Home » Sequences » Page 2

Tag: Sequences

Prove that a sequence cannot converge to two different limits

Prove that a sequence cannot converge to two different limits.


Proof. Consider a sequence \{ a_n \} such that

    \[ \lim_{n \to \infty} a_n = L \qquad \text{and} \qquad \lim_{n \to \infty} a_n = L^*. \]

We show that we must have L = L^*. By the definition of the limit we know that for all \varepsilon >  0 there exist a positive integers N_1 and N_2 such that

    \begin{align*}  |a_n - L| &< \frac{\varepsilon}{2} & \text{for all } n \geq N_1 \\  |a_n - L^*| &< \frac{\varepsilon}{2} & \text{for all } n \geq N_2. \end{align*}

Let N = \max \{ N_1, N_2 \}, then for all n \geq N and all \varepsilon > 0 we have

    \[ |a_n - L| < \frac{\varepsilon}{2} \qquad \text{and} \qquad |a_n - L^*| < \frac{\varepsilon}{2}. \]

Hence,

    \begin{align*}  &&|a_n - L| + |L^* - a_n| &< \varepsilon \\  \implies && |a_n - L + L^* - a_n| &< \varepsilon \\  \implies && |L^* - L| &< \varepsilon \\  \implies && L = L^*. \end{align*}

(The final line follows since if L \neq L^*, then |L^* - L| > 0, so setting \varepsilon = |L - L^*| would contradict that |L^* - L| < \varepsilon for all \varepsilon > 0.) Therefore, a sequence cannot converge to two different limits. \qquad \blacksquare

Find an N such that the given convergent sequence is within ε of its limit

Consider the convergent sequence \{ a_n \} with terms defined by

    \[ a_n = (-1)^n \left( \frac{9}{10} \right)^n. \]

Let L = \lim_{n \to \infty} a_n. Find the value of L and the value of N such that |a_n - L| < \varepsilon for all n \geq N for each of the following values of \varepsilon:

  1. \varepsilon = 1,
  2. \varepsilon = 0.1,
  3. \varepsilon = 0.01,
  4. \varepsilon = 0.001,
  5. \varepsilon = 0.0001.

First, we know

    \[ \lim_{n \to \infty} (-1)^n \left( \frac{9}{10}\right)^n = \lim_{n \to \infty} \left(-\frac{9}{10}\right)^n = 0 \]

since \left|\frac{-9}{10}\right| = \frac{9}{10} < 1, so by (10.10) on page 380 of Apostol we know the limit is 0.
Then we have,

    \begin{align*}  |a_n - L| < \varepsilon && \implies && \left| \left(-\frac{9}{10} \right)^n \right| &< \varepsilon \\[9pt]  && \implies && \left( \frac{9}{10} \right)^n &< \varepsilon \\[9pt]  && \implies && \log \left(\frac{9}{10}\right)^n  &< \log \varepsilon \\[9pt]  && \implies && n \log \frac{9}{10} &< \log \varepsilon \\[9pt]  && \implies && n &> \frac{\log \varepsilon}{\log \frac{9}{10}}. \end{align*}

We reversed the inequality sign in the final step since \log \frac{9}{10} < 0 since \frac{9}{10} < 1. Thus, if N > \frac{\log \varepsilon}{\log \frac{9}{10}} then for every n \geq N we have |a_n| < \varepsilon. We compute for the given values of \varepsilon as follows:

  1. \varepsilon = 1 implies N = \frac{\log 1}{\log 0.9} = 0.
  2. \varepsilon = 0.1 implies N = \frac{\log 0.1}{\log 0.9}.
  3. \varepsilon = 0.01 implies N = \frac{\log 0.01}{\log 0.9}.
  4. \varepsilon = 0.001 implies N = \frac{\log 0.001}{\log 0.9}.
  5. \varepsilon = 0.0001 implies N = \frac{\log 0.0001}{\log 0.9}.

Find an N such that the given convergent sequence is within ε of its limit

Consider the convergent sequence \{ a_n \} with terms defined by

    \[ a_n = \frac{2n}{n^3 + 1}. \]

Let L = \lim_{n \to \infty} a_n. Find the value of L and values of N such that |a_n - L| < \varepsilon for all n \geq N for each of the following values of \varepsilon:

  1. \varepsilon = 1,
  2. \varepsilon = 0.1,
  3. \varepsilon = 0.01,
  4. \varepsilon = 0.001,
  5. \varepsilon = 0.0001.

First, we know

    \[ \lim_{n \to \infty} \frac{2n}{n^3 + 1} = \lim_{n \to \infty} \frac{\frac{2}{n^2}}{1 + \frac{1}{n^3}} = 0 \quad \implies \quad L= 0. \]

So then we have,

    \begin{align*}  |a_n - L| < \varepsilon && \implies && \left| \frac{2n}{n^3+1} \right| &< \varepsilon \\[9pt]  && \implies && \frac{2n}{n^3 + 1} &< \varepsilon \\[9pt]  && \implies && 2n  &< \varepsilon (n^3+1) \\[9pt]  && \implies && \frac{2}{\varepsilon} &< n^2 + \frac{1}{n}.  \end{align*}

Thus, if N > \sqrt{\frac{2}{\varepsilon}} then for every n \geq N we have |a_n| < \varepsilon. We compute for the given values of \varepsilon as follows:

  1. \varepsilon = 1 implies N \geq \sqrt{\frac{2}{1}} = \sqrt{2}.
  2. \varepsilon = 0.1 implies N \geq \sqrt{\frac{2}{.1}} = \sqrt{20} = 2 \sqrt{5}.
  3. \varepsilon = 0.01 implies N \geq \sqrt{\frac{2}{.01}} = \sqrt{200} = 10\sqrt{2}.
  4. \varepsilon = 0.001 implies N \geq \sqrt{\frac{2}{.001}} = \sqrt{2000} = 20 \sqrt{5}.
  5. \varepsilon = 0.0001 implies N \geq \sqrt{\frac{2}{.0001}} = \sqrt{20000} = 100 \sqrt{2}.

Determine the convergence or divergence of f(n) = ne-πin / 2

Consider the function f(n) defined by

    \[ f(n) = ne^{-\frac{\pi i n}{2}}. \]

Determine whether \{ f(n) \} converges or diverges, and if it converges find its limit.


This sequence diverges.
Proof. We saw in this exercise (Section 10.4, Exercise #20) that the sequence defined by f(n) = e^{-\frac{\pi i n}{2}} diverges. We could use this to show that this sequence diverges (since for n = 4k we have f(n) = n and so \{ f(n) \} cannot approach a finite limit). For practice, we can also prove this directly from the definition by contradiction as follows. Suppose there exists a real number L and a positive integer N such that for all n > N and all \varepsilon > 0 we have

    \[ |f(n) - L| < \varepsilon. \]

Since N is positive we know 4N > N and 4N+2 > N. So, letting \varepsilon = \frac{1}{2}, we have

    \begin{align*}   && | f(4N) - L | &< \frac{1}{2} & \text{and} && |f(4N+2) - L| &< \frac{1}{2} \\[9pt]  \implies && | 4N - L | &< \frac{1}{2} & \text{and} && |-(4N+2) - L| &< \frac{1}{2} \\[9pt]   && && \implies && |4N+2+L| &< \frac{1}{2}. \end{align*}

Adding these two expressions and using the triangle inequality we have,

    \begin{align*}  && |4N-L| + |4N+2+L| &< 1 \\[9pt]  \implies && |4N - L + 4N + 2 + L| &< 1 \\[9pt]  \implies && |8N + 2| &< 1 \\[9pt]  \implies && 2 &< 1. \end{align*}

This is a contradiction. Hence, the sequence \{ f(n) \} does not tend to a limit L.

Determine the convergence or divergence of f(n) = (1 / n) e-πin / 2

Consider the function f(n) defined by

    \[ f(n) = \frac{1}{n} e^{-\frac{\pi i n}{2}}. \]

Determine the convergence or divergence of the sequence \{ f(n) \} and if it converges determine its limit.


This sequence converges since

    \begin{align*}  \lim_{n \to \infty} |f(n)| &= \lim_{n \to \infty} \left| \frac{1}{n} \cdot e^{-\frac{\pi i n}{2}} \right| \\[9pt]  &= \lim_{n \to \infty} \left( \frac{1}{n} \cdot \left| e^{-\frac{\pi i n}{2}} \right| \right). \end{align*}

But each of the limits in the product exists since

    \[ \lim_{n \to \infty} \frac{1}{n} = 0 \]

and

    \[ \lim_{n \to \infty} \left| e^{-\frac{\pi i n}{2}} \right| = \lim_{n \to \infty} 1  = 1. \]

Hence,

    \[ \lim_{n \to \infty} |f(n)| = 0 \cdot 1 = 0. \]

Therefore, \lim_{n \to \infty} f(n) = 0 (as we saw in this exercise, if f(n) is complex valued and \lim_{n \to \infty} |f(n)| = 0 then the sequence \lim_{n \to \infty} f(n) = 0 as well). Hence, the sequence \{ f(n) \} converges to 0.

Determine the convergence or divergence of f(n) = e-πin / 2

Consider the function f(n) defined by

    \[ f(n) = e^{-\frac{\pi i n}{2}}. \]

Determine whether the sequence \{ f(n) \} converges or diverges, and if it converges determine its limit.


The sequence \{ f(n) \} diverges.
Proof. First, we use DeMoivre’s theorem to write,

    \begin{align*}  f(n) &= e^{-\frac{\pi i n}{2}} \\[9pt]  &= \left( e^{-\frac{\pi i}{2}} \right)^n \\[9pt]  &= \cos \left( \frac{-n \pi}{2} \right) - i \sin \left( \frac{n \pi}{2} \right). \end{align*}

(Note: On page 380 Apostol claims (without proof) that the sequence defined by f(n) = \sin \left( \frac{n \pi}{2} \right) is divergent. If you want to accept that then this sequence will diverge since a complex-valued sequence f(n) = u(n) + i v(n) diverges if and only if the sequences defined by the functions u(n) and v(n) converge. Since it is a good exercise (and maybe Apostol wanted us to prove it ourselves) we can prove this is divergent from the definition.)

Suppose, for the sake of contradiction, that the sequence \{ f (n) \} converges to a limit L. Then we know for every \varepsilon > 0 there exists a positive integer N such that for all n > N we have

    \[ |f(n) - L| < \varepsilon. \]

Since N is positive we know 4N > N and 4N+2 > N. Hence, taking \varepsilon = \frac{1}{2} we have,

    \begin{align*}  | f(4N) - L | &< \frac{1}{2} & \text{and} && |f(4N+2) - L| &< \frac{1}{2} \\[9pt]  \implies |\cos (2N \pi) - i \sin (2N \pi) - L| &< \frac{1}{2} & \text{and} && |\cos (2N \pi + \pi) - i \sin (2N \pi + \pi) -L| &< \frac{1}{2} \\[9pt]  \implies |1 - L| &< \frac{1}{2} & \text{and} && |1+L| &< \frac{1}{2}. \end{align*}

Adding these two expressions and using the triangle inequality,

    \begin{align*}  && |1-L| + |1+L| &< 1 \\[9pt]  \implies && |1 - L + 1 + L| &< 1 \\  \implies && 2 < 1, \end{align*}

a contradiction. Hence, the sequence \{ f(n) \} diverges. \qquad \blacksquare