Namespaces
Variants
Actions

Limit

From Encyclopedia of Mathematics
Jump to: navigation, search

2010 Mathematics Subject Classification: Primary: 54A20 [MSN][ZBL]

One of the fundamental concepts in mathematics, meaning that a variable depending on another variable arbitrary closely approaches some constant as the latter variable changes in a definite manner. In the definition of a limit, the concept of nearness of the objects under consideration plays a fundamental role: only after a definition of nearness does a limit acquire an exact meaning. The following fundamental concepts of mathematical analysis are connected with that of a limit: continuity, convergence, derivative, differential, integral. One of the simplest cases of a limit is the limit of a sequence.

The limit of a sequence.

Let $X$ be a topological space. A sequence of points $x_n$, $n=1,2,\dots,$ of $X$ is said to converge to a point $x_0\in X$ or, which is the same, the point $x_0$ is said to be a limit of the given sequence if for each neighbourhood $U$ of $x_0$ there is a natural number $N$ such that for all $n>N$ the membership $x_n\in U$ is satisfied. In this case one writes $$ \lim_{n\to \infty} x_n = x_0. $$ In the case when $X$ is a Hausdorff space, the limit of a sequence $x_n\in X$, $n=1,2,\dots,$ is unique, provided that it exists. For a metric space $X$, a point $x_0$ is the limit of a sequence $\{x_n\}$ if and only if for each $\epsilon >0$ there is a natural number $N$ such that for all indices $n>N$ the inequality $\rho(x_n,x_0)<\epsilon$ is satisfied, where $\rho(x,x_0)$ is the distance between $x_n$ and $x_0$. If a sequence of points of a metric space is convergent, then it is bounded. A sequence of points of a complete metric space is convergent if and only if it is a Cauchy sequence. In particular, this is true for sequences of numbers, for which the concept of a limit of a sequence historically arose first (see Cauchy criteria). For such sequences the following formulas hold: $$ \lim_{n\to \infty} c = c, \quad \lim_{n\to \infty} cx_n = c \lim_{n\to \infty} x_n, $$ where $c$ is any given number; $$ \lim_{n\to \infty} (x_n+y_n) = \lim_{n\to \infty} x_n + \lim_{n\to\infty}y_n;\\ \lim_{n\to \infty} x_n y_n = \lim_{n\to \infty} x_n \lim_{n\to \infty} y_n; $$ and if $\lim_{n\to \infty} y_n \neq 0$, then $$ \lim_{n\to\infty}\frac{x_n}{y_n} = \frac{\lim_{n\to \infty} x_n}{\lim_{n\to\infty} y_n}. $$ These properties of sequences of numbers can be carried over to limits of sequences in more general structures, for example, the property of the limit of a sum — to sequences of points in linear topological spaces, the property of the limit of a product — to sequences of points in a topological group, etc.

If two real sequences $x_n\in\mathbb R$ and $y_n\in \mathbb R$ are convergent and if $x_n \le y_n$, $n=1,2,\dots,$ then $$\lim_{n\to \infty} x_n \le \lim_{n\to \infty} y_n, $$ i.e. non-strict inequalities are preserved under limit transition. If $$\lim_{n\to \infty} x_n = \lim_{n\to \infty} y_n = \alpha$$and if $x_n\le z_n \le y_n$, then the sequence $z_n$, $n=1,2,\dots,$ converges to the same limit: $\lim_{n\to \infty} z_n = \alpha$. These properties can be generalized to limits of sequences of points in ordered sets.

Every increasing (decreasing) sequence of real numbers $x_n$, i.e. such that $x_n \le x_{n+1}$ ($x_n \ge x_{n+1}$), $n=1,2,\dots,$ that is bounded from above (below) is convergent and its limit is the supremum (infimum) of the set of its members. For example, if $\alpha>0$, $k$ is a natural number and $\alpha_n$ is an approximate value of the root $\alpha^{1/k}$ calculated up to $n$ decimal places after the decimal point, then the $\alpha_n$, $n=1,2,\dots,$ form an increasing sequence and $\lim_{n\to\infty}\alpha_n = \alpha^{1/k}$. Another example of an increasing sequence which is bounded from above is the sequence of perimeters of regular polygons with $n$ sides, $n=3,4,\dots,$ inscribed in some circle; this sequence converges to the length of the circle.

In the theory of sequences of numbers a fundamental role is played by infinitesimal sequences or null sequences, i.e. those sequences which converge to zero. The general concept of a sequence of numbers can be reduced to that of an infinitesimal sequence in the sense that a sequence of numbers converges to a given number if and only if the differences between the terms of the sequence and the given number form an infinitesimal sequence.

The concept of an infinitely-large sequence of numbers is also useful. These are the sequences with as limits one of the infinities $+\infty$, $-\infty$, or the infinity $\infty$ without a sign. For the definition of infinite limits, the concept of $\epsilon$-neighbourhoods, $\epsilon>0$, of the symbols $+\infty$, $-\infty$ or $\infty$ in the set $\mathbb R$ of real numbers is introduced by the formulas $$ U(+\infty,\epsilon) = \left\{ x \in \mathbb R \colon x > \frac{1}{\epsilon}\right\},\\ U(-\infty,\epsilon) = \left\{ x \in \mathbb R \colon x < -\frac{1}{\epsilon}\right\},\\ U(\infty,\epsilon) = \left\{ x \in \mathbb R \colon \lvert x \rvert > \frac{1}{\epsilon}\right\},\\ $$ and the concept of the $\epsilon$-neighbourhood of $\infty$ in the set $\mathbb C$ of complex numbers is introduced by the formula $$ U(\infty,\epsilon) = \left\{ z \in \mathbb C\colon \lvert z \rvert > \frac 1 \epsilon\right\}. $$

One writes $\lim_{n\to\infty} x_n = \infty$ ($\lim_{n\to \infty} x_n = +\infty$ or $-\infty$), $x_n\in \mathbb R$, $n=1,2,\dots,$ if for each $\epsilon>0$ there is an index $N$ such that for all indices $n>N$ the membership $x_n\in U(\infty,\epsilon)$ ($x_n\in U(+\infty,\epsilon)$ or $x_n\in U(-\infty,\epsilon)$) is satisfied. The infinite limit of a sequence of complex numbers is defined similarly.

Every bounded sequence of numbers contains a convergent subsequence (cf. Bolzano–Weierstrass theorem). Every unbounded sequence contains an infinitely-large sequence.

A (finite or infinite) limit of a subsequence of a given sequence is called a subsequential limit of the latter. In the set of subsequential limits of any sequence of real numbers there is always a largest one and a smallest one (finite or infinite). The largest (smallest) subsequential limit of a sequence is called its upper (lower) limit. A sequence has a finite or infinite limit if and only if its upper limit coincides with its lower limit, and then their common value is the limit of the sequence.

Other concepts of a limit, for example, the limit of a function and of Riemann sums, can be expressed in terms of the limit of a sequence. The definition of the limit of a sequence can be generalized to directed (partially ordered) sets.

The limit of a function (mapping).

Let $ X $ and $ Y $ be topological spaces, $ E \subset X $, $ x _{0} $ an accumulation point (or a cluster point) of $ E $, and let $ f : \ E \rightarrow Y $ be a mapping from $ E $ into $ Y $. A point $ a \in Y $ is called a limit of the mapping $ f $ at $ x _{0} $( or, as one says, as $ x $ approaches $ x _{0} $), in symbols,

$$ \lim\limits _ {x \rightarrow x _ 0} \ f (x) \ = \ a \ \ \textrm{ or } \ \ f (x) \ \rightarrow \ a \ \textrm{ as } \ x \rightarrow x _{0} , $$


if for any neighbourhood $ V = V (a) $ of $ a $ in $ Y $ there is a neighbourhood $ U = U ( x _{0} ) $ of $ x _{0} $ in $ X $, such that for any point $ x \in E \cap U ( x _{0} ) \setminus \{ x _{0} \} $ the image $ f (x) $ belongs to $ V $: $ f (x) \in V $. In other words, if $ f ( E \cap U ) \subset V $. If $ Y $ is a Hausdorff space, then the mapping $ f : \ E \rightarrow Y $ can have only one limit at a given point $ x _{0} \in X $.


In the case where $ E ^{*} \subset E $ and $ x _{0} $ is an accumulation point of $ E ^{*} $, a limit of the restriction $ f \mid _ {E ^ *} $ of $ f $ to $ E ^{*} $ is called a limit of $ f $ on $ E ^{*} $, in symbols,

$$ \lim\limits _ { {x \rightarrow x _{0} \atop x \in E ^ *}} \ f (x) \ = \ \lim\limits _ {x \rightarrow x _ 0} \ f \mid _ {E ^ *} (x) . $$


If $ f : \ E \rightarrow Y $, $ E ^{*} \subset E \subset X $, $ x _{0} $ is an accumulation point of $ E ^{*} $, and $ \lim\limits _ {x \rightarrow x _ 0} \ f (x) $ exists, then the limit of $ f $ at $ x _{0} $ also exists on $ E ^{*} $ and

$$ \lim\limits _ { {x \rightarrow x _{0} \atop x \in E ^ *}} \ f (x) \ = \ \lim\limits _ {x \rightarrow x _ 0} \ f (x) . $$


If $ E _{1} ,\ E _{2} \subset X $, $ f : \ E _{1} \cup E _{2} \rightarrow Y $ and the limits

$$ \lim\limits _ { {x \rightarrow x _{0} \atop x \in E _ 1}} \ f (x) \ = \ \lim\limits _ { {x \rightarrow x _{0} \atop x \in E _ 2}} \ f (x) \ = \ a $$


exist, then the limit

$$ \lim\limits _ {x \rightarrow x _ 0} \ f (x) \ = \ a $$


also exists.

In considering the limit of a mapping (function) $ f : \ E \rightarrow Y $, $ E \subset X $, as $ x \rightarrow x _{0} \in X $, it may happen that $ x _{0} \in E $ or, on the other hand, $ x _{0} \notin E $. The case $ x _{0} \in E $ is of special interest, because it leads to the concept of a continuous function: If $ f : \ E \rightarrow Y $, $ Y $ is a Hausdorff space and $ x _{0} \in E $, then for the mapping $ f $ to be continuous at $ x _{0} $ it is necessary and sufficient that

$$ \lim\limits _ {x \rightarrow x _ 0} \ f (x) \ = \ f ( x _{0} ) . $$


If $ x _{0} $ is an isolated point of $ E $, then the limit

$$ \lim\limits _ {x \rightarrow x _ 0} \ f (x) \ = \ f ( x _{0} ) $$


always exists, for any mapping $ f : \ E \rightarrow Y $, i.e. any mapping $ f $ is continuous at all isolated points of its domain. Therefore, the concept of a limit of a mapping, in particular, of continuity, is non-trivial only for limit points of the set being mapped (cf. Limit point of a set). In the classical case of the limit of a function $ f : \ E \rightarrow Y $ it is usually assumed that $ x _{0} \notin E $, i.e. $ x _{0} $ does not belong to the set on which the limit is taken.

If the space $ X $ satisfies the first axiom of countability at the point $ x _{0} $ and the space $ Y $ is Hausdorff, then for the existence of the limit $ \lim\limits _ {x \rightarrow x _ 0} \ f (x) $ of a mapping $ f : \ E \rightarrow Y $, $ E \subset X $, it is necessary and sufficient that for any sequence $ x _{n} \in E $, $ n = 1 ,\ 2 \dots $ such that $ \lim\limits _ {n \rightarrow \infty} \ x _{n} = x _{0} $, the limit $ \lim\limits _ {n \rightarrow \infty} \ f ( x _{n} ) $ exists. If this condition holds, the limit $ \lim\limits _ {n \rightarrow \infty} \ f ( x _{n} ) $ does not depend on the choice of the sequence $ \{ x _{n} \} $, and the common value of these limits is the limit of $ f $ at $ x _{0} $.


The limit of a sequence of points $ y _{n} $ in a topological space $ Y $ is a special case of the limit of a mapping (function): in this case $ E = \mathbf N $, the set of natural numbers with the discrete topology, $ x _{0} = + \infty $, $ X = \mathbf N \cup \{ + \infty \} $, and a neighbourhood of $ + \infty $ in $ X $ is any subset $ U \subset X $ of the form $ U = \{ {n} : {n \geq n _ 0} \} \cup \{ + \infty \} $, where $ n _{0} $ is a natural number.

The concept of the limit of a multiple sequence, i.e. of a sequence the members of which are indexed by integral multi-indices, is also a special case of the limit of a mapping.

An intrinsic criterion for the existence of the limit of a mapping $ f : \ E \rightarrow Y $ at a given point $ x _{0} $( called the Cauchy criterion) in the case where the space $ X \supset E $ is first countable at $ x _{0} $ and $ Y $ is a complete metric space, is that the limit $ \lim\limits _ {x \rightarrow x _ 0} \ f (x) $ exists if and only if for each $ \epsilon > 0 $ there is a neighbourhood $ U = U ( x _{0} ) $ of $ x _{0} $ in $ X $ such that for all points $ x ^ \prime $ and $ x ^{\prime\prime} $ satisfying the condition $ x ^ \prime ,\ x ^{\prime\prime} \in E \cap U \setminus \{ x _{0} \} $, the inequality $ \rho ( f ( x ^{\prime\prime} ),\ f ( x ^ \prime ) ) < \epsilon $ holds. In particular, this criterion is valid if $ Y $ is the set of real or complex numbers.

Some properties of the limit.

If $ Y $ is a metric space, $ f : \ E \rightarrow Y $, $ E \subset X $, and the limit $ \lim\limits _ {x \rightarrow x _ 0} \ f (x) = a \in Y $ exists, then there is a neighbourhood $ U = U ( x _{0} ) $ of $ x _{0} $ such that the image of the intersection $ E \cap U $ of the set $ E $ being mapped and the neighbourhood $ U $ under the mapping $ f $ is a bounded subset of $ Y $.


If a function $ f : \ E \rightarrow \mathbf R $, $ E \subset X $, where $ \mathbf R $ is the set of real numbers, has a finite non-zero limit at a point $ x _{0} \in X $, then there exist a neighbourhood $ U = U ( x _{0} ) $ of $ x _{0} $ and a number $ c > 0 $ such that for all points $ x \in E \cap U \setminus \{ x _{0} \} $ the inequalities

$$ f (x) > c \ \ \textrm{ if } \ \lim\limits _ {x \rightarrow x _ 0} \ f (x) > 0 , $$


$$ f (x) < -c \ \ \textrm{ if } \ \lim\limits _ {x \rightarrow x _ 0} \ f (x) < 0 $$


are satisfied.

If $ Y $ is a topological group (in particular, an Abelian group with group operation written additively), $ f : \ E \rightarrow Y $, $ E \subset X $, and $ x _{0} \in X $, then the limit $ \lim\limits _ {x \rightarrow x _ 0} \ f (x) $ exists and equals $ a $ if and only if the function $ \alpha (x) = f (x) a ^{-1} $ has a limit at $ x _{0} $ which is equal to the identity element of $ Y $( respectively, the function $ \alpha (x) = f (x) - a $ has a limit at $ x _{0} $ which is equal to zero — such functions are called infinitesimal functions).

If $ Y $ is a linear topological space over a field $ P $, $ f _{1} ,\ f _{2} : \ E \rightarrow Y $, and $ E \subset X $, then the limit of a linear combination of $ f _{1} $ and $ f _{2} $ at $ x _{0} $ is equal to the same linear combination of their limits at the same point:

$$ \lim\limits _ {x \rightarrow x _ 0} \ [ \lambda _{1} f _{1} (x ) + \lambda _{2} f _{2} (x) ] \ = \ \lambda _{1} \ \lim\limits _ {x \rightarrow x _ 0} \ f _{1} (x) + \lambda _{2} \ \lim\limits _ {x \rightarrow x _ 0} \ f _{2} (x) , $$


$$ \lambda _{1} ,\ \lambda _{2} \ \in \ P . $$


If $ Y $ is the set of real or complex numbers, $ f _{1} ,\ f _{2} : \ E \rightarrow Y $( such functions are called numerical) and $ E \subset X $, then

$$ \lim\limits _ {x \rightarrow x _ 0} \ f _{1} (x) f _{2} (x) \ = \ \lim\limits _ {x \rightarrow x _ 0} \ f _{1} (x) \ \lim\limits _ {x \rightarrow x _ 0} \ f _{2} (x) ; $$


if $ \lim\limits _ {x \rightarrow x _ 0} \ f _{2} (x) \neq 0 $, then

$$ \lim\limits _ {x \rightarrow x _ 0} \ \frac{f _{1} (x)}{f _{2} (x)} \ = \ \frac{\lim\limits _ {x \rightarrow x _ 0} \ f _{1} (x)}{\lim\limits _ {x \rightarrow x _ 0} \ f _{2} (x)} , $$


and in this case, by the limit $ \lim\limits _ {x \rightarrow x _ 0} ( f _{1} (x) / f _{2} (x) ) $ is meant the limit of the restriction of $ f _{1} / f _{2} $ to the intersection of the set $ E $ being mapped and some neighbourhood of $ x _{0} $, such that on this intersection the quotient $ f _{1} / f _{2} $ is defined. If $ f _{1} (x) \leq f _{2} (x) $, $ x \in E $, and the limits $ \lim\limits _ {x \rightarrow x _ 0} \ f _{1} (x) $ and $ \lim\limits _ {x \rightarrow x _ 0} \ f _{2} (x) $ exist, then

$$ \lim\limits _ {x \rightarrow x _ 0} \ f _{1} (x) \ \leq \ \lim\limits _ {x \rightarrow x _ 0} \ f _{2} (x) . $$


If $ X $ and $ Y $ are obtained from the set $ \mathbf R $ of real numbers by completing it either with an infinity without a sign $ \infty $ or by two signed infinities $ + \infty $ and $ - \infty $, and $ E \subset \mathbf R $, $ f (E) \subset \mathbf R $, and $ x _{0} \notin E $, then the definition of the limit of a function as defined above is the classical definition of a finite or infinite limit of a real-valued function of one real variable. Similarly, if the spaces $ X $ and $ Y $ are obtained by completing the set $ \mathbf C $ of complex numbers by the infinity $ \infty $, then the definition of a (finite or infinite) limit of a function of a complex variable is obtained. On the other hand, if the space $ X $ is obtained by completing $ \mathbf R ^{n} $( $ \mathbf C ^{n} $), $ n > 1 $, by the infinity $ \infty $, then the definition of a finite or infinite limit of a function of several variables as the argument approaches a finite point or infinity is obtained.

For functions defined on subsets of the real line (or, more generally, on ordered sets), the concept of a one-sided limit is defined. Examples of functions having at least one one-sided limit at all limit points of their domain are the real-valued monotone functions: If a function $ f $ is monotone (cf. Monotone function) on a set $ E $ of the real line and the point $ x _{0} $ is a limit point of $ E $, then it is a limit point of at least one of the sets $ E _{1} = E \cap \{ {x \in \mathbf R} : {x < x _ 0} \} $ or $ E _{2} = E \cap \{ {x \in \mathbf R} : {x > x _ 0} \} $. If $ x _{0} $ is a limit point of $ E _{1} $, then $ f $ has a limit from the left at $ x _{0} $, i.e. on $ E _{1} $. On the other hand, if $ x _{0} $ is a limit point of $ E _{2} $, then $ f $ has a limit from the right at $ x _{0} $ i.e. on $ E _{2} $. If, in addition, for example, $ f $ is increasing and bounded from above, $ E _{1} \neq \emptyset $ and $ x _{0} $ is a limit point of $ E _{1} $, then the limit $ \lim\limits _ {x \rightarrow x _ 0} \ f (x) $ is finite.

A fundamental general method for finding the limit of a function is the determination of the principal parts of the function in a neighbourhood of the given point, which is often done by means of the Taylor formula. For the calculation of a limit, the l'Hospital rule is often useful.

In spite of the great generality of the concept of a limit of a mapping, it does not include all existing concepts of a limit appearing in contemporary mathematics. For example, the concept of the limit of Riemann sums (cf. Integral sum) is not included in the concept of the limit of a mapping (function). A sufficiently general concept of a limit, including all fundamental cases in a certain sense, is the concept of the limit of a mapping with respect to a filter.

The limit of a filter.

Let $ X $ be a topological space, $ X \neq \emptyset $, let $ \mathfrak U = \{ U \} $ be a base for its topology, and let $ \mathfrak F $ be a filter on $ X $( i.e. a non-empty family $ \mathfrak F $ of non-empty subsets of $ X $ such that for any $ A ^ \prime ,\ A ^{\prime\prime} \in \mathfrak F $ there is an $ A \in \mathfrak F $ such that $ A \subset A ^ \prime \cap A ^{\prime\prime} $ and such that every subset of $ X $ containing an $ A \in \mathfrak F $ belongs to $ \mathfrak F \ $). A point $ x _{0} $ is called a limit of the filter $ \mathfrak F $, or its limit point, if $ \mathfrak F $ is stronger than the filter $ \mathfrak B ( x _{0} ) $ consisting of a local base for the topology at $ x _{0} $, i.e. for any $ U \in \mathfrak B ( x _{0} ) $ there is an $ A \in \mathfrak F $ such that $ A \subset U $.


Let $ \mathbf N $ be the set of natural numbers with the discrete topology. The filter on $ \mathbf N $ consisting of the complements of all finite subsets of $ \mathbf N $ is called the natural filter on $ \mathbf N $ and is denoted by $ \mathfrak F _ {\mathbf N} $. It does not have a limit in $ \mathbf N $. The same filter on the set $ X = \mathbf N \cup \{ + \infty \} $, in which the local base $ \mathfrak B ( + \infty ) $ consists of the sets $ A _{n} = \{ {m} : {m \in \mathbf N ,\ m > n \in \mathbf N} \} $ and $ \mathfrak B (n) $ consists of the singleton $ \{ n \} $ for $ n \in \mathbf N $, has $ + \infty $ as its limit. Uniqueness of the limit of a filter on a topological space is connected with being able to separate points of the space; in order that every filter on a topological space has at most one limit it is necessary and sufficient that the space be a Hausdorff space.

Let $ X $ be a set, $ Y $ a topological space, $ \phi $ a mapping from $ X $ into $ Y $, and $ \mathfrak F $ a filter on $ X $. A point $ b \in Y $ is called the limit of the mapping $ \phi $ with respect to the filter $ \mathfrak F $, in symbols,

$$ \lim\limits _ {\mathfrak F} \ \phi (x) \ = \ b , $$


if the filter $ \phi ( \mathfrak F ) $ consisting of all sets $ \phi (A) $, $ A \in \mathfrak F $, has $ b $ as its limit in $ Y $.


If $ X = \mathbf N $ is the set of natural numbers, $ \phi $ is a mapping from $ \mathbf N $ into a topological space $ Y $, $ \phi (n) = y _{n} \in Y $, $ n \in \mathbf N $, and $ \mathfrak F _ {\mathbf N} $ is the natural filter, then the limit of $ \phi $ with respect to $ \mathfrak F _ {\mathbf N} $ in $ Y $ coincides with the usual limit of the sequence $ \{ y _{n} \} $ in $ Y $.


If in $ X = \mathbf N \times \mathbf N $, the filter $ \mathfrak F $ on $ X $ is the product of two natural filters $ \mathfrak F _ {\mathbf N} $, i.e. it consists of all sets of the form $ A \times B $, where $ A ,\ B \in \mathfrak F _ {\mathbf N} $, and if $ \phi $ is a mapping from $ \mathbf N \times \mathbf N $ into a topological space $ Y $, $ \phi ( n ,\ m ) = y _{nm} \in Y $, $ n ,\ m \in \mathbf N $, then the limit of $ \phi $ with respect to $ \mathfrak F $ in $ Y $ coincides with the usual limit of the double sequence $ \{ y _{nm} \} $ in $ Y $.


Let the elements of a set $ X $, in turn, be the sets $ x $ consisting of a partition $ \tau = \{ t _{i} \} _{i=0} ^{i=n} $ of some interval $ [ a ,\ b ] $, $ a = t _{0} < t _{1} < \dots < t _{n-1} < t _{n} = b $, and points $ \xi _{i} \in [ t _{i-1} ,\ t _{i} ] $, $ i = 1 ,\ 2 \dots $ i.e.

$$ x \ = \ \{ \tau ; \ \xi _{1} \dots \xi _{n} \} , $$


Let $ A _ \eta $( for any $ \eta > 0 $) be the subset of $ X $ consisting of all elements $ x \in X $ for which the mesh of the partition $ \tau = \{ t _{i} \} _{i=0} ^{i=n} $ appearing in $ x $ is smaller than $ \eta $, i.e.

$$ \max _ {i = 1 \dots n} \ ( t _{i} - t _{i-1} ) \ < \ \eta . $$


The system $ \mathfrak F = \{ A _ \eta \} $ is a filter. Every real-valued function $ f $ defined on $ [ a ,\ b ] $ induces a mapping $ \phi _{f} $ of $ X $ into $ \mathbf R $ by the formula

$$ \phi _{f} (x) \ = \ \sum _{i=1} ^ n f ( \xi _{i} ) ( t _{i} - t _{i-1} ) , $$


$$ x \ = \ ( \tau ; \ \xi _{1} \dots \xi _{n} ) ,\ \ \tau \ = \ \{ t _{i} \} _{i=0} ^{i=n} . $$


Therefore, $ \phi _{f} (x) $ is the Riemann sum of $ f $ corresponding to $ x \in X $.


The limit of $ \phi _{f} $ in $ \mathbf R $ with respect to $ \mathfrak F $ coincides with the usual limit of the Riemann sums of $ f $ as the mesh of the partition converges to zero. This coincidence holds in the sense that both limits simultaneously exist or not, and if they do so, then they are equal and coincide with the Riemann integral of $ f $ over $ [ a ,\ b ] $.


The limit of a mapping of topological spaces with respect to a filter.

Let $ X $ and $ Y $ be topological spaces, $ E \subset X $, let $ \mathfrak F $ be a filter on $ E $, and let $ \phi $ be a mapping from $ E $ into $ Y $. A point $ b \in Y $ is called the limit of the mapping $ \phi $ at $ a \in X $ with respect to the filter $ \mathfrak F $ if $ a $ is the limit of $ \mathfrak F $ and $ b $ is the limit of $ \phi ( \mathfrak F ) $; notation:

$$ b \ = \ \lim\limits _ {\mathfrak F} \ \phi (x) . $$


If $ \mathfrak B (a) $ is a neighbourhood base at $ a $, $ E = X \setminus \{ a \} $ and the filter consists of all "punctured" neighbourhoods $ U (a) \setminus \{ a \} $ of $ a $, $ U (a) \in \mathfrak B (a) $, then the limit $ \lim\limits _ {\mathfrak F} \ \phi (x) $ coincides with the usual limit $ \lim\limits _ {x \rightarrow a} \ \phi (x) $ of $ \phi $ at $ a $, i.e. it generalizes the classical definition of the limit of a mapping as formulated in terms of neighbourhoods. An immediate generalization of the concept of the limit of a sequence is the limit of a directed set in a topological space, i.e. a partially ordered set in which any two elements have a common successor. The concept of a limit of a mapping from one topological space into another can be formulated in terms of limits with respect to directed sets (cf. Generalized sequence; Convergence, types of).

The limit of a sequence of sets.

The topological limit. Let $ A _{n} $, $ n = 1 ,\ 2 \dots $ be subsets of a topological space $ X $. The upper topological limit $ \overline{\mathop{\rm lt}} \ A _{n} $ of the sequence $ \{ A _{n} \} $ is, by definition, the set of those points $ x \in X $ every neighbourhood of which intersects infinitely many sets $ A _{n} $. The lower topological limit $ \underline{\mathop{\rm lt}} \ A _{n} $ is the set of those points every neighbourhood of which contains points of almost-all sets $ A _{n} $. It is obvious that $ \underline{\mathop{\rm lt}} \ A _{n} \subset \overline{\mathop{\rm lt}} \ A _{n} $. If $ A = \underline{\mathop{\rm lt}} \ A _{n} = \overline{\mathop{\rm lt}} \ A _{n} $, then the sequence $ \{ A _{n} \} $ is called convergent and the set $ A $ its topological limit; one writes $ A = \mathop{\rm lt}\nolimits \ A _{n} $. The upper and lower topological limits of a sequence are closed sets.

The set-theoretical limit. There is a concept of the limit of a sequence of sets not involving topology. A sequence $ A _{n} $, $ n = 1 ,\ 2 \dots $ of sets is called convergent if there is a set $ A $, called its limit and denoted by

$$ A \ = \ \lim\limits \ A _{n} , $$


such that every element of $ A $ belongs to all sets $ A _{n} $ from some index onwards and such that every point of the union of all $ A _{n} $ not belonging to $ A $ is contained in only finitely many sets $ A _{n} $. The set $ A $ is the limit of the sequence $ \{ A _{n} \} $ if and only if the upper and lower limits of the sequence coincide and are equal to $ A $.


References

[1] V.A. Il'in, E.G. Poznyak, "Fundamentals of mathematical analysis" , 1–2 , MIR (1982) (Translated from Russian)
[2] V.A. Il'in, V.A. Sadovnichii, B.Kh. Sendov, "Mathematical analysis" , Moscow (1979) (In Russian)
[3] L.D. Kudryavtsev, "A course in mathematical analysis" , 1–3 , Moscow (1988–1989) (In Russian)
[4] S.M. Nikol'skii, "A course of mathematical analysis" , 1–2 , MIR (1977) (Translated from Russian)
[5] N. Bourbaki, "Elements of mathematics. General topology" , Addison-Wesley (1966) (Translated from French)
[6] M. Zamansky, "Introduction à l'algèbre et l'analyse modernes" , Dunod (1958)
[7] A. Krazer, W. Franz, "Transzendente Funktionen" , Akademie Verlag (1960)
[8] F. Hausdorff, "Grundzüge der Mengenlehre" , Leipzig (1914) (Reprinted (incomplete) English translation: Set theory, Chelsea (1978))


Comments

For the general theory of topological spaces one needs to study more general convergence concepts than that of sequences of points. One needs to consider limits of sets of points indexed by a directed partially ordered set (cf. Directed set). Such sets are called nets, and the convergence of nets is sometimes known as Moore–Smith convergence.

The basic (inductive) idea of "limit" is that of some object being approximated arbitrarily closely by a sequence of other objects. There are also various algebraic and categorical implementations of this idea.

Let $ {\mathcal C} $ be a category and $ A $ a partially ordered set. A diagram $ ( C _ \alpha ,\ f _ {\alpha \beta} ) $ in $ {\mathcal C} $ indexed by $ A $ consists of an object $ C _ \alpha \in {\mathcal C} $ for each $ \alpha \in A $ and a morphism $ f _ {\alpha \beta} : \ C _ \alpha \rightarrow C _ \beta $ for each $ \alpha \leq \beta $ in $ A $, such that $ f _ {\alpha \alpha} = \mathop{\rm id}\nolimits $, $ f _ {\beta \gamma} f _ {\alpha \beta} = f _ {\alpha \gamma} $ if $ \alpha \leq \beta \leq \gamma $.


The projective limit (or inverse limit) $ ( P ,\ f _ \alpha ) $ of $ ( C _ \alpha ,\ f _ {\alpha \beta} ) $ consists of an object $ P \in {\mathcal C} $ together with morphisms $ f _ \alpha : \ P \rightarrow C _ \alpha $ such that $ f _ {\alpha \beta} f _ \alpha = f _ \beta $ if $ \alpha \leq \beta $ and such that if, moreover, $ D $ and $ g _ \alpha : \ D \rightarrow C _ \alpha $ are such that $ f _ {\alpha \beta} g _ \alpha = g _ \beta $ for all $ \alpha \leq \beta $, then there is a unique morphism $ g : \ D \rightarrow P $ such that $ f _ \alpha g = g _ \alpha $. Notation: $ P = \lim\limits _ \leftarrow \ C _ \alpha $.


Dually, the inductive limit (or directed limit) $ ( I ,\ h _ \alpha ) $ of $ ( C _ \alpha ,\ f _ {\alpha \beta} ) $ consists of an object $ I $ together with morphisms $ h _ \alpha : \ C _ \alpha \rightarrow I $ such that $ h _ \beta f _ {\alpha \beta} = f _ \alpha $ for all $ \alpha \leq \beta $ and such that if, moreover, $ E $ and $ k _ \alpha : \ C _ \alpha \rightarrow E $ are such that $ k _ \beta f _ {\alpha \beta} = k _ \alpha $ for all $ \alpha \leq \beta $, then there is a unique morphism $ k : \ I \rightarrow E $ such that $ k h _ \alpha = k _ \alpha $. Notation: $ I = \lim\limits _ \rightarrow \ C _ \alpha $.


These are very general notions and include direct products and direct sums (when every pair of points in $ A $ is incomparable). In case $ A $ is directed (cf. Directed set), the idea of better and better approximations again reemerges; cf. e.g. Topological vector space.

More generally one also considers projective and inductive limits of diagrams in categories (cf. Diagram).

Still another limit concept is that of the limit of a spectral sequence: If the spectral sequence $ ( E _{p,q} ^{r} ) $ converges to, say, $ ( H _{p+q} ) $, then $ ( H _{p+q} ) $ is also called the limit of $ ( E _{p,q} ^{r} ) $.


Still other limit ideas are embodied, for example, by the concepts limit cycle; limit elements (cf. the supplementary material section of Vol. 10); limit point of a set; accumulation point; cluster point; etc.

References

[a1] R. Courant, "Differential and integral calculus" , 1–2 , Blackie (1948) (Translated from German)
[a2] K.R. Stromberg, "Introduction to classical real analysis" , Wadsworth (1981)
[a3] J.L. Kelley, "General topology" , v. Nostrand (1955) pp. 125; 127
[a4] B. Mitchell, "Theory of categories" , Acad. Press (1965) pp. 7
[a5] J. Adámek, "Theory of mathematical structures" , Reidel (1983)
How to Cite This Entry:
Limit. Encyclopedia of Mathematics. URL: http://www.encyclopediaofmath.org/index.php?title=Limit&oldid=44353
This article was adapted from an original article by L.D. Kudryavtsev (originator), which appeared in Encyclopedia of Mathematics - ISBN 1402006098. See original article