TheoremBase

Proof

Throughout, xγx_{\gamma} denotes the γ\gamma-th component of a point x∈Rlx\in\mathbb{R}^{l}, the components determining the point by claim 1 of Euclidean Points as Tuples of Real Numbers, and ∣x∣=(∑γ=1lxγ2)1/2|x|=\bigl(\sum_{\gamma=1}^{l}x_{\gamma}^{2}\bigr)^{1/2} is the Euclidean norm. By Probability Simplex, a point x∈Rlx\in\mathbb{R}^{l} lies in Δl\Delta^{l} exactly when xγ≥0x_{\gamma}\ge0 for every γ\gamma and ∑γ=1lxγ=1\sum_{\gamma=1}^{l}x_{\gamma}=1.

Claim 1. Let ee be the point with e1=1e_{1}=1 and eγ=0e_{\gamma}=0 for γ≥2\gamma\ge2, which exists by claim 2 of Euclidean Points as Tuples of Real Numbers. Then e∈Δle\in\Delta^{l}, so Δl\Delta^{l} is nonempty.

Let x∈Δlx\in\Delta^{l} and fix γ\gamma. Every other component is nonnegative, so xγ=1−∑δ≠γxδ≤1x_{\gamma}=1-\sum_{\delta\neq\gamma}x_{\delta}\le1 by claim 3 of Elementary Order Arithmetic in an Ordered Field in the nonstrict form obtained by adjoining the case of equality; thus 0≤xγ≤10\le x_{\gamma}\le1. Multiplying the inequality xγ≤1x_{\gamma}\le1 by the nonnegative number xγx_{\gamma}, using claim 10 of Elementary Order Arithmetic in an Ordered Field in the nonstrict form obtained by adjoining the case of equality, gives xγ2≤xγx_{\gamma}^{2}\le x_{\gamma}. Summing over γ\gamma and using claim 3 of Elementary Order Arithmetic in an Ordered Field repeatedly, again in the nonstrict form obtained by adjoining the case of equality,

∣x∣2=∑γ=1lxγ2≤∑γ=1lxγ=1,|x|^{2}=\sum_{\gamma=1}^{l}x_{\gamma}^{2}\le\sum_{\gamma=1}^{l}x_{\gamma}=1 ,

and therefore ∣x∣≤1|x|\le1. Indeed, suppose instead that 1<∣x∣1<|x|. From 0≤10\le1 and 1<∣x∣1<|x|, claim 2 of Elementary Order Arithmetic in an Ordered Field gives 0<∣x∣0<|x|, so claim 10 of that lemma, applied to the inequality 1<∣x∣1<|x| with the positive multiplier ∣x∣|x|, gives ∣x∣<∣x∣2|x|<|x|^{2}; combining 1<∣x∣1<|x| with ∣x∣<∣x∣2|x|<|x|^{2} by claim 2 again gives 1<∣x∣21<|x|^{2}, contradicting ∣x∣2≤1|x|^{2}\le1. Consequently every point of Δl\Delta^{l} lies within distance 11 of the origin, so Δl\Delta^{l} is bounded in (Rl,dRl)(\mathbb{R}^{l},d_{\mathbb{R}^{l}}).

For closedness we show that the complement of Δl\Delta^{l} is open in (Rl,dRl)(\mathbb{R}^{l},d_{\mathbb{R}^{l}}); by Closed Subset of a Topological Space this is exactly what is required. Let y∈Rl∖Δly\in\mathbb{R}^{l}\setminus\Delta^{l}. Then either some component of yy is negative, or all are nonnegative and s=∑γyγ≠1s=\sum_{\gamma}y_{\gamma}\neq1.

In the first case fix γ\gamma with yγ<0y_{\gamma}<0 and put r=−yγ>0r=-y_{\gamma}>0. If ∣z−y∣<r|z-y|<r then ∣zγ−yγ∣≤∣z−y∣<r|z_{\gamma}-y_{\gamma}|\le|z-y|<r by claim 4 of Elementary Properties of the Euclidean Norm on Rn\mathbb{R}^n, hence ∣zγ−yγ∣<r|z_{\gamma}-y_{\gamma}|<r by claim 2 of Elementary Order Arithmetic in an Ordered Field. The strict two-sided bound, claim 9 of Properties of the Absolute Value in an Ordered Field, then gives zγ−yγ<rz_{\gamma}-y_{\gamma}<r, and claim 1 of Elementary Order Arithmetic in an Ordered Field gives zγ<yγ+r=0z_{\gamma}<y_{\gamma}+r=0. Since every component of a point of Δl\Delta^{l} is nonnegative, the strict inequality zγ<0z_{\gamma}<0 gives z∉Δlz\notin\Delta^{l}.

In the second case put r=∣s−1∣/l>0r=|s-1|/l>0. Suppose ∣z−y∣<r|z-y|<r. By claim 4 of Elementary Properties of the Euclidean Norm on Rn\mathbb{R}^n each ∣zγ−yγ∣≤∣z−y∣|z_{\gamma}-y_{\gamma}|\le|z-y|, so by the triangle inequality for the absolute value, claim 5 of Properties of the Absolute Value in an Ordered Field, applied l−1l-1 times,

∣∑γzγ−s∣=∣∑γ(zγ−yγ)∣≤∑γ∣zγ−yγ∣≤l ∣z−y∣<l r=∣s−1∣.\Bigl|\sum_{\gamma}z_{\gamma}-s\Bigr|=\Bigl|\sum_{\gamma}(z_{\gamma}-y_{\gamma})\Bigr|\le\sum_{\gamma}|z_{\gamma}-y_{\gamma}|\le l\,|z-y|<l\,r=|s-1| .

If zz were in Δl\Delta^{l} we would have ∑γzγ=1\sum_{\gamma}z_{\gamma}=1 and hence ∣∑γzγ−s∣=∣1−s∣=∣s−1∣\bigl|\sum_{\gamma}z_{\gamma}-s\bigr|=|1-s|=|s-1| by claim 4 of Properties of the Absolute Value in an Ordered Field, contradicting the strict inequality just obtained. So z∉Δlz\notin\Delta^{l}.

In either case the open ball of radius rr about yy misses Δl\Delta^{l}, so the complement of Δl\Delta^{l} is open and Δl\Delta^{l} is closed.

Claim 2. By claim 1 and the implication from condition 2 to condition 1 in Heine-Borel Theorem in Rn\mathbb{R}^n, Δl\Delta^{l} is a compact subset of Rl\mathbb{R}^{l} for the topology determined by the Euclidean distance. By claim 2 of The Restriction of a Metric to a Subset Induces the Subspace Topology the subsets of Δl\Delta^{l} open in (Δl,dΔ)(\Delta^{l},d_{\Delta}) are exactly the sets Δl∩U\Delta^{l}\cap U with UU open in (Rl,dRl)(\mathbb{R}^{l},d_{\mathbb{R}^{l}}), that is, the topology of (Δl,dΔ)(\Delta^{l},d_{\Delta}) is the subspace topology inherited from Rl\mathbb{R}^{l}; so by Compact Topological Space and Compact Subset compactness of the subset Δl\Delta^{l} of Rl\mathbb{R}^{l} and compactness of the space (Δl,dΔ)(\Delta^{l},d_{\Delta}) are the same assertion. Hence (Δl,dΔ)(\Delta^{l},d_{\Delta}) is compact, and Δl\Delta^{l} is a sequentially compact subset of (Δl,dΔ)(\Delta^{l},d_{\Delta}) by Compactness and Sequential Compactness Agree for Subsets of a Metric Space. Separability of (Δl,dΔ)(\Delta^{l},d_{\Delta}) is claim 2 of A Totally Bounded Metric Space is Separable, applied to the compact metric space (Δl,dΔ)(\Delta^{l},d_{\Delta}).

Claim 3. By claim 3 of The Set of Controls with Values in a Compact Convex Set is Weakly Metrizable and Compact, UA\mathcal{U}_{\mathcal{A}} is a sequentially compact and a compact subset of (UA,ρ)(\mathcal{U}_{\mathcal{A}},\rho); since UA\mathcal{U}_{\mathcal{A}} is the whole underlying set of that metric space, the space (UA,ρ)(\mathcal{U}_{\mathcal{A}},\rho) is itself compact and sequentially compact. Separability is then claim 2 of A Totally Bounded Metric Space is Separable.

Claim 4. By claim 2 the set Δl\Delta^{l} is sequentially compact in (Δl,dΔ)(\Delta^{l},d_{\Delta}) and by claim 3 the set UA\mathcal{U}_{\mathcal{A}} is sequentially compact in (UA,ρ)(\mathcal{U}_{\mathcal{A}},\rho), so by claim 2 of Coordinatewise Convergence, Sequential Compactness and Density in a Product Metric Space the product X=Δl×UAX=\Delta^{l}\times\mathcal{U}_{\mathcal{A}} is sequentially compact in (X,dX)(X,d_{X}); being the whole underlying set, it makes (X,dX)(X,d_{X}) a sequentially compact space, and a compact one by Compactness and Sequential Compactness Agree for Subsets of a Metric Space.

Let DΔD_{\Delta} and DUD_{\mathcal{U}} be as in the statement. By claim 3 of Coordinatewise Convergence, Sequential Compactness and Density in a Product Metric Space the set DΔ×DUD_{\Delta}\times D_{\mathcal{U}} is dense in XX for the topology of subsets open in (X,dX)(X,d_{X}), and it is countable. Such DΔD_{\Delta} and DUD_{\mathcal{U}} exist by the separability assertions of claims 2 and 3, so (X,dX)(X,d_{X}) is separable. ■\blacksquare

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…