Two Neighborhoods, Two Different Roles
A limit compares two kinds of closeness. Inputs \(x\) are required to be close to a proposed input \(a\), and the corresponding function values \(f(x)\) are required to be close to a proposed limit \(L\). Epsilon and delta measure these two distances on the number line.
The distinction is important: delta describes a horizontal neighborhood of \(a\), while epsilon describes a vertical neighborhood of \(L\). The definition asks whether inputs in the first neighborhood (apart from \(a\) itself) have function values in the second.
Here, \(\varepsilon\) and \(\delta\) are positive distances. The implication tells us which inputs require control: those in the domain that are less than \(\delta\) units from \(a\), but are not equal to \(a\). For every such input, the output must be less than \(\varepsilon\) units from \(L\).
Open Neighborhoods on the Number Line
An open neighborhood of radius \(\varepsilon\) around \(L\) is the interval of real numbers less than \(\varepsilon\) units from \(L\). Its endpoints are not included because the required distance is strictly less than \(\varepsilon\).
For a limit, the output \(y\) in this statement is \(f(x)\). Thus the epsilon neighborhood is the interval of acceptable function values:
The delta condition describes inputs near \(a\). Because \(x=a\) is excluded, it is a deleted neighborhood: the open interval around \(a\), with its centre removed.
In interval notation, the deleted neighborhood is \((a-\delta,a)\cup(a,a+\delta)\). The limit condition says that every point in this input set that belongs to \(D\) must be sent into the epsilon neighborhood of \(L\).
Why the Interval Form Is Equivalent
The interval picture is not a different definition from the absolute-value form. It follows directly from what absolute value measures: the distance between two real numbers.
For any positive number \(r\), the inequality \(|u-c|<r\) says that the distance from \(u\) to the centre \(c\) is less than \(r\). Subtracting \(c\) from the expression inside the absolute value gives the equivalent double inequality \(-r<u-c<r\). Adding \(c\) throughout gives \(c-r<u<c+r\). Applying this first to the output and then to the input yields the neighborhood interpretation.
Combining these equivalences gives the interval version of the limit condition: for each chosen \(\varepsilon>0\), there must be a \(\delta>0\) such that every \(x\in D\cap((a-\delta,a)\cup(a,a+\delta))\) has \(f(x)\in(L-\varepsilon,L+\varepsilon)\).
Worked Example: Locating Both Neighborhoods
Worked Example: Locating Both Neighborhoods
Suppose the proposed input is \(a=-1\), the proposed limit is \(L=2\), and the tolerances are \(\delta=0.6\) and \(\varepsilon=0.25\). Write down the input and output neighborhoods.
First, the delta interval has centre \(-1\) and radius \(0.6\). Its endpoints are \(-1-0.6=-1.6\) and \(-1+0.6=-0.4\). Removing its centre gives the required input set:
Next, the epsilon interval has centre \(2\) and radius \(0.25\). Its endpoints are \(2-0.25=1.75\) and \(2+0.25=2.25\), so acceptable outputs lie in \((1.75,2.25)\).
Therefore, the condition for these particular tolerances is that every domain input in \((-1.6,-1)\cup(-1,-0.4)\) has a function value in \((1.75,2.25)\). The interval endpoints are excluded in both cases, and \(x=-1\) is excluded even though it is the centre of the input interval.
Check: The input interval is \(0.6\) units to either side of \(-1\), and the output interval is \(0.25\) units to either side of \(2\), as required.
Worked Example: A Domain Restricts the Input Neighborhood
Worked Example: A Domain Restricts the Input Neighborhood
Let \(D=[0,\infty)\), and consider a proposed limit at \(a=0\) with proposed value \(L=-2\). For \(\delta=0.5\) and \(\varepsilon=1.2\), identify the inputs that require control and the acceptable output interval.
The full open interval of radius \(0.5\) around \(0\) is \((-0.5,0.5)\). Removing the centre leaves \((-0.5,0)\cup(0,0.5)\). However, the function is defined only for inputs in \(D=[0,\infty)\). Intersecting with the domain leaves the required input set \((0,0.5)\).
For the outputs, the epsilon interval has endpoints \(-2-1.2=-3.2\) and \(-2+1.2=-0.8\). Thus acceptable values lie in \((-3.2,-0.8)\).
For these tolerances, every \(x\in(0,0.5)\) must have \(f(x)\in(-3.2,-0.8)\). There are no domain inputs to check on the left of \(0\). The conclusion is about inputs in the domain; it does not require the function to be defined at negative inputs.
Check: The input set contains only positive domain points less than \(0.5\) units from \(0\), and the output endpoints are exactly \(1.2\) units from \(-2\).
Worked Example: Checking Individual Inputs
Worked Example: Checking Individual Inputs
Suppose \(a=2\), \(L=4\), \(\delta=0.25\), and \(\varepsilon=0.5\). A table lists the following function values. Decide which listed inputs are subject to the condition, and whether their listed outputs satisfy it.
| Input \(x\) | Listed value \(f(x)\) |
|---|---|
| 1.8 | 3.7 |
| 1.9 | 4.2 |
| 2 | 9 |
| 2.1 | 3.6 |
| 2.3 | 4.4 |
The input neighborhood is \((2-0.25,2)\cup(2,2+0.25)=(1.75,2)\cup(2,2.25)\). The output neighborhood is \((4-0.5,4+0.5)=(3.5,4.5)\).
For \(x=1.8\), \(|1.8-2|=0.2<0.25\), so it is subject to the condition. Its output error is \(|3.7-4|=0.3<0.5\), so this listed value meets the requirement.
For \(x=1.9\), \(|1.9-2|=0.1<0.25\), so it is also subject to the condition. Its output error is \(|4.2-4|=0.2<0.5\), so it meets the requirement too.
The input \(x=2\) is excluded, regardless of its listed function value. For \(x=2.1\), the input distance is \(|2.1-2|=0.1<0.25\), so it is subject to the condition; its output error is \(|3.6-4|=0.4<0.5\), so it meets the requirement. Finally, for \(x=2.3\), the input distance is \(|2.3-2|=0.3\), which is not less than \(0.25\). This input is outside the required neighborhood, so its output does not affect this particular check.
Check: The three listed inputs that must be checked are \(1.8\), \(1.9\), and \(2.1\), and each of their listed output values lies in \((3.5,4.5)\). This verifies only those listed points; a table of finitely many inputs does not establish the condition for every input in the interval.
How to Read the Definition as a Mapping
It is useful to keep the direction of the requirement clear. First, an output tolerance is specified: choose an open target interval around \(L\). Then a suitable input interval around \(a\) must be found. The function must send every relevant input in that interval into the target interval.
Place \(L\) on the output number line and form the open interval \((L-\varepsilon,L+\varepsilon)\).
Place \(a\) on the input number line and form \((a-\delta,a+\delta)\), removing \(a\) itself.
Only inputs in \(D\) are relevant, so use the intersection with the domain.
Every remaining input must have an output inside the epsilon interval.
The phrase “every remaining input” matters. It is not enough for some nearby inputs to have acceptable outputs, nor is it enough for the function value at \(a\) to be acceptable. The condition concerns all domain inputs sufficiently close to \(a\), other than \(a\) itself.
Common Mistake: Including the Centre
A frequent error is to replace the deleted input neighborhood with the entire interval, then require the function value at \(a\) to satisfy the output tolerance. For example, with \(a=2\), \(L=4\), \(\delta=0.25\), and \(\varepsilon=0.5\), an incorrect statement would be:
This incorrectly includes \(x=2\). The limit condition excludes that input, so the correct input set is \((1.75,2)\cup(2,2.25)\). A limit describes nearby function values, not necessarily the value of the function at the centre. The function may be undefined at \(a\), or its value there may differ from \(L\), without changing which nearby inputs the definition tests.
Another error is to reverse the roles of epsilon and delta. Epsilon measures distance on the output number line from \(L\); delta measures distance on the input number line from \(a\). Keeping the two centres and their corresponding radii together prevents this confusion.
Why Neighborhoods Matter
The neighborhood viewpoint turns the definition into a geometric task. Instead of treating \(|x-a|\) and \(|f(x)-L|\) as isolated expressions, we can see a permitted input region and a required output region. This makes the logic of the definition visible: the chosen input region must be narrow enough that all its relevant points map inside the target region.
It also clarifies the role of the quantifiers. The output tolerance \(\varepsilon\) can be any positive number, so the target interval may be made as narrow as desired around \(L\). For each such interval, a suitable positive \(\delta\) must give an input neighborhood whose function values all lie inside it. The radius \(\delta\) may depend on \(\varepsilon\); the definition does not require the same input interval to work for every target interval.
Finally, a visual sketch or a table can help locate neighborhoods and test individual points, but it cannot by itself confirm that every point in an interval behaves as required. The neighborhood formulation identifies precisely what a complete argument must control.
Check Your Understanding
Use the neighborhood interpretation to answer these questions.
- Write the epsilon neighborhood of \(L=3\) when \(\varepsilon=0.4\).
- Write the deleted delta neighborhood of \(a=-2\) when \(\delta=0.5\).
- Why is the input \(x=a\) excluded from the limit condition?
- If the domain is \(D=[1,\infty)\) and \(a=1\), which part of a deleted neighborhood around \(a\) can contain domain inputs?
- Why does checking several points in a delta neighborhood not establish the limit condition for every input there?
