Why the Reciprocal Difference Quotient Needs Care
For many functions, finding a derivative from its definition involves expanding or rearranging an expression until a factor that matches the denominator can be cancelled. For the reciprocal function, \(f(x)=1/x\), the key first step is different: combine the two fractions using a common denominator.
That step reveals a factor of \(x-a\), which can then be cancelled when \(x\ne a\). The resulting expression has a limit as \(x\) approaches \(a\), provided \(a\) is nonzero. Keeping track of those restrictions is essential: the original quotient is not defined at \(x=a\), and the function itself is not defined at zero.
Set Up the Quotient and Its Domain
For \(f(x)=1/x\), the input \(a\) must be nonzero, since \(f(a)\) must exist. Also, the nearby input \(x\) must be nonzero so that \(f(x)\) exists, and it must differ from \(a\) so that the denominator \(x-a\) is nonzero. Thus, before simplifying, the quotient is
The common denominator of \(1/x\) and \(1/a\) is \(ax\). Subtracting the fractions gives a numerator of \(a-x\), not \(x-a\). That sign matters: \(a-x=-(x-a)\). Once the expression is combined, the shared factor can be cancelled because the quotient is being considered only for \(x\ne a\).
Derivation of the Derivative
Using the common denominator \(ax\), the numerator becomes \(a-x\).
Since \(a-x=-(x-a)\), the factor \(x-a\) is visible.
Because \(x\ne a\), cancel \(x-a\), then take the limit as \(x\to a\).
The last expression is equal to the original quotient for \(x\ne 0\) and \(x\ne a\). As \(x\) approaches the fixed nonzero number \(a\), the product \(ax\) approaches \(a^2\), which is nonzero. Therefore the simplified expression has the limit shown below.
The Same Argument in the \(h\)-Form
The difference quotient can also be written using an increment \(h\), where the nearby input is \(a+h\). In this form, \(h\ne 0\) is required for the quotient, and \(a+h\ne 0\) is required for the reciprocal function value to exist.
The cancellation of \(h\) is valid because \(h\ne 0\). For fixed \(a\ne 0\), when \(h\) is sufficiently close to zero, \(a+h\) remains nonzero. For example, if \(|h|<|a|/2\), then \(|a+h|\ge |a|-|h|>|a|/2>0\). Thus the simplified expression is defined for all sufficiently small nonzero \(h\), and
Both forms describe the same limiting process: the nearby input approaches \(a\), either by writing it as \(x\) or as \(a+h\). The common denominator is what makes the cancellation possible in either form.
Worked Examples
Worked Example: Find the Derivative at \(a=2\)
Use the \(x\)-to-\(a\) form to find \(f'(2)\) for \(f(x)=1/x\). The quotient is considered for \(x\ne 0\) and \(x\ne 2\).
Now take the limit as \(x\to 2\):
Check: substituting \(a=2\) into the derived formula gives \(-1/a^2=-1/2^2=-1/4\), matching the limit. The cancellation above was valid for \(x\ne 2\), as required by the original quotient.
Worked Example: Find the Derivative at \(a=-3\)
Use the \(h\)-form to find \(f'(-3)\). Here \(h\ne 0\), and the nearby input \(-3+h\) must not equal zero.
The cancellation of \(h\) is valid for \(h\ne 0\). Taking the limit gives
Check: the general formula gives \(-1/(-3)^2=-1/9\), the same result. For \(h\) near zero, \(h-3\ne 0\), so the reciprocal expression used in the limit is defined.
Worked Example: Find the Derivative at \(a=\frac{1}{2}\)
Use the \(x\)-to-\(a\) form to find \(f'(\frac{1}{2})\). The quotient requires \(x\ne 0\) and \(x\ne \frac{1}{2}\).
Taking the limit as \(x\to \frac{1}{2}\) gives
Check: the general formula gives \(-1/(\frac{1}{2})^2=-1/(\frac{1}{4})=-4\), which agrees with the limit. The factor \(x-\frac{1}{2}\) was cancelled only for \(x\ne \frac{1}{2}\).
A Common Mistake: Substituting Before Taking the Limit
A frequent error is to set \(x=a\) in the original difference quotient before simplifying. For example, at \(a=2\), that substitution produces
This is not a derivative value: the quotient is undefined at \(x=2\). The derivative is found from the limit of the quotient at nearby inputs, not by evaluating the quotient at the point itself. The corrected calculation first simplifies for \(x\ne 2\):
and then takes the limit of the right-hand side as \(x\to 2\), obtaining \(-1/4\). The simplified expression is not claimed to equal the original quotient at \(x=2\); it is used to compute the limit of the quotient near that point.
Why the Restrictions Matter
The two exclusions in the \(x\)-form have distinct roles. The condition \(x\ne 0\) keeps \(f(x)=1/x\) defined. The condition \(x\ne a\) keeps the difference quotient defined. Cancellation does not erase these original restrictions; it only gives a simpler expression that agrees with the quotient where the quotient exists.
The requirement \(a\ne 0\) is equally important. If \(a=0\), then \(f(a)\) does not exist, so there is no difference quotient for this function at zero. It would be incorrect to use the formula \(-1/a^2\) there: its derivation assumed \(a\ne 0\) throughout.
This calculation illustrates a useful general pattern in limits. An expression may be undefined at the point of interest, yet simplify to an expression that has a finite limit near that point. The common-denominator step exposes the factor responsible for the apparent \(0/0\) form, while the limit—not direct substitution into the original quotient—determines the derivative.
Check Your Understanding
For each question, keep track of which input values are excluded from the original quotient before taking a limit.
- For \(f(x)=1/x\), write the \(x\)-to-\(a\) difference quotient and state all restrictions on \(x\) and \(a\).
- Use a common denominator to simplify the quotient at an arbitrary nonzero input \(a\). Which factor can be cancelled, and under what condition?
- Find \(f'(4)\) from the difference quotient and check your result with the formula derived in this tutorial.
- Find \(f'(-\frac{1}{2})\) using the \(h\)-form and show why the nearby input remains in the domain for sufficiently small \(h\).
- Explain why substituting \(x=a\) into the original quotient does not give the derivative, even though the derivative limit may exist.
