COMPUTATIONAL PROBLEMS IN ABSTRACT ALGEBRA.

COMPUTATIONAL PROBLEMS IN ABSTRACT ALGEBRA. COMPUTATIONAL PROBLEMS IN ABSTRACT ALGEBRA.

thales.doa.fmph.uniba.sk
from thales.doa.fmph.uniba.sk More from this publisher
05.01.2013 Views

378 Harvey Cohn It is significant that the norm /N(y)1 = 1 except for those transformations congruent to Sr (the identity), where /N(y)1 = 2. If we refer to Fig. 3, we see a cross-section S = 1, on which many regions seem to be represented. We cannot be too sure of those represented by only one point, such as “region” 1, 10, 35, 11, 3, 5. Actually “region” 35 is a complete accident of round-off error while “region” 1 joins only at the corner point until S becomes smaller (about 0.43). To test whether points occur as accidents we have only to test cross-sections for values of S close to the one in question. FIG. 3. Pieces of floor of @ lying in cross-section of S = 1. Note the consistency with the symmetry on S = 1 in Fig. 2. (The R’ axis has inadvertently become directed downward because of the direction of the paper in the printer!) We can, however, cut and paste and rearrange the sections so that the cross-section for S= 1 is still a torus, but that there are the least number of d@erent regions showing. Clearly, region 21 is connected to region 23 by letting z become z+ 1, etc. Moreover, two regions represent the same transformation as far as @- is concerned, if for some H in I‘- we have H(Z) = &. (4.7) Algebraic topology on bicomplex manifolds 379 Thus region 23 and 2 are the same, &a = F;~ Z2 but Zz2 and Z:zs must be different since they have different denominators. It can be verified that for IN(y) 1 = 1, two transformations with the same denominator are identifiable under (4.7). It is similarly easy to attend to the transformations where j N(y) ] = 2. Thus in Fig. 3, the black lines set apart regions in which (4.7) is not valid. A dotted line is used if the trans- formation H satisfies H(z) E z+ 1 (mod 29, thus region 2 and region 6 are separated by a dotted line (2s = ZS+ 1). It is conjectured, in more general cases of f&c”), that a single piece can be put together for each value of 6 (mod yj, and this seems true from the computation here. We call IN(y)] the norm of the piece in question. Thus we have a “piece of norm 1” and a “piece of norm 2”. The piece of norm 2 is a spindle drawn in two halves in Fig. 2. It shrinks to a point at S = 0 and S = 1. We locate it for definiteness at an axis through R = 0, R’ = -+ so that transformation 31 prevails. Thus the piece is mapped into itself (under equivalence classes in P) by z. = H (z/(2+z+ 1)) (4.8) (or z. is the value symmetric to it, we shall not always repeat this). These H(z) all belong to I-;,-. j . . _ /..::‘, . . . . . . .

380 Harvey Cohn always dictated by the symmetries of the faces in Fig. 2. As S goes from 0 to 1 the cross-section becomes more and more oblique, attaching itself and detaching itself from images of the spindle (of norm 2) at critical values S=O.31 (approx.) and S=O.82 (approx.). For the piece of norm 1, we can show every transformation joins ,&, ie. for H in P zo = H(-l/z). (4.9) We use shading in Fig. 4 to show the portion which belong to r;. Thus in terms of H in r;, we have the following: “shaded portion” z. = H(- l/z), “blank portion” z. = H(- l/z+ 1). The shading is not of topological interest as much as it shows that part of the piece of norm 1 must match equivalent points in the neighboring replica of @- (formed by zf 1 = zo), if we were to match this piece in the unit square by (4.9). Actually, it is more meaningful to match it with itself. The piece of norm 1, as reassembled for Fig. 4, is matched with itself under zo = - l/z (or zi = - 1 /z) without use of the equivalence operations of (4.9). To see this consider the two-dimensional boundaries of the reassembled piece of norm 1. They consist of the simultaneous equations IIZII = 1, II Yz+a II = 1 (4.10) if ‘yz+ 6 is the denominator of a neighboring region. Under z. = - 1 /z, the boundary is mapped into another boundary given by IIZ II = 19 11 6z-y 11 = 1. (4.11) The assertion now follows from the fact that the relation z. = - l/z converts the piece into another piece of the same floor of norm 1, while the boundary was determined in an invariant fashion. Any boundary segment (4.10) determines the height of the floor uniquely. These correspond to the points A and C in Fig. 1 which are determined by fixed points rather than by any analysis of the arc ABC. Note that the pair of points A, C constitutes a O-sphere just as the boundary of the piece of norm 1 is a 2-sphere. The critical values of S are quite interesting by themselves, namely and S1 = 0.3101 . . . = 4-6: /5 c 1 S2 = 0.8165 . . . = 6’,3 Actually the points of attachment and detachment are points at which S’ takes the values believed to be the minimum for the whole floor, namely [l] S’ = +(-3+2.6’) = 0.4747 . . . . Algebraic topology on bicomplex manifolds 5. Approximate topological configuration. By using the previous information we can give an approximate description of the topological configuration. First consider @. Here @ has a manifold point at 00 from which the threedimensional base in Fig. 2 appears like a 3-sphere. It is divided into two pieces of norms 1 and 2 which are 3-cells, each folded into itself by transformations (4.8) and (4.9). (Recall that in the simple case, Fig. 1, the floor was one piece folded onto itself by z. = - 1 /z.) Next consider %. If we refer again to Fig. 1, we see that there are three replicas of the floor AC, DC, EC which transform into one another like the representatives of G/G2 in (2.7). The fact that the reassembled piece of norm 1 is transformed into itself under z. = - l/z, etc., enables us to reproduce three replicas of that piece in an analogue of Fig. 1. If we momentarily restrict ourselves to this piece (ignoring that of norm 2), we have a simple situation where the (spherical) boundaries of each of the three replicas meet in a 2-sphere analogous to the O-sphere A, C of Fig. 1. The analogy, however, is not kept because the piece of norm 2 does not map into itself under zo = - l/z, etc., hence it is not representable as three replicas lying in the replicas of the floor. The boundary of this piece of norm 2 is nestled, however, in between the various boundaries of the pieces of norm 1, and the three-dimensional piece of norm 2 bulges out of the spherical boundary into parts of the three-dimensional floor. The situation is therefore somewhat more complicated than that of lower dimensional space (as it must be since the final configuration cannot be a 4+phere!). We are confronted with the need to study the self-mapping of the floor more carefully in order to make deductions concerning the topology of the fundamental domains @and @e. There are very few cases which are analogous [5], possibly the other two involve Q(31j2), where the problem can be considered as an analogous pasting of 3 (or 4) 3-spheres. In any case, the numerical data are capable of further analysis for geometric or topological features than attempted here. REFERENCES 1. H. COHN: A numerical study of the floors of various Hilbert fundamental domains. Math. of Compuration 19 (1965), 594-605. 2. H. COHN: A numerical study of topological features of certain Hilbert domains. Math. of Computation 21 (1967), 76-86. 3. H. COHN : Conformal Mapping on Riemann Surfaces, McGraw Hill, 1967. 4. K. B. GIJNDLACH: Some new results in the theory of Hilbert’s modular group, pp. 165-180, Contributions to Function Theory (Tata Institute, 1960). 5. K. B. GUNDLACH: Die Fixpunkte einiger Hilbertscher Modulgruppen. Math. Annalen 157 (1965), 369-390. 6. W. M. W OODRUFF: The singular points of the fundamental domain for the groups of of Bianchi, Ph.D. Dissertation, University of Arizona, 1967. 381

380 Harvey Cohn<br />

always dictated by the symmetries of the faces in Fig. 2. As S goes from<br />

0 to 1 the cross-section becomes more and more oblique, attaching itself<br />

and detaching itself from images of the spindle (of norm 2) at critical<br />

values S=O.31 (approx.) and S=O.82 (approx.). For the piece of norm 1,<br />

we can show every transformation joins ,&, ie. for H in P<br />

zo = H(-l/z). (4.9)<br />

We use shading in Fig. 4 to show the portion which belong to r;.<br />

Thus in terms of H in r;, we have the following:<br />

“shaded portion” z. = H(- l/z),<br />

“blank portion” z. = H(- l/z+ 1).<br />

The shading is not of topological interest as much as it shows that part of<br />

the piece of norm 1 must match equivalent points in the neighboring replica<br />

of @- (formed by zf 1 = zo), if we were to match this piece in the<br />

unit square by (4.9). Actually, it is more meaningful to match it with<br />

itself.<br />

The piece of norm 1, as reassembled for Fig. 4, is matched with itself under<br />

zo = - l/z (or zi = - 1 /z) without use of the equivalence operations of (4.9).<br />

To see this consider the two-dimensional boundaries of the reassembled<br />

piece of norm 1. They consist of the simultaneous equations<br />

IIZII = 1, II Yz+a II = 1 (4.10)<br />

if ‘yz+ 6 is the denominator of a neighboring region. Under z. = - 1 /z,<br />

the boundary is mapped into another boundary given by<br />

IIZ II = 19 11 6z-y 11 = 1. (4.11)<br />

The assertion now follows from the fact that the relation z. = - l/z converts<br />

the piece into another piece of the same floor of norm 1, while the<br />

boundary was determined in an invariant fashion.<br />

Any boundary segment (4.10) determines the height of the floor uniquely.<br />

These correspond to the points A and C in Fig. 1 which are determined<br />

by fixed points rather than by any analysis of the arc ABC. Note<br />

that the pair of points A, C constitutes a O-sphere just as the boundary<br />

of the piece of norm 1 is a 2-sphere.<br />

The critical values of S are quite interesting by themselves, namely<br />

and<br />

S1 = 0.3101 . . . = 4-6: /5<br />

c 1<br />

S2 = 0.8165 . . . = 6’,3<br />

Actually the points of attachment and detachment are points at which S’<br />

takes the values believed to be the minimum for the whole floor, namely [l]<br />

S’ = +(-3+2.6’) = 0.4747 . . . .<br />

Algebraic topology on bicomplex manifolds<br />

5. Approximate topological configuration. By using the previous information<br />

we can give an approximate description of the topological configuration.<br />

First consider @. Here @ has a manifold point at 00 from which the threedimensional<br />

base in Fig. 2 appears like a 3-sphere. It is divided into two<br />

pieces of norms 1 and 2 which are 3-cells, each folded into itself by transformations<br />

(4.8) and (4.9). (Recall that in the simple case, Fig. 1, the floor was<br />

one piece folded onto itself by z. = - 1 /z.)<br />

Next consider %. If we refer again to Fig. 1, we see that there are three<br />

replicas of the floor AC, DC, EC which transform into one another like the<br />

representatives of G/G2 in (2.7). The fact that the reassembled piece of norm<br />

1 is transformed into itself under z. = - l/z, etc., enables us to reproduce<br />

three replicas of that piece in an analogue of Fig. 1. If we momentarily<br />

restrict ourselves to this piece (ignoring that of norm 2), we have a simple<br />

situation where the (spherical) boundaries of each of the three replicas meet<br />

in a 2-sphere analogous to the O-sphere A, C of Fig. 1. The analogy, however,<br />

is not kept because the piece of norm 2 does not map into itself under<br />

zo = - l/z, etc., hence it is not representable as three replicas lying in the<br />

replicas of the floor. The boundary of this piece of norm 2 is nestled, however,<br />

in between the various boundaries of the pieces of norm 1, and the<br />

three-dimensional piece of norm 2 bulges out of the spherical boundary into<br />

parts of the three-dimensional floor. The situation is therefore somewhat<br />

more complicated than that of lower dimensional space (as it must be since<br />

the final configuration cannot be a 4+phere!).<br />

We are confronted with the need to study the self-mapping of the floor<br />

more carefully in order to make deductions concerning the topology of the<br />

fundamental domains @and @e. There are very few cases which are analogous<br />

[5], possibly the other two involve Q(31j2), where the problem can be<br />

considered as an analogous pasting of 3 (or 4) 3-spheres. In any case, the<br />

numerical data are capable of further analysis for geometric or topological<br />

features than attempted here.<br />

REFERENCES<br />

1. H. COHN: A numerical study of the floors of various Hilbert fundamental domains.<br />

Math. of Compuration 19 (1965), 594-605.<br />

2. H. COHN: A numerical study of topological features of certain Hilbert domains. Math.<br />

of Computation 21 (1967), 76-86.<br />

3. H. COHN : Conformal Mapping on Riemann Surfaces, McGraw Hill, 1967.<br />

4. K. B. GIJNDLACH: Some new results in the theory of Hilbert’s modular group,<br />

pp. 165-180, Contributions to Function Theory (Tata Institute, 1960).<br />

5. K. B. GUNDLACH: Die Fixpunkte einiger Hilbertscher Modulgruppen. Math. Annalen<br />

157 (1965), 369-390.<br />

6. W. M. W OODRUFF: The singular points of the fundamental domain for the groups of<br />

of Bianchi, Ph.D. Dissertation, University of Arizona, 1967.<br />

381

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!