mathematics

Article
Fractal Newton Methods
Ali Akgül 1,2,3, * and David Grow 4

1 Department of Computer Science and Mathematics, Lebanese American University, Beirut 11022801, Lebanon
2 Department of Mathematics, Art and Science Faculty, Siirt University, 56100 Siirt, Turkey
3 Mathematics Research Center, Department of Mathematics, Near East University, Near East Boulevard,
99138 Nicosia, Turkey
4 Department of Mathematics and Statistics, Missouri University of Science and Technology,
Rolla, MO 65409, USA
* Correspondence: [email protected]

Abstract: We introduce fractal Newton methods for solving f (x) = 0 that generalize and improve the
classical Newton method. We compare the theoretical efficacy of the classical and fractal Newton
methods and illustrate the theory with examples.

Keywords: fractal derivative; fractal Newton methods

MSC: 49M15

1. Introduction
Newton’s method is one of the most universally known numerical algorithms. It was
first introduced in around 1669 by Isaac Newton and then modified in 1690 by Joseph
Raphson to obtain the real roots of polynomials. In 1740, Thomas Simpson extended the
method to solve nonlinear equations, and in 1879 Arthur Cayley used the method for
obtaining complex roots of polynomials [1]. Depending on the starting point, Newton’s
method may be convergent or divergent, has local quadratic convergence, and is undefined
at critical points. Many modifications and enhancements of Newton’s method exist in
the literature [2–4]. In particular, Newton’s method has recently been improved by using
Citation: Akgül, A.; Grow, D. Fractal fractional derivatives instead of the classical derivative [5], and Wang and Tao [6] introduced
Newton Methods. Mathematics 2023, a self-accelerating variable with memory into Newton’s method. In this work, we replace
11, 2277. https://doi.org/ the classical derivative in Newton’s method with an α-fractal derivative where 0 < α ≤ 1
10.3390/math11102277 and then modify the algorithm appropriately.
Fractal geometry and fractal and fractional calculus have become increasingly impor-
Academic Editors: Krzysztof
tant in mathematics when one is confronted with certain types of nonlinear problems and
Gdawiec and Agnieszka Lisowska
the modeling of non-smooth phenomena. For example, fractal theory is the theoretical
Received: 27 March 2023 basis for fractal space-time [7], and in this setting fractal calculus can deal effectively with
Revised: 9 May 2023 kinetics when fractal time replaces continuous time [8,9]. Candelario et al. [10] presented an
Accepted: 11 May 2023 optimal and low-computational-cost fractional Newton-type method for solving nonlinear
Published: 13 May 2023 equations. Golmankhaneh [11] discussed fractal calculus and its applications in detail.
Blaszczyk et al. [12] investigated the approximation and application of the Riesz–Caputo
fractional derivative of variable order with fixed memory.
Our manuscript is organized as follows: A deficiency of the classical Newton method
Copyright: © 2023 by the authors.
is illustrated in Section 2, and fractal Newton methods are introduced to remedy it. The
Licensee MDPI, Basel, Switzerland.
main definitions and theorems of fractal calculus are reviewed in Section 3. We prove our
This article is an open access article
distributed under the terms and
main convergence results for the fractal Newton methods in Section 4. We compare the
conditions of the Creative Commons
convergence properties of the fractal and classical Newton method sequences in Section 5,
Attribution (CC BY) license (https:// and in Section 6 we illustrate their similarities and differences with examples. We state our
creativecommons.org/licenses/by/ conclusions in the final Section 7.
4.0/).

Mathematics 2023, 11, 2277. https://doi.org/10.3390/math11102277 https://www.mdpi.com/journal/mathematics

Mathematics 2023, 11, 2277 2 of 13

2. Instance of Superiority of Fractal Newton Methods

Let [a, b] be a closed, bounded interval in [0, ∞). Let f be a continuous, real function
on [a, b] such that f (a) f (b) ≤ 0, so there exists at least one solution t = ξ to f (t) = 0 in [a, b].
The classical Newton method sequence {tn }, which is intended to successively approximate
such a solution, is given by selecting t1 ∈ (a, b) and then defining

f (tn )
tn+1 = tn − (n = 1, 2, 3, . . . ). (1)
f ′ (tn )
In order for this method to be robust, it is natural to assume that f is differentiable
on (a, b) and f ′ (t) ≠ 0 for all t ∈ (a, b). The intermediate value property for differentiable,
real functions then implies that either f ′ > 0 on (a, b) or f ′ < 0 on (a, b). In either case, f
is strictly monotone on [a, b], and there is precisely one solution t = ξ to f (t) = 0 in [a, b].
Therefore, without a loss of generality, we may assume f ′ > 0 on (a, b), so f is monotonically
increasing on [a, b].
This is not sufficient, however, to guarantee the convergence of the Newton method
sequence {tn } given by (1), no matter how t1 ≠ ξ is chosen from (a, b). To understand this,
consider f (t) = t1/3 on [0, 1]. If t1 ∈ (0, 1), then
1/3
f (t1 ) t1
t2 = t1 − = t 1 − = t1 − 3t1 = −2t1 ,
f ′ (t1 ) (1/3)t
−2/3
1

so t2 ∈ (−2, 0). To remedy this, we reconsider f (t) = t1/3 on [−2, 1], but then t3 = −2t2 ∈ (0, 4).
Continuing in this manner, we are ultimately forced to consider f (t) = t1/3 on (−∞, ∞).
However, if t1 ≠ 0, then the Newton method sequence {tn } = {(−2)n−1 t1 } is divergent.
To remedy this deficiency of the classical Newton method, we begin by recalling the
definition of the α-fractal derivative of a function.

Definition 1. Let α > 0 and f ∶ [0, b) → R. The α-fractal derivative of f at x0 ∈ [0, b), denoted
Dα f (x0 ), is given by
f (x) − f (x0 )
Dα f (x0 ) = lim (2)
x→x0 x α − x0α
provided the limit exists. In this case, we say that f is α-differentiable at x0 .

Fix α > 0, and let f be a real, α-differentiable function on (a, b) that is continuous on
[a, b] and satisfies f (a) f (b) ≤ 0. If Dα f (x) ≠ 0 for all x ∈ (a, b), then let x1 ∈ (a, b) and define
the α-fractal Newton method sequence {xn } by

f (xn )
α
xn+1 = xnα − (n = 1, 2, 3, . . . ). (3)
Dα f (xn )

It is clear that if α = 1, then (3) reduces to the classical Newton method sequence (1).
Equipped with fractal Newton methods, let us return to the problem of finding the
zero of f (x) = x1/3 on (−∞, ∞). If α is a rational number in (0, 2/3) of the form α = p/q with
q odd, then

x1/3−α
Dα f (x) = .
3α
If x1 ≠ 0, then the α-fractal sequence (3) is given for n ≥ 1 by

1/3 1/3−α
α
xn+1 = xnα − 3αxn /xn = xnα − 3αxnα = (1 − 3α)xnα .
n−1
Since α ∈ (0, 23 ) implies λ = 1 − 3α ∈ (−1, 1), it follows that xn = λ α x1 for n ≥ 1, and
thus xn → 0.
Mathematics 2023, 11, 2277 3 of 13

To summarize, when f (x) = x1/3 and x1 ≠ 0, the classical Newton method sequence
diverges, but if α is a rational number in (0, 2/3) of the form α = p/q with q odd, then the
α-fractal Newton method sequence converges to the solution x = 0 of f (x) = 0.

3. Preliminary Definitions and Theorems of Fractal Calculus

In this section, we formulate and prove the fractal analogues of many standard results
of classical analysis. However, these fractal results are not deducible from the corresponding
classical results, because it is possible for the fractal derivative Dα f (ξ) to exist at a point ξ,
and the classical derivative of f fails to exist at ξ. Furthermore, the proofs in the literature of
the fractal or conformable derivative results that we require for our fractal Newton method
theorems are based on the assumption that f (x) is expressible by a convergent power series
in a neighborhood of the point ξ. That is, f is assumed to be an analytic function at ξ. To
develop a robust theory, hypothesizing only the existence of the first and second fractal
derivatives of f in a neighborhood of ξ led us to include the “standard” fractal results and
their proofs. We know of no reference that states and proves these results in that generality.
As in this work, we usually insist x0 ≥ 0 in the definition of Dα f (x0 ) to avoid technical
difficulties. To illustrate these difficulties, suppose α is a positive rational number of the
form α = p/q, where q is odd and the greatest common divisor (gcd) of p and q is 1;
√
i.e gcd(p, q) = 1. Then, x α = ( q x) p is clearly well-defined and real for all real x, and
1 √
consequently Dα f (x0 ) as given by (2) is meaningful. However, the examples x 2 = x
and x π = exp(π ln(x)) illustrate why this need not be the case when x0 < 0 and α is not a
rational number of the aforementioned form.
It is clear that if f is α-differentiable at x0 , then f is continuous at x0 . Furthermore, it is easily
seen that the fractal sum rule Dα (g + h)(x0 ) = Dα g(x0 ) + Dα h(x0 ) and the fractal product rule
Dα (gh)(x0 ) = g(x0 )Dα h(x0 ) + h(x0 )Dα g(x0 ) hold when f and g are α-differentiable at x0 .
As an illustration of these facts, suppose α ∈ (0, 1], γ is a non-negative real number, and
n is a positive integer. It is easy to show using the fractal product rule and mathematical
induction that
n
Dα (tα − γα ) = n(tα − γα )n−1 (4)
for all t ≥ 0.
The proof of the next result is routine.

Theorem 1. Let α > 0, f ∶ [0, b) → R, and x0 > 0. Then, f ′ (x0 ) exists if and only if Dα f (x0 )
exists. In this case,
1
Dα f (x0 ) = α−1 f ′ (x0 ).
αx0

The hypothesis x0 > 0 is necessary in this theorem. To understand this, fix α ∈ (0, 1),
let 0 < γ < α < β ≤ 1, and define f on [0, 1) by f (x) = x α . Then, an appeal to the definition
of the (right) fractal derivative gives Dγ f (0) = 0, Dα f (0) = 1, and Dβ f (0) does not exist.
Specifically, Dα f (0) exists, but f is not differentiable at 0.
Using the previous theorem, it is easy to establish the fractal chain rule and the fractal
Rolle’s Theorem.

Lemma 1 (Fractal chain rule). Fix α > 0. Let f be continuous on [a,b], Dα f (x) exist for some
point x ∈ [a, b], g be defined on an interval I that contains the range of f , and Dα g exist at the point
f (x). If x > 0 and f (x) > 0, then the composite function

h(t) = g( f (t)) (a ≤ t ≤ b)

is α-fractal differentiable at x and

Dα h(x) = α( f (x))α−1 Dα g( f (x))Dα f (x). (5)

Mathematics 2023, 11, 2277 4 of 13

Lemma 2 (Fractal Rolle’s theorem). Fix α > 0 and [a, b] ⊆ [0, ∞). Let f ∶ [a, b] → R be
continuous, f (a) = 0 = f (b), and let Dα f (t) exist for all t ∈ (a, b). Then, there exists a point
ξ ∈ (a, b) such that Dα f (ξ) = 0.

The fractal mean value theorem is a consequence of the fractal Rolle’s theorem. For
completeness, we present a proof.

Theorem 2 (Fractal mean value theorem). Fix α > 0 and [a, b] ⊆ [0, ∞). Let f ∶ [a, b] → R be
continuous, and let Dα f (t) exist for all t ∈ (a, b). Then, there exists ξ ∈ (a, b) such that

f (b) − f (a)
= Dα f (ξ). (6)
bα − aα

Proof. We consider the function

h(t) = [ f (b) − f (a)]tα − [bα − aα ] f (t) + f (a)bα − f (b)aα (a ≤ t ≤ b).

Then, h is continuous on [a, b], Dα h(t) exists for all t ∈ (a, b), and h(a) = 0 = h(b). By
the fractal Rolle’s theorem, there exists a point ξ ∈ (a, b) such that Dα h(ξ) = 0. However,

Dα h(ξ) = f (b) − f (a) − (bα − aα )Dα f (ξ)

and the desired result (6) follows.

The proof of our main result, the fractal Newton method convergence Theorem 4,
relies on a fractal version of Taylor’s theorem. For this, we rely on the notion of higher-order
fractal derivatives.

Definition 2. If f has an α-fractal derivative Dα f on an interval, and if Dα f itself has an α-fractal

(2)
derivative, then we denote this derivative as Dα f and call it the second-order α-fractal derivative
of f . Continuing in this manner when possible, we obtain functions
(n)
f , Dα f , . . . , Dα f ,
(n)
each of which is the α-fractal derivative of the preceding one. Dα f is called the nth-order α-fractal
derivative of f .

Theorem 3 (Fractal Taylor’s theorem). Let α > 0, [a, b] ⊂ [0, ∞) and f ∶ [a, b] → R. Let n ≥ 1
f be continuous on [a, b], and Dα f (t) exist for all t ∈ (a, b). Let γ and β be
(n−1) (n)
be an integer, Dα
distinct points in [a, b], and for a ≤ t ≤ b define

n−1 Dα f (γ) α
(k)
p(t) = ∑ (t − γα )k .
k=0 k!

Then, there exists a point x between γ and β such that

Dα f (x) α
(n)
f (β) = p(β) + (β − γα )n .
n!

Proof. Let M be the number defined by

f (β) = p(β) + M(βα − γα )n

and set
g(t) = f (t) − p(t) − M(tα − γα )n
Mathematics 2023, 11, 2277 5 of 13

for a ≤ t ≤ b. Using (4), it then follows that Dα (tα − γα )m = m!, and if 0 ≤ k < m, then
(m)

Dα (tα − γα )k = 0. Consequently, Dα p(t) = 0, Dα M(βα − γα )n = Mn!, and thus

(m) (n) (n)

Dα g(t) = Dα f (t) − Mn!

(n) (n)

if a < t < b.
Next, observe that if k is an integer satisfying 0 ≤ k ≤ n − 1, then

(k) ⎛ Dα f (γ) α ⎞
n−1 (j)
= ∑ (t − γα ) j ∣t=γ
(k)
Dα p(γ) Dα
⎝ j=0 j! ⎠
n−1 Dα f (γ) (k) α
(j)
= ∑ Dα (t − γα ) j ∣t=γ
j=0 j!

Dα f (γ) (k) α D f (γ) (k) α

(k) (j)
= Dα (t − γα )k ∣t=γ +∑ α Dα (t − γα ) j ∣t=γ
k! j>k j!

= Dα f (γ).
(k)

It follows that 0 = g(γ) = Dα g(γ) = . . . = Dα

(n−1)
g(γ).
Our choice of M shows that g(β) = 0. By the fractal Rolle’s theorem, there exists
x1 between γ and β such that Dα g(x1 ) = 0. Since Dα g(γ) = 0, the fractal Rolle’s the-
orem implies that there exists x2 between γ and x1 such that Dα g(x2 ) = 0. After n
(2)

steps, we arrive at the conclusion Dα g(xn ) = 0 for some xn between γ and xn−1 . Hence,
(n)

0 = Dα g(xn ) = Dα f (xn ) − n!M; i.e.,

(n) (n)

D f (xn )
(n)
M= α .
n!

4. Fractal Newton Method Convergence Theorem

Definition 3. Let α > 0 and let f be α-differentiable at ξ ∈ (0, ∞). The α-tangent to f at (ξ, f (ξ))
is the function given by
T f ,α,ξ (t) = f (ξ) + Dα f (ξ)(tα − ξ α ).

Remark 1. If α = 1, then T f ,1,ξ (t) = f (ξ) + f ′ (ξ)(t − ξ) is the classical tangent line function to
the graph of y = f (t) at the point (ξ, f (ξ)). For the general case α > 0, if xn+1 is given by (3), then
it is easy to see using Definition 3 that T f ,α,xn (xn+1 ) = 0. Geometrically, this means that the graph
of T f ,α,xn intersects the t-axis at the point (xn+1 , 0).

Lemma 3. Let α > 0, η ∈ (a, b) ⊂ [0, ∞), and f be twice α-differentiable with Dα f ≥ 0 on (a, b).
(2)

Then, the graph of f lies on or above the graph of T f ,α,η on (a, b).

Proof. Let η and t be distinct points in (a, b). Apply the fractal Taylor’s theorem to f with
n = 2, β = t, and γ = η. Then,

Dα f (x) α
(2)
f (t) = p(t) + (t − η α )2 (7)
2!
for some x between t and η. Here,

p(t) = f (η) + Dα f (η)(tα − η α ) = T f ,α,η (t)

Mathematics 2023, 11, 2277 6 of 13

for all t ∈ (a, b). It follows from (7) that

Dα f (x) α
(2)
f (t) − T f ,α,η (t) = (t − η α )2 ≥ 0
2!
on (a, b).

Theorem 4 (Fractal Newton method convergence). Fix α > 0 and let [a, b] ⊂ [0, ∞). Let f be
twice α-differentiable with Dα f ≥ 0 on (a, b), and let f be α-differentiable with Dα f continuous
(2)

on [a, b] and positive on (a, b). If f (a) f (b) < 0 and x1 ∈ (a, b), then the α-fractal Newton method
sequence given by (3) converges to the unique point ξ in (a, b) satisfying f (ξ) = 0.

Proof. Note that f (a) f (b) < 0 and Dα f (x) > 0 for x ∈ (a, b) imply f (a) < 0 and f (b) > 0. By
the intermediate value theorem and the fractal Rolle’s theorem, there is a unique number ξ
in (a, b) such that f (ξ) = 0. Let x1 be any number in (a, b). If f (x1 ) = 0, then x1 = ξ, and it
then follows from (3) that xn = ξ for all n ≥ 1. Thus, ξ = limn→∞ xn is trivially true.
Assume f (x1 ) > 0. It clearly follows that ξ < x1 < b . We will show that the sequence
{xn } given by (3) satisfies ξ ≤ xn+1 ≤ xn for all n ≥ 1. To this end, note that when n = 1,
we have
f (x1 )
> 0,
Dα f (x1 )
so
f (x1 )
x2α − x1α = − < 0.
Dα f (x1 )

Hence, x2α < x1α and x2 < x1 . Because Dα f (t) ≥ 0 for t ∈ (a, b), it follows from Lemma 3
(2)

that the graph of f lies on or above the graph of T f ,α,x1 on (a, b). Hence, the intersection
point (x2 , 0) of the graph of T f ,α,x1 with the t-axis lies on or below the graph of f at the
point (x2 , f (x2 )). That is, f (x2 ) ≥ 0, from which it follows that ξ ≤ x2 .
Inductively, suppose ξ ≤ xm+1 ≤ xm for some integer m ≥ 1. If xm+1 = xm , then

f (xm )
= xm α − xm+1 α = 0.
Dα f (xm )

Thus, f (xm ) = 0, and it follows from (3) that xn = ξ for all n ≥ m. Similarly, if ξ = xm+1
then xn = ξ for all n ≥ m + 1. In both these cases of equality, the induction is finished.
Therefore, suppose ξ < xm+1 < xm , and hence f (xm+1 ) > 0. Consequently, (3) with n = m + 1
implies that xm+2 < xm+1 . Furthermore, it follows from Lemma 3 that the graph of f lies on
or above T f ,α,xm+1 on (a, b). Consequently, the intersection point (xm+2 , 0) of the graph of
T f ,α,xm+1 with the t-axis lies on or below the graph of f at the point (xm+2 , f (xm+2 )). Thus,
f (xm+2 ) ≥ 0 so ξ ≤ xm+2 . By induction, ξ ≤ xn+1 ≤ xn for all integers n ≥ 1 when f (x1 ) > 0.
Suppose f (x1 ) < 0. Then, clearly a < x1 < ξ and

f (x1 )
< 0,
Dα f (x1 )
so
f (x1 )
x2α − x1α = − > 0.
Dα f (x1 )
Hence x1α < x2α and x1 < x2 . As before, Lemma 3 implies that the graph of f lies on
or above the graph of T f ,α,x1 on (a, b). However, T f ,α,x1 (x2 ) = 0, so f (x2 ) ≥ 0, and thus
ξ ≤ x2 < b. The proof for the case when f (x1 ) > 0 now shows that ξ ≤ xn+1 ≤ xn for all n ≥ 2.
Therefore, if f (x1 ) ≠ 0 and n ≥ 2, the α-fractal Newton method sequence {xn } is
monotonically decreasing and bounded below by ξ. Hence, η = limn→∞ xn exists, and
Mathematics 2023, 11, 2277 7 of 13

clearly ξ ≤ η. Since Dα f is continuous on [ξ, b], it follows that B = sup{Dα f (x) ∶ ξ ≤ x ≤ b} <
∞, so
0 ≤ f (xn ) = Dα f (xn )(xnα − xn+1
α
) ≤ B(xnα − xn+1
α
)
for n ≥ 2 by (3). Consequently, f (η) = limn→∞ f (xn ) = 0, and thus ξ = η = limn→∞ xn .

Corollary 1. Fix α > 0 and let [a, b] ⊂ [0, ∞). Let f be twice α-differentiable with Dα f ≥ 0
(2)

on (a, b), and let f be α-differentiable with Dα f continuous on [a, b] and positive on (a, b). If
f (a) ≤ 0, f (b) > 0, x1 ∈ (a, b), and xn+1 is given by (3) for n = 1, 2, 3, . . . , then the sequence
{xn }∞
n=1 converges to the unique point ξ in [a, b] satisfying f (ξ) = 0.

Proof. If f (a) < 0, then f (a) f (b) < 0, so the desired result follows directly from Theorem 4.
Suppose f (a) = 0. The proof of Theorem 4 shows that {xn } is a monotone decreasing
sequence that converges to ξ = a.

It is well-known that the classical Newton method has local quadratic convergence.
The next result shows that fractal Newton methods share this property.

Theorem 5 (Fractal Newton method convergence rate). Fix α ∈ (0, 1) and let [a, b] ⊂ [0, ∞).
Let f be twice α-differentiable with Dα f , non-negative, and bounded on (a, b), and let f be
(2)

α-differentiable with Dα f , continuous, and non-zero on [a, b]. If f (a) f (b) ≤ 0 and x1 ∈ (a, b), then
the α-fractal Newton method sequence given by (3) converges with a rate of at least 2 to the unique
point ξ ∈ [a, b] satisfying f (ξ) = 0.

Proof. Suppose first that f (a) f (b) < 0. By Theorem 4, the α-fractal Newton method se-
quence {xn }∞n=1 given by (3) is convergent to ξ ∈ (a, b). If xn0 = ξ for some integer n0 ≥ 1,
then clearly xm = ξ for all m ≥ n0 , and the proof is finished. Therefore, suppose xn ≠ ξ for all
n ≥ 1. Apply the fractal Taylor’s theorem with β = ξ and γ = xn to obtain a point tn between
ξ and xn such that

Dα f (tn ) α
(2)
f (ξ) = f (xn ) + Dα f (xn )(x α − xnα ) + (x − xnα )2 . (8)
2
However, f (ξ) = 0, so rearranging (8) and using (3) yields

f (xn ) D f (tn ) α
(2)
xnα − ξ α = + α (x − ξ α )2
Dα f (xn ) 2Dα f (xn ) n

and
Dα f (tn ) α
(2)
α
xn+1 − ξα = (x − ξ α )2 .
2Dα f (xn ) n
It follows that
M α 2
∣xn+1
α
− ξα∣ ≤ ∣x − ξ α ∣ (n = 1, 2, 3, . . . )
2δ n
where ∣Dα f (x)∣ ≥ δ > 0 on [a, b] and 0 ≤ Dα f (x) ≤ M < ∞ on (a, b). Observe that
(2)

α
xn+1 − ξα M xnα − ξ α α 2
∣xn+1
α
− ξ∣∣ ∣≤ ∣ ∣∣x − ξ∣ (n = 1, 2, 3, . . . ). (9)
xn+1 − ξ
α 2δ xnα − ξ n

Since
xα − ξ α
lim = αξ α−1 ,
x→ξ x−ξ
it follows from (9) that to each e > 0 there corresponds an integer N0 = N0 (e) ≥ 1 such that

M 2 2
∣xn+1
α
− ξ∣αξ α−1 ≤ ( + e)(αξ α−1 ) ∣xnα − ξ∣
2δ
Mathematics 2023, 11, 2277 8 of 13

for all n ≥ N0 . Hence, there exists a positive constant A such that

2
∣xn+1
α
− ξ∣ ≤ Aξ α−1 ∣xnα − ξ∣ (n = 1, 2, 3, . . . ). (10)

If f (a) f (b) = 0, then the hypotheses on f imply that f (x) ≠ 0 in (a, b) and {xn }∞
n=1
converges to ξ ∈ {a, b} as in Corollary 1. If ξ = b or ξ = a > 0, then the argument leading
to (10) still holds. If ξ = a = 0, then it follows from (8) that

Dα f (tn ) 2α
(2)
α
xn+1 = x
2Dα f (xn ) n
so
M 1/α 2
∣xn+1 ∣ ≤ ( ) ∣xn ∣ (n = 1, 2, 3, . . . ). (11)
2δ
From (10), if ξ ≠ 0 and (11) if ξ = 0, we see that the α-fractal Newton method sequence
converges with a rate of at least 2.

We remark that our proof of Theorem 5 shows that for real functions f on [a, b]
satisfying the hypotheses, there exists a constant K = K( f , [a, b]), independent of α ∈ (0, 1)
and ξ ∈ [a, b], such that for all n ≥ 1,
2
∣xn+1
α
− ξ∣ ≤ Kαξ α−1 ∣xnα − ξ∣ if ξ ≠ 0, (12)

and
2
∣xn+1
α
∣ ≤ K∣xnα ∣ if ξ = 0. (13)

5. Comparison of Fractal and Classical Newton Methods

In Section 2, we encountered an instance when certain α-fractal Newton method
sequences {xn } given by (3) converge to the solution x = ξ of f (x) = 0, but the classical
Newton method sequence {tn } given by (1) diverges for every choice of t1 ≠ ξ. In Section 6,
we will encounter infinite families of such instances. In this section, we explore conditions
when the sequences {xn } and {tn } are equiconvergent.
Let α > 0 and {xn } be an α-fractal Newton method sequence given by (3). If we assume
that all xn > 0, then (3) can be rearranged to yield

1/α
α f (xn )
xn+1 = xn (1 − ) (n = 1, 2, 3, . . . ).
xn f ′ (xn )

On the other hand, assume that the classical Newton method sequence {tn } given
by (1) consists entirely of positive terms. Then, (1) can be written as

f (tn )
tn+1 = tn (1 − ) (n = 1, 2, 3, . . . ).
tn f ′ (tn )

Comparing these identities, one expects that if x1 = t1 and α is sufficiently close to 1,

then ∣tn − xn ∣ → 0 as n → ∞.
To formulate a precise version, let f be a real function on [a, b] ⊂ [0, ∞) satisfying:
(H0) f (a) f (b) < 0;
(H2) f ′′ is continuous and non-negative on the open interval (a, b);
and either
(H1+) f ′ is continuous and positive on the closed interval [a, b];
or
(H1−) f ′ is continuous and negative on the closed interval [a, b].
Mathematics 2023, 11, 2277 9 of 13

If t1 ∈ (a, b), then it is well-known that the classical Newton method sequence {tn }
given by (1) converges to the unique solution t = ξ of f (t) = 0 in the interval (a, b).
In the case when f satisfies (H0), (H1+), and (H2), the next theorem shows that the
α-fractal Newton method obeys a complete analogue of the classical Newton method
convergence theorem.

Theorem 6. Let f satisfy (H0), (H1+), and (H2). If 0 < α < 1 and x1 ∈ (a, b), then the α-fractal
sequence {xn } given by (3) converges to the unique solution x = ξ of f (x) = 0 in the interval (a, b).

Proof. Fix α ∈ (0, 1). For all x ∈ [a, b],

x1−α ′
Dα f (x) = f (x).
α
It follows from (H1+) that Dα f is continuous on [a, b] and positive on (a, b). Further-
more, one computes

x1−2α
Dα f (x) = [(1 − α) f ′ (x) + x f ′′ (x)]
(2)
α2

for all x ∈ (a, b). Hence, Dα f is continuous and positive on (a, b) by (H1+) and (H2).
(2)

Therefore, the α-fractal Newton method sequence {xn } is convergent to ξ by Theorem 4.

The next theorem shows that if f satisfies (H0), (H1−), and (H2+) ( f ′′ is continuous
and positive on the closed interval [a, b]), then there is a non-empty open interval I in (0, 1)
such that the α-fractal Newton method sequence {xn } given by (3) is also convergent to ξ
when α ∈ I. To make this precise, suppose that f satisfies (H1−). Then, 0 < M = max{∣ f ′ (x)∣ ∶
x ∈ [a, b]} < ∞. Furthermore, suppose that f satisfies (H2+). Then, m = min{ f ′′ (x) ∶ x ∈
[a, b]} > 0. Now let µ be any positive number in the open interval (1 − amM , 1).

Theorem 7. Let [a, b] ⊂ (0, ∞) and let f satisfy (H0), (H1-), and (H2+). If α ∈ (µ, 1), then the
α-fractal sequence {xn } given by (3) converges to the unique solution x = ξ of f (x) = 0 in the
interval (a, b).

Proof. If x ∈ [a, b], we have

(1 − α) f ′ (x) + x f ′′ (x) > (1 − µ)(−M) + am > 0,

and it follows that

x1−2α
Dα f (x) = [(1 − α) f ′ (x) + x f ′′ (x)] > 0.
(2)
α2
The proof of Lemma 3 shows that for any distinct points η and t in (a, b), there exists
x between η and t such that

Dα f (x) α
(2)
f (t) − T f ,α,η (t) = (t − η α )2 > 0.
2!
That is, the graph of f lies on or above the graph of T f ,α,η on the interval [a, b] for
every η ∈ (a, b) and every α ∈ (µ, 1).
Let x1 ∈ (a, b) and let {xn } be the α-fractal Newton method sequence given by (3). If
x1 = ξ, then xn = ξ for all n ≥ 1, and the proof is finished in this case. Suppose x1 ≠ ξ. Then,
either f (x1 ) > 0 and a < x1 < ξ, or f (x1 ) < 0 and ξ < x1 < b. Consider the first case, the proof
in the second case being analogous. Arguing using induction and the identity (3) as in the
Mathematics 2023, 11, 2277 10 of 13

proof of Theorem 4, we see that the sequence {xn } is increasing and bounded above by ξ.
Hence, {xn } converges to a number η in (a, b). Furthermore, it follows from (3) and

0 < a < xn ≤ xn+1 ≤ ξ

that
0 ≤ f (xn ) = −Dα f (xn )[xn+1
α
− xnα ] ≤ M(xn+1
α
− xnα ),
and thus f (η) = 0. However, then η = ξ.

6. Numerical Results
Example 1. Fix a rational number β ∈ (0, 1/2] of the form β = r/s where gcd(r, s) = 1 and s is
odd, and let f (x) = x β for x ∈ (−∞, ∞). If x1 ≠ 0 and α is a rational number in (0, 2β) of the
form α = p/q where gcd(p, q) = 1 and q is odd, then the α-fractal Newton method sequence {xn }
converges to the unique point ξ = 0 in (−∞, ∞) satisfying f (ξ) = 0. If t1 ≠ 0, then the classical
Newton method sequence {tn } diverges.

Proof. Suppose x1 ≠ 0 and let

f (xn )
α
xn+1 = xnα −
Dα f (xn )

for n ≥ 1. It follows from Dα f (x) = α x β−α for x ≠ 0 that

β

α α
α
xn+1 = (1 − )x
β n

for n ≥ 1. Setting γ = 1 − α
β ∈ (−1, 1), we have

xnα = γn−1 x1α (14)

for n ≥ 1, and the desired conclusion for {xn } follows.

Suppose t1 ≠ 0 and let

f (tn )
β
tn 1
tn+1 = tn − = tn − β−1 = (1 − )tn
f ′ (tn ) βtn β

for n ≥ 1. Setting λ = 1 − 1
β ∈ (−∞, −1], we have

tn = λn−1 t1

for n ≥ 1, and hence {tn } is divergent.

Remark 2. The observant reader will note that the example in Section 2 introducing fractal Newton
methods is the case β = 1/3 and α ∈ (0, 2β) = (0, 2/3) of Example 1. When α ∈ (0, β), the result of
Example 1 is a consequence of Corollary 1, since f (0) f (1) = 0, Dα f is continuous on [0, ∞),

β β−α
Dα f (x) = x >0
α
on (0, ∞), and
β(β − α) β−2α
Dα f (x) = >0
(2)
x
α2
for x > 0. Furthermore, it is clear in Example 1 that for 0 < α < 2β, the rate of convergence of the
α-fractal Newton method sequence is faster the closer α is to β. In fact, if α = β, then the α-fractal
Newton method converges in one step; i.e., xn = x2 = ξ = 0 for all n ≥ 2.
Mathematics 2023, 11, 2277 11 of 13

1
Example 2. Let f (x) = (x − 1) 3 on (−∞, ∞) and x1 > 0. Then:
1. The classical Newton method sequence converges if and only if x1 = 1;
2. The 13 fractal Newton method sequence converges if and only if x1 = 1;
3. The 15 fractal Newton method sequence converges if and only if there exists an integer N ≥ 1
such that x N = 1.

Proof. Let α be a positive rational number whose expression in lowest terms is α = p/q,
where gcd(p, q) = 1 and both p and q are odd. The α-fractal Newton method recursion
relation (3) in this case can be expressed as

1/α
α f (xn )
xn+1 = xn (1 − ) (15)
xn f ′ (xn )

for n ≥ 1. Because
f (x)
= 3(x − 1),
f ′ (x)
(15) reduces to
3α 1/α
xn+1 = xn (1 − 3α + ) (16)
xn
for n ≥ 1.
Suppose α = 1. Then, (16) is equivalent to xn+1 − 1 = −2(xn − 1), and hence xn+1 − 1 =
(−2)n−1 (x1 − 1) for n ≥ 1. Therefore, the classical Newton method sequence converges if
and only if x1 = 1.
Suppose α = 1/3. Then, (16) is equivalent to xn+1 = 1/xn2 , and hence

(−2)n−1
x n = x1

for n ≥ 1. If x1 > 1, then x2m → 0 and x2m−1 → ∞. If 0 < x1 < 1, then x2m → ∞ and x2m−1 → 0.
If x1 = 1, then xn = 1 for all n ≥ 1. Consequently, the 13 fractal Newton method sequence
converges if and only if x1 = 1.
Consider α = 1/5. In this case, (16) is equivalent to

2 3 5
xn+1 = xn ( + ) (17)
5 5xn

for n ≥ 1. Let
2 3 5
h(x) = x( + )
5 5x
for x > 0. Then, (17) can be expressed succinctly as

xn+1 = h(xn ) (18)

for n ≥ 1. Observe that the function h is continuous on (0, ∞) and possesses the following
properties:
1. h is decreasing on (0, 6);
2. h is increasing on (6, ∞);
3. h(x) → ∞ as x → 0+ or x → ∞;
4. 3/16 = h(6) ≤ h(x) for all x > 0;
5. There are exactly two positive solutions to h(x) = 1, namely x = 1 and an irrational
solution x = ξ that is approximately 89.9017661005;
6. x = 1 is the only positive solution to h(x) = x;
4
7. h′ (x) = 25 (1 − 6x )( 52 + 5x
3
) ;
3
8. h′′ (x) = (
36 2
25x3 5
+ 5x
3
) .
Mathematics 2023, 11, 2277 12 of 13

Suppose there exists an integer N ≥ 1 such that the 15 fractal Newton method sequence
{xn } satisfies x N = 1. Then, (18) and property (vi) of h imply that xn = 1 for all n ≥ N
and {xn } converges, to 1 in fact. Conversely, suppose x1 > 0 and the 15 fractal Newton
method sequence {xn } converges, say to the real number L. Then, Equation (18), together
with continuity and the properties of the function h, imply that L = h(L), and hence L = 1.
Suppose, by way of contradiction, that xn ≠ 1 for all n ≥ 1. By the convergence of {xn },
there exists and integer N ≥ 1 such that ∣xn − 1∣ < 1/10 for all n ≥ N.
By the classical Taylor’s theorem, to each positive real number x ≠ 1 there corresponds
a real number ξ between x and 1 such that

h′′ (ξ)
h(x) = 1 − 2(x − 1) + (x − 1)2 . (19)
2
Let n ≥ N. If 9/10 < xn < 1, then h(xn ) > 1 and xn < ξ n < 1 so h′′ (ξ n ) > 0. It follows
from (19) and (18) that
xn+1 − 1 = h(xn ) − 1 > 2∣xn − 1∣. (20)
If 1 < xn < 11/10, then 0 < h(xn ) < 1 and 1 < ξ n < xn so 0 < h′′ (ξ n ) < h′′ (1) = 36/25 and

h′′ (ξ n ) 36 1
∣−2 + (xn − 1)∣ > ∣−2 + ∣ = 1.928. (21)
2 50 10

Consequently, (19), (18), and (21) imply that

h′′ (ξ n )
∣xn+1 − 1∣ = ∣h(xn ) − 1∣ = ∣−2 + (xn − 1)∣∣xn − 1∣ > 1.928∣xn − 1∣. (22)
2

It follows from (20) and (22) that for all integers m ≥ 1,

∣x N+m − 1∣ > (1.928)m ∣x N − 1∣,

and hence {xn } diverges, a contradiction. We conclude that if x1 > 0 and the 15 fractal
Newton method sequence converges, then there exists a positive integer N such that
x N = 1.
1
Remark 3. The 15 fractal Newton method applied to f (x) = (x − 1) 3 on (−∞, ∞) in Example 2
exhibits a curious phenomenon: There is an infinite set of positive values for x1 for which the
corresponding 15 fractal Newton method sequence {xn } is convergent, necessarily to 1. To understand
this, first observe that if x1 = ξ, then x2 = h(x1 ) = h(ξ) = 1, and consequently xn = 1 for all n ≥ 2.
Next, let g denote the inverse function of h on the interval [6, ∞) (cf. property (ii) of h). If x1 = g(ξ),
then x2 = h(x1 ) = h(g(ξ)) = ξ, x3 = h(x2 ) = h(ξ) = 1, and xn = 1 for all n ≥ 3. Continuing in this
manner, if g(m) denotes the m-fold composition of g and x1 = g(m) (ξ), then xn = 1 for all n ≥ m + 2.
Thus, if
x1 ∈ {ξ, g(ξ), g(2) (ξ), g(3) (ξ), g(4) (ξ), . . . }
then the 15 fractal Newton method sequence {xn } is convergent.
In its current form, the α-fractal Newton method has deficiencies that diminish its value for
1
some practical applications. In Example 2, which considered the problem of solving (x − 1) 3 = 0,
when α = 1/5, the values of x1 for which the 15 fractal Newton method sequence converges are
isolated and differ widely from the limit 1. For instance, ξ is approximately 89.9017661, and g(ξ) is
1
approximately 8771.9667803. However, in the problem of solving x 3 = 0 near the end of Section 2,
we saw that if α is a rational number in (0, 2/3) of the form α = p/q with q odd, then for any value
of x1 , the α-fractal Newton method sequence converges to the solution x = 0. It is an important and
interesting problem for future researchers to modify the α-fractal Newton method in such a way that
it works equally well on such problems.
Mathematics 2023, 11, 2277 13 of 13

7. Conclusions
We introduced a parameterized family of α-fractal Newton methods including the
classical Newton method when α = 1. We investigated fractal Newton methods in detail,
providing general sufficient conditions for their convergence. We compared fractal and
classical Newton methods and identified general circumstances when the classical and
fractal Newton method sequences are equiconvergent. Moreover, we showed that, like
the classical Newton method, local convergence is quadratic for fractal Newton method
sequences. As a consequence of our methods, we obtained an upper bound for the error
constant of α-fractal Newton method sequences as a function of α ∈ (0, 1). We illustrated
with examples that fractal Newton method sequences can converge when the corresponding
classical Newton method sequences diverge. Further research is needed to modify and
improve the α-fractal Newton method, making it more robust in such instances. The
experimental rate of convergence of fractal Newton method sequences and their numerical
performance—e.g., CPU times and numbers of iterations—relative to the classical Newton
method are interesting problems with practical consequences that are worthy of further
investigation. The relaxation of the non-negativity hypothesis concerning the second fractal
derivative in Theorem 4 is an interesting problem. However, it is a delicate one, since a
change in the concavity of f at the root of f (x) = 0 leads easily to divergence for certain
classical Newton method sequences.

Author Contributions: Investigation, A.A.; writing—original draft, A.A. and D.G.; writing—review
& editing, A.A. and D.G. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Gdawiec, K.; Kotarski, W.; Lisowska, A. On the robust Newton’s method with the Mann iteration and the artistic patterns from
its dynamics. Nonlinear Dyn. 2021, 104, 297–331. [CrossRef]
2. Hubbard, J.; Schleicher, D.; Sutherland, S. How to find all roots of complex polynomials by Newton’s method. Invent. Math. 2001,
146, 1–33. [CrossRef]
3. Schleicher, D.; Stoll, R. Newton’s method in practice: Finding all roots of polynomials of degree one million efficiently. Theor.
Comput. Sci. 2017, 681, 146–166. [CrossRef]
4. Yau, L.; Ben-Israel, A. The Newton and Halley methods for complex roots. Am. Math. Mon. 1998, 105, 806–818. [CrossRef]
5. Akgül, A.; Cordero, A.; Torregrosa, J.R. A fractional Newton method with 2αth-order of convergence and its stability. Appl. Math.
Lett. 2019, 98, 344–351. [CrossRef]
6. Wang, X.; Tao, Y. A new Newton method with memory for solving nonlinear equations. Mathematics 2020, 8, 108. [CrossRef]
7. He, J.H. Fractal calculus and its geometrical explanation. Results Phys. 2018, 10, 272–276. [CrossRef]
8. Brouers, F.; Sotolongo-Costa, O. Generalized fractal kinetics in complex systems (application to biophysics and biotechnology).
Phys. Stat. Mech. Appl. 2006, 368, 165–175. [CrossRef]
9. Kosmidis, K.; Macheras, P. On the dilemma of fractal or fractional kinetics in drug release studies: A comparison between Weibull
and Mittag-Leffler functions. Int. J. Pharm. 2018, 543, 269–273. [CrossRef] [PubMed]
10. Candelario, G.; Cordero, A.; Torregrosa, J.R.; Vassileva, M.P. An optimal and low computational cost fractional Newton-type
method for solving nonlinear equations. Appl. Math. Lett. 2022, 124, 107650. [CrossRef]
11. Golmankhaneh, A.K. Fractal Calculus and Its Applications F α -Calculus; World Scienetific: Singapore, 2022.
12. Blaszczyk, T.; Bekus, K.; Szajek, K.; Sumelka, W. Approximation and application of the Riesz-Caputo fractional derivative of
variable order with fixed memory. Meccanica 2022, 57, 861–870. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.