MPE Mathematical Problems in Engineering 1563-5147 1024-123X Hindawi 10.1155/2019/7971495 7971495 Research Article Partial and Local Argument Properties of Holomorphic and Meromorphic Complex Functions in Several Variables https://orcid.org/0000-0002-2907-7269 Zhou Jun 1 Duan Zhaoxia 1 Loiseau Jean Jacques School of Energy and Electrical Engineering Hohai University Nanjing 211100 China hhu.edu.cn 2019 682019 2019 21 04 2019 15 07 2019 682019 2019 Copyright © 2019 Jun Zhou and Zhaoxia Duan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The study describes a general argument analysis technique for holomorphic and meromorphic complex functions in several variables, or simply n-variable complex functions with n2. Argument analytic relationships for n-variable complex functions with significance similar to the argument principle for one-variable ones are retrieved partially and locally. More precisely, argument analysis in n-variable complex functions is carried out one-by-one in terms of each and all variables, namely, partially, so that argument-principle-like relations are established in poly-disc neighborhoods of the variable domains, namely locally. The technique is applicable graphically with loci plotting, independent of Cauchy integral contour and locus orientations; it is also numerically tractable without loci plotting via argument incremental integration. Numerical examples are included to illustrate the main results.

National Natural Science Foundation of China 61573001 61703137
1. Introduction

The argument principle for one-variable complex functions is one of the fundamental concepts and theories based on the Cauchy integral, which paves the way for establishing numerous results in complex analysis, just mentioning Rouché’s theorem and the open mapping theorem as well as the maximum modulus principle, besides many others [1, 2]. As an important application of the argument principle, it is frequently employed for locating isolated zeros/poles or singularities of complex functions in terms of distribution and multiplicity; in particular, Nyquist-type stability criteria including classical and generalized ones in various system and control problems are typical results . Lately, the argument principle has further been exploited for structural and spectral aspects of linear dynamical systems in terms of controllability/observability and positive realness . When multidimensional , nonlinear [6, 13], hybrid , and functional models [15, 16] are dealt with, argument features of complex functions in several variables are becoming indispensable and inevitable. Unfortunately, however, the research situation for argument analysis in n-variable complex functions with n2 is extremely poor. As a matter of fact, no systematic studies about the related topics can be found in the complex analysis literature, to the best of the authors’ knowledge.

This note tries to contrive a slightly more general argument analysis technique for holomorphic and meromorphic n-variable complex functions. Several argument properties related to n-variable complex functions that are not known up to now are claimed and proved rigorously, which are in form similar to the argument principle for one-variable complex functions but in some partial and local sense. More specifically, by the suggested technique, argument analysis in n-variable complex functions can be completed in a variable-by-variable fashion, namely, partially in variables, so that relations in between zero/pole distribution and argument evaluations are established in various poly-disc neighborhoods, namely, locally in zero/pole distribution. It is worth mentioning that the argument analysis technique is exploitable graphically through loci plotting, independent of Cauchy integral contour and locus orientations, and it is also implementable merely via argument incremental integration, dispensable of loci plotting.

The paper is arranged as follows. Section 2 collects preliminaries, notations, and terminologies about n-variable complex functions. Novel argument relationships for holomorphic and meromorphic n-variable complex functions are claimed and proved in Section 3. Numerical examples are sketched in Section 4, whereas conclusion is given in Section 5.

2. Preliminaries to <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M12"><mml:mrow><mml:mi>n</mml:mi></mml:mrow></mml:math></inline-formula>-Variable Complex Functions 2.1. Notations and Terminologies

Notations and terminologies herein are standard, whose exact definitions can be found in most textbooks about complex analysis [1, 2, 17, 18].

In what follows, R and C denote the sets of all real and complex numbers, respectively. Rn and Cn, respectively, stand for the n-dimensional real and complex Euclidean spaces; or equivalently, Rn=R××R and Cn=C××C. N is the set of all naturals, while the set of strictly positive naturals is denoted by N+. The boundary of an open and connected set DC is denoted by D. With κN+, we denote κ_=1,,κ and thus kκ_ means that k=1,,κ. As usual, j=-1 and (·)¯ means the complex conjugate.

Notations and terminologies about discs and Cauchy contours on C are explained briefly as in Figure 1, in order to prepare us for generalizing the argument principle for one-variable complex functions to several-variable cases. For describing the disc and the Cauchy contour of Figure 1 with formulas, we write(1)DsC:s-α<γDsC:s-α=γ,where γ>0 is the radius of the disc and αC is the disc center. In other words, D is open and connected and accordingly D is the boundary of D, which is a simple and closed contour. D will be used as the integration contour in the Cauchy integral for argument analysis.

Disc and its boundary on the complex plane C.

To understand our discussion for n-variable complex functions, we need to introduce the so-called poly-contours defined in Cn as in Figure 2. Similar to those in (1), we write(2)DisC:s-αi<γiDisC:s-αi=γi,where, for in_, γi>0 is the radius of the disc Di and αiC is the center of Di on the i-th complex plane. Simply speaking, the pair of Di and Di are defined in the same way as to their counterparts in Figure 1 and (1) but on a specific complex plane accordingly. Also, {Di}i=1n are all open and connected sets with disc radius sufficiently small. All radii of the discs are not necessarily the same. To simplify our subsequent notations, we also write(3)D=D1××DnD=D1××Dn,which are called, respectively, the poly-disc defined with D1,,Dn, and their corresponding poly-circular boundaries. Clearly, D is an open and connected subset of Cn, while D is simple and closed in the individual complex plane sense.

Poly-disc and its poly-circular boundary in the space Cn.

2.2. Facts about <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M57"><mml:mrow><mml:mi>n</mml:mi></mml:mrow></mml:math></inline-formula>-Variable Complex Functions

To facilitate our statements, we recall additional notations and collect facts about n-variable complex functions from [2, 17, 18]. To those readers who are familiar with complex functions in several variables and their analysis, this part can be skipped.

Let f(s1,,sn)H(D,C) denote a n-variable scalar complex function that is a holomorphic mapping from D to C in the sense of f:DC:sf(s), with s being a short symbol for the variables s1,,sn. We also write s=[s1,,sn]TCn. The definition for holomorphic n-variable complex functions can be found in [18, p. 7, Definition 1.2.1]. In rough words, a n-variable complex function f(s) is said to be partially holomorphic, if each and all the resulting one-variable complex functions are holomorphic after one variable is viewed as variable while all other n-1 variables are fixed as constants. Exact definitions are given in [18, p. 13, Definition 1.2.21]. By Remark 1.2.28 of [18, p. 17], every partial holomorphic function is also holomorphic and vice versa. The concept of partially holomorphic n-variable complex functions is the starting point for our partial and local argument analysis.

It is said that s0=[s10,,sn0]TCn is an isolated zero of the n-variable complex function f(s) if f(s0)=0, and f(s) does not vanish identically in any sufficiently small neighborhood around s0; or equivalently, f(s10,,sn0)=0 at [s10,,sn0]TCn, while f(s1,,sn)0 in any sufficiently small neighborhood around [s10,,sn0]T.

The Cauchy line integral for one-variable complex functions is modified with regard to the n-variable complex function f(s)H(D,C) along the contour Di for a fixed in_ as follows:(4)12πjDifsdsi12πjDifs1,,si,,sndsi.To rigorously understand (4) and the ensuing discussions, let us introduce what we call the n-index order a=(a1,,an)TNn with aiN and write that(5)aa1++an,a!a1!an!. Then, we are ready to define the n-variable power multiplication and the partial derivative with the n-index order a, respectively, by(6)s-s0as1-s10a1sn-sn0an=i=1nsi-si0aiPafs0afs1,,sns1a1snans1=s10,,sn=sn0.

To see the definition and properties about the total and partial derivatives of the n-variable complex function f(s)H(D,C), let us write(7)sixi+jyi,in_.Clearly, xi,yiR are the real and imaginary portions of the complex variable si, respectively. Then, we define the total derivative by(8)dfs1,,sn=i=1nfs1,,snsidsi+fs1,,sns¯ids¯i,where(9)fs1,,snsi=12fs1,,snxi-jfs1,,snyifs1,,sns¯i=12fs1,,snxi+jfs1,,snyidsi=dxi+jdyi,ds¯i=dxi-jdyi. In particular, for every fH(D,C), it is already known by the Cauchy-Riemann equations that i=1nf(s1,,sn)/s¯ids¯i=0. Hence, it is straightforward to see that, for fH(D,C), it holds that(10)dfs1,,sn=i=1nfs1,,snsidsi.

Furthermore, it follows readily that, for a class of n-variable complex functions f1,,frH(D,C), it always holds that(11)f1sfrs/sif1sfrs=f1s/sif1s++frs/sifrs,for each in_.

2.3. Taylor Series of Holomorphic <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M110"><mml:mrow><mml:mi>n</mml:mi></mml:mrow></mml:math></inline-formula>-Variable Complex Functions

The following lemma plays a key role in establishing partial and local argument relationships for holomorphic n-variable complex functions [17, 18].

Lemma 1.

Suppose that fH(D,C) has an isolated zero at s0D. Then, there exist a neighborhood UD around s0, an identically nonvanishing holomorphic function g(s) defined on U, and a n-index order a=(a1,,an)TN+n such that f(s)=(s-s0)ag(s), for all sU.

Proof of Lemma <xref ref-type="statement" rid="lem1">1</xref>.

By the Taylor series expansion formula for n-variable complex functions, say Corollary 1.5.9 of [18, p. 36], the holomorphic assumption about f(s) ensures that(12)fs=νNns-s0νν!Pνfs0,sUholds true for any s0D and any sufficiently small neighborhood UD around s0. Here, νN2 means that the n-index power superscript ν=(ν1,,νn) changes from ν1==ν2=0 to ν1=, , νn=, for all ν1,,νnN. The above Taylor series expansion is convergent compactly and holomorphic on U.

Now let a1=minν1, , an=min{νn} with Paf(s0)0 (or equivalently, a=minν); the n-index order a=(a1,,an)T is used. Therefore, we can write(13)fs=s-s0aνNns-s0ν-aν!Pνfs0s-s0ags,where g(s) is obviously defined and thus g(s0)=1/a!Paf(s0)0. By the definition, g(s) is holomorphic and does not vanish identically on U. To complete the proof, it remains to show that a1>0,,an>0. On the one hand, if a1==an=0, then (s-s0)a=1, for all s and s0. This says in particular that (s-s0)ag(s)s=s0=g(s)s=s00, which is contradictory to the assumption of f(s0)=0. On the other hand, without loss of generality, if a1=0 and ai>0 for each i=2,,n, then (s-s0)a=(s2-s20)a2(sn-sn0)an and thus f(s)=0 at s2=s20, , sn=sn0 over all s1C; this is contradictory to the identically nonvanishing assumption of f(s) in U\s0. Cases in terms of si with i1 can be proved similarly.

Remark 2.

By Lemma 1, if the disc radii are sufficiently small and all distinctive isolated roots of f(s)=0 are located in different poly-discs, say {Dp}p, f(s) always has a Taylor expansion f(s)=(s-sp)apgp(s) over sDp, where ap=(a1p,,anp)N+n and gp(s) is holomorphic and identically nonvanishing over sDp. Here, Dp denotes the poly-disc for the p-th distinctive and isolated zero of f(s). When all isolated zeros of f(s) are examined with respect to the poly-discs sequence {Dp}p, a n-variable Taylor series sequence is eventually created for representing f(s) on the whole DCn.

In what follows, for discussion brevity and without loss of generality, only one of the isolated zeros is considered directly and explicitly. In this sense, the subscript p will be dropped.

2.4. Complex Logarithm for <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M186"><mml:mrow><mml:mi>n</mml:mi></mml:mrow></mml:math></inline-formula>-Variable Complex Functions

Now we collect some facts about the complex logarithm for one-variable functions and then extend them to n-variable complex functions. To a complex scalar z=ζejθ with ζ>0 and θR, it is natural to set logz=logζ+jθ. However, by this definition, θ is unique only up to an integer multiple of 2π. In view of this, we define the complex logarithm of 0zC as(14)logzlogz+jz+2rπ,r=0,±1,,where z=θ<π. For each r, a branch or a sheet of the complex logarithm of z is meant; in particular, the one with r=0 corresponds to the principal branch.

Accordingly, for a n-variable complex function fH(D,C) that does not vanish identically on D, the complex function logarithm is defined by(15)logfs=logfs+jfs+2rπ,r=0,±1,. In other words, the complex logarithm logf(s) can also be viewed as a complex phaser whose argument is f(s)+2rπ for some integer r.

If we treat f(s) as a one-variable complex function in the sense that si acts as variable and all other variables are fixed, Theorem 6.2 of [2, p. 100] tells us that logf(s) is holomorphic in the si-variable in each specific logarithm branch. It follows by Remark 1.2.28 of [18, p.17] that logf(s) is partially holomorphic on D and thus(16)logfssi=1fsfssi=1fs1,,snfs1,,snsiis well-defined for each and all in_.

3. Argument Analysis for <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M215"><mml:mrow><mml:mi>n</mml:mi></mml:mrow></mml:math></inline-formula>-Variable Complex Functions

In this section, we explain the partial and local argument analysis in n-variable complex functions.

Consider fH(D,C). Therefore, if we reexpress it as f(s)=(s-s0)ag(s) in some poly-disc D=D1××Dn as in Lemma 1 and Remark 2, together with (11), we can write(17)logfssi=logs-s0agssi=1s-s0agss-s0agssi=1s-s0as-s0asi+1gsgssi=s1-s10a1sn-sn0an/sis1-s10a1sn-sn0an+1gsgssi=aisi-si0+1gs1,,sngs1,,snsi,for all in_. Clearly, (17) holds in the si-variable sense for a specific in_ while all the non-si variables are fixed at some sk=ckDk (namely, kn_ but ki), over which g(s) is partially holomorphic and identically nonvanishing as appropriately.

By applying the Cauchy line integral as modified in (4) to relation (17) along the contour Di in the si-variable sense, while all the other variables sk are viewed as constants in Dk with kn_ and ki as appropriately, we have(18)12πjDidlogs-s0ags=12πjDiaisi-si0dsi+12πjDi1gs1,,sngs1,,snsidsi=ai+12πjDidloggs=ai,for all in_. In the above, we used the fact that 1/2πjDiai/(si-si0)dsi=ai by the Cauchy residue formula. Similarly, it follows that 1/2πjDid(logg(s)) is well defined and equal to zero, since g(s) is partially holomorphic in si and nonvanishing identically on D so that the argument difference of logg(s) must be zero when si runs along the specific contour Di. This is because the origin (0,j0) does not lay in the interior of the set encircled by the locus g(s1,,sn)siDi, no matter how the other variables are taken in their corresponding discs, except for the disc centers.

Furthermore, noting also that, on the same logarithm branch, the following argument increment relation is satisfied:(19)Didlogs-s0ags=Didlogfs1,,sn=fs1,,sns1,,sn+-fs1,,sns1,,sn-,where (s1,,sn)+ and (s1,,sn)- stand for the two sides of a point (s1,,sn) when si runs clockwisely along Di, while all the other variables are fixed as appropriately.

Finally, we can claim the following partial argument relation:(20)ai=12πfc1,,si,,cnc1D1p,,siDi,,cnDnp,where in_. In (20), the notation ckDk (kn_) means that the variable sk is fixed at a constant ckDk, or equivalently sk=ckDk. In what follows, f(·,,·)c1D1,,siDi,,cnDn denotes the argument incremental of f(·,,·) when the integral variable si runs clockwisely along Di and the other variables are fixed as appropriately.

Based on (20), we claim the following results.

Theorem 3.

Consider a n-variable complex function f(s)H(D,C). For an isolated zero s0 of f(s), one can always fix an open and connected poly-disc D=D1××Dn, which is a sufficiently small neighborhood around s0, such that (A1) f(s) has no zeros other than s0 in D; (A2) f(s) vanishes nowhere on the poly-circular boundary D=D1××Dn, that is, f(s)0 for any sD. Then, the n-index order a=(a1,,an)T about the zero s0 satisfies(21)a=12πjD1dlogfsDndlogfs=12πfs1,c2,,cns1D1,c2D2,,cnDnfc1,,cn-1,snc1D1,,cn-1Dn-1,snDn.

Proof of Theorem <xref ref-type="statement" rid="thm1">3</xref>.

Assertion (21) follows by rearranging relation (20) into a vector according to in_. Assumptions (A1) and (A2) guarantee that the Cauchy integrals are well defined in the sense that all the complex logarithms pointwisely throughout the contour are validated and the argument incremental can be calculated.

Remarks about Theorem 3 are summarized as follows:

In the above discussion, it is our underlying assumption that the Cauchy contour orientation is specified as clockwise, and the corresponding loci orientations are determined accordingly. Indeed, the orientations of the Cauchy contours and those of the corresponding loci can be self-defined such that each and all the argument increments are nonnegative as appropriately, when isolated zeros are concerned.

If the n-index order a possesses one or more 0-entries, this actually says that the n-variable complex function f(s) has no isolated zeros in the concerned poly-disc D. In view of this, it is reasonable to say that f(s) may or may not have isolated zeros in D, depending on how the poly-disc is defined; or alternatively, the n-index order a is D-related. In this sense, the results of Theorem 3 can only be interpreted locally.

For each entry-order of the zero s0 with respect to a specific variable, the other variables need to be treated as constants. In principle, these fixed ones can be arbitrarily taken as long as they are in the corresponding discs, excluding the disc centers as appropriately. In this sense, the argument equation (21) consists of single-variable relations; or the results of Theorem 3 need to be interpreted partially.

The argument incremental formula in the second relation of (21) can be also obtained simply via numerical integrations, without loci plotting. In this sense, Theorem 3 can be employed graphically with loci plotting or purely numerically without loci plotting. This is also the case for Theorem 4, though we will not mention this point again.

Different from the argument principle for one-variable complex functions, n loci (or n argument conditions) are involved in examining existence and multiplicity of isolated zeros for n-variable complex functions. This is also the case for Theorem 4 below.

Isolated singularities such as poles of partially meromorphic n-variable complex functions can be dealt with simiarly. This is talked about briefly in the following.

Let fM(D,C) mean that the n-variable scalar complex function f(s) is partially meromorphic (for want of better words, detailed definition can be found in ). This means that, about f(s), when each of the variables is viewed as variable, and all the other variables are fixed as constants, each and all the resulting one-variable functions are meromorphic in the one-variable sense. It is said that a singularity p0D is a pole of f(s) if p0 is an isolated zero of 1/f(s). Then, we can prove the following results.

Theorem 4.

Consider a n-variable complex function f(s)M(D,C). For a pole p0 of f(s), there is always an open and connected poly-disc D=D1××Dn, which is a sufficiently small neighborhood around p0, such that (A1) f(s) has no pole other than p0 in D; (A2) f(s) vanishes nowhere on the poly-circular boundary D=D1××Dn, that is, f(s)0 for any sD. Then, the n-index order b=(b1,,bn)T of the pole p0 satisfies(22)b=-12πjD1dlogfsDndlogfs.Here, the orientations of the Cauchy contours and the corresponding loci are specified in such a way that the n-index order a in (21) possesses nonnegative entries.

Proof of Theorem <xref ref-type="statement" rid="thm2">4</xref>.

Under the given assumptions, it remains only to show that, for the pole p0=(p10,,pn0)TDCn, we can always conclude that f(s)=(s-p0)bh(s) and(23)fs/sifs=bisi-pi0+loghssi,for some identically nonvanishing holomorphic n-variable complex function h(s)h(s1,,sn) in any small neighborhood around p0, and with bi being the order of the pole in Di.

To this end, we notice by the definition about p0 that p0 must be an isolated zero of 1/f(s). Hence, there exists a poly-disc D, which is a neighborhood of p0, such that 1/f(s) is holomorphic and by Lemma 1, 1/f(s)=(s-p0)ag(s) holds over sD for some n-index order a and g(s) that is nonvanishing identically on D. Based on these facts, we observe that (24)fs/sifs=-1/fs/si1/fs=-s-p0ags/sis-p0ags=-s-p0a/sis-p0a-gs/sigs=-aisi-pi0-gs/sigs=bisi-pi0+hs/sihs,for each in_. The last equation follows readily after letting bi=-ai and h(s)=-g(s), which yields to us the desired relation (23).

Remarks about Theorems 3 and 4 are as follows:

Since Theorem 4 is also claimed with poly-discs and the corresponding poly-circular boundaries, it may not be true if the poly-circular boundaries are replaced with the so-called toy contours defined on each different complex plane . This means in rough words that Theorem 4 only applies partially and locally, as seen by Theorem 3.

In general, isolated zeros and poles of meromorphic n-variable complex functions cannot be dealt with in the same poly-disc rigorously. This is because the proof arguments for Theorems 3 and 4 are constructed by means of Lemma 1, which is essentially true merely in some local sense. Even if a concerned zero and a pole are located in each other’s small neighborhoods, their Taylor series expansions may not be unified through analytic continuation [2, p.53], noting once again that there exist some ’frozen’ variables.

The minus “−” in (22) is not essential and has no absolute meaning. It is added in (22) simply for the sake of expressing the n-index order b in accordance with a in (21).

4. Numerical Illustrations

In this section, argument analysis for several 2-variable complex functions is illustrated numerically. The coefficients a and b in the examples are fixed at a=2 and b=3 such that b2-4a<0 and the polynomial (s2+as+b) has a pair of conjugate zeros at -1±j2. The discs D1 and D2 have the radii pair (γ1,γ2){(0.1,0.2),(0.15,0.25),(0.2,03)} and will be specified in the following; accordingly, the Cauchy contours D1 and D2 are meant as in Figure 2. The loci with (γ1,γ2)=(0.1,0.2) are plotted by black curves, while the loci with (γ1,γ2)=(0.15,0.25) and (γ1,γ2)=(0.2,03) are in blue and red curves, respectively.

As explained in the remarks about Theorems 3 and 4, zeros and poles of the examples, if any, can be examined directly in a numerical fashion without loci plotting. However, we do implement the results graphically with loci plotting so that the reader can have better understanding.

4.1. 2-Variable Complex Function with Isolated Zeros

In this case, we examine the 2-variable complex function:(25)f1s1,s2=s12+as1+bs2-12s12+s22exps1s2.Therefore, f1(s1,s2) is holomorphic in the whole space C2. Clearly, f1(s1,s2) has an isolated zero s0=(-1+j2,1)T with a=(1,2)T if the two discs are given by(26)D11=s1C:s1+1-j2<γ1D21=s2C:s2-1<γ2.Or f1(s1,s2) has the isolated zero s0=(-1-j2,1)T with a=(1,2)T if the two discs are(27)D12=s1C:s1+1+j2<γ1D22=s2C:s2-1<γ2.

The corresponding loci based on the above two poly-discs are illustrated in Figures 3(a) and 3(b), respectively. Based on Theorem 3, the loci encirclements around the origin (0,j0) in Figure 3 and the numerical argument integration exactly coincide with the isolated zero existence and the 2-index order a evaluations determined in the above.

Loci for the 2-variable complex function f1(s1,s2).

D = D 1 ( 1 ) × D 2 ( 1 )

D = D 1 ( 2 ) × D 2 ( 2 )

4.2. 2-Variable Complex Function without Isolated Zeros

In this case, we examine the 2-variable complex function:(28)f2s1,s2=s12+as1+bs1+s2-1exps1s2. Clearly, f2(s1,s2) is holomorphic in the whole space C2, and it has no isolated zero, no matter how the disc D2 is defined.

The corresponding loci with respect to the same poly-discs in Section 4.1 are illustrated in Figures 4(a) and 4(b), respectively. Based on Theorem 3, since neither of the two dash-dotted loci in Figures 4(a) and 4(b) encircles around the origin (0,j0), there is no isolated subzero in the corresponding complex plane; or f2(s1,s2) has no isolated zeros in the specified poly-disc. Our numerical argument integration also reflects this point.

Loci for the 2-variable complex function f2(s1,s2).

D = D 1 ( 1 ) × D 2 ( 1 )

D = D 1 ( 2 ) × D 2 ( 2 )

4.3. 2-Variable Complex Function with Poles

In this case, we examine the 2-variable complex fraction:(29)f3s1,s2=s12+as1+bs1+1s2-12exps1s2. It follows readily that f3(s1,s2) is partially meromorphic in the whole space C2. Clearly, f3(s1,s2) has no isolated zero, while it has a pole p0=(-1,1)T with b=(-1,-2)T.

The corresponding loci in terms of the poly-disc D=D1(3)×D2(3) with D1(3) and D2(3) being as follows are illustrated by Figure 5:(30)D13=s1C:s1+1<γ1D23=s2C:s2-1<γ2.Based on Theorem 4, the loci encirclements around the origin (0,j0) in Figure 5 and the numerical argument integration exactly coincide with the pole existence and the 2-index order b determined according to the definition.

Loci for the 2-variable complex function f3(s1,s2).

4.4. 2-Variable Complex Function with Cancellation of Pole/Nonisolated Zero

Now we examine the 2-variable complex fraction:(31)f4s1,s2=s12+as1+bs1+s2s1+1s2-12exps1s2,which is partially meromorphic in the whole space C2. Clearly, f4(s1,s2) has no isolated zero, while it seems to have a pole p0=(-1,1)T with b=(-1,-2)T by the formula. However, the loci of Figure 6 show that this is not true.

Loci for the 2-variable complex function f4(s1,s2).

Carefully examining f4(s1,s2), we can see that, for s1D1(3), limγ10(s1+s2)=-1+s2. This in turn implies that(32)limγ10f4s1,s2=limγ10s12+as1+bs1+1limγ10s1+s2s2-12exps1s2=limγ10s12+as1+bs1+11s2-1exps1s2holds true no matter how the disc D2(3) is defined. In other words, when f4(s1,s2) is investigated on D1(3)×D2(3), it reduces to the following 2-variable complex fraction partially and locally:(33)f4s1,s2=s12+as1+bs1+1s2-1exps1s2.Clearly, f4(s1,s2) has a pole p0=(-1,1)T with b=(-1,-1)T. This has been illustrated by Figure 6 graphically. If we plot the loci of f4(s1,s2) in Figure 7, the same conclusion follows.

Loci for the 2-variable complex function f4(s1,s2).

Finally, by comparing f4(s1,s2) and f4(s1,s2), it is said that pole/zero cancellation occurred in the poly-disc limit sense. However, such cancellation is not removable, since it is due to the existence of nonisolated zero s1+s2=0.

5. Conclusions

This pure mathematical note is aiming at creating a more general complex analysis technique for argument-related features in complex functions in several variables. The main contributions, namely, Theorems 3 and 4, are several argument relationships expressed in the partial and local sense for holomorphic and meromorphic complex functions in several variables. Major implementation issues are also talked about in the relevant remarks.

Carefully examining the methodological aspects of the suggested technique, one might assert that the technique can be interpreted as decomposing the original n-variable complex function to a group of one-variable complex functions, to which the one-variable complex analytical concepts and tools are applied. This study reveals for the first time to what extent the one-variable complex analysis theory can be extended to n-variable complex functions, whenever argument-related features are concerned. It is also worth doing to exploit the results for addressing control problems, where multidomain complex/frequency features are encountered, though we did not probe into any details about these points in this paper.

Data Availability

The data for the numerical simulation are available at request.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

The study is completed jointly under support of the National Natural Science Foundation of China under Grant no. 61573001 and no. 61703137.

Lang S. Complex Analysis 1999 103 4th New York, NY, USA Springer-Verlag Graduate Texts in Mathematics 10.1007/978-1-4757-3083-8 MR1659317 Stein E. M. Shakarchi R. Complex Analysis 2003 2 Princeton, NJ, USA Princeton University Press Princeton Lectures in Analysis MR1976398 Brockett R. Byrnes C. Multivariable Nyquist criteria, root loci, and pole placement: A geometric viewpoint IEEE Transactions on Automatic Control 1981 26 1 271 284 10.1109/TAC.1981.1102571 Zbl0462.93026 Fardad M. Bamieh B. An extension of the argument principle and Nyquist criterion to a class of systems with unbounded generators IEEE Transactions on Automatic Control 2008 53 1 379 384 10.1109/TAC.2007.914233 2-s2.0-56549123578 Zbl1367.93486 Vidyasagar M. Bertschmann R. K. Sallaberger C. S. Some simplication of the graphical Nyquist criterion IEEE Transactions on Automatic Control 1988 33 3 301 305 Zhou J. Qian H. M. Lu X. B. An argument-principle-based framework for structural and spectral characteristics in linear dynamical systems International Journal of Control 2016 90 12 2589 2604 10.1080/00207179.2016.1260163 Duan Z. X. Karimi H. R. Xiang Z. R. Stability and l1-gain analysis for positive 2-D systems with state delays in the Roesser model Mathematical Problems in Engineering 2013 2013 10 169713 10.1155/2013/169713 Duan Z. X. Zhou J. Shen J. Finite frequency filter design for nonlinear 2-D continuous systems in T–S form Journal of The Franklin Institute 2017 354 18 8606 8625 10.1016/j.jfranklin.2017.09.035 Zbl1380.93157 Duan Z. X. Ghous I. Akhtar J. Ali K. Jeffery M. H. Robust L2 - L∞ filter design for uncertain 2-D continuous nonlinear delayed systems with saturation IEEE Access 2018 6 73647 73658 10.1109/ACCESS.2018.2884254 2-s2.0-85058135120 Duan Z. X. Ghous I. Huang S. Fu J. Fault detection observer design for 2-D continuous nonlinear systems with finite frequency specifications ISA Transactions® 2019 84 1 11 10.1016/j.isatra.2018.09.026 Duan Z. X. Zhou J. Shen J. Filter design for discrete-time two-dimensional T–S fuzzy systems with finite frequency specification International Journal of Systems Science 2019 50 3 599 613 10.1080/00207721.2018.1564086 Duan Z. X. Shen J. Synchronization problem of 2-D coupled dynamical networks with communication delays and missing measurements Multidimensional Systems and Signal Processing 2019 30 1 39 67 10.1007/s11045-017-0545-1 2-s2.0-85038071944 Zhou J. Interpreting Popov criteria in Luré systems with complex scaling stability analysis Communications in Nonlinear Science and Numerical Simulation 2018 59 306 318 10.1016/j.cnsns.2017.11.029 MR3758390 Zhou J. Qian H. M. Stability analysis of sampled-data systems via open-/closed-loop characteristic polynomials contraposition International Journal of Systems Science 2017 48 9 1941 1953 2-s2.0-85013040850 10.1080/00207721.2017.1290298 Zbl1371.93172 Zhou J. Complex scaling circle criteria for Luré systems International Journal of Control 2019 92 5 975 986 10.1080/00207179.2017.1378439 MR3957119 Zhou J. Gao K. T. Lu X. B. Stability analysis for complicated sampled-data systems via descriptor remodelling IMA Journal of Mathematical Control and Information 2018 10.1093/imamci/dny031 Hörmander L. An Introduction to Complex Analysis in Several Variables 1990 7 3rd Amsterdam, Netherlands North-Holland Mathematical Library, North-Holland Publishing Co. MR1045639 Scheidemann V. Introduction to complex analysis in several variables 2005 Basel, Switzerland Birkhäuser Verlag MR2176976