Computing limits of sequences¶
Part 4 · Sequences · Chapter 2 · Lecture notes, volume 4 (PDF) · Exercises, volume 4 (PDF) · Exercises: Computing limits of sequences · Exercises: Additional exercises · Slides (PDF)
Prerequisites
- From this module, «Sequences and limits of sequences»: the definition of limit of a sequence.
Objectives
At the end of this chapter you will be able to:
-
compute limits with the algebra of limits, and recognize the indeterminate forms;
-
use the sign-preservation theorems;
-
compute limits with the comparison theorem.
1. Computing limits of sequences¶
-
The proofs of the basic theorems on computing limits are based on the definition of limit, on the use of inequalities, and on the use of properties that are eventually true.
-
In particular, these theorems illustrate the relation between the limit operation and the algebraic structures1 and the order structures2 present in \(\R\).
We will simply write \(a_n \rr \ell\) to mean \(a_n \rr \ell\) for \(n \rr \ip\)
Properties of the limit operation with respect to algebraic operations.
Theorem 1: Algebra of limits, case of finite limits
Hypotheses:
Claim:
Proof
We prove that:
We consider
by the triangle inequality. Since by hypothesis \(a_n \rr \ell_a~\) and \(~b_n \rr \ell_b~\), we have that
for every \(\varepsilon_a > 0\) and \(\varepsilon_b > 0\). Bounding from above the terms on the right-hand side of \(\eqref{AAA}\) we conclude that
Since \(\tilde{\varepsilon}\) is arbitrary, the claim follows. □
Proof
We prove that:
For every \(\varepsilon_a > 0\) and \(\varepsilon_b > 0\), we have:
Hence
Since \(\tilde{\varepsilon}\) is arbitrary, the claim follows. □
Proof
We prove that:
We consider
by the triangle inequality and the properties of the absolute value. Hence
Since by hypothesis \(a_n \rr \ell_a~\) and \(~b_n \rr \ell_b~\), for every \(\varepsilon_a > 0\) and \(\varepsilon_b > 0\) we have
Moreover, since
Therefore, bounding from above the terms on the right-hand side of \(\eqref{BBB}\), we conclude that
Since \(\tilde{\varepsilon}\) is arbitrary, the claim follows. □
- Moreover, the limit operation preserves the ordering
Theorem 2: Sign-preservation, first form
Hypotheses:
Claim:
Proof
We consider the case \(\ell_a > 0\). By definition of limit we have that
for every \(\varepsilon>0\), which we rewrite in the form:
Since \(\ell_a > 0\), we can choose \(\varepsilon > 0\) such that \(\ell_a - \varepsilon > 0\); then the inequality
shows that \(a_n > 0\), eventually.
The case \(\ell_a < 0\) is proved analogously. □
Theorem 3: Permanence of sign (generalized form)
Hypotheses:
Claim:
Proof
Since \(\ell_a > \lambda\), the number \(\varepsilon = \ell_a - \lambda\) is strictly positive. Applying the definition of limit with precisely this value of \(\varepsilon\), we have, eventually:
and therefore \(a_n > \lambda\), eventually. □
-
Analogously one proves that, if \(a_n \rr \ell_a \in \R\) and \(\ell_a < \lambda\), then \(a_n < \lambda\) eventually: it suffices to choose \(\varepsilon = \lambda - \ell_a > 0\) and to obtain, eventually,
\[ \ell_a - (\lambda - \ell_a) ~<~ a_n ~<~ \underbrace{\ell_a + (\lambda - \ell_a)}_{= \lambda} \] -
With \(\lambda = 0\) we recover the permanence of sign theorem, \(1^{st}\) form.
-
We can now prove the case of the quotient in Theorem Theorem 1 on the algebra of limits.
Proof
We first prove that:
The case of the quotient then follows from the case of the product, already proved, applied to the sequences \(\{a_n\}\) and \(\left\{\frac{1}{b_n}\right\}\):
Suppose \(\ell_b > 0\) (if \(\ell_b<0\) one argues in the same way on the sequence \(\{-b_n\}\)). By the permanence of sign theorem (generalized form), applied with \(\lambda = \frac{\ell_b}{2} < \ell_b\), we have
in particular \(b_n \neq 0\) eventually and the quotient \(\frac{1}{b_n}\) is well defined. Moreover, for every \(\varepsilon > 0\), we have \(|b_n - \ell_b| < \varepsilon\) eventually, and therefore, eventually:
where we used \(|b_n| > \frac{\ell_b}{2}\). Since \(\tilde{\varepsilon}\) is arbitrary, the claim follows. □
Theorem 4: Sign-preservation, second form (part I)
Hypotheses:
Claim:
Proof
It follows from the previous theorem. Indeed, if by contradiction we had \(\ell_a < 0\), from the previous theorem we would have \(a_n < 0\) eventually, which is incompatible with the hypothesis that \(a_n \ge 0\) eventually.
This case cannot occur, i.e., the opposite holds, which is the claim of the theorem. □
Theorem 5: Sign-preservation, second form (part II)
Hypotheses:
Claim:
Proof
We consider the sequence \(a_n - b_n\); by the theorem on the algebra of limits we have
Since by hypothesis
by the sign-preservation theorem, second form (part I), applied to the sequence \(a_n - b_n\) we have
□
-
This theorem tells us that in an inequality between two sequences we can pass to the limit on both sides, keeping the “\(\le\)” or the “\(\ge\)”.
-
Note, instead, that in general strict inequalities “\(<\)” and “\(>\)” are not preserved when passing to the limit.
Example 1: Passing to the limit with strict inequalities
For example, even if the \(a_n\) are strictly positive, their limit \(\ell_a\) is positive or zero, as shown by the simple example \(\frac{1}{n} \rr 0\).
Theorem 6: Comparison (squeeze) theorem
Hypotheses:
Claim:
Proof
By definition of limit we have, eventually, that
for every \(\varepsilon_a >0\) and \(\varepsilon_c >0\). Hence:
From the hypotheses of the theorem we therefore have, eventually, that
But then, eventually, we have
Since \(\tilde{\varepsilon}\) is arbitrary, the claim follows. □
Theorem 7: Comparison theorem for divergent sequences
Hypotheses:
Claim:
Proof
Since \(a_n \rr \ip\), for every \(M>0\) we have, eventually,
By hypothesis we also have \(a_n \le b_n\), eventually. Both properties hold eventually, hence, eventually, we have
Since \(M>0\) is arbitrary, we conclude that \(b_n \rr \ip\). □
-
Analogously one proves that, if \(a_n \rr \im\) and \(a_n \ge b_n\) eventually, then \(b_n \rr \im\).
-
Frequently used special cases of the comparison theorem are expressed by the following corollaries, which are very useful when studying the product of an oscillating (but bounded) sequence and one that tends to zero
Corollary 1: Of the comparison theorem (part I)
Hypotheses:
Claim:
Proof
We know that eventually we have \(-c_n \le b_n \le c_n\). Obviously
Hence by the comparison theorem (with \(a_n = -c_n\) and \(\ell = 0\)) we have that \(b_n \rr 0\). □
Corollary 2: Of the comparison theorem (part II)
Hypotheses:
Claim:
Proof
If \(\{b_n\}\) is bounded, then \(|b_n| \le M\) for some \(M>0\) and for every \(n \in \N\). We can therefore write
Since
by Corollary Corollary 1 we conclude that \(b_n \: c_n \rr 0\). □
The product of an infinitesimal sequence and a bounded one is infinitesimal.
Example 2: Application of the corollary
Consider the sequence given by a ratio of two expressions, each consisting of a sum of powers of \(n\), such as:
Factoring out the highest power both in the numerator and in the denominator we obtain:
Now, by Theorem Theorem 1 on the algebra of limits and knowing that negative powers of \(n\) tend to zero, we can state that:
hence the last sequence is convergent and consequently bounded. Now, by Corollary Corollary 2 and since
Example 3: Application of the corollary
The sequence
is the product of two sequences
hence Theorem Theorem 1 on the algebra of limits cannot be applied (the second limit does not exist).
-
However, Corollary Corollary 2 can be applied. The sequence \(\left\{\frac{1}{n}\right\}\) is infinitesimal and, since \(|\sin n| \le 1\), the sequence \(\{\sin n\}\) is bounded; therefore we have
\[ \lim_{n \rr \ip} \frac{\sin n}{n} = 0 \]
- The theorems on the algebra of limits seen so far work on pairs of sequences that are both convergent or at least bounded.
Sequences with limits \(\ip\) and \(\im\)
-
Suppose, for example, that
\[ a_n \rr \ell_a {\rm~~and~~} b_n \rr \ip \]then it is easy (and intuitive) to see that
\[ a_n + b_n \rr \ip \]We will abbreviate this as follows:
\[ \ell_a \ip = \ip \] -
Reasoning in a similar way we can summarize the rules for the limit of the sum (or difference) of two sequences, one or both of which are divergent.
Rules of partial arithmetization of the infinity symbol
Theorem 8: Partial arithmetization of the infinity symbol (addition)
Hypotheses:
Claim:
Theorem 9: Partial arithmetization of the infinity symbol (product)
Hypotheses:
Claim:
-
the sign of \(\infty\) must be determined with the usual rule of signs;
-
in claim 2. it is understood, as usual, that \(b_n \neq 0\) eventually, so that the quotient is well defined;
-
hypothesis 2., namely that \(\{b_n\}\) tends to zero from above or from below (and hence that it eventually has constant sign), is necessary: with \(a_n = 1\) and \(b_n = \frac{(-1)^n}{n} \rr 0\) we have
\[ \frac{a_n}{b_n} = (-1)^n \: n \]which is an irregular sequence (it tends neither to \(\ip\) nor to \(\im\)).
Example 4: Rule of signs
We have:
-
\({\rm if~~} a_n \rr \ell_a \in \R, \ell_a > 0 {\rm ~~and~~} b_n \rr 0^+ {\rm ~~then~~} \frac{a_n}{ b_n} \rr \ip\)
-
\({\rm if~~} a_n \rr \ell_a \in \R, \ell_a < 0 {\rm ~~and~~} b_n \rr 0^- {\rm ~~then~~} \frac{a_n}{ b_n} \rr \ip\)
-
\({\rm if~~} a_n \rr \ell_a \in \R, \ell_a > 0 {\rm ~~and~~} b_n \rr 0^- {\rm ~~then~~} \frac{a_n}{ b_n} \rr \im\)
-
\({\rm if~~} a_n \rr \ell_a \in \R, \ell_a < 0 {\rm ~~and~~} b_n \rr 0^+ {\rm ~~then~~} \frac{a_n}{ b_n} \rr \im\)
Therefore, in order to apply the rules of partial arithmetization of the infinity symbol, it is necessary to determine whether \(b_n\) tends to zero from above or from below.
Proof
We prove that:
For every \(\varepsilon >0\), since \(a_n \rr \ell_a\), eventually we have
Moreover, since \(c_n \rr \ip\), eventually we have
It follows that, eventually, we have
Since \(\tilde{\varepsilon}\) is arbitrary, the claim follows. □
The four missing operations:
are called indeterminate forms, since no rule can be established a priori to determine their result.
Example 5: Resolving indeterminate forms \(\ip\im\)
Consider the sequence
We have the difference of two sequences:
Hence we fall into the indeterminate form \(\ip \im\).
Multiplying and dividing by \(\sqrt{n+1} + \sqrt{n-1}\) we obtain
recalling that \(a^2-b^2=(a-b)(a+b)\). Now, considering the sequence in the denominator, we have
Using the rule
of the theorem on the partial arithmetization of the infinity symbol (product), we have:
-
Limits of sequences of the form:
\[ \left\{a_n^{b_n}\right\} \]can be handled by considering the sequence of their logarithms, taking for simplicity the base \(e\).
Given a sequence \(\left\{a_n^{b_n}\right\}\), we have that:
If the sequence \(\left\{b_n \log a_n\right\}\) is indeterminate (has no limit), then \(\left\{a_n^{b_n}\right\}\) is indeterminate as well.
Example 6: Computing limits by taking logarithms
taking logarithms we have
since
then
Example 7: Computing limits by taking logarithms (alternative method)
since
then
We also have the following indeterminate forms:
Taking logarithms, they correspond to the indeterminate form
since:
In the powers \(\left\{a_n^{b_n}\right\}\) the base is always (eventually) positive, hence in the form \(\infty^0\) the base is \(+\infty\).
Limits of the form
are not indeterminate forms. We have:
since, taking logarithms, we have
Example 8: Limits of the forms \(0^{\ip}\) and \(0^{\im}\)
-
Example of the case \(0^{\ip}= 0\):
\[ \lim_{n \rr \ip } \left( \frac{1}{n}\right)^{\log n} = 0^{\ip}= 0. \]Indeed we have
\[ \left( \frac{1}{n}\right)^{\log n} = e^{\log \left( \frac{1}{n}\right)^{\log n}} = e^{\log n \: \log \frac{1}{n} } \]hence
\[ \lim_{n \rr \ip } \left( \frac{1}{n}\right)^{\log n} = \lim_{n \rr \ip } e^{ \overbrace{\log n}^{\rr \ip} \: \overbrace{\log \frac{1}{n}}^{\rr \im} } = e^{\im} = 0. \] -
Example of the case \(0^{\im}= \ip\):
\[ \lim_{n \rr \ip } \left( \frac{1}{n}\right)^{-\log n} = 0^{\im}= \ip. \]Indeed we have
\[ \left( \frac{1}{n}\right)^{-\log n} = e^{\log \left( \frac{1}{n}\right)^{-\log n}} = e^{-\log n \: \log \frac{1}{n} } \]hence
\[ \lim_{n \rr \ip } \left( \frac{1}{n}\right)^{-\log n} = \lim_{n \rr \ip } e^{ \overbrace{-\log n}^{\rr \im} \: \overbrace{\log \frac{1}{n}}^{\rr \im} } = e^{\ip} = \ip. \]
2. In summary¶
In summary
-
The limit of a sum, product or quotient is the sum, product or quotient of the limits, except in the indeterminate forms.
-
If \(a_n \to \ell > 0\), then \(a_n > 0\) eventually (sign preservation).
-
If \(a_n \le b_n \le c_n\) and \(a_n, c_n \to \ell\), then \(b_n \to \ell\) (comparison theorem).
Exercises on this chapter: Computing limits of sequences · Additional exercises