Notes on one of the generators of the

S−convexity phenomenon

I.M.R. Pinheiro

PO BOX 12396, A’Beckett st, Melbourne, Victoria, 8006, AUSTRALIA

e-mail: mrpprof essional@yahoo.com

Abstract: In this one more precursor paper, we deal with one specific model
which is currently found in the literature on S−convex functions: polynomial.
We not only review this model, making it look more mathematically solid, but

we also extend it a bit further. On the top of that, side remarks of quality are
made regarding the subject.

AMS Subj. Classification:26D10(Primary), 26D15 (Secondary).

Key-words: Generator, S-convexity, Examples, Dragomir, Pierce.

1. Introduction

In this paper, we deal with a few issues related to models proclaimed to gen-

erate S−convex functions, that is, with theorems which are already found in
the literature regarding that. In the lines which follow, we investigate the

polynomial model in depth, criticizing it and fixing it, as well as extending it.
Generators

Regarding polynomial models, what is found in the literature so far is:

1.A) Generators for s2 −functions

From Dragomir and Pierce [1], page 283:
For 0 < s < 1, {a, b, c} ∈ <, u ∈ <+ , take M to be:

f (u) = a, u=0

and

f (u) = bus + c, u > 0.

1.B) Generators for s1 −functions

If, in the model just mentioned, b ≥ 0 and c ≤ a, then f ∈ Ks1 , which we shall
name A, that is, A will stand for set of conditions for which the functional
model M is a generator of examples of functions in Ks1 .

1.C) On extra assertions on the issues contained in the literature

In Dragomir and Pierce [1], p. 292, we read that if b > 0 and c < 0 then
f 6∈ Ks2 .

The same source, same page, also reveals that if b ≥ 0 and 0 ≤ c ≤ a then
f ∈ Ks2 .

Related problems found in the literature

(Problem Q) From the same source, [1], p. 283, we read that if A is found
but c < a, that is, not allowing c = a, then we have a non-decreasing function

in (0, ∞] but not necessarily in [0, ∞]. This is a severely odd remark, once

there is mathematical proof, also provided in Dragomir and Pierce [1], as to

state that any function in Ks1 will fall into the non-decreasing category for the

(0, ∞] case. We will then keep this by now and discuss this issue later on on
this paper.
 In the sections that follow, we deal with:

• Terminology;

• Definitions;

• A simple extension/fixing, in the polynomial case, for M , as well as A;

• Conclusion.

2. Terminology

We use the same symbols and definitions presented in Pinheiro [2]:

• Ks1 for the class of S-convex functions in the first sense, some S;

• Ks2 for the class of S-convex functions in the second sense, some S;

• K0 for the class of convex functions;

• s1 for the variable S, 0 < S ≤ 1, used in the first definition of S-

convexity;

• s2 for the variable S, 0 < S ≤ 1, used in the second definition of S-

convexity.

Remark 1. The class of 1-convex functions is just a restriction of the class

of convex functions, that is, when X = <+ ,

K11 ≡ K12 ≡ K0 .
 3. Definitions

Definition 3. A function f : X− > <, f ∈ C 1 , is said to be s1 -convex if

the inequality

1
f (λx + (1 − λs ) s y) ≤ λs f (x) + (1 − λs )f (y)

holds ∀λ ∈ [0, 1], ∀x, y ∈ X such that X ⊂ <+ .

Definition 4. f is called s2 -convex, s 6= 1, if the graph lies below a ‘bent

chord’ (L) between any two points, that is, for every compact interval J ⊂ I,
with boundary ∂J, it is true that

sup(L − f ) ≥ sup(L − f ).
J ∂J

Definition 5. A function f : X− > <, in C 1 , is said to be s2 -convex if

the inequality

f (λx + (1 − λ)y) ≤ λs f (x) + (1 − λ)s f (y)

holds ∀λ ∈ [0, 1], ∀x, y ∈ X such that X ⊂ <+ .

 3. Simple extension, polynomial case model

From here onwards, each time we come up with a new result, we place a F

close to it to indicate it is such.

Lemma 1. The sum of two S−convex functions is also an S−convex func-

tion.

Proof. Take f and g to be S−convex (both senses). We now have: f (ax +

by) + g(ax + by) ≤ as (f + g)(x) + bs (f + g)(y). This implies (f + g)(ax + by) ≤

as (f + g)(x) + bs (f + g)(y), that is, the sum of two S−convex functions is,
indeed, an S−convex function, as wished for.

Our first issue here appears. If it suffices that each member of the sum is
S−convex, why would it be necessary to hold the restrictions b ≥ 0 and c ≤ a,
for one of the cases, and {a, b, c} ∈ <, u ∈ <+ ?
Basically, if the domain is allowed to be negative, we could have a smaller value

for the domain, to the left side of the inequality, smaller than the first value
to the right, for instance. At least for one of the types of S−convexity (s1 ),

we have proven it to be impossible that it is decreasing (see [3], for instance).

Therefore, we’d better not allowing the members of the domain to be negative

when it comes to Ks1 .

Another important issue is that if we allow the image of the function to be

negative and positive, we may have problems with this as well. Quite trivially,

for instance taking the first point in the domain interval to hold null image,
this allowance could reduce the value to the right at a point of it being less
than the value to the left, from an earlier point at the domain...
With the image, we cannot go from positive to negative. We may remain
always on the positive values, and we may remain always on the negative

values, but no mix.

The other problem which the current literature seems to account for is the

domain being zero. Taking it to be zero will lead us to f (0) ≤ (as + bs )f (0).
That means (1 − (as + bs ))f (0) ≤ 0. If in Ks1 , that will always be true. In the
second sense, however, that would bring us to a negative value multiplied by

f (0) being non-positive. What that means is that f (0) ≥ 0, solely.

Of course, if we state that a ≥ 0 and f (0) = a, that issue is satisfied.

As to the relationship between a and c, all we know of is that we cannot
hold a smaller value for the function after zero than we hold at zero, if Ks1 is

considered. This way, bus + c ≥ a, what implies one in several items:

• c ≥ a and bus ≥ 0;

• bus ≥ a and c ≥ 0;

• or, simply, what it is: bus + c ≥ a. In other words: f (u) ≥ a whenever

u > 0.

F
µ ¶s
u
Lemma 2. c n
is an S−convex function.

Proof. For matters of proof, we know that as long as c is non-negative, c

is irrelevant to our proof. Notice that the null function will always satisfy
µ ¶s
ax+by
the inequality for S−convex functions. Consider then analyzing n
≤
µ ¶s µ ¶s
as nx + bs ny , as to its veracity. It looks trivial to infer the inequality is
correct (raise to the inverse of the power, both sides, for instance).
We here will mark with red F every theorem which fixes past mistakes in the
analysis of S−convex functions. Therefore:

Theorem 6. A model for S−convex functions, of either sense, is:

f (0) = a0

and
µ ¶s µ ¶s µ ¶s
s u u u
f (u) = a1 u + a2 + a3 + ... + am + am+1 ,
n n−1 n−m

u 6= 0, 0 ≤ m < n, {m, n} ⊂ N, 0 < s < 1, (an ) ⊂ <, for f ∈ Ks1 or f ∈ Ks2 ;

u ∈ <+ , f (u) ≥ a0 for u > 0 , Imf ⊂ <+ Y Imf ⊂ <− , for f ∈ Ks1 only,

and f (0) ≥ 0 , f ∈ Ks2 , only.

Proof. As just before the theorem, departing from what we know to be true,

for sure, about both cases.

Regarding 1.C.1, the assertion is trivially not true. Take, for instance, the
function:
f (0) = a0

f (u) = a1 us + am+1 ,

where us ≥ 1. We then have f (ax + by) = a1 (ax + by)s + am+1 . On the

other hand, as f (x) = as a1 xs + as am+1 and bs f (y) = bs a1 y s + bs am+1 . Trivially,

not mattering the value of the independent term, it is not relevant for our
judgment on S−convexity pertinence or not (if it is in Ks1 , elimination is

direct on inequality formation, if it is in Ks2 , suffices thinking the left side

overcomes the right side in regards to that parcel, so that if it is excluded

and the same happens to the other, the inequality is proved), regarding the
function. Assuming am+1 < 0 and a1 > 0, leads us to infer that (ax + by)s ≤
as xs + bs y s is always true, implying the condition of S−convexity is satisfied

in full.
The second assertion in 1.C is also not accurate, please compare with our
theorem.
Problem Q is mentioned in [3] by us. Basically, analytically, this is absurd. A

function cannot be non-decreasing in all points after zero but change its nature

for a single point, which is the attractor of the interval...if the function is there

defined (at zero), and it is decreasing as we join the image at zero, but not

before, then we hold an infinite sequence of analytical points forming a non-

decreasing sequence, with the exception of one, the own attractor. The options
for this so exotic point are then discontinuity, nothing else, for, otherwise,

whole Real Analysis could not make less sense. For a function to go from non-

decreasing to decreasing, which is the only possible difference, then it would

need to pass via more than one point where this is the case, naturally. One
cannot include one point and notice any difference, that besets continuity,
which is our assumption. Of course, in case of discontinuous functions, such
would be possible. Paying attention to the mentioned page in the source,

a = a0 , c = am+1 , and b = a1 . What is noticed then is that if discontinuity

is verified, one would have a 6= c, making odd assertion true in almost full

content, missing replacing c < a with c 6= a.

6. Conclusion

In this paper we revise and fix the generic model for S-convex functions, as well

as extend it inside of the polynomial scope. As a side result, we also review

and fix assertions which exist in the literature about the mentioned model.
 7. Bibliography

[1] S. S. Dragomir, C. E. M. Pierce, Selected Topics on Hermite-Hadamard

Inequalities and Applications. RGMIA, Monographs, located at

http : //rgmia.vu.edu.au/monographs/hermitehadamard.html, and last seen

on the 20th of August of 2007; written in 2000.

[2] M. R. Pinheiro, Exploring the Concept of S-convexity. Aequationes Math-

ematicae. 74, 3: 201-209 (2007).

[3] Pinheiro, M. R. Review Article on s1 -convexity. Submitted, 2008. Online

preprint at www.geocities.com/mrpprof essional.