 De la Vallée-Poussin criterion

On a finite measure space, a family is uniformly integrable exactly when it admits the displayed bound for some nonnegative increasing convex superlinear function $\Psi$. By the Dunford-Pettis theorem this ensures relative weak compactness in $L^1$. For the sufficient direction, choose $R_A$ so $\Psi (s)\ge As$ for $s>R_A$; then ${\int }_E|v_j|\le R_A\mu (E)+A^{-1}{\sup }_j\int \Psi (|v_j|)$ gives uniform integrability. Conversely choose increasing thresholds $R_m\to \infty$ so ${\sup }_j{\int }_{|v_j|>R_m}|v_j|\le 2^{-m}$. The convex function $\Psi (s)={\sum }_{m\ge 1}(s-R_m{)}_{+}$ is locally a finite sum, has uniformly bounded integrals by summing these tail estimates, and is superlinear because for $s\ge 2R_N$ its first $N$ terms sum to at least $Ns/2$. This proves the criterion in both directions.