• Keine Ergebnisse gefunden

Some examples of shade maps

Im Dokument detailed version (Seite 21-27)

As we already mentioned, Lemma 2.4 and its analogue for vertex-infection provide two examples of inclusion-preserving shade maps Shade. An example of an inclusion-reversing shade map comes from the theory of posets:

Example 4.9. Let E be a poset. For any F⊆E, we define F ={e∈ E | there exists an f ∈ Fwith e < f} and

ShadeF =E\F.

Then, this map Shade : P(E) → P(E) is an inclusion-reversing shade map.

Proof of Example 4.9 (sketched). We shall prove the following three statements:

Statement 0: If A and B are two subsets of E such that A ⊆ B, then ShadeB ⊆ShadeA.

Statement 1:IfF ∈ P(E)andu∈ E\ShadeF, then Shade(F∪ {u}) = ShadeF.

Statement 2:IfF ∈ P(E)andu∈ E\ShadeF, then Shade(F\ {u}) = ShadeF.

[Proof of Statement 0: Let A and B be two subsets of E such that A ⊆ B. We must show that ShadeB⊆ShadeA.

Indeed, A ⊆ B entails A ⊆ B (by the definitions of A and B). Hence, E\A ⊇ E\B, so that E\B ⊆ E\ A. But the definition of Shade yields ShadeA = E\A and ShadeB = E\B. Hence, ShadeB = E\B ⊆ E\A = ShadeA. This proves Statement 0.]

[Proof of Statement 1: Let F ∈ P(E) and u∈ E\ShadeF. We must prove that Shade(F∪ {u}) =ShadeF.

We haveu∈ E\ShadeF= F (since ShadeF =E\F). In other words, there exists a g∈ F with u< g(by the definition of F). Consider this g.

We must prove that Shade(F∪ {u}) = ShadeF. In other words, we must prove that(F∪ {u}) = F(since Shade(F∪ {u}) = E\(F∪ {u})and ShadeF= E\ F). Since F ⊆ (F∪ {u}) is rather obvious (by the definitions of F and (F∪ {u})), we only need to prove that (F∪ {u}) ⊆F.

So letv ∈ (F∪ {u}). We must show thatv ∈ F.

We have v ∈ (F∪ {u}). In other words, there exists an f ∈ F∪ {u} with v < f (by the definition of (F∪ {u})). Consider this f. If f ∈ F, then we immediately obtain v ∈ F (since v < f and f ∈ F), and thus our proof is complete. Hence, we WLOG assume that f ∈/ F. Hence, f = u (since f ∈ F∪ {u} but f ∈/ F) and thus v < f = u < g. This entails v ∈ F (by the definition of F, since g∈ F).

Forget that we fixed v. We thus have shown that v ∈ F for each v ∈ (F∪ {u}). In other words, (F∪ {u}) ⊆ F. As we explained above, this completes the proof of Statement 1.]

[Proof of Statement 2: Let F ∈ P(E) and u∈ E\ShadeF. We must prove that Shade(F\ {u}) =ShadeF.

We haveu∈ E\ShadeF= F (since ShadeF =E\F). In other words, there exists a g ∈ F with u < g (by the definition of F). Consider this g. Note that u< gentails g6=u, so that gF\ {u} (since g∈ F).

We must prove that Shade(F\ {u}) = ShadeF. In other words, we must prove that(F\ {u}) = F(since Shade(F\ {u}) = E\(F\ {u})and ShadeF = E\F). Since (F\ {u}) ⊆ F is rather obvious (by the definitions of F and (F\ {u})), we only need to prove that F ⊆(F\ {u}).

So letv ∈ F. We must show thatv∈ (F\ {u}).

We have v ∈ F. In other words, there exists an f ∈ F with v < f (by the definition of F). Consider this f. If f ∈ F\ {u}, then we immediately obtain v ∈ (F\ {u}) (since v < f and f ∈ F\ {u}), and thus our proof is complete.

Hence, we WLOG assume that f ∈/ F\ {u}. Hence, f = u (since f ∈ F but f ∈/ F\ {u}) and thus v < f = u < g. This entails v ∈ (F\ {u}) (by the definition of(F\ {u}), since g∈ F\ {u}).

Forget that we fixed v. We thus have shown that v ∈ (F\ {u}) for each v∈ F. In other words, F ⊆(F\ {u}). As we explained above, this completes the proof of Statement 2.]

Now, we have proved Statements 1 and 2. Thus, the map Shade : P(E) → P(E) satisfies the two axioms in Definition 4.2. In other words, this map is a shade map. Moreover, this map is inclusion-reversing (by Statement 0). This proves Example 4.9.

Another example of a shade map comes from discrete geometry:

Example 4.10. Let Abe an affine space overR. IfSis a finite subset of A, then anontrivial convex combinationofSwill mean a point of the form ∑

sS

λss∈ A, where the coefficientsλs are nonnegative reals smaller than 1 and satisfying

sS

λs =1.

Fix a finite subset Eof A. For any F ⊆E, we define

ShadeF={e ∈ E | e isnota nontrivial convex combination ofF}. Then, this map Shade : P(E) → P(E) is an inclusion-reversing shade map.

Proof of Example 4.10 (sketched). We shall prove the following three statements:

Statement 0: If X and Y are two subsets of E such that X ⊆ Y, then ShadeY⊆ShadeX.

Statement 1:IfF ∈ P(E)andu∈ E\ShadeF, then Shade(F∪ {u}) = ShadeF.

Statement 2:IfF ∈ P(E)andu∈ E\ShadeF, then Shade(F\ {u}) = ShadeF.

[Proof of Statement 0: Let X and Y be two subsets of E such that X ⊆ Y. We must prove that ShadeY⊆ShadeX.

Indeed, we have X ⊆ Y. Thus, any nontrivial convex combination of X is a nontrivial convex combination ofY. Therefore, anye∈ Ethat isnota nontrivial convex combination ofY also cannot be a nontrivial convex combination of X.

In other words, ShadeY ⊆ ShadeX (by the definition of Shade). This proves Statement 0.]

Assume the contrary. Thus, pis a nontrivial convex combination of F∪ {u}. In other words, p =

sF∪{u}

µss for some nonnegative reals µs smaller than 1 and satisfying ∑

sF∪{u}

µs =1. Consider these reals µs.

We haveu ∈ E\ShadeF. Hence, uEand u/ ShadeF. From u/ ShadeF, we see that u is a nontrivial convex combination of F. In other words, u =

fF

λff for some nonnegative realsλf smaller than 1 and satisfying ∑

fF Clearly, the coefficients µf +µuλf on the right hand side of this equality are nonnegative reals. Moreover, they are easily seen to be smaller than 1 13.

13Proof. Assume the contrary. Thus, µf +µuλf > 1 for some f F. Consider this f. Note that µf < 1 (since all coefficients µs are smaller than 1) and λf < 1 (for similar reasons).

Now,µf +µuλf >1 >µf (sinceµf <1), so thatµuλf >0. This entailsµuλf 6=0 and thus

Finally, they satisfy

Hence, (10) shows that p is a nontrivial convex combination of F. In other words, p ∈/ ShadeF. This contradicts p ∈ ShadeF. This contradiction shows that our assumption was false. This completes the proof of Statement 1.]

[Proof of Statement 2: Let F ∈ P(E) and u∈ E\ShadeF. We must prove that Shade(F\ {u}) =ShadeF. If u ∈/ F, then this is obvious (since F\ {u} = F in this case); thus, we WLOG assume thatu ∈ F.

We haveu ∈ E\ShadeF. Hence, u∈ Eand u ∈/ ShadeF. From u∈/ ShadeF, we see that u is a nontrivial convex combination of F. In other words, u =

fF

λff for some nonnegative realsλf smaller than 1 and satisfying ∑

fF

1−λu is a well-defined positive real.

We have

on the right hand side of this equality are non-negative reals (since 1−λu > 0). Moreover, they are easily seen to be smaller

Since all theµs are nonnegative, this subsum thus must be 6

s∈F∪{u} µf+µuλf >1. This contradiction shows that our assumption was wrong, qed.

than 1 14. Finally, they satisfy

14Proof. Assume the contrary. Thus, λf

1λu >1 for some f F\ {u}. Consider this f. From λf

1λu >1, we obtainλf >1λu (since 1λu >0), thusλf +λu>1.

However, f 6= u (since f F\ {u}), so that λf +λu is a subsum of the sum

g∈Fλg. Since all the λg are nonnegative, this subsum thus must be 6

g∈F

g∈Fλg (since all theλg are nonnegative) and therefore λf +λu+λh 6

g∈F

λg = 1 = λf +λu, so thatλh 6 0 and thereforeλh = 0 (sinceλhis a nonnegative real). Thus, we have shown that all elementsh Fexcept for (possibly) f and usatisfyλh=0. Thus, we had to rename the summation index f ashsince f means something else now.) Thus,

u=

h∈F

λhh=λff+λuu,

so that uλuu =λff. In view ofuλuu= (1λu)uand λf =1λu, this rewrites as (1λu)u = (1λu)f. We can cancel 1λu from this equality (since 1λu > 0), and thus obtainu = f. This contradicts f 6= u. This contradiction shows that our assumption was wrong, qed.

Shade(F\ {u}). Therefore,

Shade(F\ {u}) = Shade

(F\ {u})∪ {u}

| {z }

=F (sinceuF)

=ShadeF.

This proves Statement 2.]

Now, we have proved Statements 1 and 2. Thus, the map Shade : P(E) → P(E) satisfies the two axioms in Definition 4.2. In other words, this map is a shade map. Moreover, this map is inclusion-reversing (by Statement 0). This proves Example 4.10.

As a contrast to Example 4.10, let us mention a not-quite-example (satisfying only one of the two axioms in Theorem 4.3):

Example 4.11. LetVbe a vector space overR. IfSis a finite subset ofV, then anontrivial conic combination ofS will mean a vector of the form ∑

sS

λss∈ V, where the coefficientsλsare nonnegative reals with the property that at least two elementss ∈ Ssatisfy λs >0.

Fix a finite subset EofV. For any F⊆ E, we define

ShadeF={e ∈ E | e isnota nontrivial conic combination ofF}. It can be shown that this map Shade : P(E) → P(E) satisfies Axiom 1 in Definition 4.2. In general, it does not satisfy Axiom 2. Thus, it is not a shade map in general.

Im Dokument detailed version (Seite 21-27)