18.2. SEPARATION THEOREMS 489

Thus (xn,φ (xn)) /∈ B×(−∞, l +δ ) for all n. However, xn ∈ B for all n large enough. There-fore, for those values of n, it must be the case that φ (xn) /∈ (−∞, l +δ ) and so

lim infn→∞

φ (xn)≥ l +δ ≥ l

which shows, since l < φ (x) is arbitrary that

lim infn→∞

φ (xn)≥ φ (x) .

This proves the corollary.The following is a convenient fact which follows from the above.

Proposition 18.2.13 Let A be a linear operator which maps a real normed linear space(X ,∥·∥X ) to a real normed linear space (Y,∥·∥Y ) . Then xn→ x strongly implies Axn→ Axif and only if whenever xn→ x weakly, it follows that Axn→ Ax weakly.

Proof: ⇒ Define φ (x) ≡ f (Ax) where f ∈ Y ′. Then φ is convex and continuous.Therefore, if xn→ x weakly, then

φ (x) = f (Ax)≤ lim infn→∞

f (Axn) = lim infn→∞

φ (xn)

Then substituting −A for A,

− f (Ax)≤ lim infn→∞

f (−Axn) , f (Ax)≥ lim supn→∞

f (Axn)

which shows that for each f ∈ Y ′,

lim supn→∞

f (Axn)≤ f (Ax)≤ lim infn→∞

f (Axn)

and so the second condition holds.⇐ By the second condition, x→ f (Ax) satisfies the condition that if xn → x weakly,

thenf (Ax) = lim

n→∞f (Axn)

If A is not bounded, then there exists xn,∥xn∥ ≤ 1 but ∥Axn∥ ≥ n. It follows that xn/√

n→ 0and so A

(xn√

n

)→ 0 weakly. Therefore, A

(xn√

n

)is bounded contrary to the construction

which says that∥∥∥A(

xn√n

)∥∥∥≥√n. Since A is bounded, it must be continuous.

18.2.2 More Separation TheoremsThere are other separation theorems which can be proved in a similar way. The next theo-rem considers the separation of an open convex set from a convex set.

Theorem 18.2.14 Let A and B be disjoint, convex and nonempty sets with B open. Thenthere exists f ∈ X ′ such that

Re f (a)< Re f (b)

for all a ∈ A and b ∈ B.

18.2. SEPARATION THEOREMS 489Thus (xn, (%,)) ¢ Bx (—29,1 + 6) for all n. However, x, € B for all n large enough. There-fore, for those values of n, it must be the case that @ (x,) ¢ (—2°,/ + 6) and solim inf @(%,) >/+6>1n—cowhich shows, since / < @ (x) is arbitrary thatlim inf @ (x,) > @ (x).Nn—yooThis proves the corollary.The following is a convenient fact which follows from the above.Proposition 18.2.13 Let A be a linear operator which maps a real normed linear space(X, ||-||y) to a real normed linear space (Y,||-||y). Then x» — x strongly implies Ax, —> Axif and only if whenever X, — x weakly, it follows that Ax, — Ax weakly.Proof: = Define @ (x) = f (Ax) where f € Y’. Then @ is convex and continuous.Therefore, if x, — x weakly, theng (x) = f (Ax) < lim inf f (Ax,) = lim inf @ (x,)n—-eoo n—-ooThen substituting —A for A,—f (Ax) <lim inf f(—Ax,), f (Ax) > lim sup f (Ax)n—soowhich shows that for each f € Y’,lim sup f (Ax,) < f (Ax) < lim inf f (Axn)nooand so the second condition holds.< By the second condition, x + f (Ax) satisfies the condition that if x, + x weakly,then(Ax) = lim f(Axs)If A is not bounded, then there exists x;, |X, || < 1 but ||Ax,|| > 7. It follows that x, /./n > 0and so A (“s) — 0 weakly. Therefore, A a is bounded contrary to the constructionwhich says that A (2s) | > /n. Since A is bounded, it must be continuous. ff18.2.2 More Separation TheoremsThere are other separation theorems which can be proved in a similar way. The next theo-rem considers the separation of an open convex set from a convex set.Theorem 18.2.14 Let A and B be disjoint, convex and nonempty sets with B open. Thenthere exists f € X' such thatRe f (a) < Ref (b)forallac AandbeB.