Uniform boundedness explains why pointwise control of a whole family of operators is often stronger than it first appears. The theorem is a typical Baire-category result: completeness turns scalar boundedness into operator-norm boundedness.
Pointwise bounded families
Pointwise boundedness controls each vector separately. The theorem asks whether completeness upgrades this to uniform control over the unit ball.
Theorem. Let \(X\) be Banach, \(Y\) normed, and \(\mathcal T\subset\mathcal L(X,Y)\). If \(\sup_{T\in\mathcal T}\|Tx\|<\infty\) for every \(x\in X\), then \(\sup_{T\in\mathcal T}\|T\|<\infty\).
Proof. For \(n\in\mathbb N\), define \(E_n=\{x\in X:\sup_{T\in\mathcal T}\|Tx\|\le n\}\). Each \(E_n\) is closed because it is an intersection of closed sets \(\{x:\|Tx\|\le n\}\), and \(\bigcup_nE_n=X\) by pointwise boundedness.
Baire gives some \(E_N\) with nonempty interior. Thus there exist \(x_0\) and \(r>0\) such that \(B(x_0,r)\subset E_N\). Since \(x_0\in E_N\), for \(\|h\|<r\) both \(x_0+h\) and \(x_0\) lie in \(E_N\), hence \(\|Th\|\le\|T(x_0+h)\|+\|Tx_0\|\le2N\) for every \(T\). Scaling yields \(\|Tz\|\le (2N/r)\|z\|\) for \(z\ne0\). Taking the supremum over \(\|z\|\le1\) proves uniform boundedness of the operator norms.
Banach-Steinhaus convergence theorem
The convergence version is often the form used in analysis: pointwise limits of uniformly bounded operators remain bounded and linear.
Corollary. Let \(X\) be Banach and \(Y\) Banach. If \(T_n\in\mathcal L(X,Y)\), \(T_nx\) converges in \(Y\) for every \(x\in X\), then the pointwise limit \(Tx=\lim_nT_nx\) belongs to \(\mathcal L(X,Y)\) and \(\sup_n\|T_n\|<\infty\).
Proof. Pointwise convergence implies pointwise boundedness. Uniform boundedness gives \(M=\sup_n\|T_n\|<\infty\). The pointwise limit of linear maps is linear, and \(\|Tx\|=\lim_n\|T_nx\|\le M\|x\|\), so \(T\) is bounded.
Weak convergence is bounded
Weakly convergent sequences are bounded because every functional sees a bounded scalar sequence, and uniform boundedness transfers this to the norm.
Proposition. If \(x_n\to x\) weakly in a Banach space \(X\), then \(\sup_n\|x_n\|<\infty\).
Proof. View each \(x_n\) as a bounded linear functional \(Jx_n\) on \(X^*\), where \(Jx_n(f)=f(x_n)\). For each \(f\in X^*\), the scalar sequence \(f(x_n)\) converges, hence is bounded. Uniform boundedness on the Banach space \(X^*\) gives \(\sup_n\|Jx_n\|<\infty\). Hahn-Banach gives \(\|Jx_n\|=\|x_n\|\), so the norms are bounded.
Condensation of singularities
The condensation principle is the contrapositive intuition. If operator norms blow up, some vector must witness unbounded behavior.
Theorem. If \((T_n)\subset\mathcal L(X,Y)\) with \(X\) Banach satisfies \(\sup_n\|T_n\|=\infty\), then there exists \(x\in X\) such that \(\sup_n\|T_nx\|=\infty\).
Proof. This is the contrapositive of uniform boundedness. If every orbit \(\{T_nx:n\ge1\}\) were bounded, the principle would imply \(\sup_n\|T_n\|<\infty\), contradicting the hypothesis.
A Fourier-analysis consequence
Fourier analysis supplies natural families of linear operators whose norms reveal convergence behavior. Uniform boundedness turns these norms into pointwise divergence examples.
Example. Let \(S_N:C(\mathbb T)\to C(\mathbb T)\) be the \(N\)-th Fourier partial-sum operator. Its norm is the \(L^1\)-norm of the Dirichlet kernel up to normalization, which grows like \(\log N\). By condensation of singularities, there exists a continuous function whose Fourier partial sums are not uniformly bounded in sup norm, hence do not converge uniformly.
Completeness is essential
Without completeness, the category argument can fail. This example marks exactly where the Banach hypothesis enters.
Remark. The theorem fails on incomplete spaces. Completeness is used only through Baire category, but that use is decisive: it turns a countable cover by closed control sets into a genuine ball of uniform control.