INFERIOR AND SUPERIOR SETS WITH RESPECT TO AN
ARBITRARY SET
SFROM
INdUSED IN THE
MULTIDIMENSIONAL INTERPOLATION THEORY
by
Nicolae Crainic
Abstract: Among the necessary and useful notions for the elaboration of some multidimensional interpolation schemes, there is the one of inferior (superior) set with respect to an arbitrary set S from INd, d ≥1. That is why we proposed in the current paper to approach some notions related to these sets, by definitions, remarks, theorems and examples. After the definition of inferior (superior) sets, we show a disjoint decomposition of the set S
⊂
INdwith inferior (superior) sets, which by coalescences and under some conditions (theorem 1), leads to inferior (superior) sets too. Next, we define two total order relations onS
⊂
dIN which are more often used in the interpolation theory and we study the relation between the first (last) elements of Sand the inferior (superior) sets of this set (theorem 2).
Notations: Throughout this paper we use the notation INd to denote the set
} , 1 , ,
0 / ) ,..., , (
{i= i1 i2 id ik ≥ ik ∈IN k = d , we put i =i1 +i2 +...+id for any
∈
i
INd, and the pairs(
S
,
ρ
)
stand for subsets S⊂
INd together with relations"
"
ρ
.Definition 1: Given i=(i1,i2,...,id)∈INd, j=(j1, j2,..., jd)∈INd, we
say that
i
≤
j
, if0
≤
i
k≤
j
k, for anyk
=
1
,
d
. Remarks:1.1 The relation “
≤
” is reflexive, anti- symmetric and transitive, i.e. it is an order relation on INd. We use the same symbol “≤
” to denote the relationinduced on (arbitrary) subsets S
⊂
INd. Unless otherwise specified, we will consider such subsets S to be endowed with the relation “≤
”.1.2 We will also denote by
S
α the setα
Definition 2: We say thatS ⊂INdis an inferior set with respect to INd if,
for any
i
∈
S
, j∈INdwithj
≤
i
, we havej
∈
S
too. Similarly, we say that S isa superior set with respect to INd if for any i∈INd ,
j
∈
S
withj
≤
i
, we haveS
∈
i
.Remarks:
2.1. Similarly, one can define the notion of inferior (superior) sets with respect to arbitrary subsets S
⊂
INd.2.2 Often we simply say inferior, respectively superior set, without indicating the set to which it is an inferior, respectively superior, this fact being implied. In genera, in these situations, it is implied that it is with respect to eitherINd, or the set itself.
Proposition 1: If
S
⊂
IN
d and X ⊂S is inferior (superior) with respect to S, then S \X is superior (inferior) with respect to S.Proof: Let
X
be an inferior set with respect to S,i
∈
S
\
X
,j
∈
S
andj
i
≤
. Ifj
∉
S
\
X
, thenj
∈
X
, hence, sinceX
is inferior andi
≤
j
,S
X
S
⊂
∈
\
i
, it follows thatj
∈
S
\
X
. The second part of the proposition can be proven similarly.Remarks:
3.1 For any
α
=
(
α
1,
α
2,...,
α
d)
∈
S
, theS
α is superior with respect to S, and S\S
α is inferior with respect to S.3.2 The empty set is inferior, respectively superior with respect to any set S
⊂
INd.3.3 Any set S from INd is inferior, respectively superior with respect to itself.
3.4 The relation "is inferior with respect to" defines an order on the set of subsets of INd, while the relation "is superior with respect to" does not (because it is not transitive).
3.5 If
X
i⊂
IN
d andX
i is inferior (respectively superior) withrespect to
X
i+1 for anyi
=
1
,
n
−
1
, thenX
1⊂
X
2⊂
...
⊂
X
n, (respectivelyn
X
X
The remarks 3.1 and 3.2 are obvious from the definitions, while the remark 3.3 results from 3.1 for
S
'=Ø. The reflexivity of the relation “inferior with respect to”, results from 3.3 too. All other remarks result rather easily from the previous definitions.Theorem 1: If
S
1∪
S
2∪
...
∪
S
nis a disjoint cover of Sd
IN
⊂
, so thati
S
is inferior (superior) with respect toS
i∪
...
∪
S
n for anyi
=
1
,
n
, then1.
S
k∪
S
l is inferior (superior) with respect toS
k∪
S
l∪
...
∪
S
n 2.S
k∪
...
∪
S
l is inferior (superior) with respect toS
k∪
...
∪
S
n, for any 1≤k ≤l ≤n. [image:3.595.176.437.294.546.2]Proof: 1. The decomposition of S
⊂
IN
2for the inferior sets case is shown in Figure 1 a, while Figure 1b depicts the case for superior sets where each pair of curves limits points of natural coordinates from S.Figure 1b
Given
i
∈
S
k∪
S
l,j
∈
S
k∪
S
l∪
...
∪
S
n withj
≤
i
, it follows that (i
∈
S
k,j
∈
S
k∪
S
l∪
...
∪
S
n withj
≤
i
) or (i
∈
S
l,j
∈
S
k∪
S
l∪
...
∪
S
n withi
j
≤
).If (
i
∈
S
k,j
∈
S
k∪
S
l∪
...
∪
S
n withj
≤
i
), it follows that (i
∈
S
k,n l
k
S
S
S
∪
∪
∪
∈
...
j
⊂
S
k∪
S
k+1∪
...
∪
S
n withj
≤
i
) and because according to the hypothesisS
k is inferior with respect toS
k∪
...
∪
S
n, it results thatj
∈
S
k, i.e.l k
S
S
∪
∈
j
.Given (
i
∈
S
l,j
∈
S
k∪
S
l∪
...
∪
S
n withj
≤
i
), it follows that (i
∈
S
l,k
S
∈
j
withj
≤
i
), or (i
∈
S
l,j
∈
S
l∪
...
∪
S
n withj
≤
i
), and thusj
∈
S
k orl
S
∈
j
, i.e.j
∈
S
k∪
S
l. We relied also on the hypothesis thatS
l is inferior with respect toS
l∪
...
∪
S
n.In both possible cases discussed previously it results that
j
∈
S
k∪
S
l. Also according to the definition 2, it follows thatS
k∪
S
l is inferior with respect ton l
k
S
S
2. The second point of the theorem results applying successively the first one. Also the situation regarding the superior set can be proven either similarly to the previous one, using the hypothesis and the definition 2, or using the proposition 1.
Remarks:
4.1 The point 1 of the theorem can be extended to more then two "isolated" set (not necessary consecutive)
4.2 The statement "
S
i is inferior (superior) with respect ton l
k
S
S
S
∪
∪
...
∪
", is equivalent to "S
i is inferior (superior) with respect to)
...
(
\
S
1∪
∪
S
i−1S
”, which is more often used in practice.It is worthwhile to notice that the relation
"
≤
"
defined between the elements of the set INd(definition 1) and used when defining an inferior (superior) subset ofS
⊂
INd, is not a complete order relation on this set. This means that the relation"
"
≤
is not sufficient for ordering all elements of the set INdor S⊂
INd. Moreconcrete, if we use the relation
"
≤
"
, (0,0)∈IN2 can be followed by (0,1)∈IN2 and by (1,0)∈IN2 without any way of comparing these two elements between them. Nextwe will define two total order relations on the set S
⊂
INdand the first discussed one will be the lexicographic one.Definition 3: Given i=(i1,i2,...,id)∈INd,
d d IN j
j
j ∈
=( 1, 2,..., )
j we
say that i«
j
, ifi
1<
j
1, ori
1=
j
1 andi
2<
j
2… ori
1=
j
1…i
d−1=
j
d−1 andd d
j
i
≤
.Definition 4: Given i=(i1,i2,...,id)∈INd,
d d IN j
j
j ∈
=( 1, 2,..., )
j , we
say that
i
p
j
, if i < j or i = j and i«j
. Remarks5.1. On IN1, the relations from definitions 3 and 4 coincide with the relation from the definition 1.
5.2. The relations defined by 3 and 4 are total order relations on INd
and on any S
⊂
INd. In other words, with any of these two relations, the elements of d5.4. The elements of
X
⊂S are the first X elements of(
S
,
ρ
)
, with respect to the relationρ
∈
{"
p
"
,”«”}, if for any i∈X,j
∈
S
\
X
, it follows thati
ρ
j
, and the elements ofY
⊂S are the last Y elements of(
S
,
ρ
)
with respect to the relationρ
∈
{"
p
"
,”«”}, if for any i∈Y ,j
∈
S
\
Y
, it resultsj
ρ
i
.5.5. It is easy to notice that we can also define other total
relations on
dIN
The usage of one of the two previous relations depends on
S
⊂
INdand the problem to be solved.
Proposition 2: Let i,
j
∈
S
andρ
∈
{"
p
"
,”«”}.2.1. If
i
≤
j
, theni
ρ
j
, but not conversely. Ifi
p
j
, this does not imply i«j
, and conversely.2.2. The statement ”elements of
X
⊂Sare the first X elements of)
,
(
S
ρ
with respect to the relationρ
∈
{"
p
"
,”«”}” is equivalent to ”for any i∈X,S
∈
j
withj
ρ
i
, thenj
∈
X
”. The assertion ”the elements Y ⊂S are the last Y elements of(
S
,
ρ
)
, with respect to the relationρ
∈
{"
p
"
,”«”}” is equivalent to “for any i∈Y ,j
∈
S
withi
ρ
j
, thenj
∈
Y
”.Proof: 1. The first part of this proposition results easily by combining the definitions 1, 3 and 4. The other sentences are justified by counter examples. Thus, for
)
3
,
1
(
,(
2
,
1
)
2IN
∈
,(
1
,
3
)
«(
2
,
1
)
and(
1
,
3
)
is not in the relation"
≤
"
, respectively"
"
p
with(
2
,
1
)
. Also, if (3,1), (2,4)∈
IN
2, (3,1)p(2,4) and (3,1) are not in the relation"
≤
"
, respectively ”«” with (2,4).2. Assume that the elements of
X
⊂S are the first X elements of(
S
,
ρ
)
, in the relationρ
∈
{"
p
"
,”«”}, and the assertion “for any i∈X,j
∈
S
withj
ρ
i
, thenj
∈
X
” is false. Then we would have i∈X,j
∈
S
withj
ρ
i
, andX
∉
j
, i.e. we would have i∈X ,j
∈
S
\
X
withj
ρ
i
. In addition, we would have that i∈X,j
∈
S
\
X
impliesi
ρ
j
(according to the observation 5.4), that isi
=
j
("
ρ
"
being anti-symmetric). This is impossible because this would mean thatX
andX
S\ have i in common, hence the assertion can not be false. The other assertions can be similarly justified.
Theorem 2: IfS=
(
S
,
ρ
)
is a total ordered set fromINdwith the relation"
{"
p
∈
ρ
,”«”} and if the elements ofX
are the first X elements of(
S
,
ρ
)
, thenX
is inferior with respect to S, and if the elements of Yare the last Y elements of)
,
(
S
ρ
, then Y is a superior set with respect to S.Proof: Let the elements of
X
be the first elements of S in the relation"
{"
p
∈
ρ
,”«”}. Leti
∈
X
,j
∈
S
withj
≤
i
. Since, according to proposition 2, ifi
j
≤
thenj
ρ
i
, then from “i
∈
X
,j
∈
S
withj
≤
i
” it follows that “i
∈
X
,S
∈
j
withj
ρ
i
”. From this deduction according to proposition 2, we infer thatX
∈
j
(the elements ofX
are the first elements of S). We have that, ifi
∈
X
,S
∈
j
withj
≤
i
thenj
∈
X
, which according to the definition 2, is the same with the fact thatX
is an inferior set with respect to S.The last part of the theorem can be proved by analogy with the first one. Its proof can be made using the proposition 1, as well as the results of the first part. Thus, if the elements of
Y
are the last Y elements of(
S
,
ρ
)
, and becauseS Y S
Y + \ = , it results that the elements of S\Y are the first S\Y elements of
)
,
(
S
ρ
. Thus, according to the first part of the theorem, S \Y is inferior with respect to S, which according to proposition 1, implies thatS
\
(
S
\
Y
)
=
Y
is superior with respect toS.Remarks
6.1. Given
ρ
∈
{"
p
"
,”«”}, α∈(
S
,
ρ
)
, the set{
i
∈
(
S
,
ρ
)
/
i
ρ
α
}
is inferior with respect toS, but the set{
i
∈
(
S
,
ρ
)
/
α
ρ
i
}
is superior with respect toS, because the elements of these sets are the first, respectively the last elements of)
,
(
S
ρ
.6.2. In general, the previous theorem has no reciprocity. Thus, the set
)}
0
,
2
(
),
0
,
1
(
),
1
,
0
(
),
0
,
0
{(
=
X
is inferior with respect toS=
(
S
,
p
)
=(
T
22,
p
)
={(
0
,
0
),
(
0
,
1
),
(
1
,
0
),
(
0
,
2
),
(
1
,
1
),
(
2
,
0
)}
, and the four elements ofX
are not also the first four elements of(
T
22,
p
)
. AlsoX
=
{(
0
,
0
),
(
0
,
1
),
(
1
,
0
)}
is inferior with respect to S=(S,«)=(
T
22,
«)={(
0
,
0
),
(
0
,
1
),
(
0
,
2
),
(
1
,
0
),
(
1
,
1
),
(
2
,
0
)}
, and the three elements ofX
are not also the first ones of(
T
22,
«). The set)}
0
,
2
(
),
1
,
1
(
),
0
,
1
{(
=
6.3. According to observation 5.4, proposition 2.2 and theorem 2, if
S
=
(
S
,
ρ
)
is a total ordered set from
dIN
and if
S
1∪
S
2∪
...
∪
S
nis a disjoint
cover of
S⊂
IN
dso that the elements of
S
iare the first (last)
Sielements of
n i
S
S
∪
...
∪
for any
i
=
1
,
n
, then it follows that
S
k∪
S
lis inferior (superior)
with respect to
S
k∪
S
l∪
...
∪
S
nand that
S
k∪
...
∪
S
lis inferior (superior)
with respect to
S
k∪
...
∪
S
n, for any
1≤k ≤l ≤n. In the same conditions, we
have that the elements of the set
S
k∪
S
lare the first (last)
Sk ∪Slelements
of the set
S
k∪
S
l∪
...
∪
S
n, and the elements of the set
S
k∪
...
∪
S
lare the
first (last)
Sk ∪...∪Slelements of the set
S
k∪
...
∪
S
nfor any
1≤k ≤l ≤n.
REFERENCES
[1] Jia, Rong-Quing and Sharma, A., Solvability of some multivariate interpolation problems, 1990
[2] Rudolf A. Lorentz, Multivariate Birkhoff Interpolation, Springer – Verlag Berlin Heidelberg 1992