[25.10] P. C. Hammer and A. W. Wymore, Numerical evaluation of multiple integrals I, Math. Tables Aids Comp. 11.59-67 (1957).
[25.11] P. Henrici, Discrete variable methods in ordinary differential equations (John Wiley & Sons, Inc., New York, N. Y., 1961).
[25.12] F. B. Hildebrand, Introduction to numerical analysis (McGraw-Hill Book Co., Inc., New York, N.Y., 1956).
[25.13] 2. Kopal, Numerical analysis (John Wiley & Sons, Inc., New York, N.Y., 1955).
[25.14] A. A. Markoff, DifTerensenrechnung (B. G. Teubner, Leipsig, Germany, 1896).
[25.15] S. E. Mikeladze, Quadrature formulas for a regular function, SoobFrE. Akad. Nauk Grusin. SSR. 17,
[25.16] W. E. Milne, A note on the numerical integration of differential equations, J. Research NBS 43,
[25.17] D. J. Panov, Formelsammlung zur numerischen Behandlung partieller Differentialgleichungen nach dem Differenzenverfahren (Akad. Verlag, Berlin, Germany, 1955).
r25.181 E. Radau, &tudes sur les formules d’approximation qui servent B calculer la valeur d’une intBgrale dBfinie, J. Math. Pures Appl. (3) 6, 283-336 (1880).
L1
35-42 (1948).
Phys. 29,217-221 (1950).
289-296 (1956).
537-542 (1949) RP2046.
[25.19] R. D. Richtmeyer, Difference methods for initial- value problems (Interscience Publishers, New York, N.Y., 1957).
[25.20] M. Sadowsky, A formula for approximate com- putation of a triple integral, Amer. Math. Monthly 47, 539-543 (1940).
[25.21] H. E. Salser, A new formula for inverse interpola- tion, Bull. Amer. Math. SOC. SO, 513-516 (1944).
[25.22] H. E. Salser, Formulas for complex Cartesian interpolation of higher degree, J. Math. Phys. 28,200-203 (1 949).
[25.23] H. E. Salser, Formula for numerical integration of first and second order differential equations in the complex plane, J. Math. Phys. 29, 207-216 (1950).
[25.24] H. E. Salser, Formulas for numerical differentia- tion in the complex plane, J. Math. Phys. 31, 155-169 (1952).
[25.25] A. Sard, Integral representations of remainders, Duke Math. J. 15,333-345 (1948).
[25.26] A. Sard, Remainders: functions of several variables, Acta Math. 84,319-346 (1951).
[25.27] G. Schulz, Formelsammlung sur praktischen Mathematik (DeGruyter and Co., Berlin, Germany, 1945).
[25.28] A. H. Stroud, A bibliography on approximate integration, Math. Comp. 15, 52-80 (1961).
[25.29] G. J. Tranter, Integral transforms in mathematical physics (John Wiley & Sons, Inc., New York, N.Y., 1951).
[25.30] G. W. Tyler, Numerical integration with several variables, Canad. J. Math. 5, 393-412 (1953).
Tablee
[25.31] L. J. Comrie, Chambers’ six-figure mathematical tables, vol. 2 (W. R. Chambers, Ltd., London, England, 1949).
[25.32] P. Davis and P. Rabinowits, Abscissa and weights for Gaussian quadratures of high order, J. Research NBS 56, 35-37 (1956) RP2645.
[25.33] P. Davis and P. Rabinowits, Additional abscissas and weights for Gaussian quadratures of high order: Values for n=64, 80, and 96, J. Research
[25.34] E. W. Dijkstra and A. van Wijngaarden, Table of Everett’s interpolation coefficients (Elcelsior’s Photo-offset, The Hague, Holland, 1955).
[25.35] H. Fishman, Numerical integration constants, Math. Tables Aids Comp. 11, 1-9 (1957).
[25.36] H. J. Gawlik, Zeros of Legendre polynomials of orders 2-64 and weight coefficients of Gauss quadrature formulae, A.R.D.E. Memo (B) 77/58, Fort Halstead, Kent, England (1958).
[25.37] Gt. Britain H.M. Nautical Almanac Office, Inter- polation and allied tables (Her Majesty’s Stationery Office, London, England, 1956).
NBS 60,613-614 (1958) RP2875.
NUMERICAL ANALYSIS 899 [25.38] I. M. Longman, Tables for the rapid and accurate
numerical evaluation of certain infinite integrals involving Bessel functions, Math. Tables Aids Comp. 11, 166-180 (1957).
[25.39] A. N. Lowan, N. Davids, and A. Levenson, Table of the zeros of the Legendre polynomials of order 1-16 and the weight coefficients for Gauss’ mechanical quadrature formula, Bull. Amer. Math. soc.48, 739-743 (1942).
[25.40] National Bureau of Standards, Tables of La- grangian interpolation Coefficients (Columbia Univ. Press, New York, N.Y., 1944).
[25.41] National Bureau of Standards, Collected Short Tables of the Computation Laboratory, Tables of functions and of zeros of functions, Applied Math. Series 37 (U.S. Government Printing 05ce, Washington, D.C., 1954).
[25.42] P. Rabinowitz, Abscissas and weights for Lobatto quadrature of high order, Math. Tables Aids Comp. 69, 47-52 (1960).
(25.431 P. Rabinowitz and G. Weiss, Tables of abscissas and weights for numerical evaluation of integrals of the form
Math. Tables Aids Comp. 68, 285-294 (1959).
[25.44] H. E. Salzer, Tables for facilitating the use of Chebyshev’s quadrature formula, J. Math.
[25.45] H. E. Salzer and R. Zucker, Table of the zero8 and weight factors of the first fifteen Laguerre polynomials, Bull. Amer. Math. SOC. 55, 1004- 1012 (1949).
[25.46] H. E. Salzer, R. Zucker, and R. Capuano, Table of the zeros and weight factors of the first twenty Hermite polynomials, J. Research NBS 48,
[25.47] H. E. Salzer, Table of coefficients for obtaining the first derivative without differences, NBS Applied Math. Series 2 (U.S. Government Printing Office, Wmhington, D.C., 1948).
[25.48] H. E. Salzer, Coefficients for facilitating trigono- metric interpolation, J. Math. Phys. 27,274-278 (1949).
[25.49] H. E. Salzer and P. T. Roberson, Table of coeffi- cients for obtaining the second derivative without differences, Convair-Astronautics, San Diego, Calif. (1957).
[25.50] H. E. Salzer, Tables of osculatory interpolation coefficients, NBS Applied Math. Series 56 (U.S. Government Printing Office, Washington, D.C., 1958).
Phys. 26, 191-194 (1947).
111-116 (1952) RP2294.