T. Popoviciu, Asupra funcţiilor de o variabilă reală a căror mulţime de definiţie este reunirea a două submulţimi de monotonie opusă, Analele Acad. R.P.R., t. III (1950) Mem. 1, pp. 1-16 (in Romanian).
Your feedback will be used to help improve Google Translate
On functions of one real variable whose definition set is the union of two subsets of opposite monotonicity.
Prof. TIBERIU POPOVICIU
CORRESPONDING MEMBER OF THE RPR ACADEMY
1.
We will consider functions, real, finite, uniform and defined on an arbitrary linear setWe will denote bythe upper edge and withthe lower bound of this function. Finally we will use the notation
The necessary and sufficient condition thatto be monotonous onis to have, for any group of 3 pointshis/hersIt follows that if the function is not monotone, the points can be found, his, so that we have).
Let a be a real number and let us put 2 ).
Let's note withcrowdsadding, if, the point a in any one of these sets. We will then have
orif.
We always haveand we say thatis decomposed into the union of two consecutive subsets. The point a is the separation point of this decomposition. There is only one separation point decomposition, if this point does not belong toand two such decompositions if.
We will say thatis a function () (on) if there is a decompositionhis/herin two consecutive subsets such that the function is monotone on each of the sets, the monotonicity being of opposite sense on these two subsets ofMore precisely we will say thatis then a function () respectively a function () as it is non-increasing and non-decreasing respectively on(so non-decreasing respectively non-increasing on).
It is clear that any monotone function is at the same time a functionand a function (). Conversely, any function that is at the same time a function () and a function () is a monotone function. Ifis a function () respectively a function (), the function -is a function () respectively a function (). Finally, any function () onis a function () on any subset of.
2. To simplify the exposition we will make some conventions. Linearity and increase (decrease) are particular cases of non-decrease (non-increase). We also agree that any function defined on at most one point is considered monotonic, that is, regardless of whether it is increasing or decreasing.
With these conventions one of the decomposition subsets, corresponding to a function (), can be reduced to the empty set. This happens if and only if the function is monotone. The separation point a can be one of the improper points -or, which again occurs if and only if the function is monotonic.
Any function defined on at most 2 points is monotone and any function defined on at most 3 points is a function ().
Calculations with proper and improper numbers are done according to the usual rules.
3. Becauseto be a function () on the pointsit is necessary and sufficient for us to have
(1)
In the case of a crowdhaving at least 3 points, we have:
Theorem 1. Becauseto be a function () on E is necessary and sufficient for it to be a function () on each of the three-point groups of E.
This property can also be stated in the following form:
Theorem 1'. Becauseto be a function () on E it is necessary and sufficient that inequality (1) is verified for any group of 3 pointsof E.
In the case of functions () we have analogous properties, inequality (1) being replaced by
It follows that if a functionis not a function (m) we can find the pointshis/hersso that
A simple reasoning shows us that we can then find 4 pointshis/hersso that the string
(2)
to have 2 sign variations 1 ).
But on the pointsthe property of beingfunction () is expressed precisely by the fact that the string (2) has at most 1 sign variation. For functions defined on at least 4 points we therefore deduce:
Theorem 2. Becauseto be a function (m) on E is necessary and sufficient for it to be a function (m) on any group of 4 points of E.
This property can also be stated in the form:
Theorem 2'. For f (x) to be a function (m) on E, it is necessary and sufficient that the sequence (2) has at least 1 sign variation, whatever the points.of E.
4. The proof of theorem 1 has been given elsewhere 2 ). The inequalities (1) are obviously necessary. We will give a new proof
0 0 footnotetext: 1 ) The terms of the string are therefore alland alternatively positive and negative.
2 ) Tiberiu Popoviciu, Deux remarques sur les fonctions convexes. Bull. Sc. Acad. Roumaine, 20, 45-49, 1938.
of the sufficiency of these inequalities and then we will specify a little the structure of the functions ().
We have first:
Lemma 1. If inequality (1) is verified for any group of 3 pointsof E and if a is any number, the functionIt is monotonous on one of the smallest of the crowds..
For the sake of demonstration, let's assume the opposite. We can then find the points
so that we have
and it is easy to verify that on at least one of the groups of 3, pointsinequality (1) is not verified.
One can still specify Lemma 1, but the above statement is sufficient for what follows. For example, it is easy to see that in the statement one can replacebyrespectively.
We deduce from here:
Lemma 2. If inequality (1) is true for any group of 3 pointsof E , one can find a point c such that the functionto be unsightly on, and non-decreasing on.
It is enough to prove the property in the case whenit is not monotonous on).
There are pointsso thatto be monotonous onIndeed, if for the pointshis/hersHAVEthen, by virtue of the lemma, anythingcheck the property. If thenandit is monotonous on, it is also monotonic on. Eitherthe upper edge of the pointsThe pointis the only point that enjoys the property thatit is monotonous onfor anythingand it is not monotonous onifWe haveand alsobecause it is obvious that. It is also seen that there is a unique and proper pointwho enjoys their own
0 0 footnotetext: 1. See below the case whenit is monotonous.
father thatit is monotonous onfor anythingand it is not monotonous onifWe have the inequalitybecause, otherwise, the function would not be monotone on any of the sets, which contradicts Lemma 1.
Lemma 2 will be satisfied if we takeThe function is monotone on each of the setsand it remains to prove that it is effectively non-increasing onand non-decreasing onTo consolidate the ideas, let us show that the function is non-increasing onOtherwise we could find the points,so that we haveHowever, the function is not monotonic on, the points can be found,so that we haveWe should have thenIfinequality (1) is not verified on the pointsIfwe should also haveand inequality (1) would not be verified on the pointsIt is also proven thatis non-decreasing on.
5. Theorem 1 now follows easily. Leta point that satisfies Lemma 2. Ifdoes not belong to himthe decomposition sought isIf, on at least one of the sets,the function is monotonic. Indeed, otherwise, we could find the pointsso thatand inequality (1) would not be verified on the pointsTheorem 1 still follows, adding the pointto one of the crowdsand its decompositionis still in shape.
Analogous results exist for the functions ().
Ifis a function () the pointsseparation of the decompositionsin two consecutive subsets of opposite monotonicity form an interval, finite or infinite, being able to be reduced to a single point. This interval is always closed provided that we also consider improper numbersIfit is not monotonous,are precisely the finite numbers determined above. In this casemust belong to any closed interval containing three pointshis/hersso that we haveIfis monotonous, the intervalis infinite or reduces to one of the improper points,Indeed, if we look at
the function as a function () we haveif it is non-decreasing andif it is non-increasing.
If the crowdis not empty, the function reduces to a constant on this set. Indeed, ifis a functionand if, he is non-increasing onand unwaveringly onIt follows that. It is also seen that ifis a functionHAVE.
From the above it follows that, for a function (), its decompositioninto two consecutive subsets of opposite monotonicity can be done, in general, in several and even infinitely many ways. It is easy to see which are the cases in which this decomposition is unique. Assuming thatis - function, for uniqueness it is necessary that the setto be empty and furthermore it is necessary and sufficient that we have:andor, if ; at most one of the pointsto belong toandrespectivelyasrespectively, if.
6. Before going further, we will make an observation on the functions (m) that will be useful to us below.
Whethera functionthe range of corresponding separation points andan arbitrary interval (closed or not) that has no points in common with(so. Finally, eithersome subset ofUnder these conditions we have:
Lemma 3. If the intervalsand I have at least one point in common and ifis a constant such that, function
is a function () onIt is sufficient to prove the
property for. Be it thenand let's putFunctionis non-increasing on, for it is easily seen that it is non-increasing onandis a subset of this set. It is also seen thatis non-decreasing on. Lemma 3 follows.
It can still be observed that the interval of the points of separation of the decompositions ofcorresponding to the functioncontains the range, soalso contains the interval.
An analogous property holds for the functions (), the inequality from the hypothesis of lemma 3 is then.
7. We propose to generalize Theorem 2 by removing the restriction that the considered subsets ofto be consecutive.
We will say thatis a function(on) ifcan be decomposed into the union of two subsets (empty or not), on each of which the function is monotone and the monotonicity is of opposite sense on the two subsets. It is clear that any function () is a functionIn particular, any function defined on at most 3 points is a function (). Finally, it is obvious that any functiononis a function () on any subset of.
Let's first examine the functions defined on 4 distinct pointsLet's consider the string
(3)
If the functionis not univalent, it is obviously a function (). Otherwise (i.e. if the function is univalent) two cases must be distinguished:
The sequence (3) is not strictly monotone.
. The sequence (3) is strictly monotone (increasing or decreasing).
In the casewe have:
and examining all possible cases it is seen thatis a function, the decomposition of the definition set being, for example:
In caseit is sufficient to examine all the decompositions of the definition set into the union of two subsets to see that the function is not a function ().
We therefore have:
Theorem 3. Because the function, define it on the points, to be a function (M), it is necessary and sufficient that the sequence (3) is not strictly monotone.
8. Let us now consider a functiondefined on 5 distinct pointsBecauseto be a functionit is necessary that it be a functionon any group of 4
points chosen among the pointsHowever, this condition is not sufficient.
Let us seek to determine the functions that enjoy the property that they are not functions.but that they are functionson any group of 4 points chosen from the 5 points considered. We will distinguish two cases for this:
.
We will show that in the case
there are no functions of the required form. For the proof we can assumeThen the function cannot be non-increasing on the pointsand so we must haveor.
Let's assumeThe function can have one of the following 10 forms 1 ):
,
2.,
,
,
,
,
,
,
,
In cases
respectivelythe function is not a function () on groups of 4 pointsrespectivelyIn casesrespectivelyfunction is a function (), being monotone on groups of 4 points,respectivelyIn casesrespectivelyfunction is a function (), the decomposition into two subsets of opposite monotonicity being, for example,,respectively.
By symmetry it is seen that the same property occurs ifFinally, the property results the same for.
0 0 footnotetext: 1. The 10 possibilities are not mutually exclusive.
Now let it be the casefrom the beginning of this issue. We will show that in this case there are functions that are not functions (M) on the 5 points, but are functions (M) on any group of 4 points chosen from them.
Such a function cannot be monotonic on the points,and cannot take the valueat any of these points. Ifare they all greater or all smaller thanfunction is a function (), the decomposition being, for example:
It remains to examine the functions for which one of the strings:
is increasing.
It is easy to verify that such a function is a function () on any group of 4 points chosen from the points, but that it is not a function () on all these points.
Finally we have:
Theorem 4. Because the function, defined on the points, to be a function ( M ), it is necessary and sufficient that it be a function ( M ) on any group of 4 points chosen from the points, and so that it is not of the form, one of the rows (4) being increasing.
In this way, the structure of functionsdefined on no more than 5 points is specified.
9. In the following we will use the following:
Lemma 4. Letanda function defined on E that verifies the following properties:
is a functionon G and is a functionon H.
..
. Intervalsandof the separation points of the decompositions corresponding to the functionon G and H, they have at least one point in common.
Thenis a function (M) on E .
Letandthe corresponding decompositions offor the separation point c. Let us consider the sets
. Then. But the function is non-increasing onand onIfHAVEand, that isThis inequality shows that the functionis non-increasing on the set. It is also seen that it is non-decreasing on. Lemma 4 is proved.
10. It remains to examine functions defined on a sethaving at least 5 distinct points. For these functions we have:
Theorem 5. Because the functionbe a function (M) on E , it is necessary and sufficient that it be a function (M) on any group of 5 points of E.
The condition is obviously necessary. It remains to show that it is also sufficient.
Ifis a functionon any group of 5 points of hisit is, a fortiori, a functionon any group of 4 points of. -
11. To arrive at the required proof we will first establish some helpful properties.
Whethera real number and denote byset of pointsfor whichand withthe set of pointsfor whichWe have obviouslyand we deduce:
Lemma 5. Ifis a function (M) on any group of 4 points of E and if a is any real number, at least one of the following properties is satisfied:
is a functionon.
is a functiononIndeed
, in the opposite case, we could find the points,, so that we have:
It is easy to see then that ifit is not a functionon the pointsand ifis not a function () on the points.
Let us now suppose thatis a functionbut that it does not reduce to a function (). It is seen, as in the proof of the lemma, that there is a number, unique and well-determined, which enjoys
the property thatis a functiononfor anythingbut it is not a function () onif. Likewise, there is a unique and well-determined numberwho enjoys the property thatis a function () onfor anythingbut it is not a function () onif. The numbersare contained in the closed interval:
for any group of 4 pointshis/hersfor which the string (2) has 2 sign variations. We also have.
We deduce from here:
Lemma 6. IfIf there is a function (M) on any group of 4 points of E, one can find a numberjust like thatto be a functiononand a functionon.
It is enough to takeFor the property to be general we must admit thatcan also take on inappropriate values. The numberswhich satisfy Lemma 6 form a closed interval, finite or not. This interval is finite if and only if the function does not reduce to a function (). Then,are precisely the numbers determined above. Ifis a functionHAVEand ifis a function () we have.
It can still be observed that on the set of pointsso that, the function is monotonic.
Lemmas 5, 6 are satisfied, a fortiori, by any function that is a functionon any group of 5 points of his.
12. We can now deduce:
Lemma 7. Ifis a function (M) on any group of 4 points of E and if the numbercheck the conditions of the lemmais a function (M) on the set(.
Indeed, ifall assumptions of Lemma 4 are satisfied.are verified according to the definition of setsand according to Lemma 6. Hypothesisis also satisfied. Otherwise, suppose, for example,. Be it thenThe function is not non-decreasing onand onWe can therefore find the points:
so thatandwouldn't it be a functionon the points.
Lemma 7 is satisfied in particular by any function that is a function () on any group of 5 points of.
Whethera number that satisfies the conditions of Lemma 6. We can assume thatis a finite number. Otherwise the function, is a function (). We saw that the intervals [],of the separation points corresponding to the function on the sets, have at least one point in common. So we have.
We will also need the following:
Lemma 8. In the above hypotheses, if, we have.
Indeed, let's assume the opposite. For example, let's say,. Thenis not non-increasing onand it is not non-decreasing onWe can therefore find the points:
so thatand it is seen that ifis not a functionon the points, and ifis not a functionon the points.
Finally, we have
Lemma 9. In the above hypotheses, ifand if,respectivelyHAVE,respectively.
It is enough to prove the property in the caseIf we had, there would be a pointhis/herso thatWe would then haveBut from his definitionit follows that we can find the pointshis/hersso thatand thatIt would therefore follow that the function is not a function () on the points. The proof is done in the same way in the other three cases.
13. Let us now return to the proof of theorem 5, more precisely to the proof of the sufficiency of the condition expressed by this theorem.
Let's assume thatdoes not reduce to a function () (because otherwise there is nothing to prove) and eithera (finite) number
that satisfies the conditions of Lemma 6. Let us keep the notations from No. 12 and denote bythe set of points ofin whichtake the valueThen the crowdsare disjoint pairs and their union is equal to.
For the proof we distinguish 3 cases:
Case I. K is empty. The theorem then follows from Lemma 7.
Case II.has only one pointHere we will distinguish 2 subcases:
Subcase. Then the functionfromcan be extended at the pointunder the conditions of Lemma 3. So addingTOthe theorem follows from lemma 4, all the assumptions of this lemma being satisfied
Subcasedoes not belong to both intervalsTo fix the ideas, let's assume thatdoes not belong to the rangeand that, for example,Based on Lemma 8, we haveWe can extend the functionfromon the pointand then, based on Lemma 9, the interval of separation points corresponding to the functionon the crowdhas at least one point in common withSo addingTOThe theorem follows from Lemma 4, all the assumptions of this lemma being satisfied.
The proof is the same ifor ifdoes not belong to the range.
Case III. K has at least two points. Then let [the smallest interval containingWe haveHere we again distinguish two subcases:
Subcase III 1. Open intervalhas no common point with at least one of the setsFor fixing ideas either.
If, we have, based on Lemma 8. Based on Lemmais a functiononand, based on Lemma 9, the interval of separation points corresponding to this function has at least one point in common withSo addingTOthe theorem still results by applying lemma 4.
Ifis a functionon, based on Lemma 3. Based on Lemma 9, the interval of separation points corresponding to this function has at least one common pointand the theorem results as above.
The same is true if.
Subcase. The interval () has points in common with both subsets. Be it then, for fixing ideas, eitherWe haveand, based on the theorem. Be it thenthe left extremum of the smallest interval containing its pointsandthe right extremum of the smallest interval containing its points. ThenWe cannot haveOtherwise, based on the definition of the points,, we could find the points:
and then on each of the 5-point groups,we are in contradiction with theorem 4. So we haveIf, we haveI say now that the intervalsandhave at least one point in common. Indeed, we cannot havebecause then, based on the definition of the pointand based on the lemma, we would have, in contradiction with the definition of the pointWe can't have either.because, also based on Lemma 9, we would have, which again contradicts the definition of the point. It is also proven that the intervalsandhave at least one point in common. Otherwise we would come across a contradiction with the definition of the point.
But, based on a well-known property, if several intervals have at least one point in common, then they have at least one point in common. It follows that the intervals [],they have at least one thing in common.
WhetherBased on Lemma 3,is a functiononand a functionon, where, in the case, it doesn't matter which subsetwe add this pointAll assumptions of Lemma 4 are then satisfied, replacingwithrespectively. The theorem is still proven.
The same is true for the case when Theorem 5 is therefore completely proven .