Why the grid you were handed is the wrong grid
The previous guide gave you the master key: the change of variables theorem, with its Jacobian determinant that measures how much a coordinate map stretches area. This guide spends that key on the three changes of variable you will reach for again and again. The reason is simple and physical. Cartesian coordinates carve the plane into little squares, and squares are wonderful for describing a rectangle. But the world is full of disks, cylinders, and spheres, and a square grid fits a round region the way graph paper fits a dinner plate — badly, with ragged edges everywhere along the rim.
Try to set up the integral of some quantity over the unit disk in Cartesian coordinates and you feel the pain immediately. The region of integration forces the inner limits to be y running from -sqrt(1 - x^2) up to +sqrt(1 - x^2). Those curved limits then haunt every step, and an innocent factor like x^2 + y^2 fights you the whole way. The geometry of the region and the algebra of the integrand are both screaming that you brought the wrong ruler to a round problem.
The fix is to change the grid to match the symmetry. If a region is built out of circles and rays, describe its points with a radius and an angle rather than two distances. If a solid is round about an axis, use cylindrical coordinates; if it is round about a point, use spherical. In the right coordinates the curved boundary becomes a flat wall — 'radius equals 1' is a single number, not a parabola — and the region of integration turns back into a tidy rectangle in the new variables. The whole art of this guide is matching the coordinate system to the shape of the problem.
Polar coordinates and the famous r dr dtheta
In [[polar-coordinates|polar coordinates]] a point in the plane is named by its distance r from the origin and the angle theta its ray makes with the positive x-axis. The dictionary back to Cartesian is x = r cos(theta), y = r sin(theta), and the reverse is r = sqrt(x^2 + y^2), theta = the angle of (x, y). The payoff is instant: the dreaded combination x^2 + y^2 collapses to a single clean r^2, and the unit disk is simply the rectangle 0 <= r <= 1, 0 <= theta <= 2 pi. The round region became a box.
But here is the trap that catches every beginner: you cannot just write dx dy = dr dtheta. That would be a lie, and the picture tells you why. Picture the little patch swept out when r increases by a tiny dr and theta increases by a tiny dtheta. Two of its sides are radial, of length dr, but the other two are arcs. An arc subtending angle dtheta at radius r has length r dtheta — long arcs out near the rim, short arcs near the center. So the patch is roughly a rectangle of sides dr and r dtheta, and its area is the product: dA = r dr dtheta. That extra factor of r is the whole story, and it is no accident — it is exactly the Jacobian of the polar map.
You can confirm it by grinding the Jacobian directly. The matrix of partials is [partial x / partial r, partial x / partial theta; partial y / partial r, partial y / partial theta] = [cos(theta), -r sin(theta); sin(theta), r cos(theta)]. Its determinant is r cos^2(theta) + r sin^2(theta) = r. The hand-waving about arcs and the cold algebra of the determinant agree perfectly, as they must. Now watch the unit-disk integral that was a Cartesian nightmare melt: integrating 1 over the disk becomes the integral over 0 <= theta <= 2 pi and 0 <= r <= 1 of r dr dtheta, which is 2 pi times (1/2) = pi — the area of the unit disk, falling out in two trivial single integrals.
Cylindrical coordinates: polar plus a height
Step up to three dimensions and the laziest, most useful idea is to keep one Cartesian axis untouched and make the other two polar. That is [[cylindrical-coordinates|cylindrical coordinates]]: x = r cos(theta), y = r sin(theta), z = z. You describe where a point is by how far out from the z-axis it sits (r), which way around (theta), and how high (z). It is the natural language for anything round about a vertical axis — a tin can, a pipe, a tornado, a spinning shaft — because such a solid is just a polar region in the floor plan, copied straight up to some height.
Because the z-direction was left alone, the volume element is just the polar area element wearing a coat of dz. The little box has a floor of area r dr dtheta and a height dz, so its volume is dV = r dr dtheta dz. Nothing new to memorize — it is the polar r dr dtheta with a height stacked on. If you grind the full 3-by-3 Jacobian you again get exactly r, since the z-row and z-column contribute a clean factor of 1 and leave the planar 2-by-2 block to deliver its r. The arc-length intuition and the determinant shake hands once more.
Picture finding the volume of a cylinder of radius a and height h. In cylindrical coordinates it is simply the box 0 <= r <= a, 0 <= theta <= 2 pi, 0 <= z <= h, and the integral of r dr dtheta dz over it factors into three independent single integrals: (a^2 / 2) times 2 pi times h, which is pi a^2 h. Every boundary is a constant; the curved wall 'r = a' is one number; nothing curves in the limits at all. That is the signature of a well-matched coordinate system — the geometry that was hard in Cartesian has been completely absorbed into the names of the variables.
Spherical coordinates and the rho^2 sin(phi) factor
When a problem is round about a point rather than an axis — a ball, a planet's gravity, an atom's orbital, anything radiating outward equally in all directions — reach for [[spherical-coordinates|spherical coordinates]]. Here a point is named by rho, its straight-line distance from the origin; phi, the polar angle measured down from the positive z-axis (0 at the north pole, pi at the south); and theta, the same azimuthal angle around the z-axis as before. The dictionary is x = rho sin(phi) cos(theta), y = rho sin(phi) sin(theta), z = rho cos(phi). A solid ball of radius a is just the box 0 <= rho <= a, 0 <= phi <= pi, 0 <= theta <= 2 pi.
The volume element looks scarier but its three pieces are each easy to see. Increase rho by drho and you move radially a distance drho. Increase phi by dphi and you swing along a great-circle arc at radius rho, sweeping length rho dphi. Increase theta by dtheta and you swing around a circle of latitude — and here is the subtle bit — whose radius is not rho but rho sin(phi), because the circles of latitude shrink as you head toward either pole. That arc has length rho sin(phi) dtheta. Multiply the three nearly-perpendicular edges and you get dV = rho^2 sin(phi) drho dphi dtheta.
Those edge lengths drho, rho dphi, rho sin(phi) dtheta are exactly what the term [[scale-factors|scale factors]] names: the conversion rate between a change in each coordinate and the actual distance you travel. In any orthogonal system the volume element is just the product of the scale factors, which is why polar gives 1 times r, cylindrical gives 1 times r times 1, and spherical gives 1 times rho times rho sin(phi) = rho^2 sin(phi). The full 3-by-3 Jacobian determinant of the spherical map confirms this exactly: after the dust settles it equals rho^2 sin(phi), positive for phi between 0 and pi, just as the geometry promised.
AREA / VOLUME ELEMENTS (each = product of the scale factors)
polar dA = r dr dtheta scale factors: 1, r
cylindrical dV = r dr dtheta dz scale factors: 1, r, 1
spherical dV = rho^2 sin(phi) drho dphi dtheta
scale factors: 1, rho, rho sin(phi)
Volume of a ball of radius a:
integral over 0<=rho<=a, 0<=phi<=pi, 0<=theta<=2pi of rho^2 sin(phi)
= (a^3/3) * [-cos(phi)]_0^pi * (2pi)
= (a^3/3) * (2) * (2pi) = (4/3) pi a^3 <- as it must beChoosing the right system, and the fine print
- Read the symmetry of the region first, not the integrand. Round disk or anything bounded by circles and rays in the plane: go polar. A solid round about a single axis (cylinder, cone, paraboloid, a hole drilled through a ball): go cylindrical. A solid round about a single point (ball, spherical shell, ice-cream cone of a region): go spherical.
- Then check the integrand for a second opinion. A factor of x^2 + y^2 begs for polar or cylindrical (it becomes r^2); a factor of x^2 + y^2 + z^2 begs for spherical (it becomes rho^2). If the region says one thing and the integrand says another, the region usually wins, but a clean integrand can tip a borderline call.
- Rewrite the boundaries as constant limits in the new variables, never forget to attach the area or volume element, and only then integrate from the inside out. The element is part of the integrand now — leaving off the r or the rho^2 sin(phi) is the single most common mistake in the whole subject.
One more caution worth stating plainly: notation is not standardized across fields, and it will trip you in a textbook or a physics paper. Many sources swap the roles of theta and phi, or use r for the spherical distance where we used rho, or measure the polar angle up from the equator instead of down from the pole. The mathematics is identical; only the labels move. Whenever you open a new source, find its picture, pin down which letter is the radius and which angle is measured from where, and rederive the volume element from the scale factors if you have any doubt. The geometry never lies, even when two books disagree on the alphabet.
Where this leads next
You now hold the three coordinate systems that do most of the heavy lifting in physics and engineering, together with the one fact that makes each one trustworthy: the area or volume element is the product of that system's scale factors, equal to the Jacobian of the map. That single principle is general — it will reappear when you take the Laplacian in curvilinear coordinates for the heat and wave equations, where the very same scale factors decide the shape of the operator. The pattern you learned on integrals returns to govern derivatives.
More immediately, the next guide cashes these elements in for real quantities. Once you can integrate cleanly over a round body, you can weigh it: a double or triple integral of a density gives total mass, and dividing weighted-position integrals by that mass gives the center of mass. The same machinery, with a factor of distance-squared inserted, yields the moment of inertia that governs how the body resists spinning. The factor of r or rho^2 sin(phi) you just learned to respect is exactly what makes every one of those physical numbers come out right.