You’ve already met the rule dxd[ex]=ex on the derivatives of sine, cosine, eˣ, and ln x page. This page shows where it comes from: every exponential function has a slope that is a fixed multiple of its own height, and e is the one base where that multiple is exactly 1. From there you’ll get the derivative of any exponential, ax or A⋅bkt, and use it to find how fast populations, investments, and radioactive samples are changing.
The limit at the end doesn’t contain x at all. It’s just a number that depends on the base a. Call it k:
f′(x)=k⋅ax,where k=h→0limhah−1
So the slope of an exponential graph at any point is a constant times its height. In other words, the ratio f(x)f′(x) is the same constant k everywhere. Since f′(0)=k⋅a0=k, the constant is also the slope of the graph where it crosses the y-axis.
For base 2 the constant is less than 1, and for base 3 it’s more than 1. Somewhere between 2 and 3 there’s a base whose constant is exactly1. That base is the number
e≈2.71828
If you graph y=ax with a slider for a (Desmos or GeoGebra work well) and compare the slope of the tangent with the height of the graph, they match only when a=e. For f(x)=ex:
f′(x)=ex
In words: the slope of the tangent at any point equals the height of the graph at that point. At (1,e) the slope is e; at (2,e2) the slope is e2≈7.389.
At (0,1) the tangent slopes are about 0.693 for y=2x, exactly 1 for y=ex, and about 1.099 for y=3x.
Since f′(x)=(lna)⋅f(x), the graph of f′ is the graph of fstretched vertically by a factor of lna. So f′ is also an exponential function with the same base:
If a>e, then lna>1: f′ sits abovef (an expansion).
If 1<a<e, then 0<lna<1: f′ sits belowf (a compression), as for 2x in the figure.
If 0<a<1, then lna<0: f′ is a reflection in the x-axis as well. That makes sense, because a decaying exponential is always decreasing.
The derivative of y=2x is y=(ln2)⋅2x: the same curve, compressed vertically by a factor of ln2≈0.693.
Combine the rule with the chain rule whenever the exponent is more than just x:
Function
Derivative
ekx
kekx
eg(x)
eg(x)⋅g′(x)
ag(x)
ag(x)lna⋅g′(x)
A⋅bkt
A⋅bkt⋅klnb
The last line is the general growth or decay model. Notice that its derivative is (klnb) times the original function: the rate of change is proportional to the amount present. That’s the defining feature of exponential growth and decay. A population of 10000 grows twice as fast (in people per year) as the same kind of population of 5000.
At the start, the iodine is decaying at about 17.329 mg per day. After one half-life, half as much is left, and it’s decaying half as fast: about 8.664 mg per day. (The negative sign means the amount is decreasing.)
(b) From part (a),
A′(t)=8ln21⋅200(21)t/8=−8ln2A(t)≈−0.0866A(t)
So at any moment, the amount decreases at about 8.66% of the current amount per day.
(c) Set A′(t)=−2. Using part (b), −8ln2A(t)=−2, so A(t)=ln216≈23.083 mg. Now solve for t:
200(21)t/8(21)t/88tln21t=ln216=25ln22=ln(25ln22)=ln218ln(25ln22)≈24.921take ln of both sides
After about 24.921 days (just under 25 days), the iodine is decaying at 2 mg per day.
Using the power rule on an exponential.dxd[2x] is notx⋅2x−1. The power rule is for a variable base and a constant exponent (x2). Here the base is constant and the exponent is the variable, so you need axlna.
Forgetting the ln a.dxd[5x]=5xln5, not 5x. Only base e has no extra factor, because lne=1. A quick check: the slope of 5x at x=0 should be about 1.609, not 1.
Forgetting the chain rule factor.dxd[e−4x]=−4e−4x, not e−4x. Whenever the exponent is anything other than plain x, multiply by the derivative of the exponent.
Changing the exponent. The derivative of ex2 is 2xex2. The exponent stays x2; it does not become 2x or drop by one. The 2x comes out in front as a factor.
Dropping the negative sign for decay. For b<1, lnb is negative, so the derivative is negative. If your “rate of decay” comes out positive, check ln21=−ln2. In words, you can then say “decreasing at 17.329 mg per day”.
Rounding too early. Keep ln2 or the full calculator value until the last step. Rounding ln2 to 0.7 in Example 4 would give A′(0)≈−17.5, which is wrong to 3 decimal places.
(b) The values are settling down near 1.609, so k≈1.609 and f′(x)≈1.609⋅5x. This matches ln5≈1.6094.
3. (Warm-up) Explain, without a calculator, why the slope of the tangent to y=ex at the point (2,e2) is e2. Then give that slope to 3 decimal places.
Solution
For f(x)=ex, f′(x)=ex: the slope at any point equals the height of the graph there. At x=2 the height is e2, so the slope is e2≈7.389.
4. (Core) Differentiate each function.
(a) y=32x−1
(b) y=10x2
(c) y=x2e−3x
Solution
(a) The exponent 2x−1 has derivative 2:
dxdy=32x−1ln3⋅2=2ln3⋅32x−1
(b) The exponent x2 has derivative 2x:
dxdy=10x2ln10⋅2x=2xln10⋅10x2
(c) Use the product rule, with the chain rule on e−3x:
dxdy=2x⋅e−3x+x2⋅(−3e−3x)=xe−3x(2−3x)
5. (Core) Find the equation of the tangent line to y=6−2ex−1 at x=1.
Solution
Point: y=6−2e0=4, so (1,4).
Slope: dxdy=−2ex−1, so at x=1 the slope is −2e0=−2.
y−4=−2(x−1)⇒y=−2x+6
6. (Core) The population of a town is modelled by P(t)=12000(1.025)t, where t is the number of years after 2026.
(a) Find P′(t).
(b) How fast is the population growing in 2026 and in 2036? Round to the nearest person per year.
(c) Why is the growth rate larger in 2036?
Solution
(a) P′(t)=12000(1.025)tln1.025
(b) In 2026, t=0: P′(0)=12000ln1.025≈296.3, so about 296 people per year.
In 2036, t=10: P′(10)=12000(1.025)10ln1.025≈379.3, so about 379 people per year.
(c) The rate of growth is proportional to the population: P′(t)=(ln1.025)P(t)≈0.0247P(t). By 2036 the population has grown to about 15361, so it grows by more people each year.
7. (Core) You invest $2000 at 6% per year, compounded annually, so the value after t years is A(t)=2000(1.06)t dollars.
(a) Find the rate at which the investment is growing at t=0, to the nearest cent per year.
(b) When is the investment growing at $150 per year? Round to 2 decimal places.
Solution
(a) A′(t)=2000(1.06)tln1.06, so A′(0)=2000ln1.06≈116.54. The investment is growing at about $116.54 per year at the start.