Definition of a Limit

Read this section to learn how a limit is defined. Work through practice problems 1-6.

The Formal Definition of Limit

The ideas of the previous examples and practice problems can be stated for general functions and limits, and they provide the basis for the definition of limit which is given in the box. The use of the lower case Greek letters \varepsilon (epsilon) and \delta (delta) in the definition is standard, and this definition is sometimes called the "epsilon-delta" definition of limit.

Definition of \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} \mathbf{f}(\mathbf{x})=\mathbf{L}:

\lim\limits_{x \rightarrow a} f(x)=L means

for every given \varepsilon > 0 there is a \delta > 0 so that (Fig. 10)

if \quad x \quad is within \delta units of a( and x \neq a)
then f(x) is within \varepsilon units of L.
(Equivalently: |\mathrm{f}(\mathrm{x})-\mathrm{L}| < \varepsilon \quad whenever 0 < |\mathrm{x}-\mathrm{a}| < \delta).

In this definition, \varepsilon represents the given distance on either side of the limiting value \mathrm{y}=\mathrm{L}, and \delta is the distance on each side of the point x=a on the x-axis that we have been finding in the previous examples. This definition has the form of \mathrm{a} a "challenge and reponse:" for any positive challenge \varepsilon (make f(c) within \varepsilon of \mathrm{L}), there is a positive response \delta (start with \mathrm{x} within \delta of \mathrm{a} and x \neq a).

Example 4: In Fig. 11a, \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} \mathrm{f}(\mathrm{x})=\mathrm{L}, and a value for \varepsilon is given graphically as a length. Find a length for \delta that satisfies the definition of limit (so "if \mathrm{x} is within \delta of \mathrm{a}( and \mathrm{x} \neq \mathrm{a}), then \mathrm{f}(\mathrm{x}) is within \varepsilon of \mathrm{L}").


Solution: Follow the steps outlined in Example 3. The length for \delta is shown in Fig. 11b, and any shorter length for \delta also satisfies the definition.


Practice 4: In Fig. 12, \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} \mathrm{f}(\mathrm{x})=\mathrm{L}, and a value for \varepsilon is given graphically as a length. Find a length for \delta that satisfies the definition of limit.


Example 5: Prove that \lim\limits_{\mathrm{x} \rightarrow 3} 4 \mathrm{x}-5=7

Solution: We need to show that

"for every given \varepsilon > 0 there is a \delta > 0 so that

if \quad x \quad is within \delta units of 3( and x \neq 3)

then 4 x-5 is within \varepsilon units of 7".

Actually there are two things we need to do. First, we need to find a value for \delta (typically depending on \varepsilon), and, second, we need to show that our \delta really does satisfy the "if - then" part of the definition.

i. Finding \delta is similar to part (c) in Example 1 and Practice 1: assume 4 \mathrm{x}-5 is within \varepsilon units of 7 and solve for x. If 7-\varepsilon < 4 x-5 < 7+\varepsilon, then 12-\varepsilon < 4 x < 12+\varepsilon and 3-\varepsilon / 4 < x < 3+\varepsilon / 4, so x is within \varepsilon / 4 units of 3. Put \delta=\varepsilon / 4

ii. To show that \delta=\varepsilon / 4 satisfies the definition, we merely reverse the order of the steps in part i. Assume that x is within \delta units of 3. Then 3-\delta < x < 3+\delta so

\begin{array}{ll}3-\varepsilon / 4 < \mathrm{x} < 3+\varepsilon / 4 & (\text { replacing } \delta \text { with } \varepsilon / 4) \\ 12-\varepsilon < 4 \mathrm{x} < 12+\varepsilon & \text { (multiplying by 4), and } \\ 7-\varepsilon
 < 4 \mathrm{x}-5 < 7+\varepsilon & \text { (subtracting 5), so }\end{array} 

we can conclude that \mathrm{f}(\mathrm{x})=4 \mathrm{x}-5 is within \varepsilon units of 7. This formally verifies that \lim\limits_{\mathrm{x} \rightarrow 3} 4 \mathrm{x}-5=7.

Practice 5: Prove that \lim\limits_{x \rightarrow 4} 5 x+3=23.

The method used to prove the values of the limits for these particular linear functions can also be used to prove the following general result about the limits of linear functions.

Theorem: \quad \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} \mathrm{m} \mathrm{x}+\mathrm{b}=\mathrm{ma}+\mathrm{b}

Proof: Let f(x)=m x+b.

Case 1: \mathrm{m}=0. Then \mathrm{f}(\mathrm{x})=0 \mathrm{x}+\mathrm{b}=\mathrm{b} is simply a constant function, and any value for \delta > 0 satisfies the definition. Given any value of \varepsilon > 0, let \delta=1 (any positive value for \delta works). If x is is within 1 unit of \mathrm{a}, then f(x)-f(a)=b-b=0 < e, so we have shown that for any \varepsilon > 0, there is a \delta > 0 which satisfies the definition.

Case 2: \mathrm{m} \neq 0. Then f(x)=m x+b. For any \varepsilon > 0, put \delta=\frac{\varepsilon}{\mid m \mid} > 0. If x is within \delta=\frac{\varepsilon}{\mid m \mid} of \mathrm{a}, then

\mathrm{a}-\frac{\varepsilon}{\mid m \mid} < \mathrm{x} < \mathrm{a}+\frac{\varepsilon}{|\mathrm{m}|} \quad \text { so}-\frac{\varepsilon}{\mid m \mid} < \mathrm{x}-\mathrm{a} < \frac{\varepsilon}{\mid m \mid} \quad   and  |\mathrm{x}-\mathrm{a}| < \frac{\varepsilon}{\mid m \mid}

Then the distance between \mathrm{f}(\mathrm{x}) and \mathrm{L}=\mathrm{ma}+\mathrm{b} is |\mathrm{f}(\mathrm{x})-\mathrm{L}|=|(\mathrm{m} \mathrm{x}+\mathrm{b})-(\mathrm{ma}+\mathrm{b})|=\mid m|\cdot| \mathrm{x}-\mathrm{a}| < \mid m| \frac{\varepsilon}{\frac{\varepsilon}{\mathrm{m} \mid}}=\varepsilon so \mathrm{f}(\mathrm{x}) is within \varepsilon of \mathrm{L}=\mathrm{ma}+\mathrm{b}. (Fig. 13)


In each case, we have shown that "given any \varepsilon > 0, there is a \delta > 0" that satisfies the rest of the definition is satisfied.

If there is even a single value of \varepsilon for which there is no \delta, then the function does not satisfy the definition, and we say that the limit "does not exist".

Example 6: Let f(x)=\left\{\begin{array}{ll}2 & \text { if } x < 1 \\ 4 & \text { if } x > 1\end{array}\right. as is shown in Fig. 14.


Use the definition to prove that \lim\limits_{x \rightarrow 1} f(x) does not exist.

Solution: One common proof technique in mathematics is called "proof by contradiction," and that is the method we use here. Using that method in this case, (i) we assume that the limit does exist and equals some number \mathrm{L}, (ii) we show that this assumption leads to a contradiction, and (iii) we conclude that the assumption must have been false. Therefore, we conclude that the limit does not exist.

(i) Assume that the limit exists: \quad \lim\limits_{\mathrm{x} \rightarrow 1} \mathrm{f}(\mathrm{x})=\mathrm{L} for some value for \mathrm{L}. Let \varepsilon=\frac{1}{2}. (The definition says "for every ε" so we can pick this value. Why we chose this value for \varepsilon shows up later in the proof). Then, since we are assuming that the limit exists, there is a \delta > 0 so that if \mathrm{x} is within \delta of 1 then \mathrm{f}(\mathrm{x}) is within \varepsilon of \mathrm{L}.

(ii) Let x_{1} be between 1 and 1+\delta. Then x_{1} > 1 so f\left(x_{1}\right)=4. Also, x_{1} is within \delta of 1 so f\left(x_{1}\right)= 4 is within \frac{1}{2} of \mathrm{L}, and \mathrm{L} is between 3.5 and 4.5: 3.5 < \mathbf{L} < 4.5.

Let x_{2} be between 1 and 1-\delta. Then x_{2} < 1 so f\left(x_{2}\right)=2. Also, x_{2} is within \delta of 1 so f\left(x_{2}\right)= 2 is within \frac{1}{2} of \mathrm{L}, and \mathrm{L} is between 1.5 and 2.5: \mathbf{1. 5} < \mathbf{L} < \mathbf{2. 5}.

(iii) The two inequalities in bold print provide the contradiction we were hoping to find. There is no value \mathrm{L} that simultaneously satisfies 3.5 < \mathrm{L} < 4.5 and 1.5 < \mathrm{L} < 2.5, so we can conclude that our assumption was false and that f(x) does not have a limit as x \rightarrow 1.

Practice 6: Use the definition to prove that \lim\limits_{x \rightarrow 0} \frac{1}{x} does not exist (Fig. 15).



Two Limit Theorems

The theorems and their proofs are included here so you can see how such proofs proceed - you have already used these theorems to evaluate limits of functions. There are rigorous proofs of all of the other limit properties, but they are somewhat more complicated than the proofs given here.

Theorem: If \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} \mathrm{f}(\mathrm{x})=\mathrm{L}, then \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} \mathrm{k} \cdot \mathrm{f}(\mathrm{x})=\mathrm{k} \cdot \mathrm{L}.

Proof: Case \mathrm{k}=0: The Theorem is true but not very interesting: \quad \lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} 0 \cdot \mathrm{f}(\mathrm{x})=\lim\limits_{\mathrm{x} \rightarrow \mathrm{a}} 0=0 \cdot \mathrm{L}.

Case k \neq 0: Since \lim\limits_{x \rightarrow a} f(x)=L, then, by the definition, for every \varepsilon > 0 there is a \delta > 0 so that |f(x)-L| < \varepsilon whenever |x-a| < \delta. For any \varepsilon > 0, we know \frac{\varepsilon}{|k|}
 > 0 and pick a value of \delta that satisfies |f(x)-L| < \frac{\varepsilon}{|k \mid} whenever |x-a| < \delta. When 

|x-a| < \delta \quad("x is within \delta of \mathrm{a}") then |f(x)-L| < \frac{\varepsilon}{|k|} \quad\left("f(x)\right. is within \frac{\varepsilon}{|k|} of L") so |k| \cdot|f(x)-L| < \varepsilon \quad (multiplying each side by |\mathrm{k}| > 0) and |\mathrm{k} \cdot \mathrm{f}(\mathrm{x})-\mathrm{k} \cdot \mathrm{L}| < \varepsilon \quad(\mathrm{k} \cdot \mathrm{f}(\mathrm{x}) is within \varepsilon of \mathrm{k} \cdot \mathrm{L})

Theorem: If \lim\limits_{x \rightarrow a} f(x)=L and \lim\limits_{x \rightarrow a} g(x)=M, then \lim\limits_{x \rightarrow a} f(x)+g(x)=L+M.

Proof: Assume that \lim\limits_{x \rightarrow a} f(x)=L and \lim\limits_{x \rightarrow a} g(x)=M. Then, given any \varepsilon > 0, we know \varepsilon / 2 > 0 and that there are deltas for \mathrm{f} and \mathrm{g}, \delta_{\mathrm{f}} and \delta_{\mathrm{g}}, so that 

if |x-a| < \delta_{f}, then |f(x)-L| < \varepsilon / 2 ("if x is within \delta_{f} of a, then f(x) is within \varepsilon / 2 of L", and 

if |x-a| < \delta_{g}, then \lg (x)-M \mid < \varepsilon / 2 ("if x is within \delta_{g} of \mathrm{a}, then g(x) is within \varepsilon / 2 of \mathrm{M}").

Let \delta be the smaller of \delta_{f} and \delta_{g}. If |x-a| < \delta, then |f(x)-L| < \varepsilon / 2 and |g(x)-M| < \varepsilon / 2 so

\mid(f(x)+g(x))-(L+M))|=|(f(x)-L)+(g(x)-M) \mid \quad (rearranging the terms)

 \leq |f(x)-L|+|g(x)-M| \quad (by the Triangle Inequality for absolute values)

< \frac{\varepsilon}{2}+\frac{\varepsilon}{2}=\varepsilon. \quad (by the definition of the limits for \mathrm{f} and \mathrm{g}).