Intuition The one core idea
When you re-label the points of a plane with new coordinates, tiny patches of area get stretched or squeezed by a number that depends on where you are. To keep an integral honest, you must multiply by that local stretch factor — and that factor is the absolute value of the Jacobian determinant .
This page builds every symbol the parent note Change of Variables leans on, starting from nothing. If a word or squiggle appears there and confuses you, it is explained below, in the order that lets each idea rest on the previous one.
Everything here lives in two flat planes side by side.
The uv -plane : the "new" coordinate world where we choose nice coordinates.
The x y -plane : the "old" world where the region and the integral actually live.
Definition A coordinate pair
( u , v )
A pair of numbers that names a single point. The first number tells you how far along one axis, the second how far along the other. Picture a dot pinned to a flat sheet by two rulers.
We will constantly move a point from the left sheet to the right sheet. That move is called a map .
Definition Map / transformation
T
A rule that takes each point ( u , v ) on the left sheet and produces a point ( x , y ) on the right sheet. Written T ( u , v ) = ( x , y ) .
The arrow ↦ (with a little bar) reads "gets sent to ". So ( u , v ) ↦ ( x , y ) means "this input point gets sent to that output point".
The colon in T : ( u , v ) ↦ ( x , y ) reads "is the rule that ".
Concretely a map is really two ordinary functions bundled together :
x = x ( u , v ) , y = y ( u , v ) .
Read x ( u , v ) as "the x -coordinate of the output, which depends on both u and v ".
Intuition Why we need a map
We want to swap coordinates inside an integral. The map is the exact dictionary translating every new point into where it lands in the old world. Without it we could not rewrite the integrand or the region.
Two words the parent attaches to T :
The map has no rips, jumps, or sharp corners — you can take derivatives of x ( u , v ) and y ( u , v ) as often as needed, and those derivatives change gently. Picture a rubber sheet you can pull but never tear.
Definition Injective (one-to-one)
No two different new points land on the same old point. Picture: the map never folds the sheet back onto itself. This matters because if two cells piled onto the same spot, we'd count that spot's area twice.
The parent writes r ( u 0 , v 0 ) and r u . Here is what the bold r means.
Definition Position vector
r
r ( u , v ) = ( x ( u , v ) , y ( u , v )) is just the output point, viewed as an arrow from the origin to that point. Bold letters mean "this is a vector (an arrow), not a single number."
Why bother turning a point into an arrow? Because arrows can be subtracted to get displacement arrows, and displacements are exactly what we need to measure the size of a mapped patch.
This is the tool that measures how the output moves when you nudge one input .
partial derivative, not an ordinary one?
Our output x depends on two inputs u and v . An ordinary derivative asks "how does output change as the input changes" — but we have two inputs. So we ask a sharper question: "if I freeze v and wiggle only u , how fast does x change?" That restricted derivative is the partial derivative .
Definition Partial derivative notation
∂ u ∂ x = x u = lim h → 0 h x ( u + h , v ) − x ( u , v ) .
The curly ∂ (not a straight d ) is a flag saying "there are other variables held constant ".
The subscript form x u is shorthand for the exact same thing.
Plain words: the rate x changes per unit step in u , with v pinned.
Definition The velocity vectors
r u and r v
Collect both coordinates' partials into arrows:
r u = ( x u , y u ) , r v = ( x v , y v ) .
Picture: r u is the arrow the output point traces out per unit step in u . It points along the direction the image moves when you slide right on the u -axis. Likewise r v for stepping in v .
Intuition Why these two arrows are the whole game
A tiny cell in the uv -plane has two edges: one pointing along u , one along v . Under the map, those edges become r u d u and r v d v . Two edge-arrows are all you need to pin down the mapped patch's area — which is the next tool.
The link Chain Rule (Multivariable) is where these partials get combined when maps are composed; the reciprocal trick in the parent leans on it.
Definition The differentials
d u , d v
d u means "an infinitely small width along u "; d v likewise along v . Picture the width and height of a microscopic rectangle. They are the building blocks we add up in an integral.
Definition The area element
d A
d A = d x d y is the tiny area of one microscopic patch in the x y -plane. The whole point of the topic is that d A is not equal to d u d v — the map rescales it.
Why "infinitely small"? Because only when a patch is small does the smooth map look straight (linear) across it. That straightness is what lets a triangle-and-parallelogram argument work.
The mapped patch is a parallelogram (a slanted rectangle). We need its area.
Definition Parallelogram spanned by two arrows
Given edge-arrows a and b sharing a corner, the parallelogram is the slanted four-sided region they frame. Picture pushing a rectangle sideways so it leans.
Intuition Why this exact combination measures area
The quantity a x b y − a y b x is the 2D cross product — see Cross Product and Area . It equals ∣ a ∣ ∣ b ∣ sin γ , where γ is the angle between the arrows. Base × height = ∣ a ∣ × ( ∣ b ∣ sin γ ) is precisely base times perpendicular height — the honest area. The absolute value strips off a sign that only tells us whether the corner turns clockwise or counter-clockwise.
Definition Why the answer can come out negative before the bars
If the two edges are swapped (or the map flips the sheet like a mirror), a x b y − a y b x comes out negative . Area is never negative, so we wrap it in ∣ ⋯ ∣ . This is the same sign that later becomes the absolute value on the Jacobian.
The parent bundles those four partials into a matrix and takes its determinant . Both symbols are new; here they are.
2 × 2 matrix
A square grid of four numbers, written ( p r q s ) . Here each entry is one partial derivative. It records the two edge-arrows as its columns : column one is r u , column two is r v .
Definition Determinant of a
2 × 2 matrix
det ( p r q s ) = p s − q r .
Read det as "the (signed) area-scaling number of this grid". Notice p s − q r is exactly the cross-product formula from §5 — because the columns are the two edge-arrows.
Intuition Why the determinant
is the stretch factor
Feed the determinant the unit square (edges ( 1 , 0 ) and ( 0 , 1 ) ) after the linear map, and it spits out the area of the resulting parallelogram. So det literally answers "by what factor did area change?" This is the whole message of Determinant as Volume Scaling . A negative determinant = the map turned the sheet over (mirror flip); its size is still the stretch.
Now every piece assembles into the star of the topic.
Definition Jacobian determinant
∂ ( u , v ) ∂ ( x , y ) = det ( x u y u x v y v ) = x u y v − x v y u .
The stacked notation ∂ ( u , v ) ∂ ( x , y ) is one symbol meaning "the determinant of the grid of all partials of ( x , y ) with respect to ( u , v ) ". Top = output variables, bottom = input variables.
Plain words: the local area-stretch of the map T at a point.
Common mistake Reading the fraction as a normal division
∂ ( u , v ) ∂ ( x , y ) is not a numerator over a denominator you can cancel. It is a whole determinant. The only fraction-like fact allowed is the reciprocal rule with the inverse map — which comes from the Inverse Function Theorem .
The absolute value ∂ ( u , v ) ∂ ( x , y ) is what enters the integral, because (from §5–§6) area is non-negative.
Definition Double integral
∬ R f d A
Chop the region R into microscopic patches, multiply each patch's value f by its area d A , and add them all up. The two ∫ signs mean "sum over both directions". The subscript R names which region you sum over. See Double Integrals over General Regions .
R and S
S = the region in the new (uv ) plane.
R = its image in the old (x y ) plane, i.e. R = T ( S ) .
The change-of-variables theorem says ∬ R ( old ) = ∬ S ( new ) × ∣ J ∣ . You must translate both the integrand and the region.
partial derivatives x_u y_v
velocity edge arrows r_u and r_v
parallelogram area via cross product
determinant as area scaling
Jacobian dx dy over du dv
double integral over a region
change of variables theorem
Test yourself — say the answer aloud before revealing.
What does the arrow ↦ mean in ( u , v ) ↦ ( x , y ) ? "gets sent to" — the input point on the left produces the output point on the right.
What does "injective" guarantee about the map? No two different points land on the same output, so the sheet never folds onto itself and no area is double-counted.
What is x u in plain words? The rate at which x changes when you nudge u while keeping v frozen.
Why is ∂ curly instead of a straight d ? To flag that other variables are being held constant (there is more than one input).
What geometric object are r u and r v ? Arrows showing how fast and in what direction the output moves per unit step in u and in v — the two edges of the mapped patch.
What shape does a tiny uv -rectangle become under a smooth map? A parallelogram (to first order), spanned by r u d u and r v d v .
What does a x b y − a y b x compute? The signed area of the parallelogram framed by arrows a and b .
Why take an absolute value of that quantity? Its sign only records orientation (clockwise vs counter-clockwise / a mirror flip); area itself must be non-negative.
What does the determinant of a 2 × 2 matrix measure? The factor by which the linear map scales area.
In ∂ ( u , v ) ∂ ( x , y ) , which variables go on top? The output (old) variables x , y ; the input (new) variables u , v go on the bottom.
What is d A and why is it not simply d u d v ? The tiny old-plane area d x d y ; the map rescales it, so d A = ∣ J ∣ d u d v .
What must you translate when changing variables, besides the integrand? The region — the new limits S in the uv -plane.