Internat. J. Math. & Math. Sci. Vol. 8 No. 4 (1985) 641-652 641 ON CONVERGENCE OF (I,v)-SEQUENCES OF UNISOLVENT RATIONAL APPROXIMANTS TO MEROMORPHIC FUNCTIONS IN C C. H. LUTTERODT Department of Mathematics Howard University Washington, D.C. 20059 (Received June Ii, 1985) ABSTRACT. A generalization of the classical Montessus de Ballore theorem in some related theorems are discussed. Non-Montessus type of convergence in cn and 2n dimensional Lebesgue measure is presented linking it to a result of Gon6ar. KEY WORDS AND PHRASES. Unlsolvent rational approximants, meromorphic functions, Montsus-type convergence, non-Montessus type convergence. 1980 AS SUBJECT CLASSIFICATION CODE. 41A20, 32A20. O. INTRODUCTION. Let f be in the class of meromorphic functions analytic at the origin, having fixed number of codimension i polar sets over some polydisk domain in Cn. We investigate the modes of convergence of (,)-sequences of unisolvent rational approximants (URA) to f and we divide them into two main types; those associated with "horizontal rows" of (,)-sequences of URA (for which v is fixed and is free to grow infinitely) and those linked to the "slanted rows" (for which and v are related but both grow to infinity) including "diagonal rows" (cases where definition 2.1, one requirement ;i vi’ i 1,2 v). As will become clear from ,n creates a "Pad Table" of upper triangular (,)-sequences. Some study of "horizontal row" of (,v)-sequences have been carried out by Karlsson and Wallin [3] using explanations expressed in terms of homogeneous polynomials in 2. Our investigation of the convergence behavior of the "horizontal rows" of (,) sequences constructed from non-homogeneous polynomials, show that the convergence to f is uniform on compact subsets of the domain of mero- morphy except on the analytic set {z sets of n: I/f O} or on the remaining limit polar (,v)-sequences unattracted by f. These limit polar sets have measure zero. We refer to this type of convergence for (,)-sequences of URA as the Montessus type. The case of the "slanted rows" of (,v) sequences in Cn has been studied by Gon6ar [2] using diagonal sequences constructed from homogeneous polynomials. Although in one variable, it is known that for certain limited classes of functions there can be locally uniform convergence for "almost diagonal rows" of (,v) sequences, it is generally recognized that most meaningful ways to handle convergence of "slanted refers to rows" of (,)-sequences is in measure or capacity. n-Lebesgue Here the measure measure and capacity refers to any reasonably defined capacity 642 in C.H. LUTTERODT n adaptable We call this type of convergence in to this kind of problem. rows" measure (capacity) for "horizontal rows" of (,v)-sequences, the or "slanted non-Montessus type. Either type of convergence for (,) sequences gives rise to a certain rapidity For the Montessus-type over-convergence means and over-convergence (see Walsh [9]). the domain of convergence of the (,v)-sequences of URA, includes in its interior the domain of convergence of the local power series representative of f at the origin in E n. For the non-Montessus-type, according to Gontar [2] over-convergence in measure (capacity) means that convergence in measure (capacity) in any finite domain implies convergence in measure (capacity) in En for f The paper is organized to reflect the Montessus-type convergence in 3 and the non-Montessus-type in 4. In I we introduce the notations used in the paper and in 2 we introduce the definition of the URA’s and discuss the sense in which they are unique. The main theorems of the paper are theorems 3.1, 3.3, 3.4 and 4.1 (which examine the different cases associated with the horizontal rows) and theorem 4.2 (which rows). examines a case of the slanted tion of theorem 3.1. There is also theorem 3.6 which is an applica- The theorem yields a result about global analytic behaviour of f if all the horizontal rows of (,v)-sequences are constrained in a certain way. Some extension of this idea is discussed in Lutterodt [5]. (z)’s over-converge insight into the way in which Theorem 3.5 provides some in the case of Montessus-type and theorem 4.4, that of the non-Montessus type. i. NOTAT ION. (z Let z: : n (I {z. and A e z 1. n) : be an n-tuple in (v and v: Izjl I o}, i < Vn j n and : (z qn be n-tuples in < n, then A n A e n-l; Zn_l) I n-i e lq with x...x A n-times Let o n E: disk centered at origin. We denote by E the following subset of n n where is a partial ordering in IN e. IN 0 ) }, given by ’-’ 0 - <:-> 0 < J < ., 1 < < 0 is a poly- n: {% n. We adopt the following short notation as well i+...+ n Il 3z l EE Let be a domain in tions holomorphic in dz zll Z a dz ...dz 1 n znn Z ,( I n n such that Z ae n =0 0 e , al, then ;7() and continuous on holomorphic at the origin, but meromorphic in nomial Pi(z) in Kn of multiple degree or simply be written as P(z) where g gy1...y n and z 7 l Zl E n n gyz a =0 () n is the ring of all func- . is the set of all functions with finite pole sets in ’degree’ at most (%1 A poly- %n can CONVERGENCE OF (,)-SEQUENCES Let j be the class of rational functions of the form R ively; moreover (P for some P(z)/Qv(z) (z) and v respect- are polynomials of ’degree’ at most Pl(z), Q(z) (z) Q,(z)) # 0 and Qv(0) where 643 i on A n It’ except on a subvariety of codimension /> 2 0 and for all (,). I) RATIONAL APPROXIMANTS. 2. n Let U be an open neighborhood of the origin in R,(z) e is said to be a ,af.otctJ qn, E (i) E (iii) (iv) (v) 0 if to f at z (2.1) 0 P (z)) z=O an index interpolation set with the following properties: E 0 (ii) app:wma.t (Qv(z)f(x) z% for (U) with f(O) # 0, a rational function Suppose f DEFINITION 2.1: v e E , 0 ,- A E E Each projected variable has the Pad indexing set. v, For each pair (,v), [ElV < (vi) n n I where j=l =i E (v. + I) (. + i) + H IEI is the cardinality of v REMARK i. Pad index set is the index set that is used in the one variable case to define Pad Approximants. O, then If the function f used in definition 2.1 is such that f(O) depending on the order of regularity and in which variable, zj say, Weierstrass REMARK 2. Preparation Theorem can be polynomial W(,zj) employed with W(0) then apply to g in f W to write f W g where W is a Weierstrass 0 and g is a unit with g(O) # 0. Definition 2.1 may g. The rational approximants defined above are not in general unique. of uniqueness is firmly tied to the cardinality of E employed. The question Uniqueness as known v is maximal in the following sense. in Pad case seems only possible if E DEFINITION 2.2. n (. + I) + The interpolation set E n v is said to be n +l)-i j=l(J j=l It should be pointed out that there are many maximal defining unique approximants. PROPOSITION 2.1. (U) w.r.t, . Let the pair maximal E satisfies definition 2.1 w.r.t, Since f its "Pad table" equivalent. <Pi(z), Qv(z)> Suppose the same E P (z) PROOF: P*(z) define a (,)-rational approximant <P(z), Qv*(z) ’. Then is another pair that v Q(z Q*(z) w.r.t. E 4 (U) and P q are polynomials qf z Taylor development in U. can be chosen for We note that when E v is maximal, there is one-to-one correspondence between E v and to f e v that if This is a feature peculiar to several variables, unknown in the Pad case in one variable. a mxa Now by definition 2.1 and with E P (U) and has a maximal we get 644 C.H. LUTTERODT Qv(z)f(z) where gu\% P (z) \EVg z Z are the coefficients of the expansion. (2 2) <P,Q> Similarly from the pair we again have Multiplying (2.2) by * z Q(z)f(z) P*(z) en\E,0gz Q(z) and (2.3) by Qu(z) and subtracting (2.3) the former from the latter we get Q(z)P(z) Q(z)Pu(z) >" r (2 4) iz . rv where is the coefficient of the expansion after taking the desired difference. Now the R.H.S. of (2.4) is a power series regular with at least order + 1 in each 3 z.-variable. Therefore the L.H.S. of (2 4) must vanish w r t the maximal E v chosen, j since the L.H.S. is a polynomial of ’degree’ Q (z)P* (z) Recalling that maximal E Q*(z) P n) Qu(,Zn and Q(0,z_ n )’ Q*(0,z_ n Q(,O) # O. (2.6) + v. z) E w r t Thus we obtain (2.5) by construction, has maximal Pad4 indexing in each variable and thus on projecting (2.5) on z -axis n yields P (0, z where at most the uniqueness of Pad approximants then P(O,z n (2.6) Qv*(O,z_ n . do not vanish in some neighborhood U The desired result then follows w.r.t. E Q(O,O)_ U since 0 but also for each fixed holds not only for # 0 n-I U. From equation (2.1), we separate out the following linear equations" z z (Qv(z)f(z) P (z)) (Qv(z)f(z))Iz=0 z=O O; % 0; e E (2 la) . ED\E (2.1b) whose solutions for coefficients of Qv(z) and then those of P (z) give rise to a (u,) rational approximant. Now (,9)-sequences of rational approximants will be called ani5oueut if the underlying E is maximal and a certain determinatal or rank condition is satisfied for the system of linear equations that arises from (2.1b). (see also Lutterodt [4].). For the rest of this paper we shall assume that we have (,v)-sequences of unisolvent rational approximants (URA). with respect to some chosen E Quv(z) The latter is denoted by u maximal. We then normalize n(z) Qv(z), Pv(z)/Quv(z Puv(z) dividing u(z) by the modulus of the largest coefficient of Qua(z). tion leaves nu(z) unchanged but since P(z), Q(z) change we adopt the denotation of ,Dv(z) with 9(z) normalized, and in This opera- following (2.7) 3. MONTESSUS-TYPE CONVERGENCE. This section gives the generalized Montessus de Ballore theorem, some related theorems covering the Montessus-type convergence and an application of Montessus theorem to the generalized Pad Table. CONVERGENCE OF (p,v)-SEQUENCES THEOREM 3.1 (MONTESSUS). ](A n) Let p 0 and v with finite polar set defined by (Vl, Gv: 645 v {z e n) n: be fixed. Suppose qg(z) O} and f e n) where qo(z) is a polynomial of exact minimal ’degree’ v and A n n 0 Suppose (z) is an (P,)-unisolvent rational approximant ;J to f(z) with its polar set -i -i satisfying for sufficiently large, # Then as C( q Q(O) . n Q(O) (u.) rain G l<j<n (i) (ii) -i An0 Qp(0) An D G f(z) uniformly rp(z) on compact subsets of n\G REMARK: The degree of convergence in (ii) of theorem 3.1 is geometric and it rain (.). The ’degree* of in (z) has to be exactly l<j<n The following lemma is used in the proof of the theorem. depends on p’ pv(z) LEMMA 3.2. For 0 p’ < and ’ p l.qn\ E (.) rain l<j<n ’+I (i The proofs of Theorem 3.1 and Lemma 3.2 were given in Lutterodt [6]; that of Lemma 3.2 being in the appendix. The following two theorems are closely related to theorem 3.1 not only in statement but also in proof. We shall therefore state them together and then briefly in- (see, [6]). dicate where in their proofs they differ from theorem 3.1. Let THEOREM 3.3. n (i a polynomial of minimal QI(O)_ and (i) (ii) A n p n Kn is the polar set of Q-I (0) p p(z) THEOREM 3.4. Then as in qm(z) o A and G v(z). (ii) A n P n qf e C( e (An)) Then as with qw(z) where is fixed u’ n tends to a subset of A n G p f(z) unifoly on compact subsets of A_k.p Suppose the hypothesis of theorem 3.3 is satisfied with u’ (i) . O} and (z) is a (U,u)-unisolvent rational approximant to f(z) with Suppose n but ’degree’ Suppose f 0 be fixed. and p {z c a finite polar set defined by G -i np Q,(O) (z) v G s tends to a set containing as a proper subset. n f(z) uniformly on compact subsets of A except on G u Z p Lebesgue measure zero where Z q q of n_ contains the remaining limiting polar set of REMARK ON THE PROOFS OF THE THEOREMS 3.3 & 3.4. The proofs of the two theorems 3.3 and 3.4 as already indicated are essentially the same as that of theorem 3.1 in our paper [6] except for minor changes in the second half of their (i)-parts. the exceptional set has For theorem 3.4 we show further that the set G u Z w q Kn-Lebesgue measure zero. The proofs, for second half of the (i)-parts for theorems 3.3 and 3.4, which n -I n A and -i (0) n An focus on lim Qp(O) P P P p ,-=o their corresponding versions of equation (3.8) in our paper [6]. i.e. determine the relations between G Q(z)q(z)f(z) Q qa,(z)P(z). An (3.1) C.H. LUTTERODT 646 q(a)(a) codim > 2), that R.H.S. of (3.1) must give (except for some subvariety of Q n Dealing with theorem 3.3 first, we find that a 0 O. (0) implies ((z), Since (z)) we must have that q(a) (a) 0 so i for z A 0. n Here u’ and hence as -I n n A Q() Q p i (0) Anp G A n" (3.2) O The validity of equation (3.1) remains in tact on passing from subsequences to the full sequence as discussed in the proof of theorem 3.1. in A A n" n Thus the polar set of (z) I gets completely attracted by the polar set of f(z) with similar multiplicity on 0 In the case of theorem 3.4, the equation (3.1) yields the following: Suppose A O. But since q(z)f(z) # O, except on a sub-varlety of cothen q(a) n0, G a E dimension > %(a) 2 therefore 0. I An p -i Q(O) hav we must With An p (0) G as n An o p ’ (3.3) This remains true on passing to the full sequence from subsequences as done in the proof of theorem 3.1 of [6]. Next we proceed to the final part of theorem 3.4 which begins with a modified version of the inequality (3.7) from our paper [6]. The inequality in question is If(z) Let K be any compact A A and K A, Y np z get np,. (z)l < 1o<--)1 lq(z)l (xEZqn\ EB n and choose p’ > 0 such that subset of A p, p’ < Cl( p If(z) i(0) imply that n A _(G P d first note that in A np is n A n + Ap, n+n el 01 (0) n (3.4) n An (G u Zq) n A np n q(z) 0 on A which we know is not the case. So connected n A in A p" Therefore we can choose a countable sequence { mm Anp, m 1,2 centered at where sponding polydisks U m of Anp and each Um n QI(o) where d is a IR 2n from (3.3). # m 1,2,... Invoking Q {Um }m Em , An_\,Q,I(o) Jensen’s vanishes in Um cannot n P we is must be dense forms a covering inequality we have (3.5) -Lebesgue volume measure for U d We claim (0) with corre- iumlgIFd(Z)[d > that the set on which F have measure zero. (z)q(z) For if not then since A must have empty interior. P Fd(Z) and define IR2n-Lebesgue measure zero. This consequently will of IR2n-Lebesgue measure zero. To prove the result is a set of Zq) n Q-.I(0) u (i n Fd(Z) then the zero set of that B’+I ,(z)l ]-O,(z)’qo(z) (d I n _ 0 < p’ < p implies Then using Lemma 3.2 we can tighten the above inequality to K Now if we let d )" m The inequality (3.5) tells us have a positive measure, i.e. it must n the claim that cover Since there is a countable number of Um Ap, is proved and hence the desired result. Now for z > 6 and IB(z)[ C2 C 2(,C I) K\(G u Z lq(z)l q > > 0 such that and e. ’ sufficiently large we can find > 0 such that Thus on passing to sup-norm in (3.3) we can find CONVERGENCE OF (,u)-SEQUENCES (z)II < C2( P’)P’+I lf(z) 647 p’ 0 <- < 1 (3 6) From (3.6) the desired result on degree of convergence follows, showing the uniform convergence. - The next theorem compares the rate of convergence of (,) sequences of URA v(z) to f ’(A) [o(Z) polynomials is fixed, with the rate of convergence of the Taylor where The following definition shows that ’degree’ of to f. same as that of DEFINITION 3.1. A rational function if in each z.-variable, polynomials in z. las degree given by COROLLARY 3.1a. j* is said to have RIj Ru(z) follows from property (v) of E is the ’degree’ expressed as a quotient of two pseudo- (j max The ’degree’ of (,) D URA in this paper is always N. This in definition 2.1 where each pair satisfies D, making the "Pad@ Table" upper-triangular. Suppose the hypothesis of theorem 3.1 is satisfied. THEOREM 3.5. A\G compact subset of Let K be any then in terms of sup-norm on K, for sufficiently large get If(z) where 0(z) . is the Taylor polynomial PROOF. 1 where F(z) 0 e 0 ’degree’ (U) and let the V, V n G in U where ; to f at the origin. nr Anr (so that A o(Z) be the partial sum of f(z) of n and being A:) be a neighborhood of V being open and connected in A p Let the closure of U. Then o we (z)ll K z Let 0 < r < p and let U the origin such that f ’degree’ 1K (llf(z) (z) ’ FN(z) p(z)nvo(Z) 9(z). and F9(z) (U) and therefore by Cauchy’s F(t) 1 F (z) n dtl...dt n U (2i)n " 0 z.) (t J C() 6 Integral formula ; j=l U is the distinguished boundary of U. Using arguments similar to those employed in the proof of theorem 3.1 of our paper [6] we can write F (z) E z If where fl))t (2gi)n f Ft)t+l (t) i 0U dtl dtn we claim that Fv(z) where f% I )t z El+) \ E (t) 0 (t) U tX+l I (2i)n 0 1 )f x)) (3.6a) dt I .dt n (3.6b) In order to substantiate this claim, we write f(z) + g(z), where is the UO f(z) and is the remainder of the power series expansion of f in U. This makes the function analytic but regular in each variable z. with order partial sum of g(z) g(z) o(Z) C.H. LUTTERODT 648 . + Now observe from equations (2.1a) and (2.1b) that i. Dz [N)(z) no(Z) PN(zlllz=0 x :)za (blj,,(z)gl](z))I Z=0; EI X c (3.7a) and However, the regularity conditions on ;iz gN(Z)Iz=O % O, % . EIj glj(z) in each variable imply that and hence invoking Leibnitz rule in the R.H.S. of (3.7a) we get the R.H.S. to vanish, yielding the claim. . E+\EN We remark that even though for % tive formula for (3.6b): c E form for IJ+V Recall that equation (3.7b) provides an alterna- there is no real advantage gained in adopting the second ",E Qu)(t) MQ We then let l(t)], max 0U t independent of ;o(t) being the partial sum of an absolute convergent development of satisfy l[uo(t)l M, V t e 0U G since V and U Thus from (3.6a) we get using Cauchy inequality on F () Q FI v (z) IQ Thus R.H.S. of (3.8) tends to zero. Ng > . u(z) i < , G v (3.8) Thus given for z E U. n Thus on any compact subset K Now since K UR 0 ] and U (0) n R G [ An v G UR, 0 and clearly as , l’ such that R where V U #, and R we must have for D’ > NI el. , o By Theorem 3 1 Q K 0 large, we get by way of Huitz theorem in z NI > O, N i n This result holds verywhere in the complete [o (z)n[ O(z) i(z)[ Ko < 2 that l 1 Reinhardt domain of convergence of f(z) denoted by U UR on z The R.H.S. of (3.8) is the tal of a geometric series ;’ > tlus U must 0 Here M is independent U V f . v, ’degree’ is a normalized polynomial of fixed multiple it is unifoly bounded on U. H 6 .... 0 such that 0 An as n that ; %(z)l Thus for lv(z) 0 > J’ 0 for all sufficiently for all z c K 0 and furthermore I0() (z) lEo < 60 Using triangular inequality for sup-nos on K O, we get ()lK llf() and for ’ > N llf() 0 K 0 K0 0 we obtain I llf() ()iK lf(z) o(Z) IK0 + since g > 0 is arbitrary, we get on K 0 If(z) (z)ll K0 llf(z) uO (z)l K (3.9) 0 CONVERGENCE OF (,v)-SEQUENCES ’[’his inequality is not violated if K K U 0 , n, "(;,. K R &n\G possible on K is extended to any compact subset K where 0 Theorem 3.1 guarantees the L.H.S. of (3.9) to remain as small as whereas R.H.S. of (3.9) cannot be made arbitrarily small in ’I’n-\U R for sufficiently large U’. Ng ’ nD(z)llK < llf(z) -rDo(Z)IIK. IIf(z) > Thus implies Suppose f(z) is analytic at the origin and is at most meromorphic THEOREM 3.6. Cn with a finite polar set in Suppose for each fixed (z) of each (,) unisolvent rational approximant 1’ 649 (i’ to the polar set n f(z) tends to infinity as n. Then f(z) must be entire in We need tlm following lemma in order to prove the above theorem. to infinity as U’ n be fixed. and only if given any (’I Let LEbMA 3.7. if -i Quv(O) for I’ i A -i n (3.10) For fixed , Qua(O) suppose the polar set . tends to infinity as D Then it is immediate that for any given 0 > 1 and a polydisk n -1 1 I:’ > N QI,(O)_ bn,, a ’ N we have A n n 0 Now on ^n { (z)} _ ttowever, fix lim am a Q:.x)(a) . K I 0. A n 0 # . I An {U<v(z)} for which Then Now let > 1 for m m |Jv’V (z) ’degree’ (z) as so that 00 and suppose {a K 1, ’0 is a normalized sequence of polynomials with fixed and therefore we can choose a subsequence uniformly on a compact subset K a -i Qlv (0) ’. such that NI) To prove the converse, we assume its opposite, i.e. we can choose ’0 tends sufficiently large. PROOF. which U (z) Q(O) of The polar set n 0 > 0 and a polydisk A is a sequence of points in Qlv(am) 0 for ali , QI v(a) Q,(a) Q:X (0) for which m and therefore by continuity we get on K I. ConsequentIy ..,(a) O. Since it follows that A.n ’0 Thus the converse holds and this concludes the proof. The proof of theorem 3.6 was given in Lutterodt [4], so we merely provide an outline for the present paper: Suppose is analytic at the origin and meromorphic in with at most a finite polar set. By theorems 3.1 or 3.4, the polar set of f(z) attracts the whole polar set or a subset of it as the case may be. But for each fixed -i we know by hypothesis that the polar set of tends to infinity as Thus the polar set of f must be drawn to infinity, making f entire in n l(z) n.’ ,. (z), Qua(O), 650 _ C.H. LUTTERODT NON-MONTESSUS TYPE CONVERGENCE. 4. In this section, we consider convergence of (I],)-sequences of unisolvent not necessarily fixed as in 3. The convergence is rational approximants with ’ Bisho weakly expressed in terms of measure for the cases examined, using a lemma of is replaced by ) where in ( [i]. it should be noted that n (l t;e latter v is no longer an n-tuple but a mere natural number. 0) with ( I 6 Let THEOREM 4.1. e qf ’ C(on) g IN but N remains an n-tuple. g where z (i I n) < ) are such that 0 < ( and n e ’degree’ is a polynomial of minimal q0(z) IN be fixed. > 0 and 0 < I < l, and let 0, n and its finite polar set over A is G () Similarly v < q0(z) (0 rain |’ Suppose O} and ). Suppose (.). l<j<n (z) is a (,)-unisolvent rational approximant of f. Then for any > > < 1 and |0 IN, positive such that Hc O, 0 < A compact subset K Suppose for all z _ n, K\ZII where ZDvl where m is En-Lebesgue If(z)- ;](z)ll/’ {z l]v(z)qo(z) . K: ’0 < D)(z)l I/D’ If(z) m[{z e K" ’ < [;+o} />E}] and < measure. The difference between the above theorem 4.1 and the next one theorem 4.2 lies not being fixed in Theorem 4.2. in X ’ K v) is not fixed, with 0 < ), < rain (j) (,) o(D’), (z) is a (,)-unisolvent rational approximant to f(z). but n < 00 l<j<.n Suppose A ,, Suppose the hypothesis of theorem 4.1 is satisfied, except for THEOREM 4.2. ( c compact, < 1 and > O, 0 < If(z) for all z e K\Z I- where m(Z l/i) l] 0 such that ,(z)l < 6 ’ ’ > 0 - as Then for any < e and m is a E n -Lebesgue measure d LEMMA 4.3 (Bishop). Let Fd(Z be a normalized polynomial of degree (d d) in tli > -and 0 < r < 1 be given. Then there Let exists a constant c c(n,o) such that the n. set Z has n-Lebesgue {z e n ]Fd(Z)[ < N d measure satisfying m(Z) < cn 2/n. REMARK. in Narasimhan The proof of this Lemma uses an induction argument and it is presented [8]. PROOF OF THE THEOREM. q(z) same was as qf_ e . (A) i f(z) has a polar set Without loss of generality we assume that qc0(z) is 0}. .(z). n (p) and tion given in the Pl)(z) and u(z) are both therefore qf q (). first part o Since te proof G00 {z e Kn no--nalized in polynomials in En and the 6 Following the same presentaof theorem 3.1 of our paper [6], we extract 651 CONVERGENCE OF (p,)-SEQUENCES the inequality (3.7) and transplant it to this paper as IH.(z)l < ) M( >_, n %f. IN \E From the above inequality we get, where M is as before. (z)l If(z) , <-IQIz(z) qa(z)l A, INn\E (4.1) ). H n 0’ < 0 -> Let K be any compact subset of L and let 0’ > 0 be chosen so that 0 n ,n .n Then appealing to Lemma 3 2 for z e K, (4 i) becomes A and K r, " f, _ bW- P’ (p,n) .z(z)l < Zqlo(z) If(z) (4.2) qe(z)[ () where r(,,n) is a constant depending on p and n. Let z 2*. (’iven ’-- z n 0 < d} lEd(Z) e K- d) is normalized in (d X+ < 1, let d n. + 0) ( + z . + with d and The polynomial Fd(Z) Fd(Z) Thus by Lena 4.3, we must have the )(z)q03(z). of degree En-Lebesgue measure satisfying m( -I" Lrl--) cr 2/n R.H.S. being independent of the degree of F have d IFd(z)l ,,, KZ Hu. onsequentt {z t }(z) 0 and cl However, for ("i we must therefore K/Z If(z) K" (p’)’+l (4.4) we can find If(z) for z (p,n) H(z) p’ < 1 and P For any z d. O, so that lf(z Since 0 (4 3) < (SH’ 2/n < g; the set t} is included in the set Z u and (z)l I/H’ r (4.5) (.3). [rom The proof of Theorem 4.2 is identical to that of Theorem 4.1. recognize that even though because But one has to o(’), this does not pre- (0,1) for which the inequality (4.5) holds. clude us finding 6 and some generalization for the case of meromorphic maps has 4.2 The Theorem ’ as , been discussed in Lutterodt [7]. The next theorem establishes over-convergence for the non-Montessus type. (D) where D is a domain Let 0 > 1 be fixed and let f Suppose the conditions of either Theorem 4.1 or Theorem 4.2 THEORI 4.4 (Gondar). in n such that A n P D. are satisfied and for any m Then (z) f(z) on any (z) n -Lebesgue measure m as RE. f(z) in n K compact compact in D The statement ’ m (z) f(z) on K as H’ means that as f(z) as ’ and l(z) converges to o(’) C.H. LUTTERODT 652 We shall assume without loss of generality that 0 PROOF OF THEOREM 4.4. En of f in D The pole set G has n-Lebesgue measure zero (cf. final part of Theorem 3.4 proof); since D is an open connected set, (D in D. Thus we can select distinct countable disk neighborhoods A n centered at a Oj choose overlapping polydisks &nP’ n points,, A such that J n ,Aj. ,.j 0 n so al,a2,.., D\G00 in Since must be dense D\(; and poly- is compact, we j=l whose respective centers 0 a aj i are linked by a polygonal path which does not intersect the codimension 1 polar set G of f such that c A 0 n The polygonal path then becomes a path along which f n with 0 sufficiently large we must have, for g > O. cnn be analytically continued in D. from theorem 4.1, for N’ m{z e]( n ANP If(z) Here we have identified A n P (z)[ I/D > For each of the overlapping polydisk neighborhoods to (4 6) in fact holds Since =n u n Ojo so that } < An pj 1 (4 6) n ,Ap. a result identical we obtain the following: for p’ =00j sufficiently large m(z e X: If(z) (z)l I/’ > } < which gives the desired result. ACKNOWLEDGEMENT. The author is grateful to the referee for his comments. The research was partly supported by HURD grant #529053. REFERENCES i. BISHOP, E., Holomorphic Completions, Analytic continuations and interpolation of semi-norms, Ann. Math. 78 No. 3 (1963), 468-500. 2. GONCAR, A.A., A local condition for single-valuedness of analytic functions of several variables, Math. USSR Sb. 22 No. 2 (1974), 305-322. 3. KARLSSON, J., WALLIN, H., Rational approximation by an Interpolation procedure in several variables, Pad and Rational Approximation, Eds: S&V. A.P. (1977), 83-100. 4. LUTTERODT, C.H., On a theorem of Montessus de Ballore for (,)-type R.A. in Approximation Theory III (1980), 603-609. 5. LUTTERODT, C.H., On Uniform Convergence for (,v)-type rational approximants in En II Internat. J. Math. & Math. Sci. 4 No. 4 (1981), 655-660. 6. LUTTERODT, C.H., On a Partial Converse of Montessus de Ballore theorem in Approx. Theory 40 (1984), 216-225. 7. LUTTERODT, C.H., Meromorphic functions, maps and their Rational Approximants in ed. Singh Approximation Theory and Spline Functions Reidel (1984a) et. al., 379-396. 8. NAkSIMHAN, R., Several Complex Variables, Univ. of Chicago Press (1971). 9. WALSH, J.L., Interpolation and Approximation by Rational functions in Complex Domain, AMS Col. Pub. 20 (1969). En, En n, j.