Derivative of y = x^3
Here is a fact that catches nearly everyone off guard. Take a cube-shaped box —
say 1 metre along each edge — and double its side.
You do not get double the volume. You get eight times the volume:
2^3 = 8. Triple the side and the volume explodes by a factor of
27. Volume is ferociously sensitive to side length, and it
gets more sensitive the bigger the cube already is.
"How sensitive, exactly?" is a derivative question. If V = s^3, then
\tfrac{dV}{ds} tells you how fast volume responds to a nudge in the
side — and this page proves that the answer is 3s^2. At a side of
1 m, volume grows at 3 m³ per extra metre
of side; at a side of 10 m, it grows at a colossal
300 m³ per metre. Same shape, a hundred times touchier.
There is a second reason to care. You have already found
\tfrac{d}{dx}[x^2] = 2x,
and before that the derivatives of a constant and of x itself. That's
three rungs of a ladder: 0, then 1, then
2x. One more rung — the cube — and a pattern will be staring you in
the face.
f(x) = x^3
Its graph is a gentle S: it climbs steeply from the lower left, flattens for a single instant as
it slips through the origin, then climbs steeply away to the upper right. So before any algebra,
we can already sketch what the slope should do: big, then
zero, then big again — and, curiously, never negative. Keep
that prediction in mind; the algebra had better agree with it.
The proof, done honestly
The recipe is the same
difference quotient
as always — step a distance h along, measure the rise, divide, then let
h shrink. The only new algebra is expanding a cubed bracket.
Do it in two honest stages, no shortcuts: square first, then multiply by one more
(x+h):
(x+h)^3 = (x+h)\,(x+h)^2 = (x+h)\,(x^2 + 2xh + h^2) = x^3 + 3x^2h + 3xh^2 + h^3.
Read the four terms as a story about size. x^3 is the old value —
it will cancel. 3x^2h carries one factor of the small step
h: small. 3xh^2 carries two:
tiny. h^3 carries three: microscopic. Now feed the
expansion into the difference quotient:
\frac{(x+h)^3 - x^3}{h} = \frac{3x^2h + 3xh^2 + h^3}{h} = 3x^2 + 3xh + h^2.
The x^3 cancelled, and one factor of h
divided out of every remaining term — that division is what dissolved the
\tfrac{0}{0} trap. Now watch who survives the limit. As
h \to 0, the terms 3xh and
h^2 still carry a factor of h, so they die
away to nothing. Only the term that had exactly one h in the
expansion — 3x^2h, now shaved down to
3x^2 — makes it through alive:
\frac{d}{dx}\bigl[\,x^3\,\bigr] = \lim_{h \to 0}\left(3x^2 + 3xh + h^2\right) = 3x^2.
-
For f(x) = x^3, the derivative is
f'(x) = 3x^2 at every real x.
-
Equivalently: nudging x by a small step
h changes x^3 by approximately
3x^2 \cdot h — the error terms
(3xh^2 + h^3) shrink to zero faster than
h does.
Spot-check it with numbers before trusting it. At
x = 2 the formula claims a slope of
3 \cdot 2^2 = 12. Take a genuine secant from
2 to 2.1:
\tfrac{2.1^3 - 2^3}{0.1} = \tfrac{9.261 - 8}{0.1} = 12.61. Shrink the
step to 0.01:
\tfrac{8.120601 - 8}{0.01} = 12.0601. The secants are marching
straight at 12, exactly as 3x^2 + 3xh + h^2
predicts (put x = 2, h = 0.01 in and you get
12.0601 on the nose). The formula and the arithmetic tell one story.
See the algebra: a cube growing a shell
That binomial expansion isn't just symbol-shuffling — you can see every term of it.
Picture a solid cube of side x, and grow each side by a small
h. The new material forms a shell, and the shell splits into exactly
the pieces of the expansion. Step through:
The three slabs are the whole story. Each covers one face — area
x^2 — with thickness h, so together they
contribute 3x^2h of new volume. The rods and the corner cube are
crumbs: their volumes, 3xh^2 and
h^3, carry extra factors of h, so when you
divide the added volume by h and let the shell get thin, they vanish —
precisely the terms that died in the limit above. The derivative
3x^2 is not a coincidence of algebra: it is the area of the
three faces the cube grows through. Geometry and limit agree because they are the same
computation seen from two sides.
The touchiness of x^3 runs the living world. An animal's
heat production goes roughly with its volume — with x^3 for
body size x — but its heat loss goes with its skin area,
only x^2. Scale a mouse up to elephant size and the furnace inside
grows far faster than the radiator outside: volume's growth rate 3x^2
simply outruns area's 2x. That's why elephants ship enormous ears
(extra radiator), why whales can afford to live in freezing water (heat kept in), and why a shrew
must eat almost constantly (heat pouring out).
The same squeeze traps cells. A cell feeds through its surface (x^2)
but must feed its whole interior (x^3). Grow too big and the inside
outstrips the membrane's ability to supply it — so cells divide instead of ballooning. Biologists
call this the square–cube law, and it is exactly the comparison of the two
derivatives you now know: 2x versus 3x^2.
The slope is never negative
Now interrogate the formula. 3x^2 is three times a
square — and a square can't be negative. So the slope of
y = x^3 is always \ge 0:
the curve only ever rises or, for one single instant at x = 0, lies
momentarily flat. It never goes downhill. Notice the symmetry too:
3(-2)^2 = 3(2)^2 = 12, so the slope at x = -2
equals the slope at x = +2 — the S-curve climbs equally hard on both
wings.
Slide the point and watch all of this happen: the tangent tips up steeply out on the wings, eases
off as you approach the middle, kisses horizontal exactly at the origin, then steepens again —
without ever once tilting downward.
Here is a trap that snares even strong students: "the slope is zero" does not
mean "the curve turns around."
With y = x^2, the flat spot at x = 0 really
was the bottom of a valley — falling on one side, rising on the other. It's tempting to conclude
that f'(a) = 0 always marks a peak or a valley. The cubic says no.
At x = 0 the slope of x^3 is exactly
3 \cdot 0^2 = 0 — yet the curve is rising just before it
and rising just after it. It doesn't turn; it pauses. Like a cyclist cresting a bridge
without pedalling backwards, the curve flattens for one instant and carries on up.
So the honest statement is one-directional: a smooth peak or valley forces the slope to be zero
there, but a zero slope alone doesn't force a peak or valley. When you later hunt for maxima and
minima, x^3 at the origin is the standard counterexample your examiner
hopes you've forgotten. Don't forget it.
Two curves, one story
Plot the cubic f(x) = x^3 (solid) together with its derivative
f'(x) = 3x^2 (dashed) and read the pair like an instrument panel:
the dashed parabola is a continuous record of the solid curve's steepness. Wherever the
cubic is steep, the parabola is high; where the cubic flattens through the origin, the parabola
dips to touch zero — and only touches, never crosses, because the slope never goes negative.
There's a small poetry in the shapes: differentiating a cubic hands you a
parabola, just as differentiating a parabola handed you a straight line, and a straight
line handed you a constant. Each derivative is one degree simpler than the curve it came from.
For a second telling of the whole computation, watch it narrated:
The ladder so far — can you smell the pattern?
Line up everything you have proved, one power at a time, each by the same difference-quotient
recipe:
| f(x) |
f'(x) |
| c (a constant) | 0 |
| x | 1 |
| x^2 | 2x |
| x^3 | 3x^2 |
| x^4 | ? |
| x^n | ? |
Look at what happens from row to row: the exponent jumps down in front as a
multiplier, and the power drops by one.
x^2 \to 2x^1; x^3 \to 3x^2. Even the first
two rows obey it if you squint: x^1 \to 1 \cdot x^0 = 1, and a
constant c \cdot x^0 gets multiplied by its exponent
0 and vanishes.
So — before anyone tells you — write down your guess for the two missing rows.
What should \tfrac{d}{dx}[x^4] be? And
\tfrac{d}{dx}[x^n] for any whole number n?
You can even hear why the guess should be right: expanding
(x+h)^4 the same honest way begins
x^4 + 4x^3h + \cdots, and by now you know that only the
single-h term survives the quotient and the limit. Whatever
coefficient sits on that term is the derivative. The next page names this pattern —
the power rule — and proves your guess holds for every n at once.
Arrive with your prediction already written down; there's no better feeling than being told
something you had already discovered.
See it explained