In nonstandard analysis, the standard part function is a function from the limited (finite) hyperreal numbers to the real numbers. Briefly, the standard part function "rounds off" a finite hyperreal to the nearest real. It associates to every such hyperreal ${\displaystyle x}$, the unique real ${\displaystyle x_{0}}$ infinitely close to it, i.e. ${\displaystyle x-x_{0}}$ is infinitesimal. As such, it is a mathematical implementation of the historical concept of adequality introduced by Pierre de Fermat,^[1] as well as Leibniz's Transcendental law of homogeneity.

The standard part function was first defined by Abraham Robinson who used the notation ${\displaystyle {}^{\circ }x}$ for the standard part of a hyperreal ${\displaystyle x}$ (see Robinson 1974). This concept plays a key role in defining the concepts of the calculus, such as continuity, the derivative, and the integral, in nonstandard analysis. The latter theory is a rigorous formalization of calculations with infinitesimals. The standard part of x is sometimes referred to as its shadow.^[2]

Nonstandard analysis deals primarily with the pair ${\displaystyle \mathbb {R} \subseteq {}^{*}\mathbb {R} }$, where the hyperreals ${\displaystyle {}^{*}\mathbb {R} }$ are an ordered field extension of the reals ${\displaystyle \mathbb {R} }$, and contain infinitesimals, in addition to the reals. In the hyperreal line every real number has a collection of numbers (called a monad, or halo) of hyperreals infinitely close to it. The standard part function associates to a finite hyperreal x, the unique standard real number x₀ that is infinitely close to it. The relationship is expressed symbolically by writing

${\displaystyle \operatorname {st} (x)=x_{0}.}$

The standard part of any infinitesimal is 0. Thus if N is an infinite hypernatural, then 1/N is infinitesimal, and st(1/N) = 0.

If a hyperreal ${\displaystyle u}$ is represented by a Cauchy sequence ${\displaystyle \langle u_{n}:n\in \mathbb {N} \rangle }$ in the ultrapower construction, then

${\displaystyle \operatorname {st} (u)=\lim _{n\to \infty }u_{n}.}$

More generally, each finite ${\displaystyle u\in {}^{*}\mathbb {R} }$ defines a Dedekind cut on the subset ${\displaystyle \mathbb {R} \subseteq {}^{*}\mathbb {R} }$ (via the total order on ${\displaystyle {}^{\ast }\mathbb {R} }$) and the corresponding real number is the standard part of u.

The standard part function "st" is not defined by an internal set. There are several ways of explaining this. Perhaps the simplest is that its domain L, which is the collection of limited (i.e. finite) hyperreals, is not an internal set. Namely, since L is bounded (by any infinite hypernatural, for instance), L would have to have a least upper bound if L were internal, but L doesn't have a least upper bound. Alternatively, the range of "st" is ${\displaystyle \mathbb {R} \subseteq {}^{*}\mathbb {R} }$, which is not internal; in fact every internal set in ${\displaystyle {}^{*}\mathbb {R} }$ that is a subset of ${\displaystyle \mathbb {R} }$ is necessarily finite.^[3]

All the traditional notions of calculus can be expressed in terms of the standard part function, as follows.

• Adequality
• Nonstandard calculus