Canonical: https://knowlpedia.com/large-deviations/exponential-tightness/

# Exponential tightness

A compact-containment condition ensuring probabilities outside compacts decay exponentially fast.

A sequence of probability measures \((\mu_n)_{n\ge1}\) on a [topological space](https://knowlpedia.com/topology/topological-space/) \(E\) is **exponentially tight at speed** \((a_n)_{n\ge1}\) with \(a_n\to\infty\) if for every \(M>0\) there exists a [compact set](https://knowlpedia.com/topology/compact-set/) \(K_M\subseteq E\) such that
\[
\limsup_{n\to\infty}\frac{1}{a_n}\log \mu_n(K_M^{\,c}) \le -M.
\]

Exponential tightness is a strengthened form of ordinary tightness for [probability measures](https://knowlpedia.com/probability/probability-measure/): it not only forces most mass into compacts, but does so with exponentially small tails at the LDP speed. It is frequently paired with a [rate function](https://knowlpedia.com/large-deviations/rate-function/) to obtain or upgrade a [large deviation principle](https://knowlpedia.com/large-deviations/large-deviation-principle/), and it is a standard hypothesis in results like the [Gärtner–Ellis theorem](https://knowlpedia.com/large-deviations/gartner-ellis-theorem/).

## Examples

- Let \(X_1,X_2,\ldots\) be independent, identically distributed real [random variables](https://knowlpedia.com/probability/random-variable/). If their [moment generating function](https://knowlpedia.com/probability/moment-generating-function/) is finite on an open interval containing \(0\), then the laws of \(\overline X_n=n^{-1}\sum_{i=1}^nX_i\) are exponentially tight at speed \(n\); Chernoff bounds give exponential decay of both tails.
- If \(E\) itself is compact, then any sequence \((\mu_n)\) is exponentially tight at any speed, since one can take \(K_M=E\) for all \(M\).
