You are on page 1of 10

# From the Fundamental

Theorem of Algebra
to Astrophysics:
A “Harmonious” Path
Dmitry Khavinson and Genevra Neumann

T
he fundamental theorem of algebra Indeed, the harmonic polynomial h(z) = z n − z n
(FTA) tells us that every complex has an infinite number of zeros (the zero set
polynomial of degree n has precisely consists of n equally spaced lines through the
n complex roots. The first published origin). In 1992 Sheil-Small conjectured that if
proofs (including those of J. d’Alembert n := deg p > m := deg q, then h has at most n2
in 1746 and C. F. Gauss in 1799) of this conjecture zeros. In 1994 Wilmshurst found a more general
from the seventeenth century had flaws, though sufficient condition for h to have a finite num-
Gauss’s proof was generally accepted as correct ber of zeros and settled this conjecture using
at the time. Gauss later published three correct Bézout’s theorem from algebraic geometry. While
proofs of the FTA (two in 1816 and the last Wilmshurst’s bound on the number of zeros is
presented in 1849). It has subsequently been
sharp, he also conjectured a smaller bound when
proved in a multitude of ways, using techniques
the degrees of p and q differ by more than one.
from analysis, topology, and algebra; see [Bur 07],
In 2001 the first author and G. Świa̧tek [KS 03]
[FR 97], [Re 91], [KP 02], and the references
proved that the bound in Wilmshurst’s conjecture
therein for discussions of the history of FTA
held for the case of f (z) = p(z) − z. Because this
and various proofs. In the 1990s T. Sheil-Small
and A. Wilmshurst proposed to extend FTA to proof involves complex dynamics, it is natural to
a larger class of polynomials, namely, harmonic wonder whether this approach can be extended to
polynomials. (A complex polynomial h(x, y) is find a bound on the number of zeros of the ration-
called harmonic if it satisfies the Laplace equation al harmonic function f (z) = p(z)/q(z) − z. The
△h = 0, where △ := ∂ 2 /dx2 + ∂ 2 /dy 2 .) authors explored this question in 2003 [KN 06].
A simple complex-linear change of variables After posting a preprint, we learned that this
z = x + iy, z = x − iy allows us to write any com- bound settles a conjecture of S. H. Rhie concern-
plex valued harmonic polynomial of two variables ing gravitational lensing. More precisely, it gives
in the complex form the maximal possible number of images of a light
h(z) := p(z) − q(z) source (such as a star or a galaxy) that may occur
where p, q are analytic polynomials. While includ- due to the deflection of light rays by some massive
ing terms in z looks harmless, the combination of obstacles. Even more surprising, Rhie had already
these terms with terms in z can have drastic effects. shown that this bound is attained.
In this expository article we give a brief intro-
Dmitry Khavinson is professor of mathematics at the duction to gravitational lensing. We also describe
University of South Florida, Tampa. His email address necessary background concerning harmonic poly-
is dkhavins@cas.usf.edu. He gratefully acknowledges nomials and Wilmshurst’s conjecture, as well as
partial support from the National Science Foundation un-
related results. We then discuss the ideas behind
der the grant DMS-0701873.
Genevra Neumann is assistant professor of mathematics the proofs and also look at the question of sharp-
at the University of Northern Iowa, Cedar Falls. Her email ness. We close with a discussion of several possible
address is neumann@math.uni.edu. directions for further study.

666 Notices of the AMS Volume 55, Number 6

Theoretical work details about this form of the lens equation. and its history include [SEF 92] (written for physicists). This phenomenon is known as gravitational mi- crolensing. condensed galaxies or black holes) and imagine a light source S (for example. This is the basic idea lies between the observer and the light source. of S can be formed. Due to deflection of light from S by these point mass- es. see on gravitational lensing continued until the late [Wit 90] and [St 97]. We then object L between you and S. multiple images S1 . This calculation pre. One can think of the numbers. and the images are related.. it is used to find distant objects corresponds to a lensed image of S. a star or a quasar) farther away from the observer than these point masses. for example. However. [PLW 01] (written for both mathemati- cians and physicists). 1930s.. our source. P n w =z− σj /(z − zj ) . Light mass of these point masses that is orthogonal to rays emanate from S in all directions and you the line of sight between the observer and the would expect to “see” S along the shortest path center of mass of the point masses. The discovery of quasars in the 1960s If z satisfies the lens equation for a given value reawakened interest in gravitational lensing and of w . What does this have to do with counting zeros of rational harmonic functions? It’s time for us to discuss how the positions of the point masses in our lens. When we let w be the the discovery of a gravitational lensing system in position of S. L is a mapping from the lens plane to the source tonian mechanics. We will choose the origin of each plane to images at S1 and S2 in terms of the angle between lie on the line between the observer and the center the original light ray from S and the “observed” ray of mass of our point masses. resent the position of each item using complex Let’s return to Figure 1. General references on gravitational lens- ing. Light rays traveling construct a plane through S that is parallel to the past L are bent. but revisited his assumptions in 1915 using general relativity. The point masses are said to form a gravitational lens L. dicted a deflection angle that was twice as large as j=1 that predicted by Newtonian mechanics. J. This plane is between you and S. you “see” lens plane (source plane). Project each point to the observed image and calculate this angle in mass of our lens to the lens plane. of gravitational lensing observed using NASA’s Since we’re working with planes. Note that the lens plane images of S at S1 and S2 . objects. there is a massive called the lens plane or deflector plane. Note that the in the universe and determine the masses of these right hand side of the equation is often called the June/July 2008 Notices of the AMS 667 . Soldner’s calculations were based on New. terms of a potential. Instead of seeing S. . Nowadays. that they can be treated as co-planar (i. Basics of gravitational lensing the point masses are sufficiently close together (adapted from [Wa 98]). We will look at the case where Figure 1. A.. Figure 2 is an example Figure 1 illustrates this for n = 1. of gravitational lensing. Measure- ments of the deflection angle of starlight during a where σj is a nonzero real constant related to the solar eclipse in 1919 provided early experimental mass of the jth point mass in our lens. our gravitational lens will map z to the posi- it entered the mainstream in astrophysics after tion w in the source plane. and the many references therein. We construct a plane through the center of night and looking for a star S (see Figure 1). The lens equation for angle. Einstein arrived at a similar plane and is given by answer in 1911. the distance between the point masses in L is small compared to the distance between the observer and the point masses and also small as compared Astrophysics: Gravitational Microlensing to the distance between the source and the mass- Imagine that you are stargazing some clear. [Wa 98] (an online survey article). dark es). Lens equation Imagine n point masses (for example.e. each solution of the lens equation 1979. Soldner is credited with the the jth point mass of our lens L is projected to first published (1804) calculation of the deflection position zj in the lens plane. S2 . For further support for general relativity. we can rep- Hubble Space Telescope.

If the source is directly behind the lens (from the observer’s viewpoint). she showed in [Rh 03] that the conjectured bound is attained for Consequences of the lens equation every n > 1. X n ages using the lens equation given above? We first r (z) = σj /(z − zj ) + w . but one cannot really see the dim image here. instead of seeing S. a shear term (linear sible number of images produced by an n-lens is term in z) is added to the lensing map. Sticking at least 3n + 1. real polynomials What happens if n > 1? H. Witt [Wit 90] showed ( by a direct calculation (not involving Bézout’s A(x. Moreover. Sharon (Tel Aviv University) and E. By writing z = x + iy. There are actually five images (four are bright and one dim). Chwolson in 1924 and is usually called an of a complex valued harmonic polynomial h is Einstein ring. In other the Sheil-Small conjecture concerning the number words. of a rational harmonic function. Number 6 . A. This effect was predicted [Wil 98]). extra (“tidal”) gravitational pull by a distant object H. lens L. H. To model the effect caused by an images is at most n2 + 1. something very remarkable Counting Zeros occurs when n = 1. Notice that the observer. Figure 2. Ofek (Caltech). we see that the lens equation becomes FTA. NASA. y) := Re h(z) theorem) that the maximum number of observed B(x. Rhie [Rh 01] conjectured that to the very basics. 668 Notices of the AMS Volume 55. We begin with Wilmshurst’s approach to that of a circle centered about the lens. S. Putting w = 0 = z1 in the lens Let us return to the question of extending the equation. and source is identical to finding the number of zeros source S do not lie on a straight line in Figure 1. which has been lensed by the bright galaxy in the center. Figure 3 shows some Einstein rings equivalent to finding the zeros of a system of two observed using the Hubble Space Telescope. S. for an n-lens is 5n − 5. If we let What can we say about the number of lensed im. lens. Mao. The bluish bright spots towards the center are the lensed images of a quasar (a QUASi-stellAR radio source). Petters. (Photo credit: ESA.) lensing map. j=1 note that the number of lensed images depends on the relative positions of the observer. we shall omit this term in our the upper bound for the number of lensed images discussion. Witt [MPW 97] showed that the maximum pos- (a galaxy “far. far away”). y) := Im h(z). finding the number of “lensed” images of the and source. the observer would see of zeros of harmonic polynomials ([SS 02] and a circle with center L. finding the zeros by O. K.

Koopmans (Kapteyn). Moreover. and if A and B are relatively prime with deg A = m and T. and the SLACS Survey team: A. k(z) = p(z) − q(z). Bshouty. If n := deg p > m := Note that this conjecture was also proven inde- deg q. precisely n × n = n2 . then it has at most n2 zeros. applying a rotation is constant on an analytic arc (notice that f is not and a translation to each of these lines gives the required to be a harmonic polynomial). zero set of Im h(z) at exactly one point. then pendently by R. S. Suez [BHS 95]. then it through the origin. In contrast to this example where deg q = n − 1. Wilmshurst showed that if a complex harmonic An elementary calculation shows that the zero set function f (z) has a sequence of distinct zeros of Re h(z) forms a set of n equally spaced lines converging to a point in the domain of f . ESA. Figure 4 shows the zero sets also that f is entire and has an infinite number of of Re h(z) and Im h(z) for the case n = 2. thus. Moreover. there exist complex polynomials p and deg p − 1 as in Theorem 1. tinct zeros is probably too large when deg q < n−1 limz→∞ |h(z)| = ∞. Suppose zero set of Im h(z). the He combined these ideas to prove Sheil-Small’s cardinality of the zero set of h(z) is given by the conjecture and also constructed an example to total number of intersections of these lines and is show that the n2 bound is sharp. while the sharpness result was also obtained ♯{z : p(z) − q(z) = 0} ≤ m(m − 1) + 3n − 2. Bolton (Harvard/Smithsonian). Wilmshurst noted that the upper bound of n2 dis- When the degrees of p and q are different. June/July 2008 Notices of the AMS 669 . [Wil 98] that implies sharpness: if deg h = n and if h(z) = p(z) − q(z) has a finite Example. iπ iπ Further.Figure 3. W. zeros in a bounded set. Thus. then the number of common solutions We now discuss Wilmshurst’s elegant example for A = 0 and B = 0 does not exceed mn. (Photo credit: NASA. and L. If h(z) = p(z) − q(z) is k(z) := −2eiπ /4 h(z) = z n +(z −1)n +iz n −i(z −1)n . L. Let Theorem 1 ([Wil 98]).  q where deg q = n − 1 such that the upper bound n2 is attained. The sources are actually extended and that is why one sometimes sees arcs rather than complete rings. we see that deg q = n − 1 = Moreover. Wilmshurst then showed A moment’s thought shows that each line from that f must be constant on a closed loop and the zero set of Re h(z) intersects each line in the is therefore constant by the maximum principle. h can have only finitely and proposed the following many zeros and Sheil-Small’s conjecture follows. h(z) := Im(e− 4 z n ) + iIm(e 4 (z − 1)n ).) Recall that Bézout’s theorem essentially says that independently by D. Consider number of zeros. if deg p ≠ deg q. Einstein rings produced by a galaxy behind the lensing galaxy. Hengartner. Conjecture 1 ([Wil 98]). Schmid [PS 98]. Peretz and J. Moustakas (JPL/Caltech). T. then h has at most n2 zeros. Treu (UCSB). a harmonic polynomial of degree n such that k(z) has the same number of zeros as h(z). Burles (MIT). then limz→∞ |h(z)| = ∞. Writing limz→∞ |h(z)| = ∞. deg B = n.

The authors studied this case. 3n − 2 bound is sharp for all n. it is always odd. Geyer the added complication that finite poles must also [Ge 08] used complex dynamics to show that the be considered. Moreover. where p and q are relatively prime polynomials in z. pages 50-55]. the following result (well known to astronomers) of W. In 2001.5 1 1. examples when n = 4. Let n > 1. the example above shows that ages produced by a smooth mass distribution is fi- Conjecture 1 holds and is sharp. using elementary The number of lensed images by an n-mass planar complex dynamics and the argument principle lens with zero shear cannot exceed 5n − 5. Bshouty and A. Rhie [Rh 03]. Theorem 3 settles Rhie’s conjecture -1. 3 and for several Corollary 1 ([Rh 01]. [KN 06]).5 an astrophysicist. the question of sharpness for rational functions had been resolved previously by an example of S. finding the j=1 number of “lensed” images of the source is iden- tical to the situation described in Theorem 3. be a rational function. Theorem 2 ([KS 03]). n > 1} ≤ 3n − 2. The proof of Theorem showed that the 3n − 2 bound in Conjecture 2 3 for rational harmonic functions is similar. 0. If n > 1. Let us explicitly state this case as point masses and argued that the number of im- Conjecture 2 ([Wil 98]). Let r (z) = p(z)/q(z). More recently. H. Wilmshurst’s example for n = 2 (the of the bound in Theorem 3. with is sharp for n = 4. Back to gravitational lensing P n -1 If we let r (z) = σj /(z − zj ) + w . In other words. Świa̧tek and the first bound is sharp. 6. Number 6 . this case? Let p(z) b it is natural to con- Treating h as a mapping of C. Crofoot [Sa 99] our discussion of counting zeros with verified Conjecture 2 for n = 2. 5. smoothly-distributed mass. Theorem 3 ([KN 06]. we let under consideration to include rational harmonic functions and extend the methods in [KS 03] to h(z) := z − p(z). Burke [Bu 81] yields It can be found in Remark 2 of [Wil 98] and is Theorem 4 ([Bu 81]). Lyzzaik [BL 04] monic polynomials [KS 03]. G. n > 1} ≤ 3n − 2.5 and Rhie’s results settle the question of sharpness Figure 4. then ♯{z : r (z) − z = 0} ≤ 5n − 5. [Rh 03]. For m = n − 1. 8. -0. The degree of r 1 is given by deg r := max{deg p. Main Ideas Behind the Proofs We shall sketch the proof of Theorem 2 for har- In a 2004 paper D. L. Corollary 2). when the lensing is done by a set of Im h(z) is in red). The simplest of nite. rather than point masses. deg q}. and this for harmonic mappings.5 In contrast to the polynomial case. ages will be even when n is odd and odd when n is even. even when n is odd and odd when n is even. the remaining cases of Conjecture 1 is the case Rhie [Rh 01] extended this result to the case of m = 1. This also follows from the proof of ♯{z : p(z) − z = 0. in her work on gravitational lensing. where p(z) and q(z) are relatively prime polynomials in z. -0. where n := deg p > 1. Sarason and B.5 0.5 Theorem 3 ([KN 06]). and let n := deg r . Returning to Theorem 2 and the polynomial What happens if we extend the class of functions case. zero set of Re h(z) = 0 is in black and the zero Surprisingly. the number of images is author [KS 03] proved Conjecture 2 for all n > 1. ♯{z : p(z) − z = 0. Let’s summarize In the late 1990s D. If the number of lensed im- discussed in more detail in [SS 02. r (z) := b sider the domains in C in which h is respectively q(z) 670 Notices of the AMS Volume 55.

sense-preserving and sense-reversing.2 -1. Modulo the wishful thinking.8 0.2 -0.4 -0. Since zeros of h. h is n+ + n− . are finite. 12 (± 7 ± i). in the critical set. 12 (± 7±i) and n−1 = 3−1 = 2 sense-preserving sense-reversing zeros are counted with a minus zeros. These do. A moment’s thought indi- cates that h is sense-preserving (Jacobian(h) > 0) ≥ −n − (n − 1). R). positive sense) and L (in the negative sense). with the four point masses in black. There are 2n − pings in almost exactly the same way it applies 1 = 2× √ 3 − 1 = 5 sense-reversing zeros. n = 4 and ǫ = 1/100. ±1. We note that near infinity.4 0. namely. h is sense-reversing.4 0. with the small additional point mass at the origin shown in blue.8 -0. are considered in the last term. producing 5(n + 1) − 5 images. Consider consider the region Ω bounded by C(0.  sign (see [Du 04] and [SS 02]). the set of harmonic polynomials satisfying this We now employ some wishful thinking by sup.4 -0. the circle of radius R that the above argument is right on the money: centered at the origin. the argument principle can be where all combinations of plus and minus signs used to count the isolated zeros of harmonic map. The system is perturbed by removing a mass of ǫ/n from each of the n point masses and then adding a small point mass of mass ǫ at the origin.2 0. that We note that L is a lemniscate with at most n − 1 −n− = change in argument modulo 2π connected components. An example of Rhie’s construction of an (n + 1)-point gravitational lens.4 -1.4 0. all of the sense-reversing zeros done. Applying by the argument principle to Ω and noting that Ω contains only sense-reversing zeros of h. contains all of the zeros of h and does not intersect the critical set.2 -1. n+ ≤ n − 1 and the change in argument mains are separated by the critical set of h defined on L modulo 2π is −n+ ≥ −(n − 1). we may choose R sufficiently The following example inspires one to believe large such that C(0. The light source is located at the origin and is not shown in the graphs. namely.2 -0. Outside of L. We now Example.4 0.2 -0. Hence. to analytic mappings.8 -1.8 1. each point has mass 1/n. Provided that the boundary of a finitely connect. we see L := {z : Jacobian(h(z)) = 1 − |p′ (z)|2 = 0}. we are |h| → ∞ at ∞.8 -0. assumption is dense (see [KS 03]).4 0. R) (in the 1 h(z) = z − 2 (3z − z 3 ). Our function has 3n − 2 = 3 × 3 − 2 = ed region is sufficiently nice (piecewise smooth. 7 zeros: √ with no zeros on the boundary) and no zeros lie 0. In the graphs above. Here n = 3. Start with an n-point lens having 3n + 1 images. the only difference is that 0. ±1. therefore there are at most Let n+ denote the number of sense-preserving zeros of h and n− the number of sense-reversing (n − 1) + (2n − 1) = 3n − 2 zeros. Moreover. h behaves like z n . 1. Unfortunately posing that (i) no zeros of h lie on the critical set there are examples showing that the assumption and (ii) h is univalent inside each component of that h must be univalent in each component of June/July 2008 Notices of the AMS 671 .8 0.8 -0. inside each component of L and sense-reversing giving n− ≤ 2n − 1.2 Figure 5. The total number of zeros of (Jacobian(h) < 0) outside of L. Solutions to the lens equation (images of the source) are shown in red.8 1. L. R) on the critical set can be removed by showing that modulo 2π is −n. The graph on the left shows the unperturbed system. the argument increment of h along C(0.2 1. The graph on the right shows the perturbed system. hence for sufficiently Our assumption requiring that no zeros of h lie large R.

Let deg p = n. Divided into groups of at least n + 1 points and z0 a fixed point of F (F(z0 ) = z0 ). This is natural since above argument fails. Each fixed point of p(z) with |p′ (z)| < 1 attracts at least a group of n + 1 critical points of Help from dynamics Q. hence. at least one critical point of Q “runs” to a fixed point of p(z) with Theorem 5 implies that for every n > 1. the iteration of the map p(z). It essentially says that for all n and so is the upper bound 3n − 2 in the limiting behavior under the iteration of the Theorem 2. However. . For every n > 1 there exists attracting fixed point of Q. Number 6 . This picture serves as an illustration only. The proposition follows and completes our of the matter is the following proposition which sketch of the proof of Theorem 2. one cannot expect a galaxy to have a uniform mass distribution. Moreover. H. MD) and NASA. Q has n2 − 1 critical of an analytic function with itself. In other words.) L does not hold in general [Wil 94]. . Crofoot and Why is this so? First. it is not difficult to finish the ics studies iterations. the crux p(z). zn−1 with p′ (zj ) = 0 at least one critical point of Q (z is a critical point and p(zj ) = zj . |p′ (z)| < 1} ≤ n − 1. polynomials with precisely n − 1 attracting fixed The following lemma is elementary and its proof points for p. and by Fatou’s theorem a complex analytic polynomial p of degree n and (see [CG 93]. Proposition 1 is indeed sharp can be found in [KS 03]. (Photo credit: Kavan Ratnatunga (Johns Hopkins University. it can be salvaged p(z) covers the Riemann sphere n times and Q is by applying some elementary facts regarding it. Yet. in other words. With this lemma. page 59). Geyer [Ge 08] rests ♯{z : z − p(z) = 0. If F is analytic points. map Q occurs in clusters of at least n + 1 points The question of sharpness for rational functions exhibiting the same limiting behavior under the was settled by S. on fairly deep results in topological dynamics. First. Four images of a distant light source (larger blue dots) lensed by an elliptical galaxy (red). As the previous discussion shows. shows that n+ ≤ n − 1: Sharpness Results Proposition 1. it “attracts” mutually distinct points z1 . Rhie [Rh 03]: 672 Notices of the AMS Volume 55.2. if Q′ (z) = 0). . Theorem III. there are |p′ (z)| < 1 under iteration of the map Q. Baltimore. Lemma 1. Figure 6. some minimal vocabulary. Then The following theorem of L. Thus. z0 is called 2 −1 they “run” to at most nn+1 = n − 1 fixed points of an attracting fixed point of F if |F ′ (z0 )| < 1. we note that Q(z) := p(p(z)) D.2. every fixed point of p(z) with |p′ (z)| < 1 is an Theorem 5 ([Ge 08]). eration of polynomials (and iteration of rational functions) that come from complex dynamics. compositions proof of the proposition. Realistically. . Complex dynam. it was already conjectured by B. Sarason [Sa 99]: is an analytic polynomial of degree n2 . obtained by iterating p(z) twice.

In particu. an elliptic mass distribu- the perturbed system has (n + 1) point lenses and tion. sity. Wilmshurst’s conjecture (Conjecture 1 above) (If a spherical galaxy is filled with gas with such June/July 2008 Notices of the AMS 673 . Further- Corollary 2. FTA further: Rhie established Theorem 6 in the context of grav. Figure 6). multiples of π /n). Let p(z) be an analytic polynomial of itational lensing with n point masses. another image on each ray with a point mass (even the expected number of zeros tends to n [LW 08]. Petters. Li and A. the densities are assumed to be con- Extending the FTA further stant on ellipses homothetic to the initial ellipse. showing that there will be at most 5n − 5 masses has been reduced by ǫ/n.Theorem 6 ([Rh 03]). however. Wei [LW 08] proved a a gravitational lens with n equal point masses probabilistic variant of Wilmshurst’s conjecture. lar. Now perturb this system by adding a distribution sufficiently small mass at the origin. produce at most four visible images outside the Let us summarize the results concerning sharp. the the harmonic polynomial h(z) := z m − p(z)? bound she had previously conjectured in [Rh 01]. and has three images on each ray with trary mass distribution produce? argument that is an odd multiple of π /n). and the first author produces (3n + 1) + (2n − 1) = 5(n + 1) − 5 distinct [FKK 07] have shown that an elliptic uniform mass lensed images. Theorem 3 can be applied to the case of n “spher- suppose that a mass of ǫ > 0 has been added at the ically symmetric” mass distributions in the lens origin and the mass of each of the original n point plane.. If the conjectured bound In a nutshell. has two images on each ray with argument that is an even multiple Question. we also note that [KN 06] smaller than the bound of n2 given by Bézout’s includes a non-lensing example for this case. Figure 5 illustrates this for n = 4 distribution considered as a gravitational lens can 1 and ǫ = 100 . lens. Ch. there exists remains open for the cases 1 < m < n − 1. the origin and projecting it onto the lens plane. Surprisingly. One can then lensed images outside the support of the mass show that an additional 2n−1 images are produced [KN 06]. located at the vertices of a regular n-gon and a They assume that the coefficients are distribut- light source at the center of the n-gon and showed ed as Gaussian independent variables. the same upper bound holds for densities that are constant on ellipses confocal with the analytic polynomial p of degree n such that p(z)−z initial ellipse [FKK 07]. for example. How many lensed images can an arbi- of π /n. this is other lensing examples. A sharp upper estimate for the number of images is unknown in this case. degree n.. For example. there exists An even more realistic assumption would be a rational function r (z) with (finite) poles z1 . Fassnacht. p(z)−q(z) and let deg q = m and n = deg p. theorem when n > 3. other mass distribu- for ǫ sufficiently small (the perturbed system does tions are more natural. is it sharp? follows: Mao. Unfortunately. For every n > 1. . C. When Rhie analyzed the case where these point masses m = n−a where a > 0. For n ≥ 3. Consider. In particular. W. Let 1 < m < n − 1 be an integer. Question. In practice. For every n > 1 there exists a complex more. the techniques from complex dy. when m/n < α < 1 for all n.. Rhie’s construction for n ≥ 3 is as holds. One such density is called the isothermal density namics used in the proofs of Theorems 2 and 3 and is especially important. This configuration produces zeros) and the maximum of n2 given by Bézout’s 3n + 1 distinct images: one image at the origin. For most models Where Might One Go from Here? in astronomy. and two additional images on each ray with argument that is an odd multiple Gravitational microlensing by a mass of π /n. not have an image at the origin. The fol- a rational harmonic function with exactly 5n − 5 lowing question is a natural next step in extending distinct zeros. they show that the expected each have mass 1/n and are located at the vertices number of zeros asymptotically approaches n3/2 of a regular n-gon inscribed in a circle of radius when n is large. Hence. Keeton. Similarly. It arises by taking a do not seem to work as well for more general three-dimensional mass density that is inversely harmonic polynomials or rational functions with proportional to the square of the distance from the conjugate degree greater than 1. zn using a polynomial in place of the uniform den- such that the r (z) − z has precisely 5n − 5 zeros. theorem. Note that there are observations of four ness: lensed images (for example. Let h(z) = that 3n + 1 distinct lensed images can be attained. This falls between the minimum 1 1 √ a = (n−1) 2 − n / n (one of the point masses lies on number of possible zeros (there are at least n the positive x-axis). and Witt [MPW 97] studied Recently. has precisely 3n − 2 zeros. What she used a very elegant perturbation argument to is a sharp upper bound on the number of zeros of build an example with 5n − 5 distinct zeros. Wilmshurst’s conjecture predicts an She mentions that the case n = 2 is established by upper bound of 3n for the case m = 2.

Shapiro. B. [Rh 01] S. the electrostatic potential produced by n point [LW 08] W. number of such points may not exceed (n − 1)2 . Koecher. Methods Funct. Proceedings of the Conference situation has been studied extensively through on New Trends in Complex and Harmonic modeling (cf. Function Theory (1996). Sharper. [Du 04] P. 549–555. Classification of es that cannot be treated as coplanar. R. H. Basel. Math. Harmonic Mappings in the Plane. Gravita- tional lenses with images: I. of equilibrium where no electrostatic force is in Amer. G. Question. Witt. J. Microlensing by nonplanar lenses Mystery of point charges. [BHS 95] D. Anal. 244 (1981). A. Orde). Soc. in other words. Math. Carleson and T. Trends in ing that the number of images is even finite. arXiv:astro-ph/0305166. O. J. 11. J. 1494–1496. no. [KP 02] S. Fine and G. Soc. and T. L. and H. This Proc. bridge University Press (2004). Hengartner. Numbers. Keeton. A. [PLW 01] A. Maxwell conjectured the maximal [MPW 97] S. World Scientific. On the num- preliminary estimates [PLW 01]. History. Novikov and B. and the Schwarz function. Israel. preprint (2007). (1997). Astrophysical Journal 537 (2000). Petters. Ramat Gan. Wambsganss. theorem of algebra” in H. the right hand side of the lens equation son. Parks. fundamental theorem of algebra dissected. Lyzzaik. Proc. n-point gravitational lenses with (2005d:31001). Proceedings of the Eighth starting point of another exciting tale. 131 (2003). the problem of estimating the pre- functions. distributed masses. Proc. fact present. On the zero sets of harmonic polynomials. Piran. arXiv:astro-ph/0103463. Mathematical Newsletter of the Ramanujan [Re 91] R. Complex uate Texts in Mathematics (translated from Dynamics. Springer-Verlag. Soc. Mao. better ber of zeros of certain harmonic polynomials. Israel Math. 95 (2007). Amer. New York-Berlin. MR1383494. Can a gravitational quadruple lens Cambridge Tracts in Mathematics 156. Suez. second edition. Math. and J. Neumann. This mathematics. it will remain at constant temperature. produce 17 images?. were coplanar. MR 94h:30033. 134 (2006). Proc. what is the caustics. J. Cam. Boston (2002). MR2048384 [Rh 03] . as far as we ies. Bar-Ilan Univ. cise number of images is excruciatingly difficult. Undergraduate Texts in Mathematics. Świa̧tek. Gravitational lensing by elliptical galax- becomes a transcendental function. [KMW 00]) with. Khavin- However. MR2358495. London Math. Ch. Conf. Proc.) [FKK 07] C. On Crofoot. Math. 35–41. Petters has used Morse theory to make some [KS 03] D. 207–218. Birkhäuser. MR 2002m:83127.2684. The Implic- tal problem of Maxwell concerning the precise it Function Theorem. Comput. MR2196041 (2007h:26018). Peretz and J. Theory 4 (2004). Marcel Grossmann Meeting on General Rel- ativity (Jerusalem.. Soc. Duren. [KMW 00] Ch. density. S. 136 (2008). Hirzebruch. MR1476716. to appear in Anal- know. Proc. S. external Analysis. images [KMW 00]. ings of the Ashkelon Workshop on Complex [BL 04] D. Li and A. The fundamental Mathematical Society 7. Gamelin. Wei. B. arXiv:0708. Multiple gravitational imaging by H. Rosenberger. Gustafsson shear and there are models that produce up to 9 and A. New York actually seen. Math. Neukirch. Prestel. L. Grad- [CG 93] L. 67 of certain complex polynomials. Birkhäuser. Theory. Khavinson and G.. Amer.. M. On the expected number of charges vanishes. K. O. All along. The exact bound on the number of zeros [PS 98] R. in Proceed- (1995). Hermes. 443–472. Sarason’s conjecture for harmonic polynomi. 409–414. and number of points in space where the gradient of Applications. Fassnacht. Schmid. 674 Notices of the AMS Volume 55. Petters. Bshouty and A. ages? [KN 06] D. Norway. Keeton and D. 697–707. [GNS 07] A. and L1. Birkhäuser. Astrophys. though there do not seem to be [FR 97] B. F. On the num- ber of zeros of certain rational harmonic In this case. [Bur 07] R. Amer. 2007. Vasil’ev eds. Geyer. H. (1991). R. in a gravitational lens can be treated as if they [Ge 08] L. or without. Gabrielov. Khavinson and G. last question resembles the classical fundamen. “Chapter 4. Remmert. Number 6 . Voss. the German by H. Springer-Verlag. focused estimates are still far out of reach. When an n-lens consists of point mass. Ebbinghaus.-D. Krantz and H. Springer-Verlag Heidelberg (1993). 5(n − 1) images. MR 2081663. there is no proof at the present time show- ysis and mathematical physics. [Bu 81] W. 1997). W. A classical proof of the Singularity Theory and Gravitational Lensing. Rhie. Boston (2001). Properties of point mass lenses on a regu- lar polygon and the problem of maximum see [GNS 07] and the references therein for the number of images. MR2352567. 2 (2007). Burckel. we have been assuming that the masses Soc. The Fundamen- any observations where this number of images is tal Theorem of Algebra. edited by References T. D. 1077–1085. sharp upper bound on the number of possible im. MR 2003j:30015. 37–39. find the points zeros of random harmonic polynomials. 1997. Levine. (to appear). als. Mao and H. Sharp bounds for the valence of cer- tain harmonic polynomials. 203–208. Bshouty. Witt. Remmert. arXiv:astro-ph/9708111. Mainzer. Singapore (1999). Burke.

Gravitational Lenses. [SEF 92] P.K. 894–908. [Wit 90] H. Proc. 1999. [St 97] N. Math. New York-Berlin-Heidelberg (1992). arXiv:astro- ph/9703103. D. written communication to the first author. [Wil 94] A. 236 (1990).[Sa 99] D. June/July 2008 Notices of the AMS 675 . 2077–2081. Astron. Astrophys. S. Sheil-Small. Helvetica Physica Acta 70 (1997). J. Fal- co.org/ Articles/Volume1/1998-12wamb/. MR 2004b:30001. E. [Wa 98] J. thesis. (1994). Complex Polynomials. Gravitational lensing in astronomy. Phil. J. Oct. MR 98h:30029. Wilmshurst. 311–322. 126 (1998). Relativity 1 (1998) [Online Article]: cited on January 26. and E. Astronomy and Astrophysics Library. Springer-Verlag. U. Cambridge University Press (2002). Schneider. Cam- bridge Studies in Advanced Mathematics 75. Investigation of high amplification events in light curves of gravitational- ly lensed quasars. Sarason. 2000.livingreviews. Ehlers. [SS 02] T. Amer. The valence of harmonic polyno- mials. University of York. 2004. Complex formulation of lensing theory and applications. Living Rev. Straumann. Feb. Complex harmonic map- pings and the valence of harmonic polynomi- als. Witt. http://www. Soc. Wambsganss. [Wil 98] .