Differential Calculus
Derivatives of Exponential and Logarithmic Functions
Exponential functions are exactly the functions whose rate of change is proportional to their own size (on an interval, forces ; the interval hypothesis matters, since on a disconnected domain each piece may carry its own ), which is why they govern compound interest, population growth, radioactive decay and Newtonian cooling; logarithms come along as their inverses. This topic pins down the number by exactly that property, derives the four core derivatives (, , , ), and then develops logarithmic differentiation, the technique that converts a monstrous product or quotient into a sum and is one of the two standard ways to handle a variable base with a variable exponent such as (the other being the rewrite , valid when ). Everything here sits on top of the chain rule, which you will combine with every formula below.
Where the number comes from
Fix a base and let . From the limit definition of the derivative,
The factor comes out of the limit because it does not depend on . So the derivative of every exponential function is a constant multiple of the function itself, and the constant is exactly the slope of its graph at . Write . One shows (this needs a separate argument, usually via the integral definition of ) that exists for every and increases continuously in , with and as .
Definition. is the unique base for which that slope equals :
Equivalent characterisations: Once the log derivative is available we get for free, so the mysterious constant above was always a logarithm.
The four core derivatives
Natural exponential. for all real — this is the definition of applied to with .
General exponential (). Because , exists and . The chain rule with inner function , , gives
Sanity checks: gives ; gives , correct since is constant. The hypothesis cannot be dropped: for the power is not defined on any interval (only at scattered rationals), and for the function exists only on , where it is identically while does not exist at all.
Natural logarithm. is the inverse of ; is differentiable, strictly increasing, maps the real line bijectively onto , and has everywhere, so by the Inverse Function Theorem is differentiable on . Put , so with . Differentiating both sides with respect to (chain rule on the left):
Absolute-value version. For every ,
Proof by cases: for , and the derivative is ; for , and the chain rule gives . The point must be excluded — the function is not even defined there.
Other bases. For , , and , change of base gives with a constant, so
Rule table (chain-rule form, differentiable)
| Function | Derivative | Conditions |
|---|---|---|
| all | ||
| all | ||
| constant | ||
| , , | ||
| , , | ||
| constant; all for positive integer , for negative integer , when is not an integer | ||
| , differentiable | ||
With absolute values the two log rows extend to wherever — a genuinely useful widening, since is allowed to be negative.
Logarithmic differentiation
Use it when is a product or quotient of three or more factors, when factors carry awkward powers or roots, or when the exponent itself contains .
Method. Let be differentiable with near the point of interest.
- Write .
- Break the right side up with the log laws: , , and (integer ; for a general real exponent you need ).
- Differentiate both sides in . The chain rule turns the left side into .
- Multiply through by and substitute the original expression back in.
Hypotheses that matter. Step 1 requires : at a zero of the method says nothing and you must differentiate directly. The log laws split products, quotients and powers only — there is no law for , so logarithmic differentiation does not help with sums.
Worked Example 1 — a heavy quotient. Differentiate for (where all three factors are positive).
Differentiate term by term, each with the chain rule :
Multiply by :
Check at . Then and the bracket is , so . A centred difference quotient with step gives .
Worked Example 2 — variable base and variable exponent. Differentiate for .
The power rule is illegal (the exponent is not constant) and the rule is illegal (the base is not constant). Two equivalent routes.
Route A (rewrite as base ). Since , , so by the chain rule with the product rule inside,
Route B (take logs). , so , and multiplying by gives the same thing.
Check at . There , , , so ; the numerical derivative is .
Why the power rule cannot touch
is proved under the standing hypothesis that is a constant, and under the hypothesis that is a constant. In both move at once, so neither theorem applies. The correct general statement, for , comes from :
The two familiar rules are its degenerate cases: if is constant then and the formula collapses to ; if is constant then and it collapses to . For neither term dies:
At the true slope is ; the two popular wrong answers give and . Both are badly wrong, and the numerical derivative settles it.
Common mistakes
| Wrong | Right | Why |
|---|---|---|
| chain rule: the inner derivative must appear | ||
| the power rule needs variable base and constant exponent; this is the reverse | ||
| the exponent is not constant, so no theorem licenses the power rule | ||
| ; also | ||
| outer function is squaring, inner is | ||
| change of base leaves in the denominator | ||
| and are different functions | ||
| reporting as the answer | multiply back by | logarithmic differentiation produces ; the last step is not optional |
| no such law exists | logs split products, never sums | |
| using at | use the absolute-value version instead | is undefined for ; only the absolute-value form reaches negative |
Two identities students overuse: holds only for , while holds for every real .
Fast sanity checks
- Any derivative of must still contain as a factor — exponentials never disappear under differentiation.
- Any derivative of must carry that something in a denominator, since . A leftover outside that fraction usually means you differentiated the wrong layer — the honest exceptions are nested logs, where for , and powers of logs, where ; in both the surviving is correct.
- If your answer for a variable-base-variable-exponent problem has only one term inside the bracket, you probably dropped either the term or the term.
Key terms
- the number e
- natural exponential function
- natural logarithm
- change of base formula
- chain rule
- logarithmic differentiation
- implicit differentiation
- inverse function theorem
- variable base with variable exponent
- power rule (constant exponent)
- exponential rule (constant base)
- log laws (product, quotient, power)
- absolute value inside a logarithm
- domain restriction
- exponential growth and decay
Practice Problems
Differentiate . State the domain of and evaluate exactly, then give a decimal to four places.
Show hint
Differentiate term by term. Two of the three terms are not simply "the same thing back again" — remember which one picks up a of its base, and which one picks up a reciprocal.
Show answer
Domain. and are defined for every real , but requires . So the domain of is .
Differentiate term by term using the sum and constant-multiple rules, with , (valid on ) and with :
Evaluate at :
Numerically and , so
Check. A centred difference quotient gives , matching.
Answer: on , and .
Differentiate each function and state its largest domain.
(a) (b) (c)
Show hint
Each one is a standard rule with a very simple inner function; identify first and write down before you touch the outer rule.
Show answer
(a) . Outer function , inner with . The chain rule gives
Domain: all real numbers (an exponential is never undefined). Check: at , ; the numerical derivative is .
(b) . Outer , inner with :
Domain: needs a strictly positive argument, so , i.e. ; the domain is , and the derivative formula is valid exactly there. Check: at , ; the numerical derivative is .
(c) . Change of base: , and is a constant, so
Domain: . Check: at , ; the numerical derivative is .
Answer: (a) for all real ; (b) on ; (c) on .
Let . Find , factor it completely, find all critical numbers, and use a sign chart to classify them.
Show hint
Product rule first; then look for the largest common factor of the two resulting terms before hunting for zeros — one factor is never zero, which makes the sign chart short.
Show answer
Differentiate with the product rule , taking () and ( by the chain rule with inner ):
Critical numbers. is differentiable everywhere, so critical numbers are the zeros of . Since for every real , that factor never contributes a zero. Hence
Sign chart. The sign of is the sign of , and always:
| interval | sign of | sign of | sign of | behaviour |
|---|---|---|---|---|
| increasing | ||||
| increasing | ||||
| decreasing |
Classification. At the derivative is zero but does not change sign, so gives a horizontal tangent that is not a local extremum (it is a stationary inflection-type point). At the derivative changes from to , so is a local maximum by the First Derivative Test, with
Since increases on and decreases on , this local maximum is in fact the absolute maximum of over all real numbers.
Check: , and the numerical derivative at is .
Answer: ; critical numbers (no extremum) and (absolute maximum, value ).
(a) Differentiate .
(b) Differentiate and state its domain.
(c) Show that for , then differentiate both expressions and verify that the two answers agree.
Show hint
For (a) and (b) identify the inner function before applying the and rules. For (c), push everything down to base using .
Show answer
(a) Use with , , :
Check: at this is , and the numerical derivative is .
(b) Domain first: needs a positive argument, so , i.e. . On , use with , , :
Check: at this is , and the numerical derivative is .
(c) The identity. For both sides are defined, and pushing each to base :
The exponents are the same number, so .
Differentiating the left form ( with , , ):
Differentiating the right form. Here the exponent is a constant, so the ordinary power rule is legal:
Since , the two answers are identical.
Check: at we get , and the numerical derivative is .
Answer: (a) ; (b) on ; (c) both forms give .
(a) Prove that for every .
(b) Use it to differentiate and state exactly where your formula is valid.
(c) Evaluate and .
Show hint
Split the real line at and rewrite the absolute value without bars on each piece; the negative piece needs the chain rule.
Show answer
(a) Proof by cases.
Case . Then , so and .
Case . Then , so . Apply the chain rule with outer and inner , :
Both cases give , so for all . The point must be excluded: is not even defined there.
(b) Apply the chain-rule form (valid wherever ) with , :
Validity: we need , i.e. and . So the formula holds on — note this includes negative and between and , where ; the absolute value is exactly what buys us those intervals.
(c)
Check: numerical derivatives of at and give and .
Answer: (a) proved by cases; (b) for ; (c) , .
Let and .
(a) Using only the limit definition of the derivative, show that , where (you may assume this limit exists).
(b) State the property that singles out the number from all other bases.
(c) Given the rule , find the exact value of and confirm it numerically.
Show hint
In the difference quotient, use the exponent law and notice that one of the resulting factors has nothing to do with .
Show answer
(a) By definition,
where the last equality is the definition of applied at , since gives . Note the assumption that the limit exists is what lets us split the limit of a product this way.
So the derivative of every exponential function is a constant multiple of the function itself, and the constant is the slope of the graph at the -intercept.
(b) is the unique base for which that constant equals :
that is, is the unique base whose exponential graph crosses the -axis with slope exactly , and hence the unique base for which the function equals its own derivative. (Uniqueness is guaranteed because is strictly increasing.)
(c) Put into : the derivative at is . But by part (a) that derivative at is precisely . Therefore
Numerical confirmation. With , , agreeing with to six significant figures.
Answer: (a) shown; (b) is the unique base with , equivalently ; (c) the limit equals .
Let for . Find the equation of the tangent line to the curve at , and compute at .
Show hint
Product rule. Before plugging in , notice the value of — it kills exactly one term each time.
Show answer
Point of tangency. At , , so the point is .
First derivative (product rule with , , so , ):
At : .
Tangent line. Point-slope form through with slope :
Second derivative. Differentiate with the product rule again:
At :
Check. A second centred difference with gives , matching .
Answer: tangent line ; and .
Let .
(a) Find the largest domain of .
(b) Expand using log laws and differentiate.
(c) Evaluate exactly.
Show hint
Do not reach for the quotient rule inside the logarithm. Also, work out the domain from the requirement that the whole argument be strictly positive before you split anything up.
Show answer
(a) Domain. The argument of must be strictly positive.
- requires .
- requires , and for the product to be nonzero we need , i.e. .
- The numerator is then positive, so the whole quotient is positive exactly when . An odd power has the sign of its base, so this needs , i.e. .
Intersecting, (which automatically gives and ). Domain: .
(b) Expand first. On each of , and is positive, so the log laws apply cleanly:
Now differentiate term by term with :
(c) At (which is in the domain, since ):
Numerically ; a centred difference quotient on the original unexpanded gives .
Answer: (a) ; (b) ; (c) .
Use logarithmic differentiation to find for
then evaluate exactly. Say precisely where the method is valid and what happens at the points it excludes.
Show hint
Near one of the factors is negative, so take the log of the absolute value rather than of itself. Also check which factor can never vanish.
Show answer
Set up. The factor is negative near , so we must work with absolute values. Note always, so that factor causes no trouble. For and (where ),
Differentiate both sides in , using on the left () and on each right-hand term:
Multiply by and substitute back:
Evaluate at . First the function value:
Then the bracket:
Therefore
Check. A centred difference quotient with step gives .
Where the method is valid. Step 1 needs , so the derivation excludes and ; at those points the bracket has a pole and the displayed formula for is literally undefined. That is a limitation of the method, not of : is a polynomial-over-positive-polynomial and is differentiable everywhere. At the factor is a zero of order , so ; at the factor is a zero of order , so as well. (Both follow from writing with and using the product rule.)
Answer: for , and .
Let for .
(a) Explain why neither nor is correct.
(b) Find .
(c) Find the exact -value and exact -value of the minimum of on , and justify that it is a minimum.
Show hint
Look hard at the hypotheses of the two rules you already know: one demands a constant exponent, the other a constant base. Then take logs of both sides before differentiating.
Show answer
(a) Why both are wrong. The power rule is a theorem about a constant exponent — its proof (binomial expansion, or the limit definition) treats as a fixed number. The rule is a theorem about a constant base — its proof rewrites with a constant. In the base and exponent vary together, so neither hypothesis is satisfied and neither conclusion is available.
Numerically, at the true slope is (centred difference), while the first candidate gives and the second gives . Both are wrong, and interestingly the correct answer is their sum — the two bad answers are each one of the two terms the true rule produces.
(b) Logarithmic differentiation. For we have , so we may take logs:
Differentiate both sides with respect to (chain rule on the left, product rule on the right):
Multiply by :
Check: at , ; the numerical derivative is . At , ; the numerical derivative is .
(c) The minimum. Since for all , the factor never vanishes, so
Sign of : for and for , so decreases then increases — by the First Derivative Test is a local and (since it is the only critical point on the interval) the absolute minimum. The minimum value is
Second-derivative confirmation. Differentiating gives . At the first term is , so : concave up, a minimum.
Answer: (b) ; (c) minimum at with value .
Let on the interval .
(a) Explain why logarithmic differentiation is legitimate on this interval but not on .
(b) Find .
(c) Evaluate and exactly.
Show hint
Everything hinges on the sign of the base. Once you have taken logs, the right-hand side is a product of two familiar functions.
Show answer
(a) Legitimacy. The expression with a non-integer, variable exponent is defined as a real number only when the base (apart from the degenerate case with ), and taking requires strict positivity. On we have , so and makes sense. On we have , so is not a real number for most there (e.g. would need an irrational power of a negative number). The function itself does not exist there, let alone its logarithm.
(b) Differentiate. Take logs on :
Differentiate both sides in ; the left side gives by the chain rule, the right side needs the product rule and then the chain rule on :
Multiply by :
(c) Values.
At : , so , , and . Hence
(Sensible: on with equality at , so that point is a maximum and the tangent is horizontal.)
At : , , so and . Therefore
Numerically and , so . A centred difference quotient gives .
Answer: (b) on ; (c) and .
(a) Let and be differentiable. Prove that
and show that it degenerates to the power rule when is constant and to the exponential rule when is constant.
(b) Use it to differentiate the tower for , where means .
(c) Evaluate .
Show hint
For (a), the only honest starting point is . For (b) apply the formula with and — and you will need the derivative of from the previous problem.
Show answer
(a) Proof. Because , is defined and (this is the definition of a real power with a non-integer exponent). Differentiate with the chain rule, outer function with , and the product rule inside:
Degenerate case 1: constant. Then and
which is precisely the chain-rule power rule.
Degenerate case 2: constant. Then and
which is precisely the chain-rule exponential rule. So the general formula is literally "power-rule term exponential-rule term", and is the case where both terms survive.
(b) The tower. First, a notation warning: exponentiation associates upward, so , which is not . (At the first is and the second is .)
Apply the formula with (so , and as required) and , whose derivative we know:
Then
Same answer by logs. , so by the product rule , and multiplying by reproduces the line above.
(c) At . Here , so , and , . Hence
Check. A centred difference quotient of at with step gives . A second check away from the trivial point: at ,
and the numerical derivative there is .
Answer: (b) for ; (c) .
