21 Conformal mapping II: the Schwarz–Christoffel mapping
21 Conformal mapping II: the Schwarz–Christoffel mapping
21 Conformal mapping II: the Schwarz–Christoffel mapping
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>21</strong> <strong>Conformal</strong> <strong>mapping</strong> <strong>II</strong>: <strong>the</strong><br />
<strong>Schwarz–Christoffel</strong> <strong>mapping</strong><br />
Introduction<br />
In Chapter 16 we explored some of <strong>the</strong> geometrical properties of holomorphic functions,<br />
and in particular looked at <strong>the</strong> behaviour of <strong>the</strong> Möbius transformation. The key geometrical<br />
feature was that <strong>the</strong> <strong>mapping</strong> is conformal (where <strong>the</strong> derivative is non-zero) in <strong>the</strong><br />
sense that it is locally angle-preserving. In Chapter 19 we highlighted <strong>the</strong> importance of<br />
conformal <strong>mapping</strong> to <strong>the</strong> solution of Laplace's equation in two dimensions. We produced<br />
several types of solution to Laplace's equation, including several examples where<br />
<strong>the</strong> region of interest was bounded by a circle or a line in <strong>the</strong> complex plane.<br />
A question that arise naturally is how to manage matters when <strong>the</strong> region is not a<br />
half-plane or interior/exterior of a circle. Here we must draw a sharp distinction between<br />
issues of general principle and practicalities of implementation. We shall begin by<br />
stating without proof an important, but non-constructive, <strong>the</strong>orem that addresses <strong>the</strong><br />
general principle. Then we shall introduce <strong>the</strong> <strong>Schwarz–Christoffel</strong> (SC) <strong>mapping</strong> that<br />
gives an explicit construction for polygonal regions.<br />
There are very few examples of <strong>the</strong> SC <strong>mapping</strong> that can be worked out in<br />
closed-form in terms of ‘simple’ functions. A novel use of Ma<strong>the</strong>matica is to use its<br />
advanced built-in special-function capabilities, and <strong>the</strong>ir linkage to <strong>the</strong> symbolic integrator,<br />
to give evaluations of several expressions usually left as complicated integrals in<br />
more traditional treatments. We can use such evaluations to facilitate <strong>the</strong> visualization of<br />
<strong>the</strong> <strong>mapping</strong>s, and hence to confirm <strong>the</strong> correctness of <strong>the</strong> answer. As such this chapter<br />
contains an informal introduction to hypergeometric functions and elliptic functions,<br />
which arise naturally from SC <strong>mapping</strong>s associated with triangles and rectangles.<br />
Ma<strong>the</strong>matica's symbolic integrator also allows <strong>the</strong> results for triangles to be extended to<br />
regular n-gons, again in terms of hypergeometric functions. In more recent versions of<br />
Ma<strong>the</strong>matica <strong>the</strong> implementation of hypergeometric functions of two variables permits a<br />
still larger category of <strong>mapping</strong>s to be explored. It should be emphasized that this is a<br />
purely analytical treatment, best used as a teaching tool.<br />
For full professional use <strong>the</strong>re is an extensive and highly developed numerical<br />
treatment already available. We shall also comment briefly at <strong>the</strong> numerical implementation<br />
of <strong>the</strong> SC <strong>mapping</strong>. Gratitude is expressed to Professor L. N. Trefe<strong>the</strong>n F.R.S. for<br />
supplying background material on this topic. Readers interested in pursing SC <strong>mapping</strong><br />
in more detail, whe<strong>the</strong>r analytical or numerical, are encouraged to see Section 2.17 and<br />
to consult <strong>the</strong> definitive and up-to-date text on <strong>the</strong> matter: Driscoll and Trefe<strong>the</strong>n (2002).<br />
This chapter is best regarded as an educational bridge between <strong>the</strong> older and very limited<br />
analytical discussions in textbooks and <strong>the</strong> full numerical approach developed by<br />
Trefe<strong>the</strong>n, Driscoll and co-workers. Our bridge is constructed using <strong>the</strong> advanced<br />
analytical tools in Ma<strong>the</strong>matica.
4H2 =o,-+$8&;n)+>(i(&2it#&?)t#$,)tic)<br />
<strong>21</strong>.1 The Riemann <strong>mapping</strong> <strong>the</strong>orem<br />
This <strong>the</strong>orem is about simply connected domains ! in ! that are not <strong>the</strong> whole of !.<br />
7iven such a domain <strong>the</strong> <strong>the</strong>orem states that t#$r$ &i( &) 1-1 #o+o,or-#ic &/unction<br />
2 = / !3L t#)t&,)-( ! onto&t#$&int$rior&o/&t#$&unit&di(56&i7$7&to&t#$&($t&o/&co,-+$8&nu,9$r(<br />
2 (uc#&t#)t # 2 # " 1. If we fur<strong>the</strong>r specify that a given point 3 0 satisfies 0 = / !3 0L, and<br />
that a specified direction at 30 is mapped into a specified direction at 0 (for example <strong>the</strong><br />
direction along <strong>the</strong> positive real axis), <strong>the</strong> <strong>mapping</strong> is unique.<br />
The proof of this result is ra<strong>the</strong>r beyond <strong>the</strong> scope of this book. A reasonably<br />
accessible treatment is given by Eettman (196H). Iote that <strong>the</strong> result could just as easily<br />
have been stated about <strong>the</strong> existence of a <strong>mapping</strong> into <strong>the</strong> upper half-plane, by composing<br />
/ with an appropriate Möbius transform.<br />
The drawback of this <strong>the</strong>orem is that it is purely about existence M it tells us<br />
nothing about how to explicitly construct <strong>the</strong> <strong>mapping</strong>.<br />
<strong>21</strong>.2 The <strong>Schwarz–Christoffel</strong> transformation<br />
w 1<br />
"<br />
! 1 ! 2<br />
! H<br />
! 3<br />
w 3<br />
w 2<br />
w H<br />
! 4<br />
w 4<br />
w<br />
#<br />
$<br />
% 1 % 2 % 3 % 4 % H<br />
This is a very explicit constructive method for finding <strong>mapping</strong>s that will cope with<br />
regions with polygonal boundaries. We can work with various base regions, such as <strong>the</strong><br />
upper half-plane or interior of a circle, and to state <strong>the</strong> basic results we shall consider a<br />
<strong>mapping</strong> from <strong>the</strong> upper half-plane.<br />
Suppose that we have a polygon in <strong>the</strong> 2-plane with vertices at <strong>21</strong>, 22, S, 2n<br />
and interior angles at those vertices # 1, # 2, S, # n. Suppose that <strong>the</strong>se points map onto<br />
points 81, 82, S, 8n on <strong>the</strong> real axis of <strong>the</strong> 3-plane. Let ! denote <strong>the</strong> interior of <strong>the</strong><br />
polygon and : <strong>the</strong> interior of <strong>the</strong> upper half-plane. Then a transformation that maps :<br />
onto ! is given by <strong>the</strong> SchwarzMChristoffel (SC) formulaW<br />
2 = ; $ !3 $ 81L !#1%%L$1 !3 $ 82L !#2%%L$1 &S!3 $ 8nL !#n %%L$1 &' 3 + <<br />
In differential form this is just<br />
(<strong>21</strong>.1)
45.63%73*8+/.8+22'%,.99:.)-0.; so that ! moEes in a ()*+',-)./'%0$ 'his is why <strong>the</strong> ima:e in <strong>the</strong> !<br />
./ane is .o/y:ona/$ Now we need to see what ha..ens at <strong>the</strong> corners$ Ghen " .asses<br />
throu:h $'> a// <strong>the</strong> terms e=ce.t those inEo/Ein: $' remain constant> Fut <strong>the</strong> factor<br />
Ar:!" # $' "<br />
Ium.s from a Ea/ue of %> when " ( $ ' > to a Ea/ue of J> when $ ) $ ' $ 'he term<br />
$ $'<br />
!!!!!!! # #% Ar:!" # $'"<br />
%<br />
!2#$H%<br />
!2#$K%<br />
<strong>the</strong>refore Ium.s from <strong>the</strong> Ea/ue $' # % to <strong>the</strong> Ea/ue J$ 'he direction in which ! is traEe/L<br />
/in: <strong>the</strong>refore rotates in a .ositiEe sense !i$e$ antiLc/oc;wise%> Fy % # $'$ A chan:e in<br />
direction Fy this amount corres.onds to <strong>the</strong> introduction of a corner in <strong>the</strong> .o/y:on with<br />
interior an:/e $'$<br />
Ghat this discussion estaF/ishes is that <strong>the</strong> rea/ a=is ma.s to <strong>the</strong> Foundary of a<br />
.o/y:ona/ re:ion$ 'his wi// Fe sufficient for our .ur.oses$ Mne can :o fur<strong>the</strong>r and<br />
estaF/ish that <strong>the</strong> ma..in: is oneLtoLone and indeed ta;es <strong>the</strong> 1220* ha/fL./ane to <strong>the</strong><br />
'%)0*'3* of a .o/y:on$ Ge can a/ways chec; this /atter .oint Fy wor;in: out sam./e<br />
Ea/ues of .articu/ar ma..in:s> such as <strong>the</strong> ima:e of *$<br />
! 4he point at in#inity7 a simpli#ication<br />
Ge can assume that one of <strong>the</strong> .oints on <strong>the</strong> rea/ a=is> say $ % is at infinity> with it <strong>the</strong>n<br />
Fein: dro..ed from <strong>the</strong> /ist of .oints in <strong>the</strong> transformation$ 'o see this> we write> with $%<br />
finite><br />
#<br />
# "<br />
+<br />
!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!<br />
!#$%" !$%#%"##<br />
!!<br />
!!!!!!!!<br />
!" " #+ !" # $ #" !$# #%"## !" # $ 2" !$2 #%"## &$ $% # "<br />
!!!!!!!!!!!!!!!!<br />
$%<br />
% !$ %#%"##<br />
!2#$N%<br />
!2#$O%
TUT ./*0+1!2#",+3()(2$)4526,451*,4)7,<br />
!"#$#%$&'(%$&)%$*+,+&$!" ! "-$(%%.+/0$#1$f+/+&%$'/3$&)%$*'4&$f'5&"6$36".4$"7&-$*%'8+/0$74<br />
#+&)<br />
!$<br />
########<br />
!% $ #% !% & ! ;L !f ; #(L&; !% & !:L !f : #(L&; ?!% & !"&;L !f "&;#(L&;<br />
! The exponents and o<strong>the</strong>r comments<br />
9:;<br />
@%$5'/$,'(%$'$4+,.*%$0%",%&6+5'*$'607,%/&$&)'&$.7&4$"/%$5"/4&6'+/&$$"/$&)%$%A."/%/&4<<br />
!"&%$&)'&$&)%$47,$"f$&)%$%A&%6+"6$'/0*%4$"f$'/B$5*"4%3$."*B0"/$+4$:(
)*+&,-.,rm!l+m!223-g+556+789+:;8"!r$
MNO Complex Analysis with Ma<strong>the</strong>matica<br />
!a#te&ia(!)(*)#ma,!*-(c/0<br />
1#a(2e/0 3#a(2e/0 )4ti)(&///" 5!<br />
6h)8!9#a4hic&:##a3!#<br />
!a#te&ia(Ma4!< =0 1#a(2e0 3#a(2e0<br />
)4ti)(&0 >i&4,a3?-(cti)( "# @Ae(tit3"0<br />
!a#te&ia(Ma4!*-(c0 1#a(2e0 3#a(2e0<br />
)4ti)(&0 >i&4,a3?-(cti)( "# @Ae(tit3"<br />
$"0 >i&4,a3?-(cti)( "# B>i&4,a3?-(cti)("C<br />
!"#$%&'(()&*#&#+,%&-,.%#&"/*01'"&2,#+&a ! 34&5"6*''&#+*#&!a#te&ia(!)(*)#ma,&%+(2%<br />
#+"&,0*7"%&(-&#+"&."*'&*89&,0*7,8*.:&6((.9,8*#"&',8"%&;89".&#+"&0*11,874&<br />
!a#te&ia(!)(*)#ma,!%D & Ei :#c6i(!;%#(&?".#,6"%&#(<br />
@"&6(8%,9"."94&A8&,01(.#*8#&6*%"=&'"-#&-(.&:(;&#(&,8?"%#,7*#"&,8&9"#*,'&,8&B/".6,%"&C343=&,%<br />
2+"."&#+"&1(':7(8&+*%(&.,7+#-*87'"9&#;.8%&(-&(11(%,#"&%"8%"&E&#+,%&'"*9%&#(&*&0*11,87<br />
-.(0&#+"&;11".&+*'--1'*8"&#(&(8"&2,#+&*&%#"1-%+*1"9&@(;89*.:4<br />
2)*+ !riangular and rectangular boundaries<br />
<strong>21</strong> <strong>Conformal</strong> <strong>mapping</strong> <strong>II</strong>: <strong>the</strong> <strong>Schwarz–Christoffel</strong> <strong>mapping</strong> 457<br />
Show@Graphics @88GrayLevel @0.6D, Polygon@880, 0
458 Complex Analysis with Ma<strong>the</strong>matica<br />
Integrate@z ^Ha ê Pi - 1L H1 - zL ^ Hb ê Pi - 1L,<br />
8z, 0, 1 0, b > 0
<strong>21</strong> <strong>Conformal</strong> <strong>mapping</strong> <strong>II</strong>: <strong>the</strong> <strong>Schwarz–Christoffel</strong> <strong>mapping</strong> 459<br />
Now we ask about this function.<br />
? Beta<br />
Beta@a, bD gives <strong>the</strong> Euler beta function BHa, bL.<br />
Beta@z, a, bD gives <strong>the</strong> incomplete beta function BzHa, bL.<br />
Ma<strong>the</strong>matica has now given us a result in terms of <strong>the</strong> incomplete beta function. The<br />
<strong>the</strong>ory of analytic continuation <strong>the</strong>n can be used to argue that this holds everywhere. A<br />
proper discussion of this is ra<strong>the</strong>r beyond <strong>the</strong> scope of this text, but, basically, if two<br />
holomorphic functions agree on a suitable region, <strong>the</strong>y are equal everywhere, so we can<br />
use <strong>the</strong> formula in terms of a beta function, which is holomorphic, for o<strong>the</strong>r values of z<br />
besides those suggested by <strong>the</strong> output from <strong>the</strong> integrator. In any case, we can check this<br />
formula by using our visualization tool. We consider an equilateral triangle, with<br />
a = b = p<br />
ÅÅÅÅÅ<br />
3<br />
The formula for w is <strong>the</strong>n:<br />
w ê. 8a -> Pi ê 3, b -> Pi ê 3<<br />
BzH 1<br />
ÅÅÅÅ , 3 1<br />
ÅÅÅÅ L GH 3 2<br />
ÅÅÅÅ L 3<br />
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅ<br />
ÅÅÅÅ<br />
GH 1<br />
L 3 2<br />
We define a Ma<strong>the</strong>matica formula by<br />
g@z_D = Simplify@%D;<br />
(<strong>21</strong>.23)<br />
Now let us take a look at <strong>the</strong> result (this needs a fast machine to compute reasonably<br />
quickly):<br />
Cartesian<strong>Conformal</strong>@Hg@#D &L, 8-8, 8 1000,<br />
PlotRange -> All, PlotStyle Ø AbsoluteThickness@0.01DD<br />
8<br />
6<br />
4<br />
2<br />
-7.5 -5 -2.5 2.5 5 7.5<br />
0.6<br />
0.5<br />
0.4<br />
0.3<br />
0.2<br />
0.1<br />
0.2 0.4 0.6 0.8 1<br />
Slight numerical plotting effects at end-points aside, this confirms <strong>the</strong> validity of <strong>the</strong><br />
closed-form answer.<br />
Ano<strong>the</strong>r interesting exercise is to consider <strong>the</strong> region in <strong>the</strong> upper half plane<br />
above a related triangle and <strong>the</strong> real axis – this is left for you to explore in Exercise <strong>21</strong>.2.
460 Complex Analysis with Ma<strong>the</strong>matica<br />
‡ Ÿ Hypergeometric and beta functions<br />
The incomplete beta function is a special case of <strong>the</strong> hypergeometric function. You can<br />
just ask <strong>the</strong> kernel for some basic information about such functions<br />
? Hypergeometric2F1<br />
Hypergeometric2F1@a, b, c, zD<br />
is <strong>the</strong> hypergeometric function 2F1Ha, b; c; zL.<br />
More useful is <strong>the</strong> series expansion, which extrapolates <strong>the</strong> pattern shown here for <strong>the</strong><br />
first few terms:<br />
Series@ Hypergeometric2F1@a, b, c, zD, 8z, 0, 3 1 in <strong>the</strong> following<br />
example. In this case <strong>the</strong> differential SC <strong>mapping</strong> takes <strong>the</strong> form
<strong>21</strong> <strong>Conformal</strong> <strong>mapping</strong> <strong>II</strong>: <strong>the</strong> <strong>Schwarz–Christoffel</strong> <strong>mapping</strong> 461<br />
!w<br />
ÅÅÅÅÅÅÅÅ<br />
!z = A Hz + aL-1ê2Hz + 1L -1ê2 Hz - 1L -1ê2 Hz - aL -1ê2<br />
A<br />
= - ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ<br />
è!!!!!!!!!!! ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅ<br />
1 - z2 è!!!!!!!!!!!!!!<br />
a2 - z2 (<strong>21</strong>.24)<br />
It is a little more convient to write this in terms of a = 1 ê a and a new overall constant<br />
A £ as<br />
!w<br />
ÅÅÅÅÅÅÅÅ<br />
!z =<br />
A £<br />
ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ<br />
è!!!!!!!!!!! ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅ<br />
1 - z2 è!!!!!!!!!!!!!!!!!<br />
1 - a2 z2 (<strong>21</strong>.25)<br />
The integral of this is only moderately exotic, in that it defines one of <strong>the</strong> well known<br />
family of elliptic functions. We now set A £ = 1 and evaluate <strong>the</strong> integral with some<br />
suitable Assumptions:<br />
Integrate@1 ê HSqrt@1 - z ^2D * Sqrt@1 - a ^2 z ^2DL,<br />
8z, 0, z 0
462 Complex Analysis with Ma<strong>the</strong>matica<br />
Cartesian<strong>Conformal</strong>@HSCRect@#, 1 ê 2D &L,<br />
8-16, 16 200, PlotRange -> AllD<br />
15<br />
12.5<br />
10<br />
7.5<br />
5<br />
2.5<br />
-15 -10 -5 5 10 15<br />
2<br />
1.5<br />
1<br />
0.5<br />
-1.5-1-0.5 0.5 1 1.5<br />
We can see very clearly that <strong>the</strong> EllipticF function is effecting <strong>the</strong> <strong>mapping</strong> we<br />
want. The four vertices are a <strong>the</strong> points !b, !b + Â c, where, as functions of <strong>the</strong> parameter<br />
a = 1 ê a, b is given by integrating <strong>the</strong> <strong>mapping</strong> from 0 to 1:<br />
b @a_D = SCRect@1, aD<br />
KHa 2 L<br />
where <strong>the</strong> Ma<strong>the</strong>matica function just output is given by<br />
InputForm@%D<br />
EllipticK[a^2]<br />
For <strong>the</strong> case we have plotted, <strong>the</strong> value of this is:<br />
b@1 ê 2D êê N<br />
1.68575<br />
To extract <strong>the</strong> right value of c we need to be slightly careful with <strong>the</strong> branching properties<br />
(also see Wolfram (1993)) of <strong>the</strong> EllipticF function – to this end we define:<br />
c@a_, e_D = SCRect@1 ê a + e I, aD - b@aD<br />
FJsin -1 JÂ e + 1<br />
ÅÅÅÅÅ<br />
a N À a 2 N - KHa 2 L<br />
For <strong>the</strong> example at hand, <strong>the</strong> most obvious approach does not quite work:<br />
c@1 ê 2, 0D êê N<br />
-4.00238 µ 10 -9 - 2.15652 Â<br />
If we nudge <strong>the</strong> point slightly into <strong>the</strong> upper half-plane, which is, after all, where we are<br />
working, we get: