Continuity and Topology Synthesis

From Local Continuity to Global Structure

Continuity can be expressed in more than one language. At a point, the epsilon–delta definition describes how nearby inputs produce nearby outputs. Topologically, continuity is tested by preimages: the Open-Set Criterion for Continuity says that preimages of open sets are open in the domain, and the Closed-Set Criterion for Continuity gives the corresponding test using closed sets. These are not separate notions of continuity; they are different ways to track the same relationship between inputs and outputs.

Compactness adds a global ingredient. A continuous image of a compact set is compact, and a continuous real-valued function on a nonempty compact set attains its minimum and maximum. Combining these facts with closed sets gives a useful conclusion that neither local continuity nor set avoidance alone supplies: if a compact image misses a closed set, then it misses that set by a positive distance. We will prove this separation result and then show that it makes avoidance stable under sufficiently small uniform changes.

Preimages in the Relative Topology

When a function is defined on a subset \(E\subseteq\mathbb{R}\), openness and closedness of a preimage are understood relative to \(E\). A set \(A\subseteq E\) is open in \(E\) if it has the form \(A=E\cap V\) for some open \(V\subseteq\mathbb{R}\). It is closed in \(E\) if its complement \(E\setminus A\) is open in \(E\). Thus, a preimage need not be open or closed in the whole real line in order to satisfy the relevant continuity criterion.

These examples illustrate a useful discipline: first solve the condition defining the preimage, then interpret the resulting set in the domain’s topology. In particular, one should not silently replace “open in \(E\)” with “open in \(\mathbb{R}\).” Relative openness is exactly what the continuity criterion requires when the function’s domain is \(E\).

Compactness Creates Positive Separation

Suppose \(f(K)\) avoids a closed set \(F\). Avoidance alone says that no output belongs to \(F\); it does not say how close outputs may get to \(F\). On a compact domain, continuity rules out arbitrarily close approaches without an actual intersection. The next theorem expresses that fact quantitatively.

Proof. By the Continuous Image of a Compact Set Theorem, \(C=f(K)\) is compact. For \(y\in\mathbb{R}\), define its distance from \(F\) by

This infimum is finite because \(F\) is nonempty. For any \(u,v\in\mathbb{R}\) and \(z\in F\), the triangle inequality gives \(|u-z|\leq|u-v|+|v-z|\). Taking the infimum over \(z\in F\) yields \(d(u,F)\leq|u-v|+d(v,F)\). Interchanging \(u\) and \(v\) then shows

Thus \(y\mapsto d(y,F)\) is continuous. By the Extreme Value Theorem, it attains a minimum on the nonempty compact set \(C\), say at \(y_0\in C\). Since \(C\cap F=\varnothing\), \(y_0\notin F\). Because \(F\) is closed, its complement is open, so there is an \(r>0\) such that \((y_0-r,y_0+r)\cap F=\varnothing\). Consequently \(|y_0-z|\geq r\) for every \(z\in F\), and \(d(y_0,F)\geq r>0\). Set \(\eta=d(y_0,F)\). Since \(\eta\) is the minimum of \(d(y,F)\) over \(C\), every \(y\in C\) and \(z\in F\) satisfy \(|y-z|\geq d(y,F)\geq\eta\). Taking \(y=f(x)\) proves the claim. \(\square\)

The proof combines topology and metric information in a specific order. Compactness makes the image compact; the distance function is continuous; and the minimum distance is attained. Closedness of \(F\) ensures that this attained distance is positive at each point of the image. Without compactness, the infimum of the distances could be zero without any image point belonging to \(F\).

Uniform Perturbations Preserve Avoidance

Positive separation has a stability consequence. Once a compact image stays a definite distance from a closed set, a second function that remains uniformly close to the first cannot reach that closed set. The perturbed function need not be continuous for this conclusion; the uniform error bound is the relevant condition.

Proof. Fix \(x\in K\). For any \(z\in F\), the triangle inequality gives

Thus \(g(x)\neq z\) for every \(z\in F\), so \(g(x)\notin F\). This holds for each \(x\in K\); hence \(g(K)\cap F=\varnothing\). \(\square\)

To use this result when \(f\) is continuous on a compact set, first apply Positive Separation from a Closed Set to obtain \(\eta\). The stability theorem then allows any uniform error smaller than \(\eta\). The strict inequality matters: an error bound equal to the separation may be large enough to move a value onto \(F\).

Why the Compactness Hypothesis Matters

Continuity by itself does not guarantee positive separation. For example, define \(f:(0,1)\to\mathbb{R}\) by \(f(x)=x\), and let \(F=\{0\}\). The function is continuous and its image avoids \(F\), but the distances \(|f(x)-0|=x\) have infimum zero. For each \(\eta>0\), there is an \(x\in(0,1)\) with \(x<\eta\), so no positive uniform lower bound exists. The domain is bounded, but it is not compact: it omits its limit point \(0\).

This example separates three claims that are easy to conflate: every value avoids a set, every value has positive distance from the set, and all values have one common positive lower bound on that distance. The first does not imply the third on an arbitrary domain. Compactness and continuity together supply the uniform conclusion in the theorem.

The synthesis is useful in problems about zeros, signs, and perturbations. If a continuous function on a compact domain has no zero, positive separation from \(\{0\}\) gives a uniform lower bound on its absolute value. If its range avoids a closed interval or another closed set, the same reasoning supplies a uniform gap. Conversely, if compactness is missing, a sequence of inputs can drift toward an omitted boundary point while the outputs approach the forbidden set.