home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!pipex!warwick!pavo.csi.cam.ac.uk!gjm11
- From: gjm11@cus.cam.ac.uk (G.J. McCaughan)
- Subject: Re: SL(2,Z) fundamental domain
- Message-ID: <1992Sep6.230603.1395@infodev.cam.ac.uk>
- Sender: news@infodev.cam.ac.uk (USENET news)
- Nntp-Posting-Host: bootes.cus.cam.ac.uk
- Organization: U of Cambridge, England
- References: <1992Sep06.165238.20128@rz.uni-jena.de>
- Date: Sun, 6 Sep 1992 23:06:03 GMT
- Lines: 36
-
- It's asked: What semicircles (radius and centre) are there in the
- usual tesselation of the upper half-plane by copies of the standard
- fundamental region for SL(2,Z)?
- Well, every such semicircle is the image under an element of SL(2,Z)
- of Re(z)=0; in other words, of z+z*=0 (where z* is the conjugate of z).
- So we get everything of the form (az+b)/(cz+d) + (az*+b)/(cz*+d) = 0,
- which is to say, errm, (az+b)(cz*+d) + (cz+d)(az*+b) = 0,
- i.e. ac.zz* + (ad+bc)Re(z) + bd = 0,
- i.e. x^2+y^2 + (ad+bc)/ac.x + bd/ac = 0,
- i.e. (x-(ad+bc)/2ac)^2 + y^2 = (ad+bc)^2/4(ac)^2 - bd/ac
- except that I've probably made a mistake by now. (Um, when ac=0 we get
- a vertical line. This is a boring case.)
- So, the possible radii are those of the form
- root[ ((ad+bc)/2ac)^2 - bd/ac ]
- and the centres are those of the form
- (ad+bc)/2ac;
- and <centre,radius> can occur if there are a,b,c,d with ad-bc=1 giving
- that pair.
- Determining what radii can occur for given centre is the same as
- determining what bd/ac can be, given (ad+bc)/2ac.
- ... Oh, I've just noticed that the radius is equal to
- (ad-bc)/2ac, give or take a change of sign. In other words, 1/(2ac).
- So all radii are of the form 1/q, and then the centre is p/q for some
- integer q. And q must be even.
- Can all these occur? In other words, given ac, can ad+bc be anything
- we like? Well, since ad-bc=1 this is the same as: given ac, can ad be
- anything we like? Can we solve ac=x,ad=y?
- Well, let's think. We have (x,y)=(ac,ad)=a(c,d)=a. So we know what
- a has to be. Then we know also what c,d have to be. Now there will
- be an appropriate b iff ad == 1 mod c; in other words, if y==1
- mod x/(x,y).
- What are x,y in the present context? 2x=q; 2y-1=p. So we need
- (p+1)/2 == 1 mod (q/2)/(q/2,(p+1)/2)
- i.e. mod q/(p+1,q). Weird. I probably really have made a mistake
- by this point, so I'll now leave it up to those of you with pencil
- --
-