Change of variables and the Jacobian
A region bounded by two pairs of slanted lines, or by two hyperbolas and two lines, is a rectangle in the right variables. When a double integral is rewritten in those variables, what replaces dA?
Where you are 6 in Multiple integrals
Why this exists
Leah runs a tea plantation on a hillside. The bushes grow in rows along the lines 2y - x = c (meters, x east, y north), and service paths along 2x - y = c close off each block. In x and y a block is a slanted parallelogram, and an integral over it splits into three pieces. In the variables u = 2y - x and v = 2x - y it is a rectangle.
Substitution in one variable writes x = g(u) and dx = g'(u)\,du, where g'(u) measures how much g stretches length. Double integrals in polar coordinates and Cylindrical and spherical coordinates found the stretching factors r and \rho^2\sin\varphi by measuring cells by hand. This lesson gives one rule for every smooth one-to-one change of variables: the factor is the absolute value of a determinant of partial derivatives.
Definition
A transformation is a function T(u, v) = (x(u, v), y(u, v)) from a set of the uv-plane to the xy-plane. For a set S, its image is T(S) = \{T(u, v) : (u, v) \in S\}. T is C^1 on an open set if x and y have continuous first Partial derivatives there. Its Jacobian matrix, as in The multivariable chain rule, is
and its Jacobian is the determinant
The columns of DT are T_u = \langle x_u, y_u \rangle and T_v = \langle x_v, y_v \rangle; a column and the vector in angle brackets with the same components are the same vector. Joint distributions writes the same determinant as J.
In space, T(u, v, w) = (x, y, z) has the 3 \times 3 Jacobian matrix whose rows hold the partial derivatives of x, y and z, and its determinant is the Jacobian \dfrac{\partial(x, y, z)}{\partial(u, v, w)}. Regions are admissible in the sense of Double integrals as volume: closed, bounded, with boundary made of finitely many graphs of continuous functions.
Intuition
Take a small rectangle with corner (u_0, v_0) and sides \Delta u and \Delta v. A C^1 map is differentiable (Tangent planes and linear approximation), so near (u_0, v_0)
with the partials at (u_0, v_0). The image of the rectangle is therefore close to the parallelogram with sides \Delta u\,T_u and \Delta v\,T_v (Figure 1), whose area by What the determinant means is
A Riemann sum over the images of a grid then reads
a Riemann sum in u and v.
The absolute value is there because area is positive. The sign of the Jacobian says whether T keeps or reverses orientation, as the sign of any determinant does. For the polar map, x_r = \cos\theta, x_\theta = -r\sin\theta, y_r = \sin\theta, y_\theta = r\cos\theta, and the Jacobian is r\cos^2\theta + r\sin^2\theta = r, the factor of Double integrals in polar coordinates.
Results
Theorem 1 (affine changes of variables). Let T(u, v) = A \begin{bmatrix} u \\ v \end{bmatrix} + \mathbf{b} with A a 2 \times 2 matrix, \det A \ne 0. Let S be a type I or type II region of the uv-plane (Iterated integrals and Fubini's theorem) whose image D = T(S) is admissible, and let f be continuous on D. Then
In particular A(D) = \lvert \det A \rvert\,A(S).
Theorem 2 (change of variables). Let S be an admissible region of the uv-plane and T a C^1 transformation on an open set containing S. Suppose T is one-to-one on the interior of S, its Jacobian is not 0 there, and D = T(S) is admissible. Then for every f continuous on D,
So \displaystyle dA = \left\lvert \frac{\partial(x, y)}{\partial(u, v)} \right\rvert du\,dv. The polar formula is the case of the polar map on [0, b] \times [\alpha, \beta] with \beta - \alpha \le 2\pi, which fails to be one-to-one only on the boundary.
Theorem 3 (the inverse Jacobian). Let T map an open set U one-to-one onto an open set V, with T and T^{-1} both C^1. Then at (u, v) and (x, y) = T(u, v),
so neither Jacobian is 0.
Theorem 4 (three variables). Under the same hypotheses, with solids in place of regions,
where E = T(S); for T = A\mathbf{u} + \mathbf{b} the factor is \lvert \det A \rvert. The factors r and \rho^2\sin\varphi of Cylindrical and spherical coordinates are the Jacobians of the cylindrical and spherical maps.
Why it is true
Theorem 1. Three facts. (i) T sends a square of side h to a parallelogram of area \lvert \det A \rvert h^2: translation keeps area, and A sends the sides h\mathbf{e}_1, h\mathbf{e}_2 to hA\mathbf{e}_1, hA\mathbf{e}_2, so Theorem 2 of What the determinant means applies. (ii) \lVert T(\mathbf{p}) - T(\mathbf{q}) \rVert = \lVert A(\mathbf{p} - \mathbf{q}) \rVert \le K \lVert \mathbf{p} - \mathbf{q} \rVert, where K^2 is the sum of the squares of the entries of A, by the Cauchy-Schwarz inequality on each row. (iii) T is one-to-one, so T(P) \cap T(S) = T(P \cap S) for every set P.
Let S be type I, a \le u \le b, g_1(u) \le v \le g_2(u) (type II is the same with u and v exchanged). Take a square Q of side L containing S, cut it into n^2 squares Q_{ij} of side \displaystyle h = \frac{L}{n} with centers \mathbf{p}_{ij}, and let G = f(T)\,\lvert \det A \rvert on S and 0 on the rest of Q. Let f^D be f on D and 0 elsewhere; it is integrable over the parallelogram T(Q) with integral \iint_D f\,dA (Results 1 and 3 of Double integrals as volume). The parallelograms T(Q_{ij}) cover T(Q) and overlap only on edges, so \iint_D f\,dA = \sum \iint_{T(Q_{ij})} f^D\,dA. Compare term by term with \Sigma_n = \sum G(\mathbf{p}_{ij})\,h^2, a Riemann sum that tends to \iint_S f(T)\,\lvert \det A \rvert\,du\,dv.
- Outside (Q_{ij} misses S). By (iii) T(Q_{ij}) misses D, and both terms are 0.
- Inside (Q_{ij} \subseteq S). Given \varepsilon > 0, uniform continuity of f on D gives \delta. Once \displaystyle Kh\sqrt{2} < \delta, every point of T(Q_{ij}) is within \delta of T(\mathbf{p}_{ij}) by (ii), so the term differs from f(T(\mathbf{p}_{ij}))\,\lvert \det A \rvert h^2 = G(\mathbf{p}_{ij})h^2 by at most \varepsilon \lvert \det A \rvert h^2.
- Boundary (the rest). Each of the two terms is at most M \lvert \det A \rvert h^2 in size, with M = \max_D \lvert f \rvert, so they differ by at most 2M \lvert \det A \rvert h^2. As in the proof of Theorem 2 of Double integrals in polar coordinates, such a cell meets a graph v = g_i(u) or a segment of u = a or u = b. With \omega_n \to 0 the largest change of g_1 or g_2 over an interval of length h, each graph meets at most \displaystyle \frac{\omega_n}{h} + 2 cells of a column, and each segment lies in at most two columns. So there are at most \displaystyle 2n\left(\frac{\omega_n}{h} + 4\right) boundary cells.
Adding up, with nh = L,
Let n \to \infty, then \varepsilon \to 0. \blacksquare
Theorem 2. On a cell Q_{ij} of side h, T differs from its affine approximation at \mathbf{p}_{ij} by at most \eta_n h with \eta_n \to 0 uniformly in i and j: the partials are uniformly continuous on a closed bounded set, and the mean value theorem applies along segments. So T(Q_{ij}) is squeezed between two parallelograms of areas close to \displaystyle \left\lvert \frac{\partial(x, y)}{\partial(u, v)}(\mathbf{p}_{ij}) \right\rvert h^2, and the comparison above goes through. Making “squeezed between” exact needs the inverse function theorem; the complete proof belongs to Real Analysis.
Theorem 3. T^{-1} \circ T is the identity on U, and T^{-1} is differentiable because its partials are continuous (Tangent planes and linear approximation). Result 2 of The multivariable chain rule gives D(T^{-1})(x, y)\,DT(u, v) = I, and the determinant of a product is the product of the determinants (Theorem 5 of What the determinant means), so the product of the Jacobians is \det I = 1. \blacksquare
Theorem 4. For an affine T the proof of Theorem 1 runs with cubes: A sends a cube of side h to a parallelepiped of volume \lvert \det A \rvert h^3, the absolute value of the triple product of its edges (The cross product and the triple product). The general case is proved in Real Analysis, as Theorem 2 is.
Worked examples
Leah’s tea block
The block is bounded by the rows 2y - x = 0 and 2y - x = 30 and the paths 2x - y = 0 and 2x - y = 60, in meters. Leah’s leaf counts put the season’s yield at \delta(x, y) = 0.5 + 0.01(2x - y) kilograms per square meter. Find the area of the block and its total yield (Figure 2).
SolutionLet u = 2y - x and v = 2x - y. The map (x, y) \mapsto (u, v) is linear with determinant (-1)(-1) - (2)(2) = -3 \ne 0, so it is one-to-one, and the block D is the image of the rectangle S = [0, 30] \times [0, 60] under its inverse T. Solving, \displaystyle x = \frac{u + 2v}{3} and \displaystyle y = \frac{2u + v}{3}, so
The rectangle is type I, so Theorem 1 gives \displaystyle dA = \frac13\,du\,dv, and
The block has area 600 square meters and yields 480 kilograms. The sign of \displaystyle -\frac13 only says that T reverses orientation. In x and y the block takes three iterated integrals, split at the corners’ x = 10 and x = 40.
Between two hyperbolas
Let D be the region of the first quadrant bounded by xy = 2, xy = 5, y = x and y = 3x (Figure 1). Find A(D) and \iint_D y^2\,dA.
SolutionThe curves are level curves of u = xy and \displaystyle v = \frac{y}{x}, and D is 2 \le u \le 5, 1 \le v \le 3. On the open first quadrant, (x, y) \mapsto (u, v) is one-to-one with C^1 inverse \displaystyle T(u, v) = \left(\sqrt{u/v}, \sqrt{uv}\right), so S = [2, 5] \times [1, 3]. Rather than differentiate the square roots, use Theorem 3:
so \displaystyle \frac{\partial(x, y)}{\partial(u, v)} = \frac{1}{2v} > 0. Since y^2 = uv, Theorem 2 gives
In x and y the region needs three pieces, split at \displaystyle x = \sqrt{5/3} and \displaystyle x = \sqrt{2}, with a reciprocal in every limit.
A tilted ellipse
Find the area of the region D given by x^2 + 2xy + 5y^2 \le 8, and evaluate \iint_D (x^2 + 2xy + 5y^2)\,dA.
SolutionComplete the square: x^2 + 2xy + 5y^2 = (x + y)^2 + 4y^2. With u = x + y and v = 2y, D corresponds to the disk S: u^2 + v^2 \le 8. The map (x, y) \mapsto (u, v) has determinant 1 \cdot 2 - 1 \cdot 0 = 2, so its inverse \displaystyle T(u, v) = \left(u - \frac{v}{2}, \frac{v}{2}\right) has Jacobian \displaystyle \frac12 by Theorem 3. The disk is type I, and Theorem 1 gives
For the integral, Theorem 1 and then Double integrals in polar coordinates give
The tilted ellipse’s area came without finding its axes.
The ellipsoid
Let a, b, c > 0 and let E be the solid \displaystyle \frac{x^2}{a^2} + \frac{y^2}{b^2} + \frac{z^2}{c^2} \le 1, of constant density \delta and mass M. Prove that \displaystyle V(E) = \frac43\pi abc and that its moment of inertia about the z-axis is \displaystyle I_z = \frac{M(a^2 + b^2)}{5}.
SolutionT(u, v, w) = (au, bv, cw) is linear with diagonal matrix, Jacobian abc > 0, and maps the unit ball B: u^2 + v^2 + w^2 \le 1 one-to-one onto E. By Theorem 4, \displaystyle V(E) = abc\,V(B) = \frac43\pi abc, so \displaystyle M = \frac43\pi abc\,\delta. Next,
Exchanging two coordinates maps B onto itself with Jacobian -1, so by Theorem 4 the integrals of u^2, v^2 and w^2 over B are equal, and each is a third of \displaystyle \iiint_B \rho^2\,dV = 4\pi\int_0^1 \rho^4\,d\rho = \frac{4\pi}{5} (Cylindrical and spherical coordinates). So each is \displaystyle \frac{4\pi}{15}, and
With a = b = c this is the ball’s \displaystyle \frac25 Ma^2.
Sums over a triangle
Let K be the triangle x \ge 0, y \ge 0, x + y \le 1, and g continuous on [0, 1]. Prove that \iint_K g(x + y)\,dA = \int_0^1 t\,g(t)\,dt, and evaluate \iint_K e^{(x + y)^2}\,dA.
SolutionLet \Phi(u, v) = (u(1 - v), uv) on S = [0, 1]^2. It is C^1 on the whole plane, with Jacobian (1 - v)u - (-u)v = u. It maps S onto K: a point (x, y) of K with x + y > 0 has the single preimage u = x + y, \displaystyle v = \frac{y}{x + y}, both in [0, 1], while the origin is the image of the whole edge u = 0. So \Phi is one-to-one on the interior of S, where its Jacobian u is positive. Since x + y = u, Theorem 2 gives
With g(t) = e^{t^2} the integral is \displaystyle \int_0^1 t e^{t^2}\,dt = \frac{e - 1}{2}, though e^{(x + y)^2} has no elementary antiderivative in x. The shear (x, y) \mapsto (x + y, y) proves the identity too; \Phi earns its place by turning x^m y^n into a product (Exercise 20).
Boundaries
One-to-one is needed. T(u, v) = (u^2, v) maps [-1, 1] \times [0, 1] onto [0, 1]^2, of area 1, yet \iint \lvert 2u \rvert\,du\,dv = 2 over it: each point is counted once from u and once from -u. Without one-to-one, the integral of the Jacobian counts area with multiplicity.
The absolute value is needed. The swap (u, v) \mapsto (v, u) has Jacobian -1 and keeps every area. Trouble is allowed on the boundary only. The polar map at r = 0 and \Phi at u = 0 are not one-to-one and have Jacobian 0, on sets of zero area. Inside S a zero Jacobian breaks Theorem 3: T(u, v) = (u^3, v) is one-to-one, but its Jacobian 3u^2 vanishes on u = 0, where the inverse x^{1/3} has no derivative. (The integral formula survives for this map, the bad line having zero area, but the proof of Theorem 2 needs the C^1 inverse.)
Bounded regions. The theorems are for closed bounded regions. Over the whole plane, take a limit over growing regions, as in Exercise 9.
Where this goes next
Parametric surfaces and surface integrals measures a surface \mathbf{r}(u, v) in space with dS = \lVert \mathbf{r}_u \times \mathbf{r}_v \rVert\,du\,dv; for a flat piece, \mathbf{r} = \langle x, y, 0 \rangle, that length is \displaystyle \left\lvert \frac{\partial(x, y)}{\partial(u, v)} \right\rvert. Green's theorem and Stokes' theorem keep track of orientation, the sign this lesson discards. In probability, Joint distributions and The bivariate normal use Theorem 2 to find the density of transformed random variables, which picks up the same factor. The theorem in \mathbb{R}^n, with the inverse function theorem behind it, is proved in Real Analysis.
Exercises
Computing Jacobians
- Find \displaystyle \frac{\partial(x, y)}{\partial(u, v)} for x = 3u - v, y = u + 2v.
- Find \displaystyle \frac{\partial(x, y)}{\partial(u, v)} for x = e^u\cos v, y = e^u\sin v.
- Find \displaystyle \frac{\partial(x, y)}{\partial(u, v)} for x = uv, y = u^2 - v^2.
- Find \displaystyle \frac{\partial(x, y, z)}{\partial(u, v, w)} for x = u + w, y = v - w, z = 2u + v, and say what the map does to volumes.
- Let u = x + y and v = x - 2y. Find \displaystyle \frac{\partial(x, y)}{\partial(u, v)} without solving for x and y, then check by solving.
Linear changes of variables
- Evaluate \iint_D (x - y)(2x + y)\,dA, where D is the parallelogram bounded by y = x, y = x - 2, y = -2x and y = -2x + 6.
- Evaluate \iint_D e^{x + y}\,dA over the square \lvert x \rvert + \lvert y \rvert \le 1, using u = x + y and v = x - y.
- Find the area of the region 5x^2 + 6xy + 2y^2 \le 4.
- Find \iint_{\mathbb{R}^2} e^{-(5x^2 + 6xy + 2y^2)}\,dA, as the limit of the integrals over the regions 5x^2 + 6xy + 2y^2 \le R^2.
- Leah’s second block is bounded by the rows x + 3y = 0 and x + 3y = 40 and the paths 2x - y = 10 and 2x - y = 50 (meters), with yield 0.6 + 0.01(x + 3y) kilograms per square meter. Find its area and total yield.
Curved regions
- Find the area of the region in the first quadrant bounded by y = x^2, y = 2x^2, x = y^2 and x = 3y^2.
- Let D be the region of the first quadrant with 1 \le x^2 - y^2 \le 4 and 1 \le xy \le 3. Evaluate \iint_D (x^2 + y^2)\,dA.
- With K and \Phi as in Example 5, evaluate \iint_K x^2 y\,dA.
- Let a, b > 0. Find the Jacobian of x = ar\cos\theta, y = br\sin\theta, and use it to evaluate \displaystyle \iint_D \left(\frac{x^2}{a^2} + \frac{y^2}{b^2}\right) dA over the ellipse \displaystyle \frac{x^2}{a^2} + \frac{y^2}{b^2} \le 1.
Three variables
- Find the volume of the solid 0 \le x + y \le 1, 0 \le y + z \le 2, 0 \le x + z \le 3, and the integral of x + y + z over it.
- Show that x = u, y = uv, z = uvw maps the cube [0, 1]^3 onto the solid 0 \le z \le y \le x \le 1, find its Jacobian, and use it to find the volume of the solid and \iiint xyz\,dV over it.
Reasoning and proofs
- Let T(u, v) = (u^2 - v^2, 2uv) and Q = [-1, 1]^2. Compute \displaystyle \iint_Q \left\lvert \frac{\partial(x, y)}{\partial(u, v)} \right\rvert du\,dv, explain why it is not the area of T(Q), and find that area.
- Let T and S be as in Theorem 1, with A(S) > 0. Prove that the centroid of T(S) is the image under T of the centroid of S.
- Let T : (s, t) \mapsto (u, v) and R : (u, v) \mapsto (x, y) be C^1. Prove that \displaystyle \frac{\partial(x, y)}{\partial(s, t)} = \frac{\partial(x, y)}{\partial(u, v)} \cdot \frac{\partial(u, v)}{\partial(s, t)}, with the first factor taken at T(s, t).
- Prove that \displaystyle \iint_K x^m y^n\,dA = \frac{m!\,n!}{(m + n + 2)!} for nonnegative integers m and n, with K as in Example 5. First prove \displaystyle \int_0^1 v^n (1 - v)^m\,dv = \frac{m!\,n!}{(m + n + 1)!} by induction on m.
Answers
- 7. x_u y_v - x_v y_u = 3 \cdot 2 - (-1) \cdot 1.
- e^{2u}. x_u y_v = e^{2u}\cos^2 v and x_v y_u = -e^{2u}\sin^2 v, so the Jacobian is e^{2u}\left(\cos^2 v + \sin^2 v\right).
- -2(u^2 + v^2). x_u = v, x_v = u, y_u = 2u, y_v = -2v, so the Jacobian is -2v^2 - 2u^2.
- -1: the map keeps every volume and reverses orientation. Expanding along the first row of the matrix with rows (1, 0, 1), (0, 1, -1), (2, 1, 0) gives 1 \cdot (0 + 1) + 1 \cdot (0 - 2).
- \displaystyle -\frac13. \displaystyle \frac{\partial(u, v)}{\partial(x, y)} = 1 \cdot (-2) - 1 \cdot 1 = -3, and Theorem 3 gives the reciprocal. Solving, \displaystyle x = \frac{2u + v}{3} and \displaystyle y = \frac{u - v}{3}, with Jacobian \displaystyle \frac23 \cdot \left(-\frac13\right) - \frac13 \cdot \frac13 = -\frac13.
- 12. u = x - y runs over [0, 2] and v = 2x + y over [0, 6]; \displaystyle \frac{\partial(u, v)}{\partial(x, y)} = 1 + 2 = 3, so \displaystyle dA = \frac13\,du\,dv and the integral is \displaystyle \frac13 \int_0^2 u\,du \int_0^6 v\,dv = \frac13 \cdot 2 \cdot 18.
- e - e^{-1}. Since \lvert x \rvert + \lvert y \rvert = \max(\lvert x + y \rvert, \lvert x - y \rvert), the square is [-1, 1]^2 in u and v. \displaystyle \frac{\partial(u, v)}{\partial(x, y)} = -2, so \displaystyle dA = \frac12\,du\,dv and the integral is \displaystyle \frac12 \int_{-1}^{1} e^u\,du \cdot 2.
- 4\pi. 5x^2 + 6xy + 2y^2 = (2x + y)^2 + (x + y)^2, and u = 2x + y, v = x + y has determinant 2 - 1 = 1, so the region has the area of the disk u^2 + v^2 \le 4.
- \pi. As in Exercise 8 the region is the disk u^2 + v^2 \le R^2 with dA = du\,dv, so the integral is \int_0^{2\pi} \int_0^R e^{-r^2} r\,dr\,d\theta, which is \pi\left(1 - e^{-R^2}\right) and tends to \pi.
- Area \displaystyle \frac{1600}{7} \approx 228.6 square meters, yield \displaystyle \frac{1280}{7} \approx 182.9 kilograms. u = x + 3y runs over [0, 40] and v = 2x - y over [10, 50], \displaystyle \frac{\partial(u, v)}{\partial(x, y)} = -1 - 6 = -7, so \displaystyle dA = \frac17\,du\,dv. The yield is \displaystyle \frac17 \int_{10}^{50} \int_0^{40} (0.6 + 0.01u)\,du\,dv = \frac17 \cdot 40 \cdot 32.
- \displaystyle \frac19. \displaystyle u = \frac{y}{x^2} runs over [1, 2] and \displaystyle v = \frac{x}{y^2} over [1, 3], with inverse x = (u^2 v)^{-1/3}, y = (uv^2)^{-1/3}. \displaystyle \frac{\partial(u, v)}{\partial(x, y)} = \frac{4}{x^2 y^2} - \frac{1}{x^2 y^2} = 3u^2 v^2, since \displaystyle uv = \frac{1}{xy}. The area is \displaystyle \int_1^2 \int_1^3 \frac{dv\,du}{3u^2 v^2} = \frac13 \cdot \frac12 \cdot \frac23.
- 3. u = x^2 - y^2 runs over [1, 4] and v = xy over [1, 3], and \displaystyle \frac{\partial(u, v)}{\partial(x, y)} = 2x \cdot x + 2y \cdot y = 2(x^2 + y^2). So \displaystyle (x^2 + y^2)\,dA = \frac12\,du\,dv and the integral is \displaystyle \frac12 \cdot 3 \cdot 2. The map is one-to-one on the first quadrant: the point (u, 2v) has polar coordinates (r^2, 2\theta) when (x, y) has (r, \theta), and \displaystyle 0 < \theta < \frac{\pi}{2}.
- \displaystyle \frac{1}{60}. x^2 y = u^3 (1 - v)^2 v, and with the Jacobian u the integral is \displaystyle \int_0^1 u^4\,du \int_0^1 v(1 - v)^2\,dv = \frac15 \cdot \frac{1}{12}.
- abr, and \displaystyle \frac{\pi ab}{2}. x_r y_\theta - x_\theta y_r = abr\cos^2\theta + abr\sin^2\theta. The ellipse is the image of 0 \le r \le 1, 0 \le \theta \le 2\pi, where the integrand is r^2, so the integral is \displaystyle \int_0^{2\pi} \int_0^1 r^2 \cdot abr\,dr\,d\theta = 2\pi \cdot \frac{ab}{4}.
- Volume 3, integral \displaystyle \frac92. u = x + y, v = y + z, w = x + z has \displaystyle \frac{\partial(u, v, w)}{\partial(x, y, z)} = 2, so \displaystyle dV = \frac12\,du\,dv\,dw over [0, 1] \times [0, 2] \times [0, 3], of volume 6. Since \displaystyle x + y + z = \frac{u + v + w}{2}, the integral is \displaystyle \frac14 \iiint (u + v + w) = \frac14 (3 + 6 + 9).
- u^2 v; volume \displaystyle \frac16 and \displaystyle \iiint xyz\,dV = \frac{1}{48}. For x, y > 0 the inverse is u = x, \displaystyle v = \frac{y}{x}, \displaystyle w = \frac{z}{y}, and 0 \le z \le y \le x \le 1 exactly when u, v, w lie in [0, 1]; the map fails to be one-to-one only on the faces u = 0 and v = 0. The Jacobian matrix is lower triangular with diagonal 1, u, uv. The volume is \displaystyle \int u^2\,du \int v\,dv \int dw = \frac13 \cdot \frac12, and xyz \cdot u^2 v = u^5 v^3 w gives \displaystyle \frac16 \cdot \frac14 \cdot \frac12.
- \displaystyle \frac{32}{3}; the area is \displaystyle \frac{16}{3}. The Jacobian is 2u \cdot 2u - (-2v) \cdot 2v = 4(u^2 + v^2), and \displaystyle \iint_Q 4(u^2 + v^2) = 4\left(\frac43 + \frac43\right). But T(-u, -v) = T(u, v) and Q is symmetric about the origin, so T covers T(Q) twice and Theorem 2 does not apply. On the half Q^+ = [0, 1] \times [-1, 1], T squares the distance to the origin and doubles the angle, which is one-to-one for angles in \displaystyle \left(-\frac{\pi}{2}, \frac{\pi}{2}\right), so T is one-to-one on the interior of Q^+, T(Q^+) = T(Q), and the area is \displaystyle \iint_{Q^+} 4(u^2 + v^2) = \frac{16}{3}.
- Write T(u, v) = (au + bv + e, cu + dv + k). Theorem 1 with f = 1 gives A(D) = \lvert \det A \rvert A(S) for D = T(S), and with f = x it turns \displaystyle \bar{x}_D = \frac{1}{A(D)} \iint_D x\,dA into \displaystyle \frac{1}{A(S)} \iint_S (au + bv + e)\,du\,dv and so a\bar{u} + b\bar{v} + e by linearity and \iint_S u\,du\,dv = \bar{u}\,A(S). In the same way \bar{y}_D = c\bar{u} + d\bar{v} + k, so (\bar{x}_D, \bar{y}_D) = T(\bar{u}, \bar{v}). \blacksquare
- The functions are C^1, hence differentiable (Tangent planes and linear approximation), so Result 2 of The multivariable chain rule applies to R \circ T and gives the matrix product D(R \circ T)(s, t) = DR(T(s, t))\,DT(s, t). By Theorem 5 of What the determinant means the determinant of a product is the product of the determinants, which is the claim. \blacksquare
- Let I(m, n) = \int_0^1 v^n (1 - v)^m\,dv; induct on m, for all n at once. For m = 0, \displaystyle I(0, n) = \frac{1}{n + 1} = \frac{0!\,n!}{(n + 1)!}. If the formula holds for m, integrate by parts with (1 - v)^{m + 1} and \displaystyle \frac{v^{n + 1}}{n + 1}; the boundary term vanishes at both ends, so \displaystyle I(m + 1, n) = \frac{m + 1}{n + 1}\,I(m, n + 1), which by the hypothesis is \displaystyle \frac{m + 1}{n + 1} \cdot \frac{m!\,(n + 1)!}{(m + n + 2)!}, that is \displaystyle \frac{(m + 1)!\,n!}{(m + n + 2)!}. Now apply Theorem 2 with \Phi of Example 5 to the continuous x^m y^n: it becomes u^{m + n}(1 - v)^m v^n, and with the Jacobian u the integral is \int_0^1 u^{m + n + 1}\,du \cdot I(m, n), which is \displaystyle \frac{1}{m + n + 2} \cdot \frac{m!\,n!}{(m + n + 1)!} = \frac{m!\,n!}{(m + n + 2)!}. \blacksquare