Why Diagnose a Derivative?
A proposed derivative can look plausible and still contain a small but important error: a constant may have been carried through, a negative exponent may have the wrong sign, or an excluded input may have been silently restored. A reliable check does not depend on whether an answer looks familiar. It tests the expression in several independent ways.
For polynomials and rational expressions, begin by checking the terms, then check the domain, and finally use units or a numerical slope as an independent sanity check. These checks can reveal mistakes quickly. They do not all have the same logical force: a numerical estimate can expose a mismatch, but it cannot by itself prove a formula correct for every input.
The Basic Checks
For a polynomial written as a sum of powers, check every term using the power rule: for a constant \(c\) and a nonnegative integer \(n\), the derivative of \(cx^n\) is \(cnx^{n-1}\) when \(n\geq 1\), while a constant term has derivative zero. For rational expressions that can be written as sums of integer powers, use the same rule on each power, including negative powers. Keep track of the domain of the original expression while doing so.
A centered numerical slope uses values on both sides of \(a\): \(\frac{f(a+h)-f(a-h)}{2h}\). For small nonzero \(h\), this can help compare a proposed derivative value with the function’s local change. It is an estimate, not the definition itself: one finite choice of \(h\) need not give the exact derivative.
These rules explain why checking a sum term by term works. They do not give permission to ignore where the original expression is defined.
Why Term-by-Term Checking Works
The sum rule follows directly from the derivative definition. For nonzero \(h\), the difference quotient for \(f+g\) splits into two quotients:
As \(h\) tends to zero, the first quotient tends to \(f'(a)\) and the second tends to \(g'(a)\), since both derivatives exist at \(a\). The limit of their sum is the sum of their limits. Therefore the derivative of the sum is \(f'(a)+g'(a)\). For a constant multiple, the corresponding quotient is \(c\frac{f(a+h)-f(a)}{h}\); its limit is \(cf'(a)\). Applying these facts repeatedly justifies checking a finite sum one term at a time.
Worked Example: Find a Constant-Term Error
Worked Example: Find a Constant-Term Error
Consider \(f(x)=4x^3-3x^2+2x-7\). A proposed derivative is \(12x^2-6x+2-7\). Check it term by term.
For the first term, \((4x^3)'=4\cdot 3x^2=12x^2\). For the second, \((-3x^2)'=-3\cdot 2x=-6x\). For the third, \((2x)'=2\). The constant term \(-7\) differentiates to zero. Thus the derivative is
The proposed answer incorrectly retains \(-7\). At \(x=1\), the correct value is \(12-6+2=8\), whereas the proposed value is \(12-6+2-7=1\).
As a numerical check, take the centered slope at \(x=1\) with \(h=0.001\). Expanding the two values around 1 gives \(f(1+h)-f(1-h)=16h+8h^3\), so the centered slope is \(\frac{16h+8h^3}{2h}=8+4h^2\). Substituting \(h=0.001\) gives \(8.000004\), close to the correct derivative value 8 and not the proposed value 1. The term-by-term calculation establishes the formula; the numerical check corroborates its value at this particular point.
Rational Expressions: Check the Domain Too
A rational expression can often be rewritten as a sum of integer powers, making each term easier to check. But simplification does not automatically change the domain of the original function. A denominator that is zero at an input still means the original function is undefined there, even if a factor cancels algebraically.
For any claimed derivative, say explicitly which inputs are allowed. At an allowed input \(a\), the difference quotient also needs \(f(a+h)\) to be defined for sufficiently small \(h\). An excluded point is not brought back into the domain merely because a simplified formula has a value there.
Worked Example: Check a Rational Derivative and Its Domain
Let \(g(x)=2x+\frac{3}{x}\), which is defined for \(x\ne 0\). Rewrite the reciprocal term as a negative power: \(g(x)=2x+3x^{-1}\). The derivative of \(2x\) is 2. The derivative of \(3x^{-1}\) is \(3(-1)x^{-2}=-3x^{-2}\). Therefore
At \(x=1\), this gives \(g'(1)=2-3=-1\). The domain condition matters: the formula for \(g\) has no value at zero, so this calculation does not assert a derivative there.
For an independent numerical check at \(x=1\), use \(h=0.01\). The function values are \(g(1.01)=2.02+\frac{300}{101}\approx 4.99029703\) and \(g(0.99)=1.98+\frac{100}{33}\approx 5.01030303\). Thus the centered slope is approximately \(\frac{4.99029703-5.01030303}{0.02}=-1.0003\), close to \(-1\). More exactly, the centered slope simplifies to \(2-\frac{3}{1-h^2}\); substituting \(h=0.01\) gives \(-\frac{3334}{3333}\), also close to \(-1\). The estimate supports the derivative value without replacing the term-by-term calculation.
A Hole Does Not Disappear
Worked Example: Preserve a Removable Hole
Consider \(q(x)=\frac{x^2-1}{x-1}\). The original denominator is zero at \(x=1\), so the domain is \(x\ne 1\). For allowed inputs, factor and cancel:
At every point in the domain, the simplified expression has derivative 1. Hence \(q'(x)=1\) for \(x\ne 1\). It would be incorrect to state that \(q\) has a derivative at \(x=1\): the original function has no value there, and a derivative at a point requires the function to be defined at that point.
For example, at \(x=2\), the difference quotient for small nonzero \(h\) is \(\frac{q(2+h)-q(2)}{h}=\frac{(3+h)-3}{h}=1\), provided the inputs are in the domain. This checks the derivative value at 2 and illustrates why the cancellation is valid there. It does not fill in the hole at 1.
Units as a Consistency Check
When a function describes a measured quantity, its derivative has units of “output units per input unit.” A derivative of position with respect to time should therefore have units of distance per time. Unit checking is not a substitute for differentiation, but it can reveal a formula that is dimensionally inconsistent.
Worked Example: Check a Derivative Using Units and Slopes
Suppose a position function, measured in metres, is \(s(t)=\left(2\,\frac{\mathrm m}{\mathrm s^2}\right)t^2+\left(3\,\frac{\mathrm m}{\mathrm s}\right)t+1\,\mathrm m\), where \(t\) is measured in seconds. Each term has units of metres: the coefficient of \(t^2\) has units \(\mathrm m/\mathrm s^2\), and the coefficient of \(t\) has units \(\mathrm m/\mathrm s\). Differentiating each term gives
Both terms in the derivative have units \(\mathrm m/\mathrm s\). At \(t=2\,\mathrm s\), the first term is \(8\,\mathrm m/\mathrm s\), so \(s'(2\,\mathrm s)=11\,\mathrm m/\mathrm s\).
Check with a centered slope using \(h=0.01\,\mathrm s\). Direct substitution gives \(s(2.01\,\mathrm s)=15.1102\,\mathrm m\) and \(s(1.99\,\mathrm s)=14.8902\,\mathrm m\). Their difference is \(0.2200\,\mathrm m\); the time separation is \(0.02\,\mathrm s\). The slope is therefore \(\frac{0.2200\,\mathrm m}{0.02\,\mathrm s}=11\,\mathrm m/\mathrm s\), matching the derivative value. The displayed decimal values are exact for the stated inputs.
A Practical Diagnostic Sequence
When checking a proposed basic derivative, use a sequence that keeps different kinds of errors separate:
Identify denominator zeros and any other excluded inputs before simplifying.
For a rational expression, use a sum of powers when appropriate. Check coefficients, signs, exponents, and constant terms individually.
Do not include points where the original function is undefined.
Evaluate the proposed derivative and compare it with a nearby difference quotient. Treat agreement as a check, not a proof.
The derivative’s units must be output units divided by input units.
Common Pitfalls
A frequent error is applying the power rule correctly to most terms but overlooking one. For example, carrying a constant through as if it were still part of the derivative gives an incorrect extra constant, as in the first worked example. Another common error is losing the negative sign when differentiating a negative power: the derivative of \(3x^{-1}\) is \(-3x^{-2}\), not \(3x^{-2}\).
There is also a logical pitfall in numerical checking. A close estimate at one point is useful evidence, but it does not show that a formula works at every input. Conversely, an estimate far from the proposed value is a strong reason to recheck the calculation, the choice of step size, and whether the chosen inputs lie in the domain.
Check Your Understanding
For each question, identify both the calculation and any domain or interpretation check that applies.
- Differentiate \(5x^3-2x+4\) term by term. Which term contributes zero?
- For \(r(x)=\frac{4}{x}+x^2\), state the domain and find \(r'(x)\).
- A proposed derivative of \(x^2+\frac{2}{x}\) is \(2x+\frac{2}{x^2}\). Which sign is incorrect, and what is the corrected derivative?
- For \(p(x)=\frac{x^2-4}{x-2}\), what is the domain of the original function, and where is its derivative equal to 1?
- If \(d(t)\) is measured in centimetres and \(t\) in minutes, what units should \(d'(t)\) have?
- Why can a numerical slope close to a proposed derivative support the answer but not prove the formula for every input?
