The Links Between Smale’s Mean Value

Conjecture and Complex Dynamics

Hayley Miles-Leighton, UC San Diego
Abstract
In this paper we study the links between Smale’s Mean Value Conjec-
ture (SMVC) and the convergence of critical points under iteration. We
begin by introducing SMVC and discussing what progress has been made
towards proving this conjecture. From there we give a brief introduction
to a few concepts in Complex Dynamics and Complex Analysis including
conjugacy, conjugations by 1/z, and orbits. We then show that conjugacy
conserves SMVC. At this point, we go through the quadratic case to show
the patterns that we are looking for as well as to show a basic structure of
how this problem is being looked at. From there we look at the cubic case,
going through SMVC for the cubic polynomial. We then introduce the
Petal Theorem and show how it is used for the cubic case specifically. We
conclude with the connections between the SMVC in the cubic case and
the convergence of the same critical points explained and discuss future
work.

1 Introduction and Background

Complex Dynamics is the study of complex functions under iteration. We are
specifically interested in where certain points in the complex plane converge to
the origin and to infinity. It is this idea that we kept in mind when we looked
at Smale’s Mean Value Conjecture (SMVC) which states [7]:
Pn
Conjecture 1.1. Let f (z) = z + i=2 ai z i be a complex valued polynomial of
degree n ≥ 2 for which f (0) = 0 and f 0 (0) = 1. Then, there exists a critical
point c of this polynomial that satisfies the following statement:
f (c)
| |≤1
c
A variation of this conjecture has been announced for all polynomials of
degree 10 or less by Sendov and Marinov [6]. This variation replaces 1 with an
even better bound of (d − 1)/d. However, their paper has no proofs to show this
result. Several other papers use another variation on this conjecture. Namely,
this following proposition:
Proposition 1.2. Let p be a polynomial of degree N ≥ 2 over C, and supposed
that x ∈ C is not a critical point of p. Then there exists a critical point ζ of p
such that
p(ζ) − p(x)
| | ≤ 4|p0 (x)|. (1)
ζ −x

40
 In a paper written by E. Crane [2], he shows that there can be a better
bound than the bound of 4 that was proven by S. Smale in 1981 [7] in the
proposition given above. He proves that for all polynomials of degree 8 and
higher, K(d) < 4 − 2.263
√
d
where d is the degree of the polynomial and K(d) =
p(ζ)−p(x)
sup{S(p, x) : deg(p) = d, p0 (x) 6= 0} (S(p, x) = min(| (ζ−x)p 0
0 (x) | : p (ζ) = 0)).
d
−(d+1)
For 2 ≤ d ≤ 7, Schmeisser [5] proved that we can have K(d) ≤ 2 d(d−1) which
is a better bound than what Crane would have for those values of d. In 2003,
T.W. Ng proved that this original bound could be further reduced to 2 for all
nonlinear odd polynomials with a nonzero linear term [3].
Although there have been many papers published on the possibility of this
conjecture (and its variations) being true, none have been able to prove it com-
pletely. The goal of this paper is not to prove the conjecture, however. This
paper will look into possible connections between SMVC and Complex Dynam-
ics. SMVC says nothing about which critical point or points satisfy the bound.
It only states that there is one. On Pthe other hand, complex dynamics says
n
that given a polynomial f (z) = z + i=2 ai z 1 , there must exist a critical point
tending to the origin under iteration (this fact is presented formally in Propo-
sition 1.8). We hope to combine these two thoughts to see if there is a relation
between them. Specifically, we will prove the following main theorem:
Theorem 1.3. For any cubic of the form p(z) = z + a2 z 2 + a3 z 3 , there is a
critical point satisfying the SMVC bound and converging to the origin.
To begin with, we will be wanting to show that this works in general. It is
generally nice to be able to work with one form instead of many so to do this
we will use the idea of conjugacy:
Definition 1.4. If there exists a function φ(z) that is either a linear fractional
transformation (such as 1/z) or a so called afine map (such as Az + B), then
two functions are conjugate if φ ◦ f = g ◦ φ.
Below is an example of this concept:
Example 1.5. Let f (z) = z 2 − z and φ(z) = αz + β. We want to find out if this
form is conjugate to g(w) = w2 + c where c is some constant. To find out if
these two are conjugate, we will follow this chart:
z 2 −z
Z −→ C
φ↓ ↓φ
w2 +c
W −→ C
Following the chart, we see that we get α(z 2 − z) + β = (αz + β)2 + c ⇒
αz 2 − αz + β = α2 z 2 + 2αβz + β 2 + c. For two polynomials to be equal to one
another, all of the coefficients must equal one another. With this in mind, we
find that α = α2 , −α = 2αβ, β = β 2 + c ⇒ α = 1, β = −1/2, c = −3/4. Since
there exists an α and β that makes this true, we know that these two polynomials
are conjugate for c = −3/4. This means that if we let φ(z) = αz + β, we have
that φ ◦ f = g ◦ φ.

41
 From here, we need to show that we can use the conjugated form to look at
SMVC. Below is a lemma that shows this for φ(z) = αz.
Lemma 1.6. If p(z) = z + h.o.t. satisfies SMVC and if q = A ◦ p ◦ A−1 where
A(z) = αz (α 6= 0), then q also satisfies SMVC.
Proof. Suppose that c is a critical point of p(z) and that it satisfies |p(c)/c| ≤ 1.
Since they are conjugate via multiplication by α, the point x = αc is a critical
point of q. Since q = A ◦ p ◦ A−1 we can see that |q(x)/x| = |q(αc)/(αc)| =
| α(p((αc)/α))
αc | = |p(c)/c| ≤ 1. Hence, if p(z) satisfies SMVC and if q = A ◦ p ◦ A−1
where A(z) = αz (α 6= 0), then q(z) will also satisfy SMVC.
At this point, I need to introduce the concept of an orbit and a general
proposition that uses this concept as well as talk about the conjugacy of 1/z
which will be used in the cubic case.
Definition 1.7. Suppose X is any set and f : X 7→ X is any function. Let
x0 ∈ X. Then, the orbit of x0 is defined as x0 , f (x0 ), f (f (x0 )), ...
This concept is used in the following proposition:
Proposition 1.8. There exists a critical point ci such that the orbit of ci con-
verges to the origin.
From here, I will introduce the conjugacy of 1/z. In general, 1/z maps
”circles” to ”circles.” When I say ”circles” I mean that they can fit into this
general form: A(x2 + y 2 ) + Bx + Cy + D = 0. At this point, we want to see
what x and y are equal to in terms of the new coordinates u and v. After
u −v
some computations, we see that x = u2 +v 2 and y = u2 +v 2 . After we use
these to substitute into the general form, we find that in the new plane we
get D(u2 + v 2 ) + Bu − Cv + A = 0. So, lines become circles, circles become
circles, and circles can sometimes become lines. If we are mapping a circle that
is around the origin, then everything inside goes outside and vice versa. This is
because the origin goes to ∞ and ∞ goes to 0. This is the method that I will
use in the future.
Now that we have a basic background in Complex Analysis as well as Com-
plex Dynamics, we can look into the quadratic case as an example.

2 The Quadratic Case

To begin with, we will go through the quadratic case. The following Theorem
describes what we are planning to prove in this section:
Theorem 2.1. For each quadratic polynomial f (z) = z + a2 z 2 , there is exactly
one critical point. It satisfies the bound in SMVC and it converges to the origin
Proof. The form that we use is a very general form of the quadratic polynomials.
However, we want to use a simpler version by conjugating this polynomial by
φ(z) = αz using the process explained in the previous section.

42
 z+a2 z 2
Z −→ C∞
φ(z) ↓ ↓ φ(z)
z−z 2
Z0 −→ C∞
This means that we need α(z + a2 z ) = αz − α2 z 2 ⇒ z + a2 z 2 = z − αz 2 ⇒
2

α = −a2 . So, we can now do all of our calculations using f (z) = z − z 2 . First,
we need to find its critical point: f 0 (z) = 1 − 2z = 0 ⇒ z = 1/2. Referring back
to SMVC, we see that | f (1/2) 1 1
1/2 | = | 4 · 2| = | 2 | ≤ 1. Hence, SMVC is satisfied by
all quadratic polynomials of this form.
Now, we want to look at where this critical point converges to the origin.
To get an idea of this, we will refer back to Proposition 1.8. In this case,
i = 1. So, since there is only one critical point, this point must converge to the
origin. Hence, this one critical point both satisfies SMVC and converges to the
origin.
At this point, we want to see if this pattern continues for the cubic case.

3 The Cubic Case

In this section, we plan to prove the main theorem (Theorem 1.3). We will do
this both indirectly and directly. The indirect approach will use Proposition 1.8
and handles ”areas” or ”regions.” The direct approach will use Beardon’s Petal
Theorem which will be introduced later on. To begin with, we will be looking
at the following general form of cubic polynomials that satisfy the conditions in
SMVC:

p(z) = z + az 2 + bz 3 (2)
However, we would like this in a more useful form. To do this, we will
conjugate this polynomial by φ(z) = αz:

z+az 2 +bz 3
Z −→ C
φ↓ ↓φ
z− 21 (c+ 1c )z 2 + 13 z 3
Z −→ C
Following the same process described in example 1.5, we see that we have
2
z + az 2 + bz 3 = z − α2 (c + 1c )z 2 + α3 z 3 . For two polynomials to be equal, their
2
coefficients must be equal: a = − α2 (c + 1/c) and b = α3 . Hence, α = − c+1/c 2a
.
Since there exists an α we know that these two polynomials are conjugate and
we can continue with the following cubic:
1 1 1
fc (z) = z − (c + )z 2 + z 3 (3)
2 c 3
After taking the derivative, it is easy to see the two critical points: c1 = c
and c2 = 1/c. These two critical points are nice to work with since they create

43
the same map. In other words, fc (z) = f1/c (z). This means that by looking at
one of the critical points, we have a way of looking at the other. For example, if
we can show for one critical point that fcon (c1 (c)) → ∞ ⇔ f1/c
on
(c1 (c)) → ∞ ⇔
on
f1/c (c2 (1/c)) → ∞ then we see that in an area that is 1/c we have the same
thing for the other critical point. So, we will look at what happens to c1 to get
an idea what happens to c2 as well. At this point, we need to look at where c1
and c2 satisfy SMVC:
1 1 1 3c − c3 fc (c) 3 − c2
fc (c) = c − (c + )c2 + c3 = ⇒| |=| | ≤ 1 ⇒ |3 − c2 | ≤ 6
2 c 3 6 c 6
1 1 1 1 1 1 3c2 − 1 fc ( 1c ) 3c2 − 1 1
fc ( ) = − (c+ ) 2 + 3 = 3
⇒ | 1 |=| 2
| ≤ 1 ⇒ |3− 2 | ≤ 6
c c 2 c c 3c 6c c
6c c
I claim that with these two graphs combined, we see that SMVC is satisfied
for any c throughout the parameter plane. Below is a figure that shows this
fact.
3

-1

-2

-3
.. -3 -2 -1 0 1 2 3

Figure 1: The smallest area is where only c1 satisfies SMVC, the darkest area
shows where both critical points satisfy SMVC, and the area outside the larger
curve shows where only c2 satisfies SMVC. Put together, they cover the entire
complex plane.

At this point, we want to look at areas of convergence to infinity. For c1 = c,

we will prove the following lemma to show where it converges to infinity:
√
Lemma 3.1. Let δ = 43 30. If |c| > δ and |z| > |c|3 /6 then |fc (z)| > |z| and,
under iteration of f , z converges to infinity.
Proof. We know that |fc (z)| = |z||1− 21 (c+1/c)z + 13 z 2 | and that we want |z||1−
1 1 2 1 1 2
2 (c+1/c)z+ 3 z | > |z|. So, proving that |1− 2 (c+1/c)z+ 3 z | > 1 will imply the
previous inequality. However, if we prove that |− 2 (c+1/c)z + 13 z 2 | > 2 then the
1

previous inequality will be true. We can rewrite this as |z||− 21 (c+1/c)+ 13 z| > 2,
which can be implied by the following: |− 21 (c+1/c)+ 13 z| > 2/|z| > 2/(|c|3 /6) >

44
0.17. Using the triangle inequality we see that | − 12 (c + 1/c) + 31 z| ≥ || 12 (c +
3 3
1/c)| − 13 |z|| = | 13 |z| − | 12 (c + 1/c)|| > | |c| 1 |c| 1
18 − | 2 (c + 1/c)|| ≥ | 18 − 2 (|c| + |1/c|)|.
This function is monotonically increasing as a function of |c|. So, we can re-write
3
it as |c| 1
18 − 2 (|c| + |1/c|) > 1.67. Since 1.67 > 0.17 we see that all of the previous
inequalities hold. Hence, under iteration of f , z converges to infinity.
√
This means that for all values of |c| greater than 3 430 , c1 will converge to
infinity. Based on the earlier explanation,
√ we know that c2 will also converge
to infinity in the circle of |c| > 43 30 mapped by 1/c in the parameter plane.
This will give us the area where c2 converges to infinity. This circle will be
|c| < 3√430 . Now that we have two areas where c1 and c2 converge to infinity,
we want to refer back to Proposition 1.8. In this case, i = 2. This means that,
for example, in the area where c1 converges to infinity, we know that c2 must
converge to the origin under iteration. The same happens in the area where c2
converges to infinity. From here we can combine these circles with Figure 1 to
create five distinct areas:

..

Figure 2: There are five different points representing the five different areas
created by four graphs.

We are interested in what happens within each of those areas. Below are
five charts that give us an idea of what is going on:

Region a c1 c2 Region b c1 c2 Region c c1 c2

SMVC y n SMVC y n SMVC y y
→0 y n →0 ? ? →0 ? ?
Region d c1 c2 Region e c1 c2
SMVC n y SMVC n y
→0 ? ? →0 n y

As you can see, only regions a and e have full charts. We can see that each
of those have a column of yeses. This is the pattern that we are looking for.

45
This means that the chart for region c does satisfy this pattern automatically
based on Proposition 1.8. Since in this area both c1 and c2 satisfy SMVC
and we know that one of those critical points will converge to the origin under
iteration, we know that we will have a column of yeses. Hence, the two regions
that we are interested in are b and d. However, since our critical points have
the same map, we can study the area that contains point d and be able to have
an understanding of the area that contains point b. To be able to get a good
understanding of what is happening in region d, we will need to look at the
annulus 1.75 ≤ |c| ≤ 4.11 which encompasses the entire region of d. In this
annulus, we will be looking at the following theorem:

..

Figure 3: This shows what is meant in the following theorem by ”petal.” In our
specific case, we will only have one petal.

Theorem 3.2. (The Petal Theorem) Suppose that the analytic map f has a
Taylor Expansion

f (z) = z − z p+1 + O(z 2p+1 ) (4)

at the origin. Then for all sufficiently small t:
1. f maps each petal Πk (t) into itself;
2. f n (z) → 0 uniformly on each petal as n → ∞;

3. argf n (z) → 2kπ/p locally uniformly on Πk asn → ∞;

4. |f (z)| < |z| on a neighborhood of the axis of each petal;
5. f : Πk (t) → Πk (t) is conjugate to a translation. [1]
Since our fc (z) is not in the form that is needed to be able to use this
theorem, we will need to conjugate it to create the form that is needed:

46
 z− 12 (c+ 1c )z 2 + 13 z 3
Z −→ C
φ↓ ↓φ
z−z 2 +c0 z 3
Z −→ C
where φ(z) = αz. This shows that we need to find what α is as well as c0 :

1 1 1 1 1 1
α(z− (c+ )z 2 + z 3 ) = αz−α2 z 2 +c0 α3 z 3 ⇒ z− (c+ )z 2 + z 3 = z−αz 2 +c00 α2 z 3
2 c 3 2 c 3
So, for two polynomials to be equal, their coefficients must be equal. This
means that α = 21 (c + 1c ) and that 13 = c0 ( 12 (c + 1c ))2 . So, we now know that
α = 12 (c + 1c ) and c0 = 31 ( c22c 2
+1 ) . Since we are now working in a new plane,
which we will call the u-plane, we need to convert our annulus into the new
coordinates. The following lemma shows this:
Lemma 3.3. Let R = {c : 1.75 ≤ |c| ≤ 0.96} and let R0 = {c0 : 0.07 < |c0 | <
0.96}. Then c ∈ R ⇒ c0 ∈ R0 .

Proof. To show this, we need to reduce c0 to a more useful form: |c0 | = | 13 ( c2c2 2
+1 ) | =
4 1 2
3 | c+1/c | . To begin with, we want to look at the denominator of the main frac-
tion. Using the triangle inequality, we see that |c + 1c | ≥ ||c| − |c| 1
|. To be
1
able to get a better bound on this, we need to show that |c| − |c| is monotoni-
cally increasing as a function of |c|. To do this, we want to make sure that the
derivative is always positive for values between 1.75 and 4.11: 1 + |c|12 is always
positive. This means that the smallest value will happen at the bound. So, we
1
have that ||c| − |c| ≥ | 47 − 47 | = | 33 1 28
28 |. This means that |c+1/c| ≤ | 33 |. Combining
this with the rest of the function, we get an upper bound: |c0 | ≤ 43 | 33 28 2
| < 0.96.
For the lower bound, we will follow a similar process. By the triangle in-
equality, we see that |c + 1c | ≤ |c| + |c| 1
. Before we can continue, we need to
show that this function is monotonically increasing as a function of |c|. The
derivative is 1 − |c|12 which is always positive between 1.75 and 4.11. This
means that the greatest value will happen at the boundary. So, we have that
1 1
|c| + |c| ≤ 4.11 + 4.11 < 4.35. Putting this into the rest of the function, we find
that the lower bound is |c0 | > 0.07.
Now that we have bounds in our new coordinate system, we want to go back
to the Petal Theorem. We will show that this theorem works for our case which
means that p = 1 and k = 0 (assuming that we are ignoring the points c = ±i).
To do this, we want to prove the following lemma:
Lemma 3.4. Given z in the dynamical plane of fc , let z 0 = (1/2)(c + 1/c)z be
the corresponding point in the dynamical plane, and put w = 1/z 0 . If c ∈ R and
Re(w) > 4.36 then under iteration of fc , the point z converges to the origin.

47
Proof. To begin with, we want to conjugate p(z) = z − z 2 + c0 z 3 to the w-plane
by the conjugation σ : z 7→ 1/z. First, we will look at what 1/p(z) is. Using
polynomial division, we find:
1 1
= + 1 + Az + ν(z), (5)
z − z 2 + c0 z 3 z
0 2 0 0 3
)z −c (1−c )z
where A = 1 − c0 and |ν(z)| = | (1−2c 1−z+c 0 z2 | ≤ B|z|2 , B > 0.
At this point we need to find bounds on both B and A by proving this
corollary:
Corollary 3.5. If c0 ∈ R0 and |z| < 0.23 then |A| < 1.96 and |ν(z)| ≤ B|z|2
where B = 4.66.

Proof. First, we will look at finding a bound on A. We know that A = 1 − c0 so

|A| = |1 − c0 |. Using the triangle inequality, we see that |A| = |1 − c0 | ≤ 1 + |c0 |.
Since c0 ∈ R0 , we know that |c0 | < 0.96. So, 1 + |c0 | < 1.96 and hence |A| < 1.96.
Now we want to find B and so we need a bound on ν(z). By definition, |z| <
0.23. This choice will be explained later. First, we will look at the denominator
of ν(z). By the triangle inequality, we see that |1 − z + c0 z 2 | ≥ ||1 − z| − |c0 z 2 || =
|1 − z| − |c0 ||z|2 ≥ 1 − |z| − |c0 ||z|2 ≥ 1 − 0.23 − (0.96)(0.23)2 ≈ 0.719.
Now that we have a bound on the denominator of ν(z), we can continue to
get a bound on the full function which will become our B. To do this, let’s
first look at the numerator separately: |(1 − 2c0 )z 2 − c0 (1 − c0 )z 3 | = |z|2 · |(1 −
2c0 ) − c0 (1 − c0 )z|. Since the bound on |ν(z)| is B|z|2 , we can disregard the |z|2
term. Using the triangle inequality and the fact that |z| < 0.23 we see that
|(1 − 2c0 ) − c0 (1 − c0 )z| ≤ |1 − 2c0 | + |c0 (1 − c0 )||z| < |1 − 2c0 | + 0.23|c0 (1 − c0 )|.
Earlier we were able to find a bound on the denominator, so if we let B =
|1−2c0 |+0.23|c0 (1−c0 )|
0.719 , we can guarantee that |ν(z)| ≤ B|z|2 .
|1−2c0 |+0.23|c0 (1−c0 )|
At this point,we want to find a bound on B. Since B = 0.719 ,
0 0 0
1+2|c0 |+0.23|c0 |(1−|c0 |)
we know by the triangle inequality that |1−2c |+0.23|c
0.719
(1−c )|
≤ 0.719 .
0 0 0 1+2|c0 |+0.23|c0 |(1−|c0 |)
Using the fact that |c | < 0.96 since c ∈ R , we have that 0.719 <
4.66.
At this point we want to complete the proof by showing the following corol-
lary:
Corollary 3.6. Given c in R, the critical point c2 = 1/c converges to the origin.
Proof. I claim that we can show that c2 converges by using the second iterate.
If we can show that the second iterate converges to the origin, then we can
2
conclude that c2 converges to the origin as well. Recall that fc (c2 ) = 3c6c−1
3 .

Iterating one more time, we find that

6c(6c3 )2 (3c2 − 1) − 3c2 (6c3 )(3c2 − 1)2 − 3(6c3 )(3c2 − 1)2 + 2c(3c2 − 1)3
fc (fc (c2 )) = .
6c(6c3 )3

48
We need to convert this into the new coordinate system. To do this, multiply the
2
above function by α = c 2c +1
. After working through the computation, the result
5 5 1 1 1 3
is 48c2 − 72c4 + 72c6 − 1296c8 − 1296c 10 + 16 . From here, we want an upper bound
5 5 1 1 1
on 48c2 − 72c4 + 72c6 − 1296c8 − 1296c10 to create a circle whose radius is that upper
3
bound, centered at 16 . To get an upper bound on this, we will use the triangle
5 5 1 1 1 5 5 1 1
inequality: | 48c2 − 72c4 + 72c 6 − 1296c8 − 1296c10 | ≤ | 48c2 |+| 72c4 |+| 72c6 |+| 1296c8 |+
1 5 5 1 1 1
| 1296c10 | ≤ | 48(7/4)2 | + | 72(7/4)4 | + | 72(7/4)6 | + | 1296(7/4)8 | + | 1296(7/4)10 | ≈ 0.042.

6
0.2

0.1
2

0.0 0

-2

-0.1

-4

-0.2 -6

.. -0.2 -0.1 0.0 0.1 0.2 .. -6 -4 -2 0 2 4 6

Figure 4: The graph on the left shows the areas where the denominator of ν(z)
does not vanish and the values of the second image of c2 . The graph on the
right are these same graphs but mapped by 1/z. The half-plane in the right
hand graph is the half-plane that we want to show is forward invariant.

Now that we have most of our information, we can continue with the mapping
by finding g(z) in the w-plane. To do this, we can take what we have for the
1/p(z) function and replace z with w = 1/z. This creates g(w) = w+1+ A w +θ(w)
−1 −1 2 B
where |θ(w)| = |ν(σ (w))| ≤ B|σ (w)| ≤ |w|2 and B and A are the same as
earlier. Figure 4 shows where all the values of the second iterate of c2 go under
the mapping of 1/z as well as where the circle of |z| < 0.23 goes under the same
mapping. There is also a half-plane at Re(w) > 4.36. We want to show that
this half-plane is forward invariant. To do this, we need to make sure that when
we move the real values of w by w 7→ w + 1 + A w + θ(w), it moves forward. To
do this, we need the most negative vales that A and θ(w) can take.
Recall that |A| < 1.96. This means that the most negative A can be is -1.96.
Hence, the most negative A w can be is about -0.45 since w = 4.36. Now, we
B B
know that |θ(w)| ≤ |w| 2 which means that −|θ(w)| ≥ − |w| 2 . Earlier, we also
1
showed that B < 4.66 hence −B > −4.66. This means that since |w| = 0.23,
the most negative θ(w) can be is -0.25. Combining all of these together, we
see that 4.36 7→ 4.36 + 1 − 0.45 − 0.25 = 4.66. This leads us to the following
proposition:
Proposition 3.7. Under iteration of w 7→ w + 1 + A w + θ(w), every point w in
the right half-plane Re(w) > 4.36 converges to infinity: the real parts of points
in its orbit converge to +∞.

49
Proof. We will show this using induction. The first step has already been shown
above. If we let w1 = 4.36, we will start with the base case n = 1. Following
the same process as above, we found that w1 7→ 4.66 = w2 . Hence, w1 < w2
and so the base case is true.
Now, we want to assume that n holds. This means that wn−1 7→ wn and
hence wn−1 < wn . So, we need to look at where wn gets mapped to: wn 7→
wn +1+A/wn +θ(wn ). Since we know that wn > wn−1 , we can say that 1/wn <
1/wn−1 . So, we have that A/wn < A/wn−1 which means that the most negative
it can be is −A/wn−1 . Now, we need to look at a bound on θ(wn ). By definition,
|θ(wn )| ≤ B/|wn |2 < B/|wn−1 |2 which means that −|θ(wn )| > −B/|wn−1 |2 .
So, with these replacements, we see that wn 7→ wn + 1 + A/wn + θ(wn ) >
wn + 1 − A/wn−1 − B/|wn−1 |2 . However, since wn > wn−1 , we know that
wn−1 + 1 − A/wn−1 − B/|wn−1 |2 < wn + 1 − A/wn − B/|wn−1 |2 . We also know
that wn−1 +1−A/wn−1 −B/|wn−1 |2 = wn . So, if we let wn+1 = wn +1+A/wn +
θ(wn ) > wn +1−A/wn−1 −B/|wn−1 |2 > wn−1 +1−A/wn−1 −B/|wn−1 |2 = wn ,
we have that wn < wn+1 which completes the proof.
What this proposition has told us is that the real values in the half-plane in
the w-plane converges to infinity. Since this line is mapped backwards by 1/z,
when this line converges to infinity, we have that the resulting circle converges
to the origin. To see this, consider the half-planes given by x > 4.36 and
x > 4.66. Using the method explained in section 1, we can see that these
two lines map to these two circles, respectively: (x − 0.115)2 + y 2 < 0.013
and (x − 0.107)2 + y 2 < 0.012. When these two circles are graphed together,
we see that the second is mapped inside the first but both are tangent to the
origin. Hence, as the real value of the half-plane converges to infinity, these
circles converge to the origin. This means that since this area that represents
all values of the second iterate of c2 within our annulus is contained within the
half-plane, we can see that the second iterate c2 must converge to the origin.
Hence, c2 must converge to the origin as well and this completes our proof of
Corollary 3.6.
Combining the proofs of Corollaries 3.5 and 3.6 as well as the proof for
Proposition 3.7, we can see that Lemma 3.4 has been proven.

Since we were able to show that c2 converges to the origin within our annulus,
we can say something about c1 in the other region under consideration. Because
fc (z) = f1/c (z), we can say that fc (c2 (c)) ⇔ f1/c (c2 (c)) ⇔ f1/c (c1 (1/c)). Hence,
within the other area (since it is where c2 is mapped by 1/c) we can say that c1
converges to the origin within region b. With these new pieces of information,
we can look back to the charts for regions b and d:

Region b c1 c2 Region d c1 c2
SMVC y n SMVC n y
→0 y ? →0 ? y

50
 As you can see, we can now say that c1 converges to the origin in region b
and that c2 converges to the origin in region d which completes the pattern that
we were looking for. Hence, we know that for any cubic, there is a critical point
satisfying the SMVC bound and converging to the origin.

4 Concluding Remarks
This paper has shown that Complex Dynamics can help give an idea of which
critical points satisfy SMVC for quadratic and cubic polynomials. It began with
a brief introduction the problem as well as some basic background information
in Complex Dynamics and Complex Analysis. From there, it described the
quadratic case as an example for the cubic case. In the cubic case, we showed
that the intuition that convergence of a critical point and SMVC do have a
relation by using a specific instance of the Petal Theorem. From here, if we
can show this intuition is true for higher order polynomials, we may have a
different possibility for proving that this form of SMVC is true. There is also
the possibility of looking into the following stronger form of SMVC:
Pn
Conjecture 4.1. Let f (z) = z + i=2 ai z i of degree n ≥ 2 be a complex valued
polynomial for which f (0) = 0 and f 0 (0) = 1. Then, there exists a critical point
c of f that satisfies: | f (c) d−1
c | ≤ d where d is the degree of the polynomial.

This has been shown to be true for polynomials of degree 2, 3, and 4 by

algebraic methods in a paper by Q. Rahman and G. Schmeisser [4]. However,
the link between this version of SMVC and the convergence of the orbits of the
critical points does not seem to have been looked into.

5 Acknowledgments
I would like to thank Professor Kevin Pilgrim for is support and guidance these
past 8 weeks. I would also like to thank him for organizing this great opportunity
for all of us.

References
[1] Alan Beardon. Iteration of Rational Functions. Number 132 in Graduate
Texts in Mathematics. Springer-Verlad, New York, 1991.
[2] Edward Crane. A bound for smale’s mean value conjecture for complex
polynomials. Bulletin of the London Mathematical Society, 39(5):781–791,
2007.
[3] T.W. Ng. Smale’s mean value conjecture for odd polynomials. Journal of
the Australian Mathematical Society, 75(3):409–411, 2003.

51