Linearizing Reciprocal And Rational Functions

Why Linearize These Functions?

A reciprocal or rational function can be awkward to evaluate mentally, even when the input is close to a number that gives an easy calculation. A tangent line replaces the function, near a selected base point, with a simpler linear expression. The estimate is most useful when the base point is both convenient to calculate at and close to the input of interest.

For reciprocal functions, a small change in the input does not produce the same change in the output at every point. For instance, the slope of \(1/x\) is steep near zero and much flatter when \(x\) is large. A useful linearization must therefore account for both the function value and its slope at the chosen point.

The Tangent-Line Model

For a function \(f\) differentiable at a point \(a\), its linearization at \(a\) is the tangent-line function

The value \(f(a)\) anchors the line at the known point, and the derivative \(f'(a)\) gives its slope. For inputs \(x\) close to \(a\), we use \(f(x)\approx L_a(x)\). The approximation is local: the formula for \(L_a\) exists regardless, but it is generally a useful model only near the base point.

For the reciprocal function \(f(x)=1/x\), the domain is \(x\ne 0\), and its derivative is \(f'(x)=-1/x^2\). Thus, at any nonzero base point \(a\),

The base point and the input must be in a region where the reciprocal function is defined. In particular, if an interval between \(a\) and \(x\) crosses zero, the reciprocal function is not defined throughout that interval and the tangent line at \(a\) should not be treated as a local model across the singularity.

Why the Reciprocal Model Works

The tangent-line formula is more than a convenient recipe. For \(a\ne 0\) and \(a+h\ne 0\), the difference between the reciprocal function at \(a+h\) and its linearization there can be calculated exactly. First, the model at \(x=a+h\) is

Subtract this from the actual reciprocal value and combine the terms:

In the numerator, \(a^2-(a-h)(a+h)=a^2-(a^2-h^2)=h^2\), which gives the final expression. Consequently,

The first two terms on the right are exactly the tangent-line model. The remaining term is the difference between the actual value and the model. For a fixed nonzero \(a\), as \(h\) approaches zero, that difference approaches zero; indeed, its numerator is \(h^2\), while its denominator approaches \(a^3\ne 0\). This exact identity explains why the tangent line gives a good local estimate and also lets us check the direction and size of the error.

Choosing a Convenient Base Point

For a reciprocal function, values such as \(1/2\), \(1/3\), and \(1/4\), along with their squares in the derivative, are easy to work with. A nearby input can then be written as \(a+h\). For a rational function, inspect its denominator as well: a base point that keeps the denominator simple can make both \(f(a)\) and \(f'(a)\) straightforward.

Convenience does not override domain restrictions. If a rational function is \(f(x)=p(x)/q(x)\), where \(p\) and \(q\) are differentiable and \(q(a)\ne 0\), then the quotient rule gives

The condition \(q(a)\ne 0\) is essential: at a point where the denominator vanishes, the function is not defined by this quotient, so there is no tangent-line model there. Once the base point is chosen, the same linearization formula applies as for any differentiable function.

Linearizing a Simple Rational Function

A rational function need not be a pure reciprocal to use this method. Find its value and derivative at the base point, taking care with the quotient rule, then substitute them into the tangent-line formula. A rational expression may also sometimes be rearranged into a constant plus a reciprocal term, which can reveal simple values and slopes. Either way, check that the chosen point is not a zero of the denominator.

A Common Sign Error

For \(f(x)=1/x\), the derivative is negative, not positive. If the input increases, the reciprocal decreases; a tangent line with a positive slope would predict the opposite direction. For example, an incorrect derivative \(f'(x)=1/x^2\) would give the following wrong estimate for \(1/4.1\):

This is already larger than \(1/4\), even though \(4.1>4\) and \(1/4.1<1/4\). The correction is to use \(f'(x)=-1/x^2\):

A quick direction check—whether the function should increase or decrease as the input changes—can expose a sign error before it affects later arithmetic.

Practical Checks and Limitations

A reliable linearization uses three checks. First, confirm that the base point belongs to the function's domain and that the derivative exists there. Second, compute the function value and slope at that same point. Third, make sure the target input is near the base point. A tangent line is a local model, not a replacement formula for the function at arbitrary distances.

For \(1/x\), the exact error identity shows why proximity matters, but it also highlights the danger of approaching zero: the denominator \(a^2(a+h)\) can be small. For a rational function, a nearby input close to a point where its denominator vanishes may likewise give a poor approximation. Never use a model to cross a point where the original function is undefined.

These methods turn a difficult evaluation into a calculation involving a known value, a known slope, and a small input change. That makes them useful for quick estimates, while the checks above keep the estimate tied to the function's actual domain and behavior.