Peel an onion. It doesn't come apart into flat slices — it comes apart into layers: thin curved sheets wrapped one around the next, each hugging the one inside it. A tree trunk is built the same way, one ring of new wood laid around last year's ring. Nature, it turns out, is very fond of storing volume in nested shells — and so is calculus.

The disc method cut a solid of revolution into flat circular slabs perpendicular to the axis. That works beautifully — until it doesn't. Rotate the region under y = f(x) about the y-axis and try to slice horizontally: every slab's radius is an x-value at a given height, so you must invert the function, solving y = f(x) for x. For y = x^2 that's a mild nuisance (x = \sqrt{y}). For y = 2x - x^2 it's a quadratic-formula headache with two branches. For something like y = x + \sin x it's flatly impossible in closed form.

The shell method sidesteps the whole fight. Instead of slicing across the axis, it peels the solid around the axis, onion-style: each thin vertical strip of the region, swung around the axis, sweeps out one thin-walled tube — a cylindrical shell. The shell at position x has radius x, height f(x), and wall thickness dx — everything stays written in the variable you were given. Shells integrate with the function instead of against it:

Where does 2\pi x come from? Unroll one shell

The factor 2\pi x looks mysterious until you slit a shell open and flatten it — the same move as unrolling a toilet roll or a rolled-up carpet.

Step 1 — picture one shell. Take a thin vertical strip of the region, sitting at distance x from the y-axis, of width dx and height f(x). Rotate it about the axis. It sweeps out a hollow tube: radius x, height f(x), wall thickness dx.

Step 2 — slit it and unroll. Cut the tube straight down one side and press it flat. It becomes a thin rectangular sheet. Its width is the circumference of the circle it used to be, 2\pi x; its height is still f(x); its thickness is still dx. (Because the wall is thin, the inner and outer circumferences are essentially equal, so no volume is lost or gained in the flattening — that's the limit doing its quiet work.)

Step 3 — the shell's volume is the sheet's. A flat sheet is just a very thin box: volume = (width) × (height) × (thickness):

Step 4 — nest the shells and refine. Stack the shells one inside the next, from radius a out to radius b, and let the walls grow infinitely thin. The Riemann sum becomes an integral:

Unroll the shell yourself

On the left, the curve y = x^2 with one shell drawn as a vertical strip at radius x. On the right, that same shell unrolled into a flat rectangle of width 2\pi x (its circumference) and height f(x) = x^2. Slide the radius and watch two things compete: pushing the strip outward makes the rectangle wider (bigger circumference) and taller (the parabola climbs). The readout shows the lateral area 2\pi x\, f(x) — the integrand itself, growing like x^3.

The volume integral simply adds up all of these rectangles, one for every radius from 0 to 1, each carrying its thin slab of thickness dx.

Worked example 1: y = x^2 about the y-axis

Rotate the region under y = x^2, 0 \le x \le 1, about the y-axis.

Step 1 — identify radius and height. A shell at position x has radius x (its distance to the axis) and height f(x) = x^2.

Step 2 — build the integrand 2\pi x\, f(x):

Step 3 — integrate and evaluate.

The bowl-shaped solid swept out by y = x^2 around the y-axis has volume \pi/2. (Notice how the extra factor of x from the radius nudged the power from x^2 up to x^3 — shells always integrate one power higher than the function you started with.)

Worked example 2: when washers really fight back

Rotate the region under the parabolic arch y = 2x - x^2, 0 \le x \le 2, about the y-axis.

Try washers first — and watch it go wrong. To slice horizontally you need x as a function of y. Solving y = 2x - x^2 with the quadratic formula gives x = 1 \pm \sqrt{1 - y} — two branches. Every washer's outer radius comes from one branch and its inner radius from the other, so you'd be integrating \pi\big[(1 + \sqrt{1-y})^2 - (1 - \sqrt{1-y})^2\big]. Doable, but square roots everywhere and easy to fumble.

Now shells. Radius x, height 2x - x^2 — no inverting, no branches:

Integrate term by term and evaluate:

Three lines, all in the variable the problem handed you. This is the shell method's home turf: a curve that is easy to read but hard to invert, rotated about a vertical axis.

Worked example 3: one solid, both methods — do they agree?

Trust, but verify. Take the triangular region under y = x, 0 \le x \le 1, and rotate it about the y-axis. The result is a cylinder of radius 1 and height 1 with a cone drilled out of the top.

By shells (radius x, height x):

By washers: slice at height y. The region occupies y \le x \le 1 there, so the washer has outer radius 1 and inner radius y:

And by pure geometry: cylinder minus cone, \pi r^2 h - \tfrac{1}{3}\pi r^2 h = \tfrac{2}{3}\pi. Three completely different computations, one answer. Shells aren't an approximation or a rival theory — they are the same volume, carved up a different way. When both methods are easy, computing both is the cheapest error-check in calculus.

Worked example 4: a shifted axis, x = c

Nothing says the axis must be the y-axis. Rotate the region under y = x^2, 0 \le x \le 1, about the vertical line x = 2 — an axis standing to the right of the region.

Step 1 — the radius is the distance to the axis. The strip at position x is now 2 - x away from the axis (the axis is at 2, the strip at x, and x < 2). Sanity-check the endpoints: the strip at x = 0 has radius 2 ✓, the strip at x = 1 has radius 1 ✓. The height is unchanged: still x^2.

Step 2 — assemble and expand.

Step 3 — evaluate.

Compare with example 1: same region, same height, but every shell now travels a longer circle (radius 2-x instead of x), so the volume jumps from \pi/2 to 5\pi/6. Had the axis stood to the left of the region — say x = -1 — the radius would be x - (-1) = x + 1. Distance, always distance.