Zariski

Studies in Equisingularity I Equivalent Singularities of Plane Algebroid Curves
Author(s): Oscar Zariski

Source: American Journal of Mathematics, Vol. 87, No. 2 (Apr., 1965), pp. 507-536
Published by: The Johns Hopkins University Press
Stable URL: https://www.jstor.org/stable/2373019
Accessed: 21-03-2019 18:28 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
The Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and
extend access to American Journal of Mathematics
This content downloaded from 200.130.19.153 on Thu, 21 Mar 2019 18:28:57 UTC
All use subject to https://about.jstor.org/terms
STUDIES IN EQUISINGUARITY I
EQUIVALENT SINGUARITIES OF PLANE ALGEBROID CURVES.
By OSCA,R ZARISIu.1
With this paper we initiate a series of investigations on the concept of

equisingularity of an algebraic variety V, along an irreducible (singular)
subvariety W of V, at a given point of W. The actual definition of this
concept, at least in certain special situations (and, in particular, in the case
codvW =1), was given by us in an earlier paper ("Equisingular points on
algebraic varieties," seminari dell'Istituto Nazionale di Alta Matematics 1962-
63, Roma, 1964, pp. 164-177), and will be presented and discussed in full
(i. e., including proofs) in our next paper in this series. The present paper
is preliminary in nature. It deals first of all with various ways of defining
equivalent singularities of plane algebroid curves (?? 1-4 bis). We define equiv-
alence by induction on the number of locally quadratic transformations which
are necessary to resolve the given singularity. No explicit (or implicit) use
of Puiseux expansions is made, and, in fact, the case of non-zero characteristic
(where Puiseux expansions are not available) is included in our treatment.
We give 3 different definitions of equivalence, and we prove that they are all
equivalent to each other. [It is our belief, however, that in the case of non-
zero characteristic our treatment is not definitive, as it lumps together, in
one equivalence class, singularities which, on the basis of some internal evi-
dence, should be further subelassified; we propose to come back to the case
p7 =O in a subsequent paper in this series.]
In ? 5 we establish some technical results on derivations in complete

integrally closed, domains (Theorems 3 and 4) and we obtain, as a conse-
quence, a criterion for such a domain to be a power series ring over a subring.
In ?? 6-7, the geometric results of ??1-4bis and the purely algebraic

results of ? 5 are applied toward the derivation of a discriminant criterion
which enables one to decide whether or not, given an analytic family of alge-
broid curves Ct: f(x, y; t) = 0, the special member CO of the family has a
singularity which is equivalent to the singularity of the generic member Ct.
Received October 26, 1964.

* This research was supported in part by the Air Force Office of Scientific Resea
and in part by the National Science Foundation.
507
508 OSCAR ZARISKI.
This criterion will be used in our next paper dealing with equisingularity
in codimension 1.
1. Algebroid plane curves. Let kc be an algebraically closed field (of

arbitrary characteristic) and let k[[x,y]] be the ring of power series of two
independent variables x, y, with coefficients in 7k. By an algebroid plane
curve C we mean a local ring of the form kc[[x,y]]/(f), where f= f(x,y) is
a non-unit element of Ic [ [x, y] ], free from multiple factors. We shall say
that f(x, y) = 0 is an equation of the curve C (the power series f is deter-
mined by C uniquely, to within an arbitrary unit factor in Ic [ [x, y]]). The
point P: x = y = 0 will be referred to as the origin of C.
If f(x, y) is an irreducible element of kI[[x, y]], C is said to be an
irreducible algebroid curve. If f is reducible, f = 102 . . . Oh, where each
qi is an irreducible element, then the h irreducible algebroid curves y,: =i 0
are the irreducible components or branches of C.
Assume that the power series f begins with terms of degree s: f = f, (x, y)
+ terms of degree > s, where f, is a homogeneous polynomial of degree s
(f8 =leading form of f). Then one says that P is an s-fold point of C,
h
and we write s-rmp (C). Clearly mp (C) = ,mp(y).

j=1.
It is known that if f is irreducible then f5 is the power of a linear form:

= f (ax + by)S (Hensel's lemma). The line (or direction) ax + by = 0 is
then called the tangent line of C. If C is reducible, then the tangent lines
of the various irreducible components yi of C shall be, by definition, the
tangent lines of C (the number of distinct tangent lines of C is therefore ? h).
2. The quadratic transform of C. Let 1h: aix + by -0 be the tangent

line of yi (i = 1, 2, * * *, h). Without loss of generality we may assume that
the bi are all different from zero. We set x'=x, y' =y/x, and
f (xI, x'y) - x'sf' (x', y).
Then f' is a power series in x', y', and in fact, f' is a power series in x', whose
coefficients are polynomials in y' (with coefficients in k):
( 1 ) f' (x , y) =fs (1, y') + x'fS+l (1, y') + X'2fS+2 (1, y') +.
Let t be the number of distinct tangent lines Pv of C, and let y - a,xv =0

(v= 1, 2, ., t) be the equations of these lines (so that for each i 1, 2, , h
the ratio - ai/bi is equal to one of the x,). Let I, be the set of integers i
(1 ? i_? h) such that - ai/bi == a, and let C, be the union of the irreducible
branches yi, i C I,. If we set Fl,(x, y) I= H cj(x, y), then C, is the algebroid
i E lp
STUDIES IN EQUISINGULARITY I. 509
curve defined by the equation F,(x, y) 0. The irreducible branches of C,

are the yi, with i C Iv, and they all have the same tangent line p,. We have
t
f(x, y) =- H F(x, y), so that C is the union of the t algebroid curves C,.
V=1
We call the C, the tangential components of C.
Let ri ==mp(y),sV =mp(C.)( E ri),

i E Iv
Oi (X,Y) (r(X, Y) + Oi,,+l (X, Y) +***

Fv (x, y) Fv,s (x, y) + Fv,sv+l (x, y) +--
Since pir6(X,y) = (y - aVX) r for iC Iv and Fv,8,(x, y)= (y avx)sv, it

follows that
,4i(X', xy') - X ri '(X, y'),

F, (x', x'y') - x'svF,`(x', y'),
where, for i C IV:
i' (X'7 y') (y' -av) r + X'(fi,rt+l(1l y') + *

FVI(XI' y') 17 4i'(x', y') = (y'- zv) sv + x'Fv,sv+1(1, y') +**
i E Iv
We regard qi' and Fv' as power series in x' and y'- av and we denote by 7i'
and Cv' the algebroid curves, with origin at the point Pv'(0, av), defined
respectively by the equations Oi'(x', y') = 0 and F/'(x', y') = 0. The curves
yi' (i C IV) are irreducible branches of Cv'. We note that
t
f'(x', y') II1 Fv' xl 7 y').

v=l
Hence the equation f'(x', y') = 0 may be regarded as representing the set of
(disjoint) algebroid curves C1', C2', * , Ct', having respectively as origins
the points Pl', P2', ,Pt'. We denote by T the locally quadratic trans-
t
formation (with center P) defined by x' = x, y' y/x, we set C' U Cv

p=1
and we call C', CV' and y( the proper quadratic transforms of C, Cv an

respectively:
C' T(C)~
-/i' T (yi) .
We shall refer to C,', C2'y , Ct' are the connzected components of C'.
Note. Although the above definitions are based on a specific (and

explicit) choice of regular parameters x, y of the point P of the (x, y) -plane,
they are easily seen to have an intrinsic invariantive meaning, as they could
510 OSCAR ZARISIZI.
have been formulated in terms of local algebra (see Zariski-Samuel, Commu-

tative Algebra, v. 2, Appendix 5). In such an intrinsic formulation one
starts with a complete, noetherian, equicharacteristic local ring o, of dimen-
sion 1, whose residue field is algebraically closed and whose maximal ideal
m has a basis of two elements. If one, furthermore, assumes that o has no
proper nilpotent elements, then-for any given choice of a field k of repre-
sentatives in o and of a basis ($, q) of m-the local ring o will be the local
ring of a unique plane algebroid curve, defined over Ic. The notion of the
quadratic transform of o is, however, independent of the choice of Ic, t and 7q.
Similarly, given two local rings o, o, satisfying the above conditions,
aind having the same (or isomorphic) residue fields, then the definition of
equivalence of o and o given in ? 3 below, remain meaningful and are
independent of the choice of k, e and r.
3. Three definitions of equivalence of algebroid curves. Let D be

another plane algebroid curve, with some origin Q. We assume that C
cand D have the same number h of irreducible branches, and we denote by
81& 82 . . . 7 87h the irreducible branches of D.
DEFINITION 1. A (1, 1) mapping 7r of the set of branches y1, 72, . . , y7h

of C onto the set of branches 81, 82, , 8h of D is said to be a tangentially
stable patrtng 7r: C -* D between the branches of C and tlhose of D, if the
following condition is satisfied: given any two branches yi and yj of C, the
corresponding branches -r(yi) and r(yj) of D have the same tangenlt if and
only if 7yi and yj have the same tangent.
Assume that there exists a tangentially stable pairing 7r: C-* D between
the branches of C and the branches of D. Then it is clear that C and D have
the same number t of distinct tangent lines and that -r induces a (1, 1) mapping
of the set {pl, p2, P * pt} of tangent lines of C onto the set {ql, q27 * * *, pt}
of tangent lines of D. We choose our indexing of these tangent lines in
such a way that p, and q, are paired in this induced mapping, and we deno
by C, (resp. D,) the tangential component of C (resp., D) associated with p,
(resp., q,). Then it is clear that for each v = 1, 2, * * *, t, 7r induces a (1,
mapping 7r,: C0 -* D,, of the set of branches of C, onto the set of branch
of D, (the pairing 7r, is trivially tangentially stable, since both C7r and D,
have only one tangent line).
Let 7r and lr,1 be as above (7r-tangentially stable), let T be a locally
quadratic transformation with center at the origin P of C and let S be a
locally quadratic transformation with center at the origin Q of D. Let
C' T(C), Cv' =T(1Cv), D' =S(D), Dv' =S(D,) be the proper transfor
It is clear that 7r, induces a (1, 1) mapping 7rv' of the set of branches of C/'
onto the set of braniches of Dv'. Namely, if we assume that the branches of
C and D have been so inidexed that -rT(ye) = Si, for i = 1 2, . , h, then, in
the notations of ? 2, we set rv' (yi') = 8' for all i E Iv, where yt' T (yj) alnd
V =S (8s). The pairing 7rv': Cv'-- D,' between the branches of C,' and the
branches of Dv' is, however, not necessarily tangentially stable.
An algebroid curve C is regular if its origin P is a simple point of C,
i.e., if mp(C) =1. If P is a singular point (i.e., if mp(C) >1), then we
can resolve the singularity of C at P by a finite number of locally quadratic
transformations. By a sequence of successive quadratic transforms of C' we
mean a sequence {C, C', C", * , C(t, * } of algebroid curves C(t) such that
for each i, C(i+') is a connected component of the proper transform of C(t)
under a locally quadratic transformation whose center is the origin of
C(M)(C() =C). The fact that the singularity of C can be resolved can then
be stated as follows: there exists an integer N such that in any sequence O01
successive quadratic transforms of C, the curves C(4) are regular if i ? N.
We denote oa(C) the smallest integer N with the above property (a (C) 0
if and only if C itself is a regular curve).
It is clear that if C1', C,', - ., Ct' are the connected components of the
proper quadratic transform T(C) of C, and if cro(C) > 0, then or(Cv') < v(C)
for v = 1,2, *, t. Our first definition of equivalence of algebroid curves
(or-xwhat is the same-of equivalence of algebroid singularities) proceeds by
induction on o(C).
Let r: C -* D be a pairing between the branches of C and the branches
of D (it is already assumed that C and D have the same number h of
branches). If C is regular (whence ar(C) = 0), then C (and therefore also D)
has only one branch, r: C-- D is uniquely determined, and we say that 7r
is an (a)-equivalence if also D is a regular curve. Assume that for all pairs
of algebroid curves r, A with the same number of branches and such that
ar(r) <ov(C) it has already been defined what is to be meant by saying that
a pairing r -> A between the branches of r and the branches of A is an (a) -
equivalence. Then we define an (a) -equivalence between C and D as follows
(we use the notations introduced earlier in this section):
DEFINITION 2. An (a) -equivalence 7r: C -> D is a pairing r between the

branches of C and the branches of D having the following properties:
1) Xr is tangentially stable.
2) If X, 7r(y) (i==1,2,. * *,h), then mp( y)=mp(Si).

3) The pairing 7rv': Cv'-> Dv' (v = 1, 2, ,t) is an (a) -equivalence.
512 OSCAR ZARISKI.
Example. P is said to be an ordinary s-fold point of C if s = mp(C)

and if the number t of distinct tangents of C is exactly equal to s. It follows
that if P is an ordinary s-fold point of C, then C has exactly s branches
yi, 72, * Ys, these branches have distinct tangents pl, p ,* p., and each
yi is a regular algebroid curve (since s E mp (ye) and since np (y) ? 1).

j=1
The proper quadratic transform C' = T(C) of C has then exactly s connected
components Cl', C2', , Cs0 where Ci' = yi' = T (yj), and each Ci' is a
t
regular curve (since it follows, quite generally, from (1), ? 2, that E mp,, (Cv')
jV=l
_mp(C), for any algebroid curve C). Now, suppose that P is an ordinary
s-fold point of C and that a pairing r: C-* D between the branches of C
and D satisfies conditions 1) and 2) of Definition 2. Then it is clear that
also Q is an ordinary s-fold point of D. Condition 3) is now automatically
satisfied, since C.' and D,' are regular curves. Then 7r is an (a)-equivalence.
Conversely, it is clear that if P is an ordinary s-fold point of C and if Q
is an ordinary s-fold point of D, then any pairing 7r: C -- D between the s
branches of C and the s branches of D is an (a)-equivalence.
We now proceed to our second definition of equivalence between algebroid
singularities. If T is our quadratic transformation, with center P (see ? 2),
then T blows up P into the line x'- 0 of the (x', y')-plane. We denote this
line by (E and we refer to &' as the exceptional curve of T. If C, is a
tangential component of C and C,' = T (C,) is the proper T-transform of C>,
then 6' contains the origin P.' of C,', but 6' is not a component of C,'.
We denote by C,'* the algebroid curve C,' U &' and we call C,'* the total T-
transform of C>; in symbols: C>'* = T{C }. We set C'* = T(C) U &' and
we call C' the total T-transform of C. Note that mnp, (Cv'*) is always ; 2.
It is known that after a finite number of successive quadratic trans-
formations one can reach a stage where the total transform of C has only
ordinary double points. More precisely: there exists an integer N ? 0
(depending on C) with the following property: if {C, C,*, C",* , C()
is any sequence of algebroid curves such that for any i we have C(i+1)*
- c(i+1) U S+1 , where C(U+1) is a connected component of the proper quadratic
transform T(M) (O(t)*) of C(i)* (T(M) being a qualdratic transformation with
center at the origin P(M) of C(t)*) and 6(i+1) is the exceptional curve of T(M,
then for ? N the origin PM() of C(t) is an oridinary dlouble point of C(iMN.
We denote by ar' (C) the smallest integer N having the above property.
It is clear that a: (C) =0 if and only if the origin P of C is an ordinary
double point of C. If C is a regular curve then a strict interpretation of our
definition of v (C) would require to set oa* (C) = 1. However, we agree to

set u* (C) 0 also if C is a regular curve (this could also have been achieved
by a slight change in our general definition of o* (C)). It is easily seen that
u* (C)= 1 if and only if P is an ordinary s-fold point of C and s > 2. If P
is a tacnode of the first kind (i.e., if C consists of two regular branches
having simple contact), then the total transform C*= T{C} of C has an
ordinary triple point, and hence r* (C) = 2 in this case. On the other hand,
if P is an ordinary cusp of C, then C'* T-'C} has a tacnode of the first
kind (while the proper transform T(C) of C is regular), and thus a* (C) 3
in this case.
Let C and D have the same number of branches and let -r:: C -> D be a
pairing of the branches of C with the branches of D. If u* (C) =- 0, i. e.,
if P is either a simple point or an ordinary double point of C, then we shall
say thatr is a (b) -equivalence between C and D if and only if also u* (D) = 0,
i. e., if and only if the origin Q of D is a simple point or an ordinary double
point of D according as P is a simple point or an ordinary double point of C.
Assume that for all pairs r, A of algebroid curves, with the same number
of branches, such that a* (r) <,* (C), it has already been defined what is
meant by saying that a pairing r -- A between the branches of r and the
branches of A is a (b)-equivalence. Then we define a (b)-equivalence between
C and D as follows:
DEFINITION 3. A (b)-equivalence r: C->D is a pairing r between the

branches of C and the branches of D having the following properties:
1) 7r is tangentially stable.
2) The pairings 7-,,': CV' >D,' (v== 1,2, . ,t) are (b) -equivalences.
3) If S' and E' are the exceptional curves of the quadratic trans-
formations T and S respectively (having centers at P and Q), if
CvC'* CvIUel Dv'* = D,' U E', and if we extend the pairing 7rv'
to a pairing 7r,'*: C,'*---D,'* by setting v (8') =E', thlen r,'*
is a (b)-equivalence.
Note that conditionls 1) and 2) of this definition are identical with the
condtions 1) and 3) of Definition 2; condition 2) of Definition 2 has been
deleted and has been replaced in Definition 3 by conditionl 3). Thus the
equality of the multiplicities of corresponding branches under 7r is not
explicitly postulated in Definition 3.
We now give a third definition of equivalence of algebroid singularities,
which we shall refer to as formal equivalence. Again we proceed by induc-
tion on u* (C), where we agree that if v* (C) 0 formal equivalence coincides
with (b) -equivalence.
17
514 OSCAR ZARISKI.
DEFINITION 4. Given two algebroid curves C, D, having the same nutmber

of branches, we say thtat C and D are formally equivalent if there exists a
tangentially stable pairing 7r: C -*19 between the branches of C and the
branches of D such that (in our previous notations):
1) C07 and D^' are formally equitvalentt (v = 1, 2, , t).

2) C.'* and D.'* are formally equivalent (v = 1, 2, t).
Note that this definition does not say anything about the nature of the
pairings 7rv': C0'7*D,1' and 7*: Cv': -> D,'* induced by -r. Condition 1)
merely requires that there exist, for each v = 1, 2, *,t, some tangentially
stable pairing p^': Cv0 -D,,9 satisfying the conditions of the above inductive
definition; and similarly, condition 2) requires that there exists a tangentially
stable pairing p,'*: Cv* - D,,'* satisfying similar conditions. It is not even
required that p,'* be an extension of p7'. For this reason, Definition 4 is
the most subtle (and also the weakest) of our three definitions of equivalence.
The fact (established below) that these three definitions are all equivalent to
each other is therefore not devoid of interest.
W shall use the following notations:
If 7r: C-> D is a pairing between the branches of C and the branches of
a b
D, then we write 7r: C- >D or 7r: C >D according as 7r is an (a)-
equivalence of a (b) -equivalence. We shall write C = D if C and D are

formally equivalent.
4. Proof of equivalence of Definitions 2, 3 and 4. We recall first a

few (well-known) elementary facts about the intersection number (C, D)
of algebroid curves (having the same origin P) and about the behaviour of
this number under a quadratic transformation T centered at P. We assume,
of course, that C and D have no branches in common.
If n =mp(C) and m mp((D) and if we assume that the line x 0
is not tangent to either C or D, then, by the Weierstrass properation theorem,
we may assume that C and D are defined by f(x,y) =0 and g(x,y) = 0,
where f and g ar e monic polynomials in y, of degree n and m respectively, w
coefficients which are power series in x. If y1, y2, , y,y are the roots of
f (x, y) and k1, 2,. . 9, , are the roots of g (x, y) (in an algebraic clos
of the field k{{x}} of meromorphic functions of x), then (C, 1), is defined
n M
as the order of the power series II TI (ytT- ij) in x (this power series is an
j=1 j=1
element of k[[x]] since the yi and ij are integral over lc[[x]]). Since
and m are the degrees of the leading forms of f and g, it follows that
f _ yn + xal (X)ynl1 + * * . + Xnan

g ==ym + xbl(x)ym-l +* + xmbn (x),
with ai(x) and bj(x) in kc[[xl]. If f'(x',y')-O 0 and g'(x',y') =0 are the
equations of the proper transforms C'` T(C) and D' T(D) of C and D
(see ?2), then
ft y/n + a, ( x') y/n-1. + ***+ a, (xZ )
g' y'mn + bbi(x')y'm-l + + bm(x')

where x'= x. Thus, if Yi', Y2', * Yn' and 91',92?', , im' are the roots
of f'(x',y') and g'(x',y') (regarded as polynomials in y'), then the yi' and
gj' are still integrally dependent on 7c[[x]]. Since yi xyi' and - =4j',
it follows that if Cj, C2', ,Ct' are the connected components of C' and
D1', D2', . ., D/ are the connected components of D', then
t r
(2) (C, D) mn + E E ( (Cv', D,').

v=1 A=1
Here (CV, D,D') > 0 if and only if C0/ and D,' have the same origin. By (2)
if follows that (C, D) > inn, with equality if and only if C and D have no
common tangents. From the above expression of f' it follows also that if
6': x' 0 is the exceptional curve of T, then
t
(3 ) ( ',,' ( Cv', 6') n (~ MP ( C)).

v=1
These are all the facts that we shall need in the sequel.
Let now C and D be algebroid curves, with origins P and Q, and with
the same number of branches.
a
LEMMA 1. Let T-: C->D be an (a)-equivalence between C and D
and let r be a regular curve through P and A a regular curve through Q.
Assume that r is not a branch of C, that A is not a branch of D, and that
for every pair (y'i, 8) of corresponding branches under 7r we have
(4) (7yi,r) (8j,5), A) i 1,2, ,h.

Then the pairing p: C U rP - D U A betwee
branches of D U A which is the unique exte
(a) -equivalence.
Proof. (by induction on &* (C U r)).
If & (C u r) = o, theni C is a regular curv

and r have distinct tangei Is. I Ience, by
516 OSCAR ZARISKI.
C and D are (a) -equivalent, also D is a regular curve, and it follows by (4)
that (D, A) = 1, whence o* (D U A) 0. Therefore p is an (a) -equivalence.
In the general case, we observe that (yi, r) mp (y), with equality if
and only if yi and r do not have the same tangent. Similarly, for (8s, A)
and mQ (v). Since r is an (a) -equivalence, we have mp (yi) = mQ (8). Hence
it follows from (4) that yj and r have the same tangent if and only if As and
A have the same tangent. Therefore p is a tangentially stable pairing. Since
mp (r) = mQ (A) = 1, it is clear that any two corresponding branches under
p have the same multiplicity. Thus p satisfies conditions 1) and 2) of the
definition of an (a) -equivlance (Definition 2, ? 3). We now check condition
3) of that definition.
Let T(r) ==P', S(A) =A'. We consider two cases.
First Case. r is not tangent to any of the branches of C (and hence A

is not tangent to any of the branches of D). In this case, C1', 0p', , Ct'
and r' are the connected components of the proper transform T(C U r),
and D',,D2'< D ,Dt' and A' are the connected components of S (D U A).
If pp': U9' D,' (v 1,2, , t), pt+':' r A' are the pairings induced by
p, then p,' r^is an (a) -equivalence (since Xr is an (a) -equivalence), and
pt+i' is trivially an (a) -equivalence (since r' and A' are regular curves).
Thus condition 3) of Definition 2 is satisfied.
Second Case. The tangent of r coin1cides with one of the t tangents

Pi p2, * * *, pt of C, say pi is the tangent of F. In that case, the corres-
ponding tangent line q1 of D (under 7r) is the tangent of A. The tangential
components of C U r are now C] U r, C2,< * *, Ct, and the corresponcling tan-
gential components (under p) of D U A are DU A, D2, , Dt. The COnl-
nected components of T(C u r) and S(D U A) are now Cl' U r', C21, ,Ct
and Dl' U A', Dt, , Dt'. Consider the iliduced pairing
pl': C1 u r-> Di' u A', pv': C'-* D,' (v =2, 3, t).
We have p'=- 7r' for v > 2, and p,' is an (a) -equivalence for
Consider now pl'. It is an extension of 7rw', with pl' ((r) = A'
branch of Cl' (i C 11) and let 8! be the corresponding branch
7ri'). Then yj'==T(yj), 8j'=S(8j), with 8=w7r(yj). By (2), we have
(yi, r) mp(yi) + (y/', rI), (8, A) == mp(8c) + (8i', A'). Since mp(yi) mp(si)
and (yi, r) = (S, A), it follows that (yi', r') (si',i?A"'). Since a* (Cl' U r')
< u-(C U r) (if o-* (C U r) #z 0), it follows, by our induction hypothesis, that
pl' is an (a) -equivalence. This completes the proof of the lemma.
LEM-MA 2. If ,: C->D and S(i) s (i=1,2 * *,h), then

Y,yi) = (8j,8j) for all i=/=j.
Proof. We have (yi, yi) nmp(y) mp(yj) + (yi, yjf), where yi'=T(yi)
(8, ) =mQ (8i) mQ (8i) + (8,' 8j'). Here (yi', yj') is to be replaced by zero
if /i' and yj' have distinct origins (and similarly for (8i', 8j') ). Since
mp(75) = mQ (8t) for all i, the lemma follows by iniduction on ao(C).
THEOREM 1.
a
(a) If 7r: C ->D is an (a)-equivalence, then Xr is also a (b)-equiv-
alence, and conversely.
(b) If there exists an (a) -equivalence (or a (b)-equivalence) r: C-- D,

then C and D are formally equivalent.
Proof. (a) The assertion is true if v* (C) = 0. We therefore use

induction on au (C).
Assume that 7r: C _ D is an (a) -equivalence. Using the notations of

Definition 2, we have the (a) -equivalences -r,': C,'> D,v', v =- 1, 2 . . t.
Therefore, by our induction hypothesis, the 7-r' are also (b) -equivalences.
So we have only to show that (using the notations of Definition 3) the
pairings 7rw'*: CV U C Dv' U E' are (b) -equivalences. Since O,* (C' U 6')
< (C) (if u* (C) A 0), it is sufficient to show (by our induction hypothesis)
that 7r,'* is an (a)-equivalence. Now rv': Cv' - >D>' is an (a)-equivalence,
the curves C' and E' are regular, and 7rT"' is an extension of 7rV', with
7v'* (C') E= . So, by Lemma 1, it is sufficient to show that if y/=rv'(y)
are corresponding branches of Cv' and Dv', under 7rv', then (yi', C') = (8', E').
Now, y(' and 81' are the proper transforms of the branches yi, 8A of C and D
respectively, and 3i ,7r(ys). Since 7r is an (a)-equivalence we have mTp(yi)
mp(8t), and since, by (3), we have mp(yi) E(7,C') and nQ(8) - (8', E'),
the equality (y/', C') = (8i', E') is proved.
Coniversely, assume that r: C -> D is a (b) -equivalence. Then, by
induction, the (b)-equivalences lrT': Cv'-->Dv' are also (a)-equivalences.
There thus remains only to show that if 8 ,7r (yt) then mp (y.)=mQ(8t).
Now, consider the (b)-equivalences 7rv': Cv' U l'- D' U E' (which are
extensions of the 7rv'). By induction, Wv" is also an (a)-equivalence. -leilce,
by Lemma 2, we have (y,C') = (8, E') for any i =,2, ,h. This
implies, by (3), that mp(yi) = mQ (8), as asserted.
From now on we mav replace the terms (a) -equivalence and (b) -equiv-
518 OSCAR ZARISKI.
alence by the single term of equivalence (to be distinguished, for the moment,
from formal equivalence).
(b) Assume that there exists an equivalence r: C = D. We have

then the equivalences 7rp: C,' > D,' and 7r'p*: CG' U S' > D-I U E'. Hence,
by induction on Ov(C), it follows that C,' and D,' are formally equivalent
and that also C,' + S' and D,' + E' are formally equivalent. Thus C and D
are formally equivalent.
f
4 bis. Continuation: proof that formal equivalence C -D implies
the existence of a pairing equivalence =: C -> D.
LEMMA 3. Let C and D be algebroid curves having the same number

of branches, let 662< * *nm denote certain branches of C and assume that
these branches c0 are regular and have distinct tangents. Similarly, let
el, e2, . *, em be regular branches of D, with distinct tangents. Let
c = r u 61 u e2 U . . . U Em, D=AUeU e2 U . . . U e,,.
Assume that r and A are equivalent and that for each a = 1, 2, ,n m also
r u 'E u C2 U ea and A U el U e2 U . . . U ea are equivalent. Then there
exists a pairing equivalence 7r: C -->D such that 7r(E) ea, for a 1, 2,
. M.
Proof. We consider two cases, according as m 1 or rm> 1.
Case I. m = 1. We divide this case into two subcases, according as

the tangent of E (=--=) is not or is one of the tangents of C.
Case Ia. C -r u , D = A U e, and the tangent of E is not a tangent

of r. Since r --A, r and A have the same number of distinct tangent lines.
Similarly, r u E = A U e implies that r u E and A U e have the same number
or distinct tangent lines. But since the number of distinct tangents of r u E
is one greater than the number of distinct tangents of r, the same is true
of A U e and A. Hence the tangent of e is different from any tangent of A.
It follows at once from the definition of (a)-equivalence that given any
equivalence -r: r - A, the extended pairing 7r: r u s_ A U e defined by
7r() =--e is also an eqiivalence.
Case lb. The tangent of E coincides with one of the tangents of r (and
hence the tangent of e coincides with one of the tangents of A, by the
reasoning of the Case Ia).
Let r1, r2, , rt and A1, A2, , At be the tangential components of
r and A respectively, arranged in some arbitrary order, except that we assume
that the tangent of e coincides with the tangent of r1 and the tangent of e
coincides with the tangent of A1. Then the tangential components of C are
r1 U s, r2, ***, rt, and the tangential components of D are A1 U e, A2, , At.
We shall first show that
(5) r1 A1 and r1UesA1Ue.
Let a be the number of curves in the set {r1 U r2, P , rt} which are
equivalent to r1 (a ? 0). Since r1 u c r1, it follows that a + 1 is the
number of curves in the set {r1, r2, * , rt} which are equivalent to r1.
Since r =-A, any r* is equivalent to at least one Aj, and any Ai is equivalent
to at least one rj. Since equivalence of algebroid curves is obviously a transi-
tive relation it follows that a + 1 is also the number of curves in the set
{A1, A, * * *, At} which are equivalent to r1. Similarly, from r u c = A +
follows that a is the number of curves in the set {A1 + e, A2,, ' -, At} which
are equivalent to rl. Therefore, we must have necessarily rP -A1.
Similarly, if we denote by b the number of curves in this set
{r1, r2, , Prt} which are equivalent to r1U 8 (b 0), then b + 1 is the
number of curves in the set {rl U c, r2, . . *, rt} which are equivalent to rT U 8
Since r = A and r u 8 - A + e, it follows, by a similar argument as above,
that rPU AlUe.
We have thus shown that the assumptions of the lemma are satisfied when
r and A are replaced by rP and A1 respectively (and thus C and D are
replaced by r1 U E and A1 U e respectively). With r and A replaced by rT
and A1 we haves a special case of the lemma, namely the case in which
mn :1 and each of the two curves C and D has-only one tangent. Assume
that the lemma has already been proved in this special case. Then we can
assert that there exist an equivalence r1: rP U e -* A1 U e such that 7r1 (e) = e.
We shall show now that 7r1 can be extended to an equivalence 7r: C-> D.
Consider the set {r1 U 8, r2, . , rt} of tangential component of C and the
set {A1 U e, A2,, , At} of tangential components of D. Let M denote the
set of those tangential components of C which are equivalent to r1 U 8 (M
contains at least one element, namely r1 U c). Similarly let N be the set
of those tangeneial components of D which are equivalent to A1 U e. Since
r1 u E 8=-A1 U e and since equivalence is transitive, it follows that 111 and N
have the same number of elements. Now, by assumption, we have C = D i. e.,
there exists some equivalence p: C _> D. Fix such an equivalence p. For
v 1, 2, . ., t, p induces an equivalence p,: C, -> D, where C, = Pr u ,
C=T r for v -2, * ,t, and D1, D,y * Dt are the curves A1 U e, 2, *, At,
in some order. We have already our equivalence 7r1: rP U 8 h-* 1 U e. We
define lrp, for v =2,- - *,t as follows: a) If C, (= r) / Ji (i.e., if
520 OSCAR ZARISKI.
rP r1U 8), then clearly also D, ,4N (since J' D,, and rFU c lUe,
whence D Al U e) ; in this case we set 7rv-pp: rJv- ->Dv). b) We set
up an arbitrary (1, 1) correspondence between the set of curves r, in 31.
other than rP u c, and the set of curves A. in N, other than Al + e.

Assuming that corresponding curves are furnished with the same index 1x,
we have that r,F1 =- for all , such that ri + c rA C M and Al + e 7 z, C 1N
(since rJ, uF ri , p AA AlU e and ri U E Al U e). We then fix an
arbitrary equivalence 7rA,: r, AA. This defines 7rv for all V = 1, 2, , t.
Since r1 U c, J'2p , rt and Al U e, L2. . , At are the tangential compo-
nents of C and D respectively, it follows that the v equivalences 7rT patch up
together to an equivalence 7r: C .> D (use for instance the definition of
(a) -equivalence).
Thus, in older to complete the proof of the lemma (in the case m = 1)
we have only to show that (in the case lb under consideration) there exists
an equivalence 7r1: rJ u ' -U A1 U e such that ri(e) =- e. Let us apply our
quadratic transformations T and S (with center P and Q respectively) to
r1 U E and Al1 U e. If 6' and E' are the exceptional curves of T and S,
will have the total transforms
T{Jri U } l rl' U ' U S',

S{z 1 U e}l A1' U e' U E',
where rl1'==T(rJ), e'=T(e) ; l1' =S(j1), e'==S(e). Since ri Alp it

follows that ri',- Al' (by the definition of (a) -equivalence). Since rF U E
= U e, it follows that rl'U e'-= Al' U e' and that
r1' u el u (E = Al u e' U E'
(by the definition of (b) -equivaleilce). Furthermore, -' and S' are regular
curves, with distinct tangents; similarly, e' and E' are regular curves, with
distinct tangents. We have here therefore the case m = 2 of our lemma.
Since &- (rI U e) < c* (C) (unless e* (C) 0, in which case the lemma is
trivially true), we may assume, by induction, that there exists an equLivalence
r1*: r'l U ' U s _ Al' U e' U E'
such that 7rl* (Z) = e' and vl* (s') = E'. Then 7r-* induces an equivalence
7rl: r,U - > Al U e such that 7ri (?) e.
Case 11. m > 1. Let Cj be the tangenitial component of C determined

by the tangent of ej (Cj may be ej itself). Similarly let DJ be the tangential
component of D determined by the tangent of ej. Then Cl, C2, , C?n are
distinct tangential components of C, and similarly for D1, D, , Dm and D.

Since r U U U . . Uej- A U e, U * * * U e1_1 and
rU , U .U U cj=A U e, U e,U U ej, U ej,
it follows, by the case m 1 that there exists an equivalence
j: ir u 1 + u U cu cj_ > A u el U * * * U ej_l U ej
such that j (,j) =- ej. Then crj induces an equivalence 7rj: Cj_> Dj. In the
course of the proof of the lemma in the case I we have shown that if CE D
and if 7r1: C1 _> D1 is a given equivalence between two tangential components
C, and D1 of C and D respectively, then -r1 can be extended to an equivalence
7r: C >) D. In a similar fashion it can be shown that the m equivalences
7rj: Cj > Di can be extended to (i. e., are induced by) a single equivalenc
7r: C >D. Since wrj(ej) =ej, the proof of the lemma is complete.
COROLLARY. If two algebroid curves C and D are formally equivalent,

they are equivalent.
In the notations of Definition 4 we have, by hypothesis, that C,' and

D,' are formally equivalent, and also that C0' U 8' and D,' U E' are formally
equivalent (v = 1, 2, * * *, t). By induction on c* (C) we may therefore
assume that C.'-= D,' and C' U 8'i- Dv' U E'. HIence, by the preceding
lemma, there exists an equivalence 7rv': Cv' U 8' - DV' U E' such that
7rv'(8') E'. But that means that C and D are equivalent (namely (b) -
equivalent).
Note. In a Lincei Note ("La risoluzione delle singolarita delle superficie

algebriche immerse," Nota I. Accadenmia Nazionale dei Lincei, Rendiconti,
vol. XXXI, fasc. 3-4, Settembre-Ottobre (1961)) we gave the following
inductive definition of equivalence of algebroid singularities (we use the
notatianis of Definition 2, ?3):
C D if the following three conditions are satisfied:
1) mp (C) == mQ (D).
2) C and D have the same number t of distinct tangents.
3) For a suitable ordering of the connected components C1', C2', * , Ct'

and D1', D2', *,Dt' of the proper quadratic transforms of C and
D, it is true that C'= Dv' (v= 1, 2, * *, t) .
This definition is not "correct." In fact, even the following, muore exacting,
definition of equivalence would not be "correct":
522 OSCAR ZARISKI.
C = D if there exist a tangentially stable pairing 7r: C-> D such that
1') mp(y/) =mQQ(S), if 7r = (yi).

2') C0' D,' for v 1,2, , t, where C07 and Dv' are the proper
transforms of tangential components C, and D, which are asso-
ciated with each other under 7r.
Here is a counter-example.
Let C consist of two branches -y, -2, having the same tangent (y = 0), where
7i and 72 are. defined by the following power series:
: x t4, y1 t-j + t1l;

72: t6 + t9, y -tl? + t13.
Let also D consist of the following two branches 81, S,, with the sam
x = O :
8, : x tl? + t13, Y-t4;

82 : x t'? + t'l, y =t6 + t7.
We have imp (71) mp (81) 4 and mp (72) mp (82) = 6. Hence the pairing
7r: C-> D defined by 7r(yi) =Si and 7r(y2) = 82 satisfies condition (1') above.
Applying to C the quadratic transformation T: x' x, y' y/x, we get that
C'===T(C) =yl'U 72', where 71' =T(71) and 72' T(y2) are defined by
yi': x' t4, y' t6- + t7;

72': X' t6 + t9, y' = t4.
Similarly, applying to D the quadratic transformation S: x' = x/y, y' =y,

we get D'==S(D) =81'U 82', where 81'==-S(81) and 82'=S(82) are defined
by
Si': x' t6 + t9, y'-t4;
8': x' t4 yI== t6 + t7.
Thus 7i' =S2' and y2' 81', whence C' = D'. However, C and D are not
equivalent in any reasonable algebro-geometric sense. This is so, because
already the corresponding branches 7i, 81 (both having a 4-fold point a
are not equivalent. This can be seen by observing that the 4th quadratic
transform /1(4) of yi is a regular curve, while the 4th quadratic transfo
81(4) of 81 has a cusp. (Similarly it can be seen that 72, 82 are not equivalent).
5. Some auxiliary results on derivations. In this section we deal

with an integrally closed (Noetherian) local domain R, which we assume to
STUDIES IN EQUISINGULARITY 1. 523
be pseudo-geometric (see Nagata, Local Rings, p. 131), and with the quotient
field K of R.
We consider a finite separable algebraic extension K' of K, and we
denote by R' the integral closure of R in K'. Thus R' is a finite R-module.
(Later on, in this section, we shall consider the more general case of a ring
R' which is a finite R-module, is integrally closed in its total ring of quotients
and is free from nilpotent elements.)
THEOREM 2. Let D be a derivation of K, twith values in K, which is

regular in R (i. e., such that DR C R), and let D' be the unique extension
of D to K'. Assume the following:
A. If p' is any minimal prime ideal of R' (i. e., 4' is a prime ideal of
hight 1 in R'), then p' is tamely ramified over R.
B. If 4' is any prime ideal of R', of hight 1, which is ramified over R
and if ,' R = R , then there exists an element x in j such that x f p(2) and
Dx= 0.
Under these assumptions we have DIR' C R'.
[Note. Following Abhyankar (" Tame coverings and fundamental groups

of algebraic varieties, Part I," American Journal of Mathematics, vol. 81, no. 1
(1959), p. 53) we say that p' is tamely ramified over R if the following is true:
Let p =' pln R, let (S,2) and (S', 2)') be respectively the completions
of the localizations Rp and R,', and let F, F' be the fields of quotients of S
and S' respectively. (Note that since R is pseudogeometric, S and S' are
integral domains.) Let F* be the least Galois extensions of F which contains
F' (note that since K' is separable algebraic over K, also S' is separable
algebraic over S). Let (S*, 92*) be the integral closure of S in F*. (Note
that since S is complete, St is a local domain). Then
[F*:F][S*/92T*: S/9]fl_ X O (modp),

where p is the characteristic of S/M. Note that this condition is vacuous
if p = 0 and that " ' unramified" implies "p' tamely ramified."]
Proof. Since R' n R0 ' where the intersection symbol is extended to all
minimal prime ideals p' of R', it is sufficient to show that D'R'P, C R'P for
every minimal prime ideal p' of R'. Now if p' n R- = , then p is a minimal
prime ideal in R; and if p is any minimal prime ideal in R, then the integral
closure of Rp in K' is the intersection of all the R'l such that p' n R = p.
Since assumptions A and B remain valid if we replace R by R]X (p-minimal
in R) and R' by the integral closure of R]X in K' we may assume that R is
a regular local ring, of dimensiont 1. In this case, R' is a Dedekind domain
524 OSCAR ZARISKI.
having only a finite number of prime ideals #p (0), all lying over the maximal
ideal , of B.
Let p' be any of the prime ideals of R', different from (0). We wish
to show that D'R'p, C R'p,.
The case in which t' is unramified over R is trivial. Namely, we have
in this case: [F*: F] [S'/19': S/] -[S/I': S/I] = (say) n, and
S'IZ' = R'/4p'. If t is any element of R' whose W-residue is a p
element of S'/1' over S/P, then it is immediate that S'= S[t] (since S'`,
is the maximal ideal of S'). The derivation D has a unique extension to S,
which we shall continue to denote by D, and D has a unique extension to the
field of quotients ' of S' (since $ is separable algebraic over F), which we
shall continue to denote by D'. If f (X) is the minimal polynomial of t over
F, then f(X) E S[X] and f'(t) 9)'. Since f'(e)D Y -fD(4) =0, where
fD is obtained by applying D to the coefficients of f, it follows that D'$ C S'.
Thus D' is regular on S'. The assertion that D'R'pr C R'p now follows
from the fact that S' n K'= R'p, and that D'K' C K' (if a, b C R'p, then
S'b n R'D, = R'r, b, and hence b divides a in S' if and only if b divides a
in RID,).
Assume now that p' is ramified over R (hence tamely ramified). Again,
S*/WI* is a finite separable algebraic extension of S/19. Upon replacing R
by the integral closure of R in the inertia field of W*, and using the un-
ramified case already settled above, we may assume that S*/I* - S/1P.
In this case we have S*9ft 9X*e, e0 (modp), and [F*: F] =e. Let x
be as in assumption B. We have then Sx = WI and S*X - 9)?*e. Since the
Galois group of F*/F is abelian, the proof can be reduced to the case of a
cyclic extension F*/F, of degree e. The adjunction to F and F* of a
primitive e-th root of unity does not destroy the validity of assumptions A
and B. We may therefore assume that F contains a primitive e-th root of
unity. Since S and S* are complete rings, it is well-known then that S*
is of the form S[I], where t is an element of S* such that te - ax, a-a unit
__ teDa Da
in S. We have ete-lD't - xDa = a,a
i.e., D't
ea - Et S*. * Hence
D'S* C S*, a
We now generalize Theorem 2 as follows:
Let R be as above, and let R' be an overring of R which is integral
over R and is a finite R-module (whence R' is a semi-local ring). We assume
that
(a) R' is integrally closed in its total quotient ring;
(b) R' has no nilpotent elements.
STUDIES IN EQUISINGULARITY I. 52a5
As a consequence of (a) and (b), R' is a direct sum of local, integrally

closed domains:
R' = '1 RI, '2 O * ** R'h,
where, if 1 = e1 + e, + * * - + e7, is the decomposition of 1 into idempotents

e, (with eej =- 0, if i = j), then R' =R'e,. Let R, = Re,, and let K', be the
field of quotients of R'*. Then R'- is the integral closure of R1 in K', and
is a finite R1-module. Furthermore, K'i is a finite algebraic extension of the
field Ke,. We assume furthermore that
(c) K'i is a separable extension of Ke, (i l ,2y * *,h).
THEOREM 3. Let D be a derivation of K, regular on R, and let D' be the

unique extension of D to the total quotient ring K' (= K'1 0 K'2 e ... ** K'h)
of R' (the uniqueness of D' follows from (c)). Then under the same assump-
tion A and B of Theorem 1, we have D'R' C R'.
Proof. We have the decomposition of D' into the " direct" sum of
derivations e,D' of the K'i:

D= elD' 0 e2D' 0 0 D ehD'.
If we set D', = e,D' and Di= eiD, where Di is intended as a derivation

of e,K (Di (e,u) - ejDu), then D' is the unique extension of Di to K'i.
Applying Theorem 2 to Ret, R'e,, we see at once that D'i is regular on R'e,
and this implies that D' is regular on R'.
The following theorem is a special case of Theorem 3:
THEOREM 4. Let R k= c [ [xi, x2, , * xr] ] be a power series ring, over

a field I, let K be the field of quotients of R and let.
K'=K E'10 K'02 0 * * * G K'h D K
be a direct sum of fields K'- - K'ej, with K'l a finite separable algebraic
extension of Kei. Let R' be the integral closure of R in K'. Assume the
following:
A. Every priime ideal of R', of hight 1, is tamely ramified over R.
B. If A is a discriminant of a basis of K'/K consisting of elements of

R', then A-up to a unit factor in R-is a power series which is independent
of x1.
Under these assumptions, the derivation of K' is regular on R'.

ox,
For the proof it is only necessary to observe that if an ideal p' of B',
526 OSCAR ZARISKI.
of hight 1, is ramified over R, th
by some irreducible factor $ of A. We have then p(2) and - =0, and
thus assumption B of Theorem 3 is satisfied.
COROLLARY. Let R=k[[x1,x2, *,xc]] (as in Theorem 4), let f(Z)

be a monic separable polynomial in R[Z], free from multiple factors, let K'
be the total ring of quotients of the ring R[Z]/f (Z), and let R' be the
integral closure of R in K'. Assume that condition A of Theorem 4 is
satisfied. Assume furthermore that the discriminant A0 of f(Z) (AO C R) is
of the form eh, where e is a unit in R and h is a power series independent
of x,. Then the derivation-a ils regular on R'.

Obvious consequence of Theorem 4, since the discriminant A of K'/K is
a divisor of AO in R.
We have in mind a certain application of Theorem 4 to complete local
rings. For that application we need a lemma:
LEMMA 4. Let (o,m) be a complete semi-local ring of characteristic

zero (with m denoting the intersection of the maximal ideals of o) and let
D be a derivation of o with values in o. Assume that there exists an element
x in m of o such that Dx is a unit in o. Then o contains a ring ol of
representatives of the (complete) local ring o/ox, having the following
properties: (a) D is zero on oi; (b) x is analytically independent on Jo;
(c) o is the power series ring o1[[x]]. [It follows that x is not a zero advisor
of o, and hence dimo1 dimo -1.]
Proof. Without loss of generality we may assume that Dx =- 1 (replace
D by I D). We consider the operator

x2 x3
e--$D = I -xD + 2!D D2 ! D3+ *
where I is the identity map of o and D(n) denotes the n-th iterate of the
derivation D. It is immediate that: (1) e-xD is an endomorphism of o;
(2) if o - Im e-$D then D is zero on o; (3) the kernel of e-xD is the
principal ideal ox (it is obvious that kernel C ox; the opposite inclusion
follows from e-D (x) = 0). From (2) follows that: (4) the restriction of
e-D to ol is the identity map. From (1), (3) and (4) it follows that oi
is a ring of representatives of o/ox. From ol n ox = (0) all the remaining
assertions of the lemma follow at once.
COROLLARY. The notations and assumption being as in Theorem 4 (or
as in its Corollary), assume furthermore that k is of characteristic zero. Then

R1' is a powver series ring in xI, with coefficients in a subring R'1 of R' such
that R', is a ri-ng of representatives of R' R'x1 and such that is zero on R'1.
There is one essential complement to this corollary in the case dealt
with in the corollary to Theorem 4. We have namely the following:
THEOREM 5. The notations and the assumptions being as in the

Corollacry to Theorem 4, assume furthermore that k is of characteristic zero.
Let A'=R[z] be the subring R[Z]/(f(Z)) of R' (here z is the f-residue
of Z), let Rl1=k[[X2,X3, . .x7 . , Xr c-xD(Z) (where D= - ) and

Ox,
A'1 - R1[e] (=- D (A')). Then: (a) the ring R', (=- e-$DR') is the
integral closure of A'1 (in the total quotient ring of A'1) ; (b) if f0(Z) is the
polynomial obtained from f(Z) by reduction module xl, then f0(Z) is the
minimal polynomial of g over k{{x2,x3,* ,Xr}}.
Proof. (a) Since R' is integral over A' and e-xDi is an endomorphism
of R', R'1 is integral over A'1. From the fact that xl is analytically indepen-
dent over R'1 and that R' =R',[[xJ1], follows at once that the total quotient
ring K'1 of R'1 is a subring of the total quotient ring K' of R' and that
RI' n K'1 = R'i. Hence, since R' is integrally closed in K', it follows that R'1
is integrally closed in K'1. To complete the proof of part (a) it remains
only to show that K'1 is also the total quotient ring of A'1. This, however,
will follow once (b) is proved. In fact, (b) implies that the total quotient
ring of A'1, as a vector space over the field k{ {x2, x3, *. .. } }, has dimension
n, where n is the degree of f. On the other hand, also K' has dimension n
over the field k{ {XI, X2, , * * Xr}}. This implies, in particular, that any n + 1
element of R'1 are liinearly dependent over the ring k [ [xl, x2, *, X.r] .
Since xl is analytically independent over RI' (and since kc [ [x2, * * *, xr] I C R'i),
it follows that any n + 1 elements of R'1 are linearly dependenlt over the ring
k [ [X2, X3, * xr]]. Hence dim K'1/k{ {x2, X3, * Xr} } ? n, and since wve
have just seen that the total quotient rings of A'1 has exactly dimension n
over IX{ {X9, X3, * Xr} }, it follows that this total quotient ring coincides
with K'1.
(b) Clearly, fo (g) = 0, and hence the minimal polynomial 0 (Z) of

g, over k{ {X2, X3, ** Xr} }, divides fo (Z). On the other hand, we have
o- + (g) e-iD (+ (z)), whence + (z) c Ker e-iD, i. e., + (z) C R' x1. Now,
if Ao is the Z-discriminant of f(Z), then AoR' C A'. Hence Ao, (z) = x1g (z),
where g(Z) c lc[[xl, x2,. . Xr]] [Z], and therefore we have an iclentity
Ao0 (Z) -x1g(Z) =A (Z) f (Z), where again, A (Z) C k [[x1i x, ,Xr]] [Z]
528 OSCAR ZARISKI.
Setting xl - 0 in this identity, and observing A, (0, x,, * x,X) 74 0, we see

that +b(Z) is divisible by fo(Z). Hence k(Z) ==fo(Z), as asserted.
THEOREM 6. The assumptions and notations being as in Theorem 5, let

f (Z) s 6 1(Z) 02(Z), * *, Oh(Z) be the factorization of f in ir-reducible
factors in R[Z], antd let ,oi(Z) ( C R1[Z]) be the polynomial obtained from
j by reduction modulo x1. Then:
(a) fo(Z) 1(il(Z)020 (Z) 0710(Z) is a factorization of f0(Z) in

irreducible factors in RJ[Z].
(b) If kI is an algebraically closed field (always of characteristic zero)

and if P1 is an algebraic closure of lc{{xl}} then the cp(Z) are also irre-
ducible in Pl{{X2,X3,. ' ,Xr}}[Z].
Proof. (a) The integer h is the number of direct summands of the total
quotient ring K' of R[z]. If 1 = el+ e2+ + eh is the decomposition
of 1 into idempotents, then from e,2 = e, follows 2eiDet, = De,, whence De- = 0
and e--Dei = e,, showing that the ek belong to the total quotient ring K'l o
A', (we use here the notations of the proof of Theorem 5). Hence also K'1
is a direct sum of h fields, and this implies that fo (Z) is a product of h
distinct factors in Rl[Z]. This proves (a).
(b) Let F=kI{{x1,x2,* *, ,X}}. We have K' = [z] (where z is

the f-residue Z) ==total quotient ring of the integral closure R' of R. Since
R' R'l [[xl]] and z C R', it follows at once that K' = F[g] (note that the
total quotient ring K'l of R'1 is k{ {x2,x3, , x } }[a]). We now pass to
any of the h direct summands K'e, of K'. We set y zei, get and
(for simplicitv) we identify F ei with F. Then F[y] F[Z]1(0i(Z)),
k{ {X2X3 * **,Xr}} [q]-k{ {X2,,X3, * **,Xr}}[Z] /(so0(Z)), and we conclude
at once that F[y] =F[f] . Let f denote the field F{{x2,x3, ,Xr } Our
claim that the polynomial +1 (Z) is irreducible over F is equivalent to the
claim that the tensor product F[y] OF P is an integral domain. Since
F [y] i F (), this claim is equivalent to asserting that the polynomial
O (Z) is irreducible over P. Now, the coefficients of 4io are in the fiel
Ie{{x2, X3, * , Xr}}, and (i, is irreducible over that field (by part (a) o
the theorem). The irreducibility of fio over P follows now from that fac
that k{{x2, x3,~ * ,Xr}} is maximally algebraic in .'*
* By a simple induction, it is sufficient to pr-ove the following: if k and K are fields,

k C K, and k is maximally algebraic in K, then the power series field k{{t}} is maxi-
mally algebraic in K{{t}}. Proof. The natural valuation V of K{{t}} (with residue
field K) is the extension of the natural valuation v of k{{t}} (with residue field k).
STUDIES IN EQUISINGULAIRITY I. 529
6. Specializations of algebroid curves and equivalence. Iff (x, y) is

a power series in x, y, with coefficients in a field kc, and such that f(O, 0) = 0
and if this power series is free from multiple factors and is regular in y,
then, by the Weierstrass preparation theorem, the total quotient ring K' of
Ic[[x, y] (f) is a direct sum of fields which are finite algebraic extensions
of the field k{{x}}. If n is the dimension of K', regarded as a vector space
over Ic {x} }, then 1, y, y2, ** yn-1 (or-more precisely-their f-residues)
form a basis of K'; we shall denote by Ayf the discriminant of that basis, and
we shall refer to Avf as the y-discriminant of f. The discriminant Aef is a
power series in x, hence up to a unit factor-is simply a power of x. If, by the
Weierstrass preparation theorem, we write f (x, y) = E(x, y) + (x, y), where
E(x,y) is a unit in kl[[x,y]] and (x, y) is a polynomial in y, of degree n,
with coefficients in k [ [x]], then Avf is simply the y-discriminant of the poly-
nomial 4.
Let now f(x, y; t) be a power series in x, y, t, with coefficients in an
algebraically closed field kI, of characteristic zero. We assume that f(0, 0; t)
is identically zero and the the power series f is regular in y. We denote
by Pt the algebraic closure of the quotient field k{{t}} of lc[[t]], and we
regard f as a power series in x, y, with coefficients in Pit. We denote by
A(x, t) the y-discriminant of f(A (x, t) C lk[[x, t]i]) and we assume that
A (x, t) is not identically zero. In that case f has no multiple factors (in
Pt [ [x, y]]), and we can interpret the equation f 0 as defining an algebroid
curve Ct over the algebraically closed field Pt, with origin at P: x = y = 0.
Since f(x,y;t) has been assumed to be regular in y,f(x,y;0) is not
identically zero. We set fo(x,y) ===f(x,y;0), so that fO(x,y) is also regular
in y. We shall also assume that the discriminant A (x, t) is not divisible by t.
In that case, the y-discriminant A (x, 0) of fO is not identically zero, fO has
no multiple factors (in k [ [x,y]]), and the equation fo (x, y) = 0 definles an
algebroid curve CO over Ic (and hence, a fortiori, also over Pt), with the same
origin P as Ct. We shall say that CO is a specialization of Ct over t -> 0.
Our principal object in this section is the proof of the following theorem.
THEOREM 7. Let k be of characteristic zero (and algebraically closed),

and let CO be a specialization of Ct over t -0. Then:
Let 2 be a finite algebraic extension of k{{t}} in K{{t}}, and let w be the restriction
of V to M. Then w is an extension of v, with the same value group as v, and the
residue field of w is an algebraic extension of k. Since this residue field is contained
in K (= residue field of V) and since k is maximally algebraic in K, it follows that
w and v have the same residue field. Since k{{t}} is a complete field (with respect
to the valuation v), it follows that 2 = k{{t}}.
18
530 OSCAR ZARISKI.
(a) A sufficient condition that CO and Ct be equivalent (in the sense

of ?3) is that the y-discriminant A(x,t) of f(x,y;t) be of the form E(X,t)XN
wvhere E (x, t) is a unit in e [ [x, t] ].
(b) If the line x 0 is not a tangent of CO, then the above condition
on A(x, t) is also necessary for the equivalence of Ct and CO.
Proof. We shall need two lemmas which we shall now state.. In order
not to interrupt the proof of the theorem, we shall assume for the moment
these lemmas and will prove them in the next section.
LEMMA 5. Let C: f(x,y) =--0 and D: g(x,y) O0 be two algebroid

curves, over an algebraically closed ground field k of characteristic zero (with
common origin P: x = y = 0). We assume that the power series f and g
are regular in both x and y, that C and D have the same number h of
irreducible branches, and that the line x O is neither a tangent of C nor a
tangent of D. We assume that Axf = Axg - y- (apart from unit factors in
7c [ [y] ]). We furthermore assume that there exists a pairing r: C->D
between the branches 71, 72, , yh of C and the branches zL.& , 8h O of
D having the followting properties:
a) If 7r(yi) = j then mp(,yi) = mp(81), and the intersection nurnbers

(l,y), (l,8i)are equal (i=1,2, ,h) ; here I denotes the line y =O.
b) If cp (x,y) 0 is an irreducible equation of /I and qj(x,y) is (tn

irreducible equation of 8i (= 7(yj) ), then Ax$+ = A = ylit (apart from
unit factors in k [ [y] ] ) .
Then Ayf = Ayg (apart from unit factors in kc [ [x]]).
Note that the assumption (1, yi) = (1, 8) in a) can also be expressed as
follows: if we assume as we may-that the +1 and tp in b) are monic
polyniomial in x, then for each i 1, 2, , h, the two polynomials
are of the same degree.
LEMMA 6. Let C: f(x,y) =0 and D: g(x,y) =0 be two algebroid

curves, over an algebraically closed field k of char-acterictic zero (wit
origin P: x= y 0). We assume that the power series f and g are regular
in y and that there exists an equivalence 7r: C -).D having the following
property: if 7r(yj) &=- (i=1,2, ,h) then (m,y) (m, 8t); here m
,denotes the line x =0. Then Avf Avg (apart from unit factors in kc[[x]]).
Note the following immediate Corollary of Lemma 6:
COROLLARY. If two algebroid curves C: f(x, y) =- 0, D: g(x, y) =- 0 (over

an algebraically closed ground field k, of characteristic zero) are equivalent
and if the line x - 0 is neither a tangett of C nor a tangent of D (whence
f and g are certainly regular in y, i.e., neither f nor g is divisible by x),
then Avf f-i Avg (apart from unit factors in k[[x]1]).
For if 7r: C = D is any equivalence between C and D and ?r (y,) =Si

then (m,,yi) =Mp(y) =-mp(Si) (m,&8) (since the line m: x=- 0 is neither
a tangent of yi nor a tangent of Si).
We now proceed with the proof of the theorem.
(a) Our basic assumption in this part of the theorem was that the
y-discrimninant A(x, t) of f(x, y; t) is of the form ,E(X, t)xAN, with E(x, t)
unit in k- [ [x, t] ]. This was also the basic assumption made in the corollary
to Theorem 4 (with r = 2, x- t, X2 = x ald h =XN). Thus Theorems 5
anid 6 are applicable. Theorem 6 tells us that-upon assuming that f (x, y; t)
is a polynomial in y-the factorization f cf= 1 .. of f into irreducible
factors in k [ [t, x] ] [y] is also a factorizatioln of f in irreducible factors inr
Pt[[x]] [y]. Thus the irreducible branches y4 of C are given by yi: p (x, y; t)
0. Theorem 6 also tells us that if , (x, y) O. (x, y; 0), then the irre-
ducible branches 8 of CO are given by Si: + o(x, y) = 0. This yields then a
natural specialization pairing 7r: C' -> CO between the branches of Ct and the
branches of CO: 7r (y) 8i. Upon replacing y by y + cx, where c is a suitable
element of k (this substitution has no effect on the discriminant Aef), we
may assume that the line y = 0 is not tangent to CO (whence fo, and hence
also f, is regular in x). We shall now show that all the assumptions of
Lemma 5 are satisfied if C and D are replaced by Ct and CO respectively and
if the roles of x and y in Lemma 5 are interchanged.
Let y: 0 (x, y; t) 0 be any of the irreducible branches of Ct and let

8=7r(y) : ip(x,y) =0, where ip(x,y) = (x,y;0). We have of course
=== &/'g (apart from unit factors in c [ [x, t]] ), silnce AYJp, as a factor of
nsf, is still of the form E1(x, t) xM, with e, a unit in kI[[x, t]], and so
=-- E, (x,0)xM. We pass to the local domain kI[[x,y]]/(iji) of 8 and we
denote by -q0 the V residue of y and by D (a) the field of quotients kc{{x}l} [qOi
of that local ring. Similarly, we denlote by X the +-residue of y and by Pt(y)
the field of quotients Ft{{x}} [,] of the local ring Pt[[x, y]]/Q() of -y.
Fromn the proof of part (b) of Theorem 6 follows that Ft(y) is obtained
from 7c (8) by the ground field extension kI -- Ft, i. e.,
t{ {x}}[D] kI{{x}} [o]

k{{a}}
F P{{X} }
532 OSCAR ZARISKI.
If [k(8)): k{{x}}] =n, then k1(8) =k0{{x1In}} and Pt(y) =Ft{{Xl1fl} }. Let
us denote by v, and v the natural valuations of the complete field ic (8) and
t(y) respectively. Setting - ln, we have v,(t) v(t) 1, and v is the
extension of vo.
We next show that
(6) v (D)- vo 6'9).
We know from Theorem 5 that D is g
(7) D ?0 + uit + u2t2 + * **
where the ui are elements of 1k (8) which

the u, are elements of k [ [f]]. Taking conj
t{ {x}}; this amounts to taking conjugates of qZu1o, u2, * over k{ {x}}),
we observe that, for o i, , s-o divides Avf in Pt[[t]]. SO is-
apart from a unit in Ft[[f]]-a power of t. Since -,,? O is a non-
in k[[411, it follows at once that for each ui a1, u- u must be d
by W2- fiO in k[[f]] (here u1, U,2,. * *,uin are the conjugates of
k{{x}}). This implies that v0(ui) ? min{n,vo(q0)}. (Recall that we
assumed that f(O, 0; t) is identically zero; this implies that all
are non-units in Ic [ [f]]. Let ui = a,0 i- + a ,1~dl,+ + , v > 0, a0
(aj C Dc). If v, < n, then there exists a conjugate ui2 such that v0 (uf
vj. So, in this case, viV0( V0lo- 2?) ?:VO(D).). Since the line y
is not tangent to C, we have v0(D?) ?vo(x) =n. Hence vo(ui)
and so v(D) =vo( 0), as asserted.
On the other hand, we have v0(x) =v(x) n. Since
mp(y) =min * {v(x), v(D) } and mp(S) min{vo(x), v0(oO)}

it follows from (6) that
(8) mp(y) mpQ)M
This holds for any two branches

condition a) of Lemma 5. Note als
y = 0 is not tangent to any branc
to be a tangent of CO). This is on
(with x and y interchanged).
Note also that since v(x) =v0(x) =n, we have (1, y) = (1,8), where
now I is the line x = 0. Thus both parts of condition (a) of Lemma 5 are
satisfied.
It has already been pointed out above that Avp = Av5r (apart from unit
factors in kc[ [x, t] ]). Thus also condition b) of Lemma 5 is satisfied.
We can therefore conclude that Axf = xg (apart from unit factors in

AI[y, t]]).
Since y = 0 is not tangent to either Ct or CO, the above conclusion tells
us that in order to prove that Ct and CO are equivalent, we may add to our
original assumption Avf = Avg the assumption that the line x = 0 is not a
tangent of CO (and therefore also not a tangent of Ct, in view of the fact that
we have v (x) = vO (x) for any two corresponding branches y and 8 = 7- (y)).
The rest of the proof is fairly simple and will consist in showing, by
induction on the exponent M of x in the discriminant Avf, that the natural
specialization pairing 7r: Ct->CO introduced above is an equivalence (if
M 0, then both Ct and CO are regular curves).
First of all we show that 7r is tangentially stable. If we set Avop = Ao0
-XMi (apart from unit factors), then the expression of the discriminant
as the square of the product of the differences of the roots of the poly-
nomial and-on the other hand-the definition of intersection multiplicity
h
given in ? 4, show that M =, M1 + E (y yJ)2, where Avf = xM. Similarly,

h i=1 i<j
M = M, + Y
t=1 tij
over t -- 0, that (yi, yj) C (i, j).

for all i, j = 1, 2, . . ., h, i #7 j. This shows that 7r is tangentially stable,
since yj and yj (respectively Si and Sj) have the same tangent if and only if
(yi, yj) > mp (yi) mp (yj) (respectively, if and only if (&8j, Sj) > mp () mp (8))2
and we have just shown that mp(yi) = mp(S) for all i =1,2, *, h.
We now apply a quadratic transformation T, with center P. Since the
line x 0 is not tangent of either Ct or CO, we may take
x' x, y' =y/x
as the equations of T. Let CU't, C,'? be the connected c

and T(CO) respectively. Let (0, c,) and (0, a,,) be the
Cv,'O respectively. The a,,o are, of course, elements of th
field k. As to the a,, they are-a priori-elements of Pt.
is an important point-the a, are actually elements of the power series ring
kc [ [t] ]. In fact, let yj: *i (x, y; t) = 0 be any of the irreducible branches
of the tangential component CU,t of Ct, let s* mp (yi) and let 0j (x, y; t)
be the leading form of p(x, y; t). . This form has coefficients in kl[[t]], and
the coefficient of yqi is a unit CE in 1[[t]]. Since , =1(y :xjxs, it
follows that aCx Ek [[t]], as asserted.
It follows that if we set yr,' '- Va, then the equation of C/tt will be
of the form
f '(x,y"';t) =0,
534 OSCAR ZARISKI.
where f/ (x, y/; t) is an element of 1c [ [x, t]] [y'], -regular in Yv' Tfhe
equation of C0'/ will be
f vo (X Y,vo) Yvo, =0. y'- av(O)
where fvo' (x, Y) == f/ (x, Y; 0 ), and Cv'0 is a specialization of C/t1 over t-* 0
The product of the discriminants Av'vf' is a divisor of the discriminant Q?Jf,
since, if we set
f (xn xy,; t) ==- xsf' (Xn y'; t),
where s - mp (Ct), then Avf - xs(8l)A}y'f '. Thus, the discriminant AYvf ' is
a power of x (apart from a unit in 1c[ [x, t] ]. The pairing 7rw': Cp' - CV/O
between the branches of C/'t and the branches of C'0'/, induced by our original
specialization pairing ir, is obviously still a pairing by specialization t -* 0.
It follows, by our induction hypothesis, that each 7rW' is an equivalence. This
proves that 7r is an equivalence.
(b) This part of Theorem 7 is an imnmediate consequence of the

corollary to Lemma 6. For, let Avf -A(x,t)xN, with A(0,t) 740. Then
-yfo A (x, O)xN. Since the line x = 0 is not tangent to C0, it is certain
not taiigent to Ct (the assumption that Ct _ CO implies that mp(CO) -w
whence the leading form of f0 is the specialization of the leading form of f,
over t -> 0). It follows therefore, by the corollary to Lemma 6, that xN is
the highest power of x which divided Ayf . Hence A (0, 0) 74 0, i. e., A (x, t)
is a unit in 1c[[x,t]].
This completes the proof of Theorem 7.
7. Proofs of the Lemmas 5 and 6.
Proof of Lemma 5. We consider the equalities used already in the proof

of Theorem 7:
h
M E, Mi + EY (-yi y1j)2
j=1 i<j
AI yM, + (8,' j) 2
i<j
They imply that
(9) E ~~~~~~~~(-y.~ -yj) 2 E(.a

9<j iKj
Let now (apart from unit factors in
Av'j = XN$ A1g =fxN;N
. =- xNi', Xg-T
STUDIES IN EQUISINGULARITY 1. 535
From
h
(10) N-,N,+ (/ 1j)2

j=1 j<j
(10 ) N' ''N,' + '(48, j)2

j=1 i<j
and from (9), it follows that in order to prove that N = N' it will be
sufficient to prove that N, Ni', i = 1, 2, * * , h. Now, for fixed i, let v
and v' denote the natural valuations of the fields of quotients of the local
rings of yi and 8 respectively, and let dx, dy and d'x and d'y denote the
differentials of x and y on Yi and 8 respectively. We have
(11) v (dx) v'(d'x) (=- mp (-i) 1 =MP (81) - ),
since the line x = 0 is not tangent to yj, nor to Si. We also have
(12) v(dy) =v'(d'y) (= (l,y) - 1= (,) 1).

Finally,
(13) (ax)
vN - at:N' v'()
The equality Ni=N,' f

relations
axi dx[+ ah- dy- 0 (on y),
dx + d dy 0 (on a).
ax ay
Proof of Lemma 6. This time (9)

(see ? 4, Lemma 2). Hence, by (1
that N -N i' (i 1,2,* ,h). Hence, we may assume that C and D are
irreducible curves. We shall now proceed by induction on the numerical
character ar(C) ( a(D) ; see ?3).
First case. The line m: x = 0 is not tangent to C (and therefore als

not tangent to D, since we have assumed that (m, C) = (m, D)). Withou
loss of generality, we may assume that y = 0 is tangent to both C and D.
536 OSCAR ZARISKI.
Then the quadratic transformation T, with cented P: x-- y 0, gives irre-

ducible proper transforms
C: f'(x,Y') =0
D': g'(x,y') 0,
having origins at x = y' = 0 (here y' =). We have C' D', and if m'
denotes the line x = 0 in the (x, y') -plane (m' is the exceptional curve of T),
then we know (see (3), ? 4) that (m', C') (m', D') mp (C) (= mp(D)).
Hence, by our induction hypothesis, we have Ay/f' - Av'g'. Since
AJf -= xs(S-')Ay'fl
Azg - X8(8-1)AV'gf.
where s= np (C) (= mrp(D)), the proof in this case is complete.
Second Case: the line x =0 is tangent to C (and hence also to D).
We may assume that the line y =0 is not tangent to C, nor tangent to

D. Then by the first case we have Axf = A&g (apart from unit factors in
k [ [y] Tf). Using the relations
f dx + Of dy 0 on C
a_xI ag
ig dx + ag dy 0 on D,
a_x ay
and relations analogous to (11), (12) and (13), with yi, X% replaced by C
and D, and with x and y interchanged, we conclude that Avf = Avg.
HARVARD UNIVERSITY.

Zariski

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Zariski

Uploaded by

Copyright:

Available Formats

Studies in Equisingularity I Equivalent Singularities of Plane Algebroid Curves

Author(s): Oscar Zariski

EQUIVALENT SINGUARITIES OF PLANE ALGEBROID CURVES.

With this paper we initiate a series of investigations on the concept of

In ? 5 we establish some technical results on derivations in complete

In ?? 6-7, the geometric results of ??1-4bis and the purely algebraic

Received October 26, 1964.

1. Algebroid plane curves. Let kc be an algebraically closed field (of

and we write s-rmp (C). Clearly mp (C) = ,mp(y).

It is known that if f is irreducible then f5 is the power of a linear form:

2. The quadratic transform of C. Let 1h: aix + by -0 be the tangent

f (xI, x'y) - x'sf' (x', y).

Let t be the number of distinct tangent lines Pv of C, and let y - a,xv =0

curve defined by the equation F,(x, y) 0. The irreducible branches of C,

We call the C, the tangential components of C.

Let ri ==mp(y),sV =mp(C.)( E ri),

Oi (X,Y) (r(X, Y) + Oi,,+l (X, Y) +***

Since pir6(X,y) = (y - aVX) r for iC Iv and Fv,8,(x, y)= (y avx)sv, it

,4i(X', xy') - X ri '(X, y'),

i' (X'7 y') (y' -av) r + X'(fi,rt+l(1l y') + *

f'(x', y') II1 Fv' xl 7 y').

formation (with center P) defined by x' = x, y' y/x, we set C' U Cv

and we call C', CV' and y( the proper quadratic transforms of C, Cv an

Note. Although the above definitions are based on a specific (and

have been formulated in terms of local algebra (see Zariski-Samuel, Commu-

3. Three definitions of equivalence of algebroid curves. Let D be

DEFINITION 1. A (1, 1) mapping 7r of the set of branches y1, 72, . . , y7h

DEFINITION 2. An (a) -equivalence 7r: C -> D is a pairing r between the

2) If X, 7r(y) (i==1,2,. * *,h), then mp( y)=mp(Si).

Example. P is said to be an ordinary s-fold point of C if s = mp(C)

yi is a regular algebroid curve (since s E mp (ye) and since np (y) ? 1).

definition of v (C) would require to set oa* (C) = 1. However, we agree to

DEFINITION 3. A (b)-equivalence r: C->D is a pairing r between the

DEFINITION 4. Given two algebroid curves C, D, having the same nutmber

1) C07 and D^' are formally equitvalentt (v = 1, 2, , t).

equivalence of a (b) -equivalence. We shall write C = D if C and D are

4. Proof of equivalence of Definitions 2, 3 and 4. We recall first a

f _ yn + xal (X)ynl1 + * * . + Xnan

g' y'mn + bbi(x')y'm-l + + bm(x')

(2) (C, D) mn + E E ( (Cv', D,').

(3 ) ( ',,' ( Cv', 6') n (~ MP ( C)).

(4) (7yi,r) (8j,5), A) i 1,2, ,h.

Proof. (by induction on &* (C U r)).

If & (C u r) = o, theni C is a regular curv

Let T(r) ==P', S(A) =A'. We consider two cases.

First Case. r is not tangent to any of the branches of C (and hence A

Second Case. The tangent of r coin1cides with one of the t tangents

pl': C1 u r-> Di' u A', pv': C'-* D,' (v =2, 3, t).

LEM-MA 2. If ,: C->D and S(i) s (i=1,2 * *,h), then

(b) If there exists an (a) -equivalence (or a (b)-equivalence) r: C-- D,

Proof. (a) The assertion is true if v* (C) = 0. We therefore use

Assume that 7r: C _ D is an (a) -equivalence. Using the notations of

(b) Assume that there exists an equivalence r: C = D. We have

LEMMA 3. Let C and D be algebroid curves having the same number

c = r u 61 u e2 U . . . U Em, D=AUeU e2 U . . . U e,,.

Proof. We consider two cases, according as m 1 or rm> 1.

Case I. m = 1. We divide this case into two subcases, according as

Case Ia. C -r u , D = A U e, and the tangent of E is not a tangent

(5) r1 A1 and r1UesA1Ue.

other than rP u c, and the set of curves A. in N, other than Al + e.

T{Jri U } l rl' U ' U S',

where rl1'==T(rJ), e'=T(e) ; l1' =S(j1), e'==S(e). Since ri Alp it

r1' u el u (E = Al u e' U E'

r1*: r'l U ' U s _ Al' U e' U E'

[F:F][S/92T*: S/9]fl_ X O (modp),