JFSA Journal of Function Spaces and Applications 1758-4965 0972-6802 Hindawi Publishing Corporation 549845 10.1155/2013/549845 549845 Research Article Relative Infinite Determinacy for Map-Germs Shi Changmei 1, 2 Pei Donghe 1 Hencl Stanislav 1 School of Mathematics and Statistics Northeast Normal University Changchun 130024 China nenu.edu.cn 2 School of Mathematics and Computer Science Guizhou Normal College Guiyang 550018 China gzhnc.edu.cn 2013 22 10 2013 2013 05 05 2013 08 09 2013 09 09 2013 2013 Copyright © 2013 Changmei Shi and Donghe Pei. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The infinite determinacy for smooth map-germs with respect to two equivalence relations will be investigated. We treat the space of smooth map-germs with a constraint, and the constraint is that a fixed submanifold in the source space is mapped into another fixed submanifold in the target space. We study the infinite determinacy for such map-germs with respect to a subgroup of right-left equivalence group and finite and infinite determinacy with respect to a subgroup of contact group and give necessary and sufficient conditions for the corresponding determinacy.

1. Introduction

This work is concerned with the singularity theory of differentiable maps. Singularity theory of differentiable maps is a wide-ranging generalization of the theory of the maxima and minima of functions of one variable, and it is now an essential part of nonlinear analysis. This theory has numerous applications in mathematics and the natural sciences; see, for example, [1, 2].

In differentiable analysis, the local behavior of a differentiable map can be determined by the derivatives of the map at a point. Hence we have the theories of finitely and infinitely determined map-germs. We know that every finitely determined map-germ is equivalent to its Taylor polynomial of some degree, and infinite determinacy is a way to express the stability of smooth map-germs under flat perturbations. The analysis of the conditions for a map-germ to be finitely or infinitely determined involves the most important local aspects of singularity theory. Therefore, the study of finite and infinite determinacy of smooth map-germs is a very important subject in singularity theory.

Now there are several articles treating the question of infinite determinacy with respect to the most frequently encountered and naturally occurring equivalence groups, for instance, the right-equivalence group , left-equivalence group , group 𝒞, contact group 𝒦, and right-left-equivalence group 𝒜. In [3, 4], Wilson characterized the infinite determinacy of smooth map-germs with respect to one of the groups , , 𝒞, and 𝒦 and the infinite determinacy for finitely 𝒦-determined map-germs with respect to 𝒜. Besides, for the case of 𝒜, Brodersen  showed that the results of  hold for map-germs without assuming 𝒦-finiteness. We can see  for detailed survey of all of these results. Recently, there appeared the notion of relative infinite determinacy for smooth function-germs with nonisolated singularities. Sun and Wilson  treated the smooth function-germs with real isolated line singularities, and this work was later generalized and modified in the case where germs have a nonisolated singular set containing a more general set; for instance, see [8, 9]. In addition,  studied the relative versality for map-germs and these map-germs with the constraint that a fixed submanifold in the source space is mapped into another fixed submanifold in the target space.

Inspired by the aforementioned papers, we will study the relative infinite determinacy for map-germs with respect to a subgroup of the group 𝒜 and relative finite and infinite determinacy with respect to a subgroup of the group 𝒦 by means of some algebraic ideas and tools, and our main results extend the part of the works in [3, 4].

Now, we consider the following map-germs.

Let N and P be submanifolds without boundary of n and p, respectively, both containing the origin. Since this paper is concerned with a local study, without loss of generality, we may assume that (1)N={0}×n-n0n,P={0}×p-p0paaaaaaaaaaaaaaaaaaaaaaaaaaa(n0,p01).

Denote by Cnp the space of map-germs f:(n,0)p, with f(N)P. Such map-germs are quite common in singularity theory and geometry.

Example 1.

Let f:(2,0)3 be given by (right helicoid) (2)  f(u,v)=(ucosv,usinv,cv). It is clear that f(N)P for N={0}×12 and P={0}×13.

Example 2.

Let f:(4,0)4 be given by (3)f(x1,x2,x3,x4)=(x1,x2,x32+x1x4,x42+x2x3).

Then f(N)P for N=P={0}×24.

In this paper we want to characterize the relative infinite determinacy for such map-germs. To formulate the main results we need to introduce some notations and definitions.

Let En denote the ring of smooth function-germs at the origin in n, and let mn denote its unique maximal ideal. For a germ f, let jkf(x) denote the Taylor expansion of f of order k at x. In the case k=, jf(x) can be identified with the Taylor series of f at x.

Let denote the group of germs at the origin of local diffeomorphisms of n, and let (4)  N={ϕ:ϕ|NidN}, where id denotes the identity. Then N is a subgroup of .

Let CN be the local ring {fEn:f|Nconstant}, and let mN denote the maximal ideal of CN.

Similarly, we can define the corresponding notation for (p,P).

Let N={MM:(n,0)GL(p,)a  C  map-germand  M|N=Ip}, where GL(p,) denotes the general linear group, and Ip denotes the (p×p) identity matrix.

Now, we define two groups (5)𝒜N,P=N×P,𝒦N=N×N.

Obviously, 𝒜N,P and 𝒦N are subgroups of the groups 𝒜 and 𝒦, respectively. In particular, when N=P={0}, then 𝒜N,P=𝒜. The two groups act on the space Cnp in the following way.

If fCnp, let (ϕ,ψ)𝒜N,P and (M,h)𝒦N; then (ϕ,ψ)·f and (M,h)·f are defined by (6)(ϕ,ψ)·f=ψfϕ-1,(M,h)·f=M·(fh).

Remark 3.

( 1 ) Let f,gCnp. If f and g are 𝒜N,P-equivalent or 𝒦N-equivalent, then f|Ng|N. Set (f,N)={gCnp:f|Ng|N}.

( 2 ) In general, if f and g are 𝒜-equivalent, then f and g also are 𝒦-equivalent. However, this result does not hold for 𝒜N,P and 𝒦N. For example, let f(x1,x2)=(x1,x22) and g(x1,x2)=(x1,x1+x22); where N=P={0}×12. Set  ϕ=id and ψ(y1,y2)=(y1,y1+y2), then (ϕ,ψ)𝒜N,P and (ϕ,ψ)·f=g. So f is 𝒜N,P-equivalent to g. But f and g are not 𝒦N-equivalent.

Let e1,,ep be the canonical basis of the vector space p, and they define a system of generators of CN-module (7)  (CN)×p=CN{e1,,ep}.

For any fCnp, then the germ f induces a ring homomorphism (8)f*:CPCN defined by f*(h)=hf, for any hCP.

This allows us to consider every CN-module as an CP-module via f*.

Let f*mP=f1,,fp0 be the ideal generated by the components f1,,fp0, and let f*(mP) denote the image of mP under f, which is not (in general) an ideal of CN.

For a map-germ fCnp, define (9)T𝒜N,Pf=mNfx1,,fxn+f*(CP){e1,,ep},T𝒦Nf=mNfx1,,fxn+f*mp·(mN)×p.

The notation T𝒜N,Pf and T𝒦Nf are very nearly the tangent spaces to the orbit of germ f under 𝒜N,P-equivalence and 𝒦N-equivalence, respectively.

Definition 4.

Let fCnp; let G be a group acting on Cnp. We say that the map-germ f is k-G-determined if for any germ g(f,N) with jkf(0)=jkg(0), g is G-equivalent to f. If f is k-G-determined for some k<, then it is finitely G-determined. If k=, we say that f is -G-determined.

The purpose of this paper is to characterize the -𝒜N,P-determinacy and -𝒦N-determinacy for map-germs. The main results in this paper are stated in Theorems 5 and 10.

The rest of this paper is organized as follows. In Section 2, we will give a suffcient condition for -𝒜N,P-determined map-germs. In Section 3, we study the 𝒦N-determinacy of map-germs. We will give necessary and sufficient conditions for a map-germ to be finitely 𝒦N-determined or -𝒦N-determined.

Throughout the paper, all map-germs will be assumed smooth.

2. The <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M158"><mml:mrow><mml:mi>∞</mml:mi></mml:mrow></mml:math></inline-formula>-<inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M159"><mml:mrow><mml:msub><mml:mrow><mml:mi>𝒜</mml:mi></mml:mrow><mml:mrow><mml:mi>N</mml:mi><mml:mo mathvariant="bold">,</mml:mo><mml:mi>P</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>-Determinacy of Map-Germs

In this section, the main result is the following theorem.

Theorem 5.

Let fCnp be a map-germ. Suppose that f satisfies the following conditions.

For some r<, mnrmNmNJ(f)+f*mP·mN, where J(f) denotes the ideal in En generated by the determinants of (p×p) minors of the Jacobian matrix df of f.

(mnmN)×pT𝒜N,Pf.

Then f is -𝒜N,P-determined.

In order to prove this theorem, we need the following results.

Lemma 6 (see [<xref ref-type="bibr" rid="B10">10</xref>]).

Let fCnp and M be a finitely generated CN-module. Then M is finitely generated as a f*(CP)-module if and only if dim(M/f*mP·M)<.

Lemma 7.

Let f,gCnp. Let M be a finitely generated (f,g)*(CP×P)-module. Then  M  is finitely generated as a g*(CP)-module if and only if (10)dimMg*mP·M<.

Proof.

The proof is essentially the same as that of Corollary 1.17 in .

Lemma 8.

Let f,gCnp and g-f(mnmN)×p. Then (11)(f,g)*(CP×P)f*(CP)+mnmN.

Proof.

For any hCP×P, then (f,g)*(h)-(f,f)*(h) is in mnmN. So, (12)(f,g)*(CP×P)(f,f)*(CP×P)+mnmN.

Let ϕ:(p,P)(p×p,P×P) be given by ϕ(y)=(y,y).

Since (f,f)*(h)=f*(ϕ*(h)), it follows that (13)(f,f)*(CP×P)f*(CP).

Thus, (11) holds.

Proof of Theorem <xref ref-type="statement" rid="thm2.1">5</xref>.

Suppose that f~(f,N) and jf~(0)=jf(0). Let u=f~-f; then u(mnmN)×p. For any t0[0,1], define (14)g:(n×,(0,t0))p by g(x,t)=f(x)+tu(x) and gt(x)=g(x,t).

Let F(x,t)=(f(x),t) and G(x,t)=(g(x,t),t) denote the map-germs at (0,t0); then F (or G) induces a ring homomorphism: (15)F*:CP×{t0}CN×{t0},hhF=F*(h), where CN×{t0} (resp., CP×{t0}) denotes the ring of function-germs at (0,t0) which are constant when restricted to N×{t0} (resp., P×{t0}).

Set mNkCN×{t0}=mTk and mPkCP×{t0}=mTk.

Let T-𝒜N,PF=tF+ωF, where tF denotes mTf/x1,,f/xn and ωF denotes F*(CP×{t0}){e1,,ep}.

We are trying to show that f~ is 𝒜N,P-equivalent to f. It suffices to show that there exist germs ϕ:(n×,(0,t0))(n,0) and ψ:(p×,(0,t0))(p,0) satisfying the following conditions.

ϕ(x,t)=x and ψ(y,t)=y, for any t sufficiently close to t0, xN and yP.

ϕt0=idn and ψt0=idp.

gt0=ψtgtϕt-1, for any t sufficiently close to t0.

By the method initiated by Mather in , it suffices to show that (16)(mnmT)×pmTgx1,,gxn+G*(mT){e1,,ep}.

To see this, it remains to show that

(mnmT)×pT-𝒜N,PF.

T-𝒜N,PF=T-𝒜N,PG.

If (i) and (ii) hold, then (17)(mnmT)×pT-𝒜N,PG.

Multiplying (17) by G*mT, we get (18)G*mT·(mnmT)×pmTgx1,,gxn+G*(mT){e1,,ep}.

On the other hand, using condition (1) in Theorem 5, and (19)  J(f)·(mN)×pmNfx1,,fxn, we get (20)(mnrmN)×pmNfx1,,fxn+f*mP·(mN)×p.

Obviously, we have (21)(mnmN)×pmNfx1,,fxn+f*mP·(mN)×p.

Since gt-f(mnmN)×p, for each t, this gives that (22)mNgx1,,gxn+gt*mP·(mN)×pmNfx1,,fxn+f*mP·(mN)×p,mNfx1,,fxn+f*mP·(mN)×pmNgx1,,gxn+gt*mP·(mN)×p+mn(mnmN)×p.

Hence, by Nakayama’s lemma we have (23)mNfx1,,fxn+f*mP·(mN)×p=mNgx1,,gxn+gt*mP·(mN)×p.

From (21), it follows that (24)(mnmN)×pmNgx1,,gxn+gt*mP·(mN)×p, for any given t.

Multiplying (24) by mnCN×{t0}, we get (25)(mnmT)×pmTgx1,,gxn+G*mT·(mnmT)×p.

Thus, (16) follows by substituting (18) in (25).

Now we prove the assertion (i).

By hypothesis, we have (26)mnrmNmNJ(f)+f*mP·CN, for some r<.

This means that ideal mNJ(f)+f*mP·CN has finite codimension in CN. Let (27)CN(mNJ(f)+f*mP·CN)={a-1,,a-s}, where aiCN and a-i is the projection of ai in the quotient space, i=1,,s.

Now, set M=CN×{t0}/mTJ(f); then M is a finitely generated CN×{t0}-module. Since F*mP×{t0}·CN×{t0}=f*mP·CN×{t0}+(t-t0)CN×{t0}; (28)MF*mP×{t0}·MCNmNJ(f)+f*mP·CN.

Thus, by Lemma 6, M is finitely generated F*(CP×{t0})-module; that is, (29)M=F*(CP×{t0}){a1,,as}.

In particular, (30)CN×{t0}=mTJ(f)+F*(CP×{t0})CN.

Since (31)T-𝒜N,PF(mNfx1,,fxn+F*(CP×{t0}){e1,,ep}fx1)·F*(CP×{t0})T𝒜N,Pf·F*(CP×{t0})(mnmN)×p·F*(CP×{t0}), and mNJ(f)·(CN×{t0})×pmTf/x1,,f/xn, it follows that (32)T-𝒜N,PF(mnmN)×p·[mTJ(f)+F*(CP×{t0})CN]=(mnmT)×p.

So (i) holds.

Proof of Assertion (ii). Since g-f(mnmT)×p, by (i) we get (33)T-𝒜N,PGT-𝒜N,PF=T-𝒜N,PG+(mnmT)×p.

Applying (25) and (i), we have (34)(mnmT)×pmTgx1,,gxn+G*mP×{t0}·(mnmT)×pmTgx1,,gxn+G*mP×{t0}·T-𝒜N,PF.

By (33) and (34), we have (35)T-𝒜N,PGT-𝒜N,PF=T-𝒜N,PG+G*mP×{t0}·T-𝒜N,PF.

Set E=T-𝒜N,PF/tG, and (35) implies (36)E=T-𝒜N,PGtG+G*mP×{t0}·E.

Now it remains to show that

(a) E is a finitely generated G*(CP×{t0})-module.

For if this holds, then by Nakayama’s lemma, (37)E=T-𝒜N,PGtG.

Hence, T-𝒜N,PF=T-𝒜N,PG.

To prove (a), by Lemma 7, it suffices to show that

(b) E is a finitely generated (F,G)*(CP×{t0}×P×{t0})-module.

(c) dim(E/G*mP×{t0}·E)<.

Set R=(F,G)*(CP×{t0}×P×{t0}). By Lemma 8 and (i), we get (38)R·T-𝒜N,PF[F*(CP×{t0})+mnmT]·T-𝒜N,PFT-𝒜N,PF.

Thus, T-𝒜N,PF is a R-module. Hence, E is a R-module.

Obviously, ωF is a finitely generated F*(CP×{t0})-module; hence ωF is also finitely generated as a R-module.

Moreover, since tF/tGtF is a finitely generated CN×{t0}-module and annihilated by mTJ(gt). Thus, tF/tGtF is a finitely generated CN×{t0}/mTJ(gt)-module. By the argument as equality (29), it follows that CN×{t0}/mTJ(gt) is a finitely generated G*(CP×{t0})-module. Hence, tF/tGtF is a finitely generated R-module.

The earlier argument shows that E is finitely generated as a R-module.

Besides, by (35), we have (39)dimEG*mP×{t0}·E=dimT-𝒜N,PG+G*mP×{t0}·T-𝒜N,PFtG+G*mP×{t0}(T-𝒜N,PG+T-𝒜N,PF)=dimωG+(tG+G*mP×{t0}·T-𝒜N,PF)G*mP×{t0}·ωG+(tG+G*mP×{t0}·T-𝒜N,PF)dimωGG*mP×{t0}·ωG<.

So (b) and (c) hold. This completes the proof.

3. The <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M312"><mml:mrow><mml:msub><mml:mrow><mml:mi>𝒦</mml:mi></mml:mrow><mml:mrow><mml:mi>N</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>-Determinacy of Map-Germs

In this section, by a similar way as , we will give necessary and sufficient conditions for finitely 𝒦N-determined map-germs and -𝒦N-determined map-germs.

Theorem 9.

Suppose that fCnp. The following conditions are equivalent.

f is finitely 𝒦N-determined.

(mnrmN)×pT𝒦Nf, for some r.

Proof.

Let Jr+1(n,p) denote the set of (r+1)-jets at 0 of elements in (f,N). If f is r-𝒦N-determined, then for any g(f,N) with the same r-jet as f, the 𝒦N-orbit of f contains g; that is, (40){g(f,N):jrg(0)=jrf(0)}{g(f,N):g  and  fare  𝒦N-equivalent}.

Taking (r+1)-jets on both sides, we have (41){jr+1g(0)Jr+1(n,p):jrg(0)=jrf(0)}{jr+1g(0)Jr+1(n,p):g  and  fare  𝒦N-equivalentjr+1}.

Taking tangent spaces at jr+1f(0) on both sides, we have (42)(mnrmN)×pT𝒦Nf+(mnr+1mN)×p.

By Nakayama’s lemma, this implies (43)  (mnrmN)×pT𝒦Nf.

Then (1) implies (2).

Conversely, let t0[0,1] be fixed and g(f,N) with jr+1f(0)=jr+1g(0). Define (44)F:(n×,(0,t0))p by F(x,t)=f(x)+t(g(x)-f(x)) and Ft(x)=F(x,t).

We are trying to show that g is 𝒦N-equivalent to f. Since F0=f and F1=g, and [0,1] is connected, it suffices to show that Ft is 𝒦N-equivalent to Ft0, for all  t  sufficiently close to t0.

Since Ft0(x)-f(x)=t0(g(x)-f(x)) and g-f(mnr+1mN)×p, from condition (2), it follows that (45)T𝒦NFt0T𝒦Nf, and T𝒦NfT𝒦NFt0+mn·(T𝒦Nf).

Since T𝒦Nf is a finitely generated En-module, and mn is the maximal ideal of En, by Nakayama’s lemma, we get (46)T𝒦NFt0=T𝒦Nf.

Hence T𝒦NFt0 also satisfies condition (2). So (47)(mnrmNEn+1*)×pT𝒦NFt0·En+1*, where En+1* denotes the ring of smooth function-germs in variables (x,t) at the point (0,t0).

Let mn+1* denote the maximal ideal of En+1*.

Let T-𝒦NF=mNEn+1*F/x1,,F/xn+F*mp·(mNEn+1*)×p. Obviously, T-𝒦NF is a finitely generated En+1*-module.

By the same argument as (46), we have (48)T-𝒦NF=T𝒦NFt0·En+1*.

By (47) and (48), we get (49)(mnrmNEn+1*)×pT-𝒦NF.

Since F/t=g-f(mnr+1mN)×p, this means that there exist germs Xi in mNEn+1*, i=1,,n, such that (50)(g-f)+i=1nXi(x,t)FxiF*mp·(mNEn+1*)×p.

Thus, we can find a germ of vector field X in n× of the following form: (51)t+i=1nXi(x,t)xi, such that DF(X)F*mp·(mNEn+1*)×p.

That is, we can find a (p×p) matrix A(x,t) with entries in mNEn+1* such that (52)DF(X)=A(x,t)·F(x,t).

By integrating the vector field X, we get a one-parameter family of local diffeomorphisms ϕt in N. Thus (53)ddtF(ϕt(x),t)=tF(ϕt(x),t)+i=1nXi(ϕt(x),t)Fxi(ϕt(x),t)=A(ϕt(x),t)·F(ϕt(x),t).

Hence, for fixed xn, F(ϕt(x),t) is a solution of the differential equation y˙=A(ϕt(x),t)y with initial condition y(x,t0)=Ft0(x).

Since the solution of this differential equation is unique and of the form (54)y(x,t)=B(x,t)·y(x,t0), with B(x,t) as an invertible matrix, B(x,t0)=Ip and B(x,t)=Ip for each xN. Thus (55)Ft(x)=B(ϕt-1(x),t)·(Ft0ϕt-1(x)).

Thus, Ft and Ft0 are 𝒦N-equivalent for all t sufficiently close to t0, which completes the proof.

Theorem 10.

Suppose that fCnp. Then f is -𝒦N-determined if and only if (56)  (mnmN)×pT𝒦Nf.

Proof.

“Only if”: if f is -𝒦N-determined, by the definition, we get (57)f+(mnmN)×p𝒦N·f.

Taking tangent spaces at f on both sides, we have (58)Tf(f+(mnmN)×p)Tf(𝒦N·f).

Note that (mnmN)×pTf(f+(mnmN)×p), and Tf(𝒦N·f)T𝒦Nf. Hence, (59)  (mnmN)×pT𝒦Nf.

“If”: the proof is the same as that of Theorem 9.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Grant no. 11271063) and supported in part by Graduate Innovation Fund of Northeast Normal University of China (no. 12SSXT140). The authors would like to thank the referee for his/her valuable suggestions which improved the first version of the paper.

Arnol'd V. I. Goryunov V. V. Lyashko O. V. Vasil'ev V. A. Singularity theory II: classification and applications Encyclopaedia of Mathematical Sciences 1993 39 1 235 Wang Z. Pei D. Chen L. Kong L. Han Q. Singularities of focal surfaces of null cartan curves in minkowski 3-space Abstract and Applied Analysis 2012 2012 20 823809 http://dx.doi.org/10.1155/2012/823809 MR2975355 ZBL1253.53005 Wilson L. C. Infinitely determined map germs Canadian Journal of Mathematics 1981 33 3 671 684 10.4153/CJM-1981-053-3 MR627650 Wilson L. C. Mapgerms infinitely determined with respect to right-left equivalence Pacific Journal of Mathematics 1982 102 1 235 245 MR682054 10.2140/pjm.1982.102.235 ZBL0454.58005 Brodersen H. On finite and infinite Ck-A-determinacy Proceedings of the London Mathematical Society 1997 75 2 369 435 10.1112/S0024611597000397 MR1455861 Wall C. T. C. Finite determinacy of smooth map-germs The Bulletin of the London Mathematical Society 1981 13 6 481 539 10.1112/blms/13.6.481 MR634595 ZBL0451.58009 Sun B. Wilson L. C. Determinacy of smooth germs with real isolated line singularities Proceedings of the American Mathematical Society 2001 129 9 2789 2797 10.1090/S0002-9939-01-06068-3 MR1838804 ZBL0971.58024 Grandjean V. Infinite relative determinacy of smooth function germs with transverse isolated singularities and relative Łojasiewicz conditions Journal of the London Mathematical Society 2004 69 2 518 530 10.1112/S0024610703004861 MR2040619 ZBL1054.58030 Thilliez V. Infinite determinacy on a closed set for smooth germs with non-isolated singularities Proceedings of the American Mathematical Society 2006 134 5 1527 1536 10.1090/S0002-9939-05-08253-5 MR2199202 ZBL1090.58022 Bulajich R. Gómez J. A. Kushner L. Relative versality for map germs Unione Matematica Italiana 1987 1 7 305 320 Gaffney T. du Plessis A. More on the determinacy of smooth map-germs Inventiones Mathematicae 1982 66 1 137 163 10.1007/BF01404761 MR652651 ZBL0489.58004 Mather J. N. Stability of C mappings III: finitely determined map-germs Institut des Hautes Études Scientifiques. Publications Mathématiques 1969 35 127 156 MR0275460 Brodersen H. A note on infinite determinacy of smooth map germs The Bulletin of the London Mathematical Society 1981 13 5 397 402 10.1112/blms/13.5.397 MR631095 ZBL0504.58009