First Note on the Denition of s1convexity

I. M. R. Pinheiro P.O. Box 12396 ABeckett St, Melbourne, Victoria, Australia, 8006, illmrpinheiro@gmail.com.

Abstract In this note, we analyze a few major claims found in the scientic liter1 . As a consequence, we nullify one major theorem and ature about Ks a couple of remarks of relevance.

Mathematics Subject Classication: 26A51 Keywords: Analysis, Convexity, denition, S -convexity, geometry, shape

Introduction

1 Ks is a very interesting component of S convexity, not to say exotic: It diers, 2 substantially, from Ks , yet, in a certain sense, seems to supplement it.

1.1

Notation

We use the symbols from [1] here:

1 Ks for the class sconvex functions in the rst sense, where s /0 < s 1; 2 Ks for the class sconvex functions in the second sense, where s /0 < s 1;

K0 for the class convex functions; s1 for the variable s, 0 < s1 1, used for the rst type of sconvexity; s2 for the variable s, 0 < s2 1, used for the second type of sconvexity. Remark 1. The class 1convex functions is simply a subclass of the class convex functions. If we make the domain of the convex functions be inside of the set of the non-negative real numbers, we then have the class 1convex 2 1 K0 . K1 functions: K1

I. M. R. Pinheiro

1.2

Denition

We use the denition from [1] here: Denition 2. A function f : X > is said to be s1 convex if the inequality f ( s x + (1 ) s y ) f (x) + (1 )f (y ) holds / [0, 1]; x, y X ; X + . Remark 2. If the inequality is obeyed in the reverse1 situation by f , then f is told to be s1 concave.
1 1

1.3

Theorems that we discuss here

Dragomir and Pearce (in [2]) state that Hudzik and Maligranda, in [3], told us that:
1 Theorem 1.1. Let 0 < s < 1. If f Ks , then f is nondecreasing on (0, ) and limu>0+ f (u) f (0).

We can infer, from the above theorem, that: 1 (1) (Claim X) Any function in Ks , with domain contained in (0, ), s specied, s = 1, is nondecreasing; 1 (2) (Claim Y) limu>0+ f (u) f (0) for f Ks , s = 1. In this paper, we prove that (1) is not true and (2) is incomplete, controversial, or unnecessary.

2
2.1

Preliminary Notes
Analyzing Claim X

[2] presents the following sequence of implications as proof of the claim X: Proof. We have, for u > 0 and [0, 1], 1 1 (PROBLEM 1) f [( s + (1 ) s )u] f (u) + (1 )f (u) = f (u). The function 1 1 h() = s + (1 ) s
1 is continuous on [0, 1], decreasing on [0, 2 ], increasing on [ 1 , 1] and 2 1 1 1 (PROBLEM 2) h([0, 1]) = [h( 2 ), h(1)] = [2 s , 1]. This yields that
1

Reverse here means >, not .

First Note on the Denition of s1 convexity

1

(5.147) f (tu) f (u) for all u > 0, t [21 s , 1]. 1 1 1 If now t [21 s , 1], then t 2 [21 s , 1], and therefore, by the fact that (5.147) holds for all u > 0, we get 1 1 1 f (tu) = f (t 2 (t 2 u)) f (t 2 ) f (u) for2 all u > 0. By induction, we therefore obtain that (5.148)(PROBLEM 3) f (tu) f (u) for all u > 0, t (0, 1]. Hence, by (taking ) 0 < u v and applying (5.148), we get f (u) = f
u .v v

f (v ), which means that f is non-decreasing on (0, ).

We prove that PROBLEM 1, PROBLEM 2, and PROBLEM 3 will make the proof not be a mathematical proof. There are more problems with the proof, however. From [4], we learn that we cannot have x = y in the denition of S convexity (it is all tied to the geometric denition of convexity). This way, PROBLEM 1 should, per se, nullify the proof that we have just presented. 1 Notwithstanding, notice that 21 s goes as close as we wish (PROBLEM 2) to 0 (limit when s > 0) and actually assumes the value 1 (when s = 1), 1 what then makes the interval [21 s , 1] be a degenerated interval, or not be an interval, but just a point instead. Besides, we have dierent intervals, depending on the value of s we choose (s = 0.5 = t [0.5, 1], s = 0.7 = t [0.7433 , 1]). We then know that we cannot generalize this to t (0, 1]. 1 1 That is a very serious mistake. Also notice that t replaces s + (1 ) s and is what they themselves have called h(), therefore a function, what then would have to mean that t is not a constant. That means that we can replace t with numerical values, but those would have to be each and every value of the 1 1 function s + (1 ) s for us to claim that we have h(). We do notice that s < 1 instead of the usual, s 1, in the theorem, so that 1 [21 s , 1] does not run the risk of being a degenerated interval or an improper interval. Even if the result were true, and we are obviously entitled to try to get it using the reasoning contained in the above proof, we cannot use the just-exposed lines as a proof. As for PROBLEM 3: Notice that, when s = 0.5, t [0.5, 1]. This means that t [0.5, 1] instead of what is written there at least when s = 0.5. With that, we must have 0 < v 2u 2v , for instance, in the next line, not 0 < u v , when s = 0.5. Notice that this problem will appear for all values of s.
2

Notice that the second member of the inequality should have been f (t 2 u), not f (t 2 ) as This is an approximation. The original value was something like 0.742997

well
3

I. M. R. Pinheiro

In this situation, just like in the original situation, in the proof, v and u cannot be variables for f because whatever be a variable for f spans (0, ) or its domain set, with no discrimination. We select only the values that obey the rule that we have created, which is, after due xing in what regards s = 0.5, 0 < v 2u 2v . This way, we are using only a few selected values of the original variable of f . Because of that, we should create a new variable, which we could call w, which is then going to be equal to u , with u and v assuming v only the values that we have selected, what means that the simplication is not valid in the proof. Consequently, the inference is not valid. On the other hand, notice that w = u is a function that is totally dierent v from 1 1 h() = s + (1 ) s in shape, so that even if the rst coordinates of the functions are the same, say both u and v vary from 0 to 1, the second coordinates are completely dierent. We must respect the original function, the one from the denition, when proving something in Mathematics or replace it with a completely equivalent function, what then means that this step is unacceptable (replacing h() with u v and then applying the rule that should apply when we have f (h()u) for when u)). we have f ( u v We have then just proven that there is no proof of f being nondecreasing on (0, ). It is then the case that we either have to nd a proper proof for the claim or a suitable counter-example/proof of the contrary. 1 1 Notice that any convex function is s1 convex (Ks is supposed to extend K1 , so that this should be valid for any s1 we choose, provided that 0 < s1 1) if its domain is in + , in principle. Besides, when s = 1, we have precisely the class convex functions. This way, it suces that we nd a convex function that be decreasing to prove the claim to be false. Easy enough, the quadratic function f (x) = x2 8x + 16, in the piece of domain (0, 4], is one of those. This claim is, therefore, FALSE.

2.2

Analyzing Claim Y

f (0) must exist because it appears in the theorem. Therefore, 0 is part of the domain of the function and we can replace the domain interval (0, ) with [0, ) at least when stating the second part of the theorem. Because every sconvex function is continuous, we know that limu>c+ f (u) = limu>c f (u) = f (c) for each c in the domain of f , therefore limu>0+ f (u) = limu>0 f (u) = f (0). As a consequence, limu>0+ f (u) f (0) and limu>0+ f (u) f (0) are both

First Note on the Denition of s1 convexity

true. Since it is never true that limu>0+ f (u) < f (0) or limu>0+ f (u) > f (0), we should at most write that limu>0 f (u) = f (0). 1 1 The denition of Ks implies any f such that f Ks is continuous (simply imagine that it be not. Imagine a discontinuity of rst type, right at the vertex of a parabola that has a point of minimum value, which we know is the image of a convex, therefore S convex in both senses, function. Now let (x1 , f (x1 )) be a point that is suciently distant from the vertex, which is fully detached from the rest of the graph and lies miles below it. Now make f (y ), from the 1 inequality of denition for Ks , be the image of x1 , which is the rst Cartesian coordinate for the vertex (that should be in the graph but is not). Notice that we will unavoidably nd points in the graph that are above the limiting line, and that will make the function be both not convex and not S convex, what is absurd). Please call this paragraph (*). We will also try to nd the mistake in the proof presented in [2]. The proof in [2] is: Proof. For u > 0, we have (PROBLEM 3) f (u) = f (u + 0) s f (u) + s f (0) and making u > 0+ , we obtain (PROBLEM 4) limu>0+ f (u) limu>0+ f (u) s limu>0+ f (u) + s f (0). Hence, lim f (u) f (0).
u>0+

PROBLEM 3 is that the assertion is only true if and satisfy the conditions 1 of the denition of Ks , that is, if s + s = 1, what, as we know, implies 0 1 and 0 1. Even if such a piece of information is not relevant to the proof, we can only accept the proof as a proof if such constraints are mentioned. PROBLEM 4 is that it is missing explaining where the information f (u) f (u) came from, for instance. Since u u and their assumption was that the function, in this situation, does not decrease (see (5.147)), what implies that f (u) f (u), not the opposite, things are unacceptable from this point onwards in the proof.

2.3

Supplementary Remarks

Still in [2], we nd a remark that is told to be in [3]:

I. M. R. Pinheiro

1 Remark 3. If 0 < s < 1, then the function f Ks is nondecreasing on (0, ) but not necessarily on [0, ).

From Real Analysis, we know that this remark is absurd. It is not possible that one point, in a continuous function, change the nature of the function from nondecreasing to decreasing. That can only happen to a function with a discontinuity on x = 0. Suppose that f : X + > is continuous and nondecreasing in (0, ) but not in [0, ). Then, for 0 < x1 < x2 < x3 , we have f (x1 ) f (x2 ) f (x3 ) and f (0) > f (x1 ) ( (f (0) f (x1 ))). We can then nd x1 , x1 extremely close to zero but dierent from it, such that f (x1 ) < f (0) but x1 > 0. That would mean that > 0 such that x1 (0 , 0 + ) and f (x1 ) = f (0) , > 0. That implies f (x1 ) (f (0) 1, f (0) + 1) and f (x1 ) (f (0) n , f (0) ), n N, n > 3. Therefore, limx>0 f (x) = f (0). 2 n 1 If we then say that Ks functions do not have to be continuous, sustaining that they extend the class convex functions will be a likely-to-be-impossible task (see (*)). 1 Besides, the denition of Ks does not allow us to do that.

Main Results

In this paper, we have completely nullied the following theorem:

1 Theorem 3.1. Let 0 < s < 1. If f Ks , then f is nondecreasing on (0, ) and limu>0+ f (u) f (0).

This theorem appears in [2] and is there reported to have originated in [3]. We have proved that there are plenty of decreasing functions that are in 1 Ks and if limu>0+ f (u) f (0), then limu>0+ f (u) f (0) as well since limu>0 f (u) = f (0) for any sconvex function that be dened on (0, ) or in one of its subsets and also on zero. It does not make sense making a theorem to tell the just-written information because we aim objectivity and clarity in Mathematics and the only actual piece of information that we need to know regarding this is contained in the denition in an almost explicit way (continuity).
1 is nondecreasing on (0, ) Remark 4. If 0 < s < 1, then the function f Ks but not necessarily on [0, )

has also been nullied in this paper in what regards continuous functions and, therefore, unless we change the denitions of convex and sconvex functions in order not to have only continuous functions inside of our sets, when one

First Note on the Denition of s1 convexity

could then think of reassessing this remark, this remark has been nullied in full.

References
[1] M. R. Pinheiro. Convexity Secrets. Traord Canada, 2008. ISBN 1-4251-3821-7. [2] C. E. M. Pearce; S. S. Dragomir. Selected Topics on Hermite-Hadamard Inequalities and Applications. RGMIA monographs, http : //www.rgmia.vu.edu.au, 2002. Accessed in December of 2002. [3] H. Hudzik; L. Maligranda. Some remarks on sconvex functions. Aequationes Mathematicae, 1994. Vol. 48, pp. 100 111. [4] M. R. Pinheiro. Minima Domain Intervals and the S convexity, as well as the Convexity, Phenomenon. Advances in Pure Mathematics, 2013. Vol. 3, no. 1.

Received: Month May, 2013