Choose a Theorem by Matching Its Job to the Proof
The Covering Strategy showed how compactness can turn local neighborhoods into finite or uniform control. The next question is practical: when a problem offers several plausible tools, how should you decide which one to use? The answer begins with the conclusion you need. A proof should be organized around the form of that conclusion, not around a theorem whose name happens to look relevant.
For example, a conclusion about every point may invite a direct argument from the definition. A conclusion that something is false may be easier to prove by its contrapositive or by contradiction. If a sequence is involved and the desired conclusion concerns a limit, subsequences may expose an obstruction. If local neighborhoods or point-dependent estimates appear, a covering argument may be the right way to obtain finite or uniform control. These are not competing recipes: each addresses a different logical shape.
A useful first pass is to separate the statement into two parts: what is assumed, and precisely what must be shown. Then ask whether the assumptions already match a theorem from earlier in the course. If they do, cite and apply it. If they do not, identify the missing hypothesis before trying to force the theorem to fit. In particular, compactness can convert local information into a global conclusion, but it cannot supply a property that the local information does not establish.
Is it a pointwise assertion, a uniform estimate, existence of a limit, or a statement that a proposed conclusion cannot occur?
Look for compactness, continuity, a sequence, a dense set, local neighborhoods, or a sign or order condition.
Use a theorem whose hypotheses are actually present; choose a proof strategy that addresses the remaining logical gap.
Pointwise control need not be uniform, local estimates need not be global, and a theorem’s conclusion may be weaker than the problem asks for.
A Compactness Test for Local Information
The covering approach is particularly useful when a property is available near each point, but the desired conclusion must hold across an entire compact space. The first theorem gives a simple test case: local boundedness supplies an open cover, and compactness reduces that cover to finitely many neighborhoods.
Proof. If \(K=\varnothing\), the function is bounded vacuously. Suppose \(K\neq\varnothing\). For each \(x\in K\), choose a neighborhood \(U_x\) and a finite bound \(M_x\) as in the definition. The family \(\{U_x:x\in K\}\) is an open cover of \(K\), so compactness gives a finite subcover \(U_{x_1},\ldots,U_{x_m}\). Define \(M=\max\{M_{x_1},\ldots,M_{x_m}\}\), which is finite because it is the maximum of a nonempty finite set of finite numbers. Given any \(y\in K\), choose \(i\) with \(y\in U_{x_i}\). Then \(|f(y)|\leq M_{x_i}\leq M\). Thus \(|f(y)|\leq M\) for every \(y\in K\), as required. \(\square\)
Notice what the proof does and does not use. It needs neither continuity nor a single bound that works on every neighborhood at the outset. It uses compactness to select finitely many local bounds and then takes their maximum. This is a useful theorem-selection pattern: when the conclusion asks for one global bound and the assumptions provide bounds on neighborhoods, check for compactness before attempting a pointwise estimate over the whole space.
Worked Example: Bounding a Discontinuous Function on a Compact Interval
On \(K=[0,1]\), define \(f(x)=1\) when \(x\) is rational and \(f(x)=0\) when \(x\) is irrational. This function is not continuous, so a theorem requiring continuity would not be a suitable choice. But for every \(x\in K\), take \(U_x=K\) and \(M_x=1\). Then \(|f(y)|\leq1\) for every \(y\in U_x\), so \(f\) is locally bounded. Since \(K\) is compact, the Local Boundedness on a Compact Space theorem applies and gives a global bound.
In fact, the local bounds in this example already use the same number everywhere. That makes the conclusion immediate, but the theorem is more useful when the available bound depends on the neighborhood and no common bound is apparent. The essential step is not continuity; it is the finite-subcover reduction.
From Local Lipschitz Estimates to One Global Estimate
A stronger local estimate can also become global on a compact metric space. The result below is a useful guide when a problem gives a different Lipschitz constant in different neighborhoods. The covering strategy handles small distances; boundedness handles pairs that are not close.
Proof. If \(K=\varnothing\), the global inequality is vacuous. Suppose \(K\neq\varnothing\). First we show that \(f\) is bounded. For each \(x\in K\), intersect the neighborhood in the local Lipschitz condition with \(B_K(x,1)\), and call the resulting neighborhood \(V_x\). For \(y\in V_x\), the local estimate gives
Thus \(f\) is locally bounded. By the Local Boundedness on a Compact Space theorem, there is a finite \(M\geq0\) such that \(|f(y)|\leq M\) for every \(y\in K\).
The neighborhoods \(U_x\) from the local Lipschitz condition cover \(K\). Compactness gives a finite subcover \(U_{x_1},\ldots,U_{x_m}\). Apply the Lebesgue Number Lemma to this cover: there is a \(\delta>0\) such that every nonempty subset of \(K\) with diameter less than \(\delta\) is contained in some \(U_{x_i}\). Put \(L_0=\max\{L_{x_1},\ldots,L_{x_m}\}\).
Take any \(u,v\in K\). If \(d(u,v)<\delta\), the set \(\{u,v\}\) has diameter less than \(\delta\), so it lies in some \(U_{x_i}\). The local estimate there yields
If \(d(u,v)\geq\delta\), boundedness gives
Therefore both cases satisfy \(|f(u)-f(v)|\leq Ld(u,v)\) for \(L=\max\{L_0,2M/\delta\}\). This is one finite constant valid for every pair in \(K\), so \(f\) is globally Lipschitz. \(\square\)
The proof illustrates why theorem choice often involves more than spotting one hypothesis. The Lebesgue Number Lemma puts sufficiently close pairs inside one neighborhood, where a local estimate applies. It says nothing about distant pairs. The separate boundedness argument controls those pairs. Omitting either part would leave a gap: local Lipschitz estimates alone only apply when both points lie in a suitable neighborhood.
Worked Example: A Local Estimate for the Cubic Function
Let \(K=[-1,1]\) with the usual distance and \(f(x)=x^3\). For any \(x\in K\), take \(U_x=K\) and \(L_x=3\). For \(u,v\in K\), factor the difference:
because \(|u^2|\leq1\), \(|uv|\leq1\), and \(|v^2|\leq1\). Thus the local Lipschitz hypothesis holds, and the compact-space theorem guarantees a global Lipschitz estimate. Here the calculation already supplies the global constant \(3\); the theorem is useful as a general method when only neighborhood-by-neighborhood estimates are available.
Worked Example: Why the Compactness Hypothesis Matters
On \(\mathbb{R}\), let \(f(x)=x^2\). For each \(x\in\mathbb{R}\), the neighborhood \(U_x=(x-1,x+1)\) gives a local Lipschitz estimate. If \(u,v\in U_x\), then \(|u|<|x|+1\) and \(|v|<|x|+1\), so
The constants depend on the neighborhood, as permitted by local Lipschitz continuity. But there is no global Lipschitz constant. For each positive integer \(n\), take \(u=2n\) and \(v=n\). Then
Since \(3n\) is unbounded as \(n\) increases, no single finite constant can bound these ratios. The domain is not compact, so the compact-space theorem does not apply. This example helps diagnose a common error: local estimates, even when valid at every point, do not by themselves provide one estimate for all pairs.
Common Selection Errors
A theorem should be chosen because its hypotheses and conclusion fit the problem, not because its name seems close. Several recurring checks help prevent mismatches:
- Pointwise versus uniform: A bound or radius that can vary with the point is not yet one common bound or radius. Look for an additional hypothesis, such as compactness, that justifies uniformity.
- Local versus global: A local estimate applies only to points in the neighborhood where it was proved. Explain how relevant points enter one such neighborhood, or handle pairs outside that range separately.
- Missing assumptions: If a theorem requires compactness, continuity, completeness, or boundedness, verify that the problem supplies it. Do not infer a missing hypothesis from the desired conclusion.
- Empty cases: If a proof selects a point, takes a maximum over a finite subcover, or chooses a pair of points, first ensure the relevant set is nonempty.
A useful final check is to ask whether each line of the proposed proof answers one of two questions: why does the chosen theorem apply, and how does its conclusion give exactly what is required? If a step quietly changes “for each point” into “for all points with one constant,” the proof needs an explicit uniformity argument. If a compactness theorem is available, an open cover and finite subcover may provide precisely that argument; otherwise, a counterexample may reveal that the desired conclusion is false.
Check Your Understanding
For each question, focus on the match between the assumptions, the conclusion, and the tool being used.
- Why does a finite subcover allow the local bounds in the Local Boundedness on a Compact Space theorem to be replaced by one bound?
- In the local Lipschitz theorem, what role does the Lebesgue Number Lemma play, and which pairs does it help control?
- Why does the local Lipschitz proof need a separate estimate for pairs whose distance is at least the Lebesgue number?
- Which hypothesis fails when the function \(f(x)=x^2\) on \(\mathbb{R}\) is used to test the compact-space theorem?
- When a problem asks for one constant valid everywhere but gives point-dependent estimates, what should you check before choosing a proof strategy?