Third Note on the Shape of S convexity

I.M.R. Pinheiro April 22, 2013

Abstract As promised in Second Note on the Shape of sconvexity, we now discuss the exponent of the denition of the sconvex class of functions that deals with negative images of real functions. We also present a severely improved version of the denition of the phenomenon sconvex ity.

MSC(2000): 26A51 Key-words: Analysis, Convexity, Denition, S -convexity, sconvexity, geometry, shape.

I. Introduction
In First Note on the Shape of S convexity, we have presented more evidence on our re-wording of the piece of denition of the phenomenon sconvexity that deals with non-negative real functions being of fundamental importance for Mathematics and have proposed a geometric denition for the phenomenon. In Second Note on the Shape of S -convexity, we have once more proved that our proposed modications to the denition of the phenomenon sconvexity, this time for the negative share of the real functions, constitute a major step towards making the phenomenon be a proper extension of Convexity. We have also proposed a geometric denition for the negative case. We imagine that our limiting curve for sconvexity, when the real function is negative, may be a bit bigger than the limiting curve for sconvexity when the real function is not negative in terms of length, what means that
Postal address: P.O. Box 12396 ABeckett St, Melbourne, Victoria, Australia, 8006. Electronic address: [email protected].

our lift may not be the same for both cases. Our perimeters seem to be close enough in dimension, however, the dierence being noticed by the rst decimal digit only and being less than 0.5 (considering our approximation for pi, our manual calculations, and the approximation for the perimeter via elliptical curve). In Third Note on the Shape of sconvexity, the present note, we study the perimeter of our limiting curve for the phenomenon sconvexity aiming equal perimeters for geometrically equivalent situations. As a consequence, we propose new renements to the denition of the phenomenon at the end of this note. The limiting curve originates in the application of the denition of the phenomenon. Piece of the analytical denition of the phenomenon sconvexity that we here deal with (see [Pin11] and [Pin13], for instance) Denition 1. A function f : X > , where |f (x)| = f (x), is told to 2 if the inequality belong to Ks f ((1 )x + (x + )) (1 ) s f (x) + s f (x + ) holds / [0, 1]; x/x X ; s = s2 /0 < s2 1;X/X + X = [a, b]; /0 < (b x). Piece of the geometrical denition that we here deal with Denition 2. A real function f : X > Y , for which |f (x)| = f (x), is called sconvex if and only if, for all choices (x1 ; y1 ) and (x2 ; y2 ), where {x1 , x2 } X , {y1 , y2 } Y , Y = Imf , and x1 = x2 , it happens that the line drawn between (x1 ; y1 ) and (x2 ; y2 ) by means of the expression (1 1 1 ) s y1 + s y2 , where [0, 1], does not contain any point with height, measured against the vertical Cartesian axis, that is inferior to the height of its horizontal equivalent in the curve representing the ordered pairs of f in the interval considered for the line in terms of distance from the origin of the Cartesian axis.
1 1

II. Chasing equal lengths for both cases

tions): p

1 + [s(1 )s1 y1 + ss1 y2 ]2 d. the expression for the Arc

In Second Note on the Shape of sconvexity, Length was (negative real functions): 1 1s 1 p 1 + [ (1 ) s y1 + s 0

1 1s 2 s y2 ] d. s

Our tables, containing some of the allowed values for s and their respective Arc Lengths, had been put together through approximating values to no decimals in the third listed value for s, 0.25, and through approximating the value of pi to 3.141516 plus the result of the calculation to two decimal digits in the second listed value for s, 0.5. There is some chance that the negative case give more rope than the nonnegative case, for instance. That would imply that we would get more functions in each s-group if we chose to make the non-negative functions limiting line be the same as the negative functions limiting line (in a relative manner). To chase equal lengths for both cases, we rst observe that if the percentage we select inside of the brackets, when the denition of the sconvexity limiting line is used, is the same, what commands the result is the value of the function, rst of all, and then the value of the exponent we raise the percentage to. This way, to compare both cases and chase the same result, we should start by nding functions that hold the same value in modulus in the interval we choose to consider. Whilst in the non-negative case we increase the value of the percentage when raising it to a fractionary exponent, given that it is always at most 100% and at least 0%; in the negative case, we decrease it if doing the same because increasing the portion we take from a negative value is making the result smaller, not bigger. That tells the reader why we have guessed 1 s for the second part of our definition. To chase equal lengths, we use the convexity limiting line as a reference line for both cases and equate the moduli of the dierences between the limiting lines for the extension of convexity and the limiting lines for convexity. We choose the constant functions to work with because they are the nicest functions in the group of functions available to us inside of the phenomenon. With this, suppose that g (x) = A, A > 0 (notice that A = 0 implies that the 3

limiting line for convexity is equal to the limiting line for sconvexity), and h(x) = A. Also suppose that f (cp ) = (1 )A + A, f (scp ) = (1 )1 A + 1 A, 1 1 f (cn ) = ( 1)A A, f (scn ) = (1 ) s A s A, f1 (scp ) = (1 )s A + s A, and f1 (scn ) = (1 )2 A 2 A. We then have |f (scp ) f (cp )| = |f (scn ) f (cn )| and |f1 (scp ) f (cp )| = |f1 (scn ) f (cn )|. With this: 1 1 A) |(1 )1 A + 1 A [(1 )A + A]| = | (1 ) s A s A [( 1)A A]|; and B) |(1 )s A + s A [(1 )A + A]| = | (1 )2 A 2 A [( 1)A A]| What follows is: 1 1 |(1 )1 + 1 1| = | (1 ) s s + 1| and |(1 )s + s 1| = | (1 )2 2 + 1|. Making = 1 2 for practical purposes, we have: 1 1 1 1 |2 +2 1| = | 2 s 2 s +1| and |2s +2s 1| = | 22 22 +1|. What follows is two possibilities for each case, that is, four possibilities: 1 A) |2 21 1| = | 2 2 s + 1|. Then: 1 1 (A1) 2 21 1 = 2 2 s + 1 or (A2) 2 21 1 = 2 2 s 1. B) |2 2s 1| = | 2 22 + 1|. Then: (B1) 2 2s 1 = 2 22 + 1 or (B2) 2 2s 1 = 2 22 1. (A1) returns 1 = log2 1 1 and (A2) returns 1 = 1 s. (B1) returns 2 = log2 11 and (B2) returns 2 = s. 2s When we use Mathematica and the Arc Length function, we notice that 1 s wins over s up to a value (more than 0.5) and loses after that, and, if we consider the variation of the dierences, we will prefer 1 s over s, like it will give us more rope. The main problem that appears here is that we cannot extend a function that is basically founded on a geometric idea (under a line or over it, line that is symmetric if we consider its middle part), in a way to get a non geometrically-alike limiting line, and we have just become suspicious of this line, like we have just started to think that that (absence of symmetry) might be the case with both s and 1 s. Testing such a hypothesis is not dicult. We just have to get one example for each case, exhibit their limiting lines, and prove that the rst derivative applied to points that are equally distant from the center of the limiting line ( = 0.5) will generate dierent results (in terms of modulus). 4
12
s

One of the most trivial examples of sconvex functions of the sort |f (x)| = f (x) is f (x) = x. x is 0.5convex, for instance. Accompany the proof: Proof. A = f ((1 )x + (x + )) = (1 )x + (x + ). B = (1 )0.5 f (x) = (1 )0.5 x. C = 0.5 f (x + ) = 0.5 x + . A2 = (1 )x + (x + ). (B + C )2 = (1 )x + (x + ) + 2[(1 )0.5 x](0.5 x + ). A2 (B + C )2 because A, B , and C are all non-negative. For the same reason, A2 (B + C )2 = A B + C . Now, we will worry about our limiting line only, the 0.5-line. We call it L(). This way, L() = (1 )0.5 x + 0.5 x + . L () = 0.5(1 )0.5 (1)y1 + 0.50.5 y2 . With this, L (0.3) = 0.5(0.7)0.5 y1 + 0.5(0.3)0.5 y2 and L (0.7) = 0.5(0.3)0.5 y1 + 0.5(0.7)0.5 y2 . Notice that everything looks promising at this stage. It suces that we consider a function that is constant, as we have done, and our limiting curve will be symmetric. However, if y1 diers from y2 , then we do not have symmetry. In the case of our chosen function here, lets choose = 1 and x1 = 4 to make it easier. This way, x1 + = 5. y1 = 2 and y2 = 5. L (0.3) = 0.846013 and L (0.7) = 0.489436. The moduli are so dierent that one is almost twice the other. Symmetry is obviously not a possibility here. It is interesting how we never have questioned, this far, the capability of Hudzik and Maligranda in terms of using the mathematical terms correctly. An extension of a set in Mathematics is, for instance, the real numbers. The Real Numbers Set extends the Integer Numbers Set. How do we prove that? Do we simply prove that the Real Numbers Set contains the Integer Numbers Set and that the dierence between the Real Numbers Set and the Integer Numbers Set is non-null? That is all we need to have an extension of a set, or of a class, according to the linguists (see http : //dictionary.ref erence.com/browse/ex tension, for instance).

One should notice that there is a dierence between a bigger set and an extension of a set. In Mathematics, when we say that we have extended something, we usually mean that we have uniformly extended something, contrary to what we read from http : //www.math.utah.edu/on line/1010/numbers/, for instance. Hudzik and Maligranda, who blame Breckner for that, but were the ones to rst come up with the denitions we work with since 2001, thought that, to extend the set of convex functions, they had to preserve the analytical shape of the limiting line for convexity, what then made them think of the current shape of the denition (we have preserved one hundred percent of their chosen shape for the non-negative case). It did not occur to them that percentage is a linear operation. In Vector Analysis with Analytical Geometry, at the university, we learn that some operators are linear operators (see http : //mathworld.wolf ram.com/LinearOperator.html, for instance), and that is the case with the percentage operation. However, exponential operators are not linear. When Hudzik and Maligranda used the exponent s to extend convexity, they changed a linear operator into a non-linear one. That gave us something that is not symmetric and grows in a very dierent manner. They made the line depend on the function values, but the linear operator we had that far returned always a straight line. They did get a set that included the set of convex functions as a result, also a set that gave us a dierence in its favor when compared with the set of the convex functions, so that everything was looking OK in those regards. If we do not care about symmetry or uniformity, however, zillions of sets would extend the set of convex functions, for it suces having one point more that we have an extension. Uniformity of our extensional elements, however, will allow us to derive nice theorems in Real Analysis, theorems of the sort of those that we have in convexity. Notice that every unit interval in the Real Numbers Set is equal to any other unit interval in the Real Numbers Set shape-wise. This repetition, these patterns, is what allow us to have books of more than 200 pages with theorems involving real numbers (Real Analysis books, for instance). Even though the name of the numbers change from unit interval to unit interval, the moduli of the dierences between the elements found in similar geometric situations and the rst element of the unit intervals are equal. We should be after the same sort of situation here so that we have the largest number of analytical theorems as possible. Also notice that convexity is essentially a geometric concept, so that Hudzik and Maligranda should have privileged the geometrical description of the class of functions over the analytical description when creating their extension. If we care about the geometrical description more than we care about the 6

analytical description, we will have a linear operator, or at least a symmetric limiting curve instead of what we have this far. If we x the limiting curve suggested by Hudzik and Maligranda in a way to make it be symmetric, we will still not have the desired uniformity in the extension, since we will not interpolate values uniformly for each unit interval of each limiting line, like we will not do it in the same way for each limiting line we draw, for instance. Notice that we can easily get examples in which the inclination of the limiting curve diers if we consider the middle point (even for the same function in the same sort of situation). Notice that the convex limiting line is always a straight line connecting starting to nishing point in the interval under consideration instead. That is what we would like to get as a limiting line for sconvexity: The same geometric shape always for all complete limiting curves (drawn from starting to nishing point). That would be mimicking the work of the convex limiting line, the geometry involved. We can however nish our work on the denitions proposed by Hudzik and Maligranda by presenting the most acceptable result when the analytical description of the phenomenon convexity is privileged over its geometric description. And that is what we shall do.

III. Conclusion
There might be exponents that are nicer than our chosen 1 and 2 and are still acceptable. We have decided to deal with the phenomenon sconvexity as if it were an exclusively extensional concept for issues that have to do with practicality and accuracy (we now forbid s to assume the value 1 in our denition). We have extended the domain of the sconvex functions to because the denition should only bring necessary limitations, and we have found problems only with the image of the functions so far in what regards the current shape of the denition of the phenomenon we study, not the domain. We have decided to swap the coecients in our denition because = 0 should bring f (x) to life, trivially, not f (x + ). We have added the interval of denition of s to our geometric denition to make it be independent from the analytical denition. Because of our new ndings and decisions, we have produced a new update for our denition of the phenomenon sconvexity, and we present it here, after the third paragraph from this one.

This paper should leave the readers with certainty that we cannot have anything suitable for inclusion in Real Analysis coming from the mathematical work of Hudzik and Maligranda in what regards sconvexity. Their idea, however, of extending convexity through drawing a bigger limiting line, is highly interesting. We promise presenting a nal suggestion, this time one that is fully acceptable for inclusion in the body of Real Analysis, in future work. Update on our renements of the denitions proposed by Hudzik and Maligranda

1) Analytical Denition

We have two possibilities so far for each piece of the analytical denition of the phenomenon if we think of including the denition of Hudzik and Maligranda, and those stated that the denition had actually been created by Breckner, properly in Mathematics.

1.A) Possibility 1

Denition 3. A function f : X > , where |f (x)| = f (x), is told to 2 if the inequality belong to Ks f ((1 )x + (x + )) (1 )s f (x) + s f (x + ) holds / [0, 1]; x/x X ; s = s2 /0 < s2 < 1;X/X X = [a, b]; /0 < (b x). Denition 4. A function f : X > , where |f (x)| = f (x), is told to 2 if the inequality belong to Ks f ((1 )x + (x + )) ( ) ( (1 )
log2
1 12s

) f (x + )

log2

f (x) + 8

1 12s

holds / [0, 1]; x/x X ; s = s2 /0 < s2 < 1;X/X X = [a, b]; /0 < (b x). Remark 1. If the inequalities are obeyed in the reverse1 situation by f , then f is said to be s2 concave. 1.B) Possibility 2

Denition 5. A function f : X > , where |f (x)| = f (x), is told to 2 if the inequality belong to Ks f ((1 )x + (x + )) (1 )
log2
1 1 12 s

f (x) +

log2

1 1 12 s

f (x + )

holds / [0, 1]; x/x X ; s = s2 /0 < s2 < 1;X/X X = [a, b]; /0 < (b x). Denition 6. A function f : X > , where |f (x)| = f (x), is told to 2 if the inequality belong to Ks f ((1 )x + (x + )) (1 ) s f (x) + s f (x + ) holds / [0, 1]; x/x X ; s = s2 /0 < s2 < 1;X/X X = [a, b]; /0 < (b x). Remark 2. If the inequalities are obeyed in the reverse2 situation by f , then f is said to be s2 concave. 2) Geometric Denition
1 1

We then have two possibilities for each piece of the geometric denition as well.
1 2

Reverse here means >, not . See the previous footnote.

2.A) Possibility 1

Denition 7. A real function f : X > Y , for which |f (x)| = f (x), is called sconvex3 if and only if, for all choices (x1 ; y1 ) and (x2 ; y2 ), where {x1 , x2 } X , {y1 , y2 } Y , Y = Imf , and x1 = x2 , it happens that the line drawn between (x1 ; y1 ) and (x2 ; y2 ) by means of the expression (1 )s y1 + s y2 , where [0, 1], does not contain any point with height, measured against the vertical Cartesian axis, that is inferior to the height of its horizontal equivalent in the curve representing the ordered pairs of f in the interval considered for the line in terms of distance from the origin of the Cartesian axis. Denition 8. A real function f : X > Y , for which |f (x)| = f (x), is called sconvex4 if and only if, for all choices (x1 ; y1 ) and (x2 ; y2 ), where {x1 , x2 } X , {y1 , y2 } Y , Y = Imf , and x1 = x2 , it happens that the line drawn between (x1 ;( y1 ) and ( ) ) (x2 ; y2 ) by means of the expression (1 ) y1 + y2 , where [0, 1], does not contain any point with height, measured against the vertical Cartesian axis, that is inferior to the height of its horizontal equivalent in the curve representing the ordered pairs of f in the interval considered for the line in terms of distance from the origin of the Cartesian axis. Remark 3. If all the points dening the function are located above the limiting line instead, then f is called sconcave. 2.B) Possibility 2 Denition 9. A real function f : X > Y , for which |f (x)| = f (x), is called sconvex5 if and only if, for all choices (x1 ; y1 ) and (x2 ; y2 ), where {x1 , x2 } X , {y1 , y2 } Y , Y = Imf , and x1 = x2 , it happens that the line drawn between (x1 ; y1 ) and (x2 ; y2 ) by means of the expression (1
s must be replaced, as needed, with a xed constant located between 0 and 1 but dierent from 0 and 1. For instance, if the chosen constant is 0.5, then the function will 1 be 0.5-convex or 2 -convex and s will be 0.5 in the expression that denes the limiting line. 4 See the previous footnote. 5 See the previous footnote.
3

log2

1 12s

log2

1 12s

10

(
log2
1 1 12 s

)
log2

(
1 1 12 s

) y1 + y2 , where [0, 1], does not contain any point with height, measured against the vertical Cartesian axis, that is inferior to the height of its horizontal equivalent in the curve representing the ordered pairs of f in the interval considered for the line in terms of distance from the origin of the Cartesian axis. Denition 10. A real function f : X > Y , for which |f (x)| = f (x), is called sconvex6 if and only if, for all choices (x1 ; y1 ) and (x2 ; y2 ), where {x1 , x2 } X , {y1 , y2 } Y , Y = Imf , and x1 = x2 , it happens that the line drawn between (x1 ; y1 ) and (x2 ; y2 ) by means of the expression (1 1 1 ) s y1 + s y2 , where [0, 1], does not contain any point with height, measured against the vertical Cartesian axis, that is inferior to the height of its horizontal equivalent in the curve representing the ordered pairs of f in the interval considered for the line in terms of distance from the origin of the Cartesian axis. Remark 4. If all the points dening the function are located above the limiting line instead, then f is called sconcave.

See the previous footnote.

11

References
[Pin11] M. R. Pinheiro. First Note on the Denition of S2 convexity. Advances in Pure Mathematics, 2011. Vol. 1, pp. 1 2. [Pin13] 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.

12