Multivariable calculus

Double Integral of 1/x Between x = y and x = y^2

The double integral of 1/x over the region between x = y and x = y^2 for y from 1 to 2 equals 2 ln 2 - 1, about 0.386294. The inner integral collapses to ln(y^2) minus ln(y), which is just ln y, and integrating ln y from 1 to 2 gives 2 ln 2 - 1.

f(x,y)=1xf(x,y) = \frac{1}{x}
12yy21xdxdy=2ln21\int_1^2 \int_{y}^{y^2} \frac{1}{x}\,dx\,dy = 2\ln 2 - 1

Numerically 0.386294, confirmed by quadrature on every build.

The inner integral collapses to a single logarithm

On 1y21 \le y \le 2 the curve x=y2x = y^2 lies to the right of x=yx = y, and the whole region sits in x1x \ge 1. So 1x\frac{1}{x} is continuous there and nothing is singular.

yy21xdx=ln ⁣(y2)lny=2lnylny=lny\int_{y}^{y^2} \frac{1}{x}\,dx = \ln\!\left(y^2\right) - \ln y = 2\ln y - \ln y = \ln y

That collapse is the point of the problem. What survives is the standard integral of lny\ln y, done by parts with u=lnyu = \ln y and dv=dydv = dy.

12lnydy=[ylnyy]12=(2ln22)(01)=2ln21\int_1^2 \ln y\,dy = \left[y\ln y - y\right]_1^2 = \left(2\ln 2 - 2\right) - \left(0 - 1\right) = 2\ln 2 - 1

Numerically 2ln210.3862942\ln 2 - 1 \approx 0.386294.

The mistake: a difference of logs is not a log of a difference

ln ⁣(y2)lny\ln\!\left(y^2\right) - \ln y is ln ⁣(y2y)=lny\ln\!\left(\frac{y^2}{y}\right) = \ln y. It is not ln ⁣(y2y)\ln\!\left(y^2 - y\right). The subtraction happens to the logarithms, not to their arguments.

The faulty version is not merely a different expression, it is undefined at y=1y = 1, where y2y=0y^2 - y = 0. A formula that blows up at an endpoint the region genuinely includes is a signal to back up one step.

  • lnalnb=ln ⁣(ab)\ln a - \ln b = \ln\!\left(\frac{a}{b}\right), never ln(ab)\ln(a-b).
  • ln ⁣(y2)=2lny\ln\!\left(y^2\right) = 2\ln y for y>0y > 0, which is where the factor of 22 in the answer comes from.
  • dxx=lnx\int \frac{dx}{x} = \ln|x|, and the absolute value is safe to drop here only because x1x \ge 1 throughout the region.

The other slip is losing the y-y in lnydy=ylnyy\int \ln y\,dy = y\ln y - y. Without it the answer would read 2ln21.3862\ln 2 \approx 1.386, more than three times too large.

Where the region is, and why 1/x is safe here

At y=1y = 1 the two curves meet at the single point x=1x = 1, so the region tapers to a corner. It widens as yy increases, reaching 2x42 \le x \le 4 at y=2y = 2.

area=12(y2y)dy=7332=56\text{area} = \int_1^2 \left(y^2 - y\right)\,dy = \frac{7}{3} - \frac{3}{2} = \frac{5}{6}

Because xx never approaches 00, the integrand stays between 14\frac14 and 11 on the region. Multiplying those bounds by the area brackets the answer between 14560.208\frac14 \cdot \frac56 \approx 0.208 and 1560.8331 \cdot \frac56 \approx 0.833, and 0.3862940.386294 sits inside.

Had the region reached x=0x = 0, the same integrand would have made the integral improper, and none of this bracketing would apply.

Frequently asked questions

Why is this a type 2 region rather than type 1?

Both boundaries are given as xx in terms of yy, and each horizontal slice crosses the region once. As a type 1 region it splits at x=2x = 2: for 1x21 \le x \le 2 the slice runs from y=xy = \sqrt{x} to y=xy = x, and for 2x42 \le x \le 4 from y=xy = \sqrt{x} to y=2y = 2. Two integrals in place of one.

Does ln here mean the natural logarithm?

Yes. Throughout calculus ln\ln is the logarithm base ee, the one whose derivative is 1x\frac1x. Since ln20.693147\ln 2 \approx 0.693147, the answer is 2(0.693147)10.3862942(0.693147) - 1 \approx 0.386294.