Hey there, let's work through your questions about the function ( F(x) = \mu(\text{supp}f_x) ) step by step, using concrete examples and standard measure theory arguments.

可积性：不一定成立，需额外假设 #

First off, ( F ) is not automatically integrable over ( X ). Let's build a simple counterexample:

• Take ( X = \mathbb{R} ) with the standard Borel σ-algebra, and let ( \mu ) be the Lebesgue measure.
• Define ( f_x(y) ) as a continuous function whose support is the open interval ( (x, x+1) ) (for example, a bump function centered at ( x+0.5 ) with width 1). Then ( \text{supp}f_x = (x, x+1) ), so ( F(x) = \mu((x, x+1)) = 1 ) for all ( x \in \mathbb{R} ).
• The integral ( \int_{\mathbb{R}} F(x) d\mu(x) = \int_{\mathbb{R}} 1 dx = \infty ), which is clearly not finite.

保证可积性的常见假设 #

To ensure ( F ) is integrable, you can add conditions on either ( X ), ( \mu ), or the family ( {f_x} ):

• Compact ( X ) + finite ( \mu ): If ( X ) is compact, then ( \text{supp}f_x \subseteq X ), so ( F(x) \leq \mu(X) < \infty ) for all ( x ). Since ( F ) is measurable (supports of continuous functions are open, hence Borel sets), it's a bounded measurable function on a finite measure space, so integrable.
• σ-finite ( \mu ) + bounded ( F ): If ( F(x) \leq C < \infty ) for all ( x ), and ( X ) can be covered by countably many sets of finite ( \mu )-measure, then ( F ) is integrable.
• Dominance condition: If there exists an integrable function ( g: X \to \mathbb{R}_{\geq 0} ) such that ( |F(x)| \leq g(x) ) for all ( x ), then ( F ) is integrable via the Dominated Convergence Theorem (though you'd first need to confirm ( F ) is measurable, which holds here since supports are Borel sets).

连续性：不一定成立，需集值映射的连续性 #

Just like integrability, ( F ) is not automatically continuous. Let's construct another counterexample:

• Take ( X = [0, 2] ) with the Borel σ-algebra, and let ( \mu = \delta_1 ), the Dirac measure concentrated at the point ( 1 ).
• Define the family ( {f_x} ):
  • For ( x \in [0,1) ), let ( f_x(y) ) be a continuous function with support ( [0,1) ) (e.g., a bump function on ( [0,1) ) that vanishes at ( y=1 )). Then ( F(x) = \delta_1(\text{supp}f_x) = 0 ).
  • For ( x = 1 ), let ( f_1(y) ) be a continuous function with support ( [0,2] ) (e.g., a constant function 1 on ( [0,2] )). Then ( F(1) = \delta_1([0,2]) = 1 ).
  • For ( x \in (1,2] ), let ( f_x(y) ) be a continuous function with support ( (1,2] ). Then ( F(x) = \delta_1(\text{supp}f_x) = 0 ).
• Here, ( F(x) ) jumps from 0 to 1 at ( x=1 ), so it's discontinuous at that point.

保证连续性的常见假设 #

To get continuity of ( F ), you need the supports ( \text{supp}f_x ) to "vary continuously" with ( x ), paired with regularity of ( \mu ):

• Hausdorff-continuous supports + regular ( \mu ): If the map ( x \mapsto \text{supp}f_x ) is continuous with respect to the Hausdorff distance on subsets of ( X ), and ( \mu ) is a regular Borel measure (e.g., a Radon measure on a locally compact Hausdorff space), then ( F(x) = \mu(\text{supp}f_x) ) will be continuous. Regularity ensures that ( \mu ) behaves nicely with respect to approximating sets, which pairs well with Hausdorff continuity.
• Compact-open continuity of ( {f_x} ): If the family ( {f_x} ) is continuous in the compact-open topology (i.e., for every compact ( K \subseteq X ) and open ( U \subseteq \mathbb{C} ), the set ( {x \in X : f_x(K) \subseteq U} ) is open), then the supports ( \text{supp}f_x ) tend to vary continuously. Combined with a regular ( \mu ), this implies ( F ) is continuous.
• Uniformly controlled supports: If for every ( x_0 \in X ), and every ( \epsilon > 0 ), there exists a neighborhood ( U ) of ( x_0 ) such that ( \mu(\text{supp}f_x \triangle \text{supp}f_{x_0}) < \epsilon ) for all ( x \in U ) (where ( \triangle ) denotes symmetric difference), then ( F ) is continuous at ( x_0 ).

备注：内容来源于stack exchange，提问作者s.h