CAPITULO VI Estudio Legal y Organizacional
6.1 Estudio Legal
6.1.2. Registro de marcas y patentes
Now, suppose that none of the measuresµkorνkis purely atomic. As noted in Section 4.2.1, we
can restrict our attention to functions supported where every dνk
dµk >0, since non-zero values off that set only increaseρ(f), notξ(f), and thus take us further fromC= supf∈L+(X)
ξ(f)
ρ(f). This can be effected by restricting our measures, after which we have dνk
dµk > 0µk-a.e. and, consequently, either bothµk andµk are purely atomic or neither of them is.
In this case, it is not only necessary that every two-variable subinclusion hold to have
Lρ ⊂ Lξ, but also sufficient. Furthermore, as we learned from studying the two-variable case, with no purely atomic measure, this occurs when the exponents satisfy a condition based on Minkowski’s inequality.
Theorem 7.6.1. Suppose that, for each k ∈ {1, . . . ,n}, neitherµk norνk is purely atomic and
that there is a constant Ck < ∞such thatξk ≤Ckρk. Then the following are equivalent.
1. For any1≤i< j≤ n there exists some constant C0< ∞such that
ξj◦ξi ≤C0ρτ(j)◦ρτ(i)
whereτis the identity ifσ−1(i)< σ−1(j)orτswaps i and j ifσ−1(i)> σ−1(j).
2. For any1≤i< j≤ n, ifσ−1(i)> σ−1(j)then qi ≤ pj.
3. There is a finite C< ∞such that
ξ≤Cρ.
Furthermore, if these are true, then the least possible values of C and C1, . . . ,Cn satisfy
C = C1· · ·Cn.
Proof. Suppose the first condition holds and we have every two-variable inclusion. In every case where σ−1(i) > σ−1(j) while i < j, the two-variable inclusion is permuted, so Theorem 4.2.16 provides that min(pi,qi) ≤ max(pj,qj). Furthermore, Theorem 2.1.8 provides thatqi ≤
piandqj ≤ pj, so that the Minkowski criterion on exponents isqi ≤ pj. This proves the second
condition.
Assuming the second condition, we have the Minkowski sufficient condition described in Section 3.6, so every permuted (non-identityτ) inclusion in the first condition holds. Of course, whenτis the identity, unpermuted inclusion follows from single-variable inclusions, by Theo- rem 3.4.1. Therefore we have every two-variable inclusion required for the first condition.
Corollary 7.4.5 says that the third condition implies the first, so all that remains is to es- tablish, with the assumed single-variable inclusions and atomless measures, that the second condition implies the third. The aim is to prove that all subinclusions (each corresponding to someS ⊂ {1, . . . ,n}) hold, by induction on the size of the subset. While this is a superficially stronger result, Theorem 7.4.2 tells us that it is actually equivalent to the overall inclusion.
Claim: For any S ⊂ {1, . . . ,n}, let s = |S|and writeS = {k1, . . .ks} = {l1, . . . ,ls}, where
k1 <· · ·< ksandσ−1(l1)<· · · < σ−1(ls). Then the least constantC ∈[0,∞] such that
7.6. Minkowski criterion with no purely atomic measure 99
isC =Ck1· · ·Cks.
The proof is by induction on s. The base case s = 1 is in the hypothesis. Assume that all subinclusions for subsetsS with|S|< shold, each with the specified constant, the product of all single-variable constants. (The case of factorable functions fk1· · · fks, each fki ∈ L+(Xki), givesC ≥Ck1· · ·Cks, so it remains only to establish (7.9) withC=Ck1. . .Cks .)
Define a permutation τon {1, . . . ,s} by, for each i, j ∈ {1, . . . ,s}, τ(i) = j if and only if
ki =lj. Leti0= τ−1(1), soki0 = l1. The single-variable inclusionξl1 ≤Cl1ρl1 implies that
ξki0 ◦ · · · ◦ξk1 ≤Cl1ρl1 ◦ξki0−1 ◦ · · · ◦ξk1. (7.10)
For anyi< i0,τ(i)> 1=τ(i0), soli occurs afterli0 in thelordering ofS, i.e.σ
−1(i)> σ−1(i 0). By the second condition, qki ≤ pki0 = pl1. Repeatedly apply Minkowski’s integral inequality
(Proposition 7.2.3) to alli< i0.
ρl1 ◦ξki0−1 ◦ · · · ◦ξk1 ≤ξki0−1 ◦ρl1 ◦ξki0−2 ◦ · · · ◦ξk1
...
≤ξki0−1 ◦ · · · ◦ξk1 ◦ρl1
Composition withξks ◦ · · · ◦ξki0+1 gives
ξks ◦ · · · ◦ξki0+1 ◦ρl1 ◦ξki0−1 ◦ · · · ◦ξk1 ≤ ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1 ◦ρl1,
whereξcki0 indicates that this term is missing from the composition. With (7.10), this means that
ξks ◦ · · · ◦ξk1 ≤Cl1ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1 ◦ρl1. (7.11)
Apply the inductive hypothesis toS \ {l1}=S \ki0 to find that
ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1 ≤Cls· · ·Cl2ρls ◦ · · · ◦ρl2
which, together with (7.11), gives
ξks ◦ · · · ◦ξk1 ≤Cl1ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1 ◦ρl1 ≤Cl1· · ·Clsρls ◦ · · · ◦ρl1,
(7.12)
as desired.
The case of nonatomic measures is not the only time when the multivariable problem re- duces to two-variable subproblems. There is a similar result involving counting measure.
Theorem 7.6.2. Suppose that, for each k ∈ {1, . . . ,n}, both of µk and νk are (unweighted)
counting measure and that there is a constant Ck < ∞ such that ξk ≤ Ckρk. (When this
happens, the least constant is Ck = 1.) Then the following are equivalent.
1. For any1≤i< j≤ n there exists some constant C0< ∞such that
ξj◦ξi ≤C0ρτ(j)◦ρτ(i),
whereτis the identity ifσ−1(i)< σ−1(j)orτswaps i and j ifσ−1(i)> σ−1(j).
2. For any1≤i< j≤ n, ifσ−1(i)> σ−1(j)then pi ≤qj.
3. There is a finite C< ∞such that
ξ≤Cρ. (When this is true, the least constant is C= 1.)
Proof. The proof proceeds much as in Theorem 7.6.1. One difference to note is that each single-variable inclusion`pk(X
k)⊂`qk(Xk) holds if and only if pk ≤ qk, and in this caseCk = 1,
by Corollary 2.6.10.
Assume the first condition. Wheneveri< jandσ−1(i)> σ−1(j), the permuted two-variable inclusion requires pi ≤qj by Proposition 5.3.5. This proves the second condition.
Assume the second condition. The hypothesis provides one-variable inclusions and, con- sequently, pi ≤ qi and pj ≤ qj. Theorem 3.4.1 shows that the one-variable inclusions suffice
in the first condition’s unpermuted case, withC0 = CiCj = 1. The permuted case, with non-
identityτ, has σ−1(i) > σ−1(j) and therefore pi ≤ qj. Proposition 5.3.5 then gives permuted
two-variable inclusion withC0 = 1, finishing the proof of the first condition.
Subinclusions are necessary by Corollary 7.4.5, so the third condition implies the first. It remains only to prove that the second condition implies the third. As in Theorem 7.6.1, the following claim is established by induction on s, assuming the second condition.
Claim: For any S ⊂ {1, . . . ,n}, let s = |S|and writeS = {k1, . . .ks} = {l1, . . . ,ls}, where
k1 <· · ·< ksandσ−1(l1)<· · · < σ−1(ls). Then the least constantC ∈[0,∞] such that
ξks ◦ · · · ◦ξk1 ≤Cρls ◦ · · · ◦ρl1 (7.13)
isC =1.
The hypothesis includes the base case s = 1. Assume that all subinclusions for subsetsS
with|S|< shold with constant 1. (Factorable functions giveC ≥Ck1· · ·Cks =1.)
Define a permutation τon {1, . . . ,s} by, for each i, j ∈ {1, . . . ,s}, τ(i) = j if and only if
ki =lj. Leti0= τ−1(s), soki0 =ls. The single-variable inclusionξls ≤ ρls implies that
ξks◦ · · · ◦ξki0 ≤ξks ◦. . . ξki0+1 ◦ρls. (7.14) For any j > i0,τ(j) < n = τ(i0), solj occurs beforeli0 in thelordering ofS, i.e. σ
−1(j) <
σ−1(i
0). By the second condition, pls = pki0 ≤ qkj. Repeatedly apply Minkowski’s integral inequality (Proposition 7.2.3) to all j> i0.
ξks◦. . . ξki0+1 ◦ρls ≤ξks ◦. . . ξki0+2 ◦ρls ◦ξki0+1
...
≤ρls ◦ξks ◦ · · · ◦ξki0+1
Composition withξki0−1 ◦ · · · ◦ξ1gives
ξks ◦ · · · ◦ξki0+1 ◦ρls ◦ξki0−1 ◦ · · · ◦ξk1 ≤ ρls ◦ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1,
whereξcki0 indicates that this term is missing from the composition. With (7.14), this means that
7.7. Multiple-variable summary 101
Apply the inductive hypothesis toS \ {ls}= S \ki0 to find that
ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1 ≤ ρls−1 ◦ · · · ◦ρl1
which, together with (7.15), gives
ξks ◦ · · · ◦ξk1 ≤ρls ◦ξks ◦ · · · ◦ξcki0 ◦ · · · ◦ξk1 ≤ρls ◦ · · · ◦ρl1.
(7.16)
In each of these cases, the multivariable problem has been reduced to two-variable subprob- lems. Where these two-variable subproblems are permuted, inclusion in either case requires the Minkowski criterion. Additionally, each one-variable subproblem requires a particular or- der of exponents. For measures which are not purely atomic, one-variable inclusions demand
qk ≤ pk so that the H¨older criterion for inclusion can apply; with this, the Minkowski case of
min(pi,qi)≤ max(pj,qj) is reduced toqi ≤ pj.. For unweighted`p, pk ≤ qk is both necessary
and sufficient, and it reduces the Minkowski case to pi ≤ qj. Without these special conditions,
it is not clear that just having the Minkowski condition for each permuted two-variable subin- clusion suffices to establish the full inclusion. However, this has been proven to suffice in all cases with three or four variables, and computation has furthermore ruled out a counterexample in five or six variables. Therefore, the following conjecture is suggested.
Conjecture 7.6.3. For each k ∈ {1, . . . ,n}, let Ck ∈[0,∞]denote the least constant so thatξk ≤
Ckρk. Assume that every permuted two-variable subinclusion satisfies the Minkowski criterion.
That is, assume that for any1 ≤ i < j ≤ n, ifσ−1(i) > σ−1(j), thenmin(p
i,qi) ≤ max(pj,qj).
Then the least constant C such that
ξ≤Cρ is C =C1· · ·Cn.