Why the Graph Orientation Matters
A curve can often be described either by \(y=f(x)\) or, when the curve is invertible on the interval being considered, by \(x=g(y)\). These are two descriptions of the same profile, but they need not lead to equally convenient surface-area integrals.
For a surface of revolution, the integrand combines the distance to the axis of rotation with the length of a small piece of the profile. Changing the graph orientation changes how that length is written and may also change the limits of integration. The useful question is therefore not simply “Which variable is usual?” but “Which description gives the more manageable radius, derivative, interval, and integral?”
Throughout this tutorial, consider profiles in the first quadrant rotated about the \(x\)- or \(y\)-axis. The profile is assumed to be a single smooth branch on the interval used. The surface-area formulas below are the standard arc-length-based formulas for these rotations.
The Two Orientations and Their Formulas
For revolution about the \(x\)-axis, the radius is the height \(y\). For revolution about the \(y\)-axis, the radius is the horizontal distance \(x\). Multiplying circumference \(2\pi r\) by the arc-length element gives the corresponding surface-area formulas.
These formulas describe the same surface when they use the same profile segment and the same axis. The radius in each formula is measured from that axis; it is not automatically the dependent variable. The expressions assume the profile lies on one side of the axis, so the radius is nonnegative.
Why the Two Descriptions Agree
Here is the change-of-orientation argument when the inverse derivative is well behaved. Suppose \(y=f(x)\) on \([a,b]\), with \(f(a)=c\), \(f(b)=d\), and \(f'(x)>0\) throughout the interior. Then \(f\) is increasing and has an inverse \(x=g(y)\) on \([c,d]\). At corresponding interior points, the inverse-function derivative formula gives \(g'(y)=1/f'(x)\), and \(dy=f'(x)\,dx\).
Thus the two expressions for \(ds\) agree. Since the radius to a fixed axis is also the same geometric distance in either description, integrating \(2\pi r\,ds\) gives the same surface area.
For a decreasing profile with a nonzero derivative, the inverse formula also applies on the interior, but \(y\) decreases as \(x\) increases. Reversing the \(y\)-limits keeps the surface area positive. In practice, write the limits in increasing order for each integral and use the corresponding positive arc-length element.
A Practical Comparison
Before integrating, write down what each orientation changes. A convenient choice often makes one feature especially simple, but the whole integral matters: a simpler derivative is not helpful if the radius or limits become awkward.
| Feature to compare | Question to ask |
|---|---|
| Radius | Does either description make the distance to the axis simpler? |
| Derivative | Is \(dy/dx\) or \(dx/dy\) simpler inside the square root? |
| Limits | Are the \(x\)- or \(y\)-limits easier to state and use? |
| Integral | After combining the radius and arc-length factor, which integrand is easier? |
| Branch and endpoints | Is the curve a single-valued graph on the interval, and are any endpoint singularities integrable? |
There is no rule that \(dx\) is preferable for a graph \(y=f(x)\), or that \(dy\) is preferable whenever the inverse exists. The formulas are alternatives for describing the same curve segment. Choose the one that produces the clearest valid setup.
Worked Example: A Parabola Revolved About the \(y\)-Axis
Worked Example: Compare Both Orientations
Revolve the profile \(y=x^2\), \(0\le x\le 1\), about the \(y\)-axis. First use \(x\) as the variable. The radius is \(x\), and \(dy/dx=2x\), so
Set \(u=1+4x^2\), so \(du=8x\,dx\), or \(x\,dx=du/8\). The limits change from \(x=0,1\) to \(u=1,5\). Therefore
Now use \(y\) as the variable. Since \(y=x^2\), the inverse branch is \(x=g(y)=\sqrt y\), for \(0\le y\le1\). On the interior \(y>0\), \(g'(y)=1/(2\sqrt y)\). The radius to the \(y\)-axis is \(g(y)\), so
The \(dy\) description simplifies the integrand after the radius and derivative are combined. Although \(g'(y)\) is unbounded as \(y\) approaches \(0\), the combined integrand has the finite limit \(1/2\). The integral is therefore understood as the limit from positive \(y\), and it converges. Both orientations give the same exact area.
Worked Example: A Line Revolved About the \(y\)-Axis
Worked Example: Check That Changing the Limits Preserves the Area
Take \(y=2x+1\), \(0\le x\le2\), and revolve the segment about the \(y\)-axis. In the \(dx\) description, \(dy/dx=2\), the radius is \(x\), and
For the \(dy\) description, solve for \(x\): \(x=(y-1)/2\). The \(y\)-limits are \(1\) and \(5\), and \(dx/dy=1/2\). Thus
The limits changed because the same segment runs from \(y=1\) to \(y=5\). The matching result checks both the radius and the derivative factor in the \(dy\) setup.
Worked Example: A Quarter-Circle Profile
Worked Example: Compare Descriptions with an Endpoint Singularity
Use the quarter-circle \(x^2+y^2=1\) in the first quadrant, from \((0,1)\) to \((1,0)\), and revolve it about the \(x\)-axis. As a graph of \(x\), it is \(y=\sqrt{1-x^2}\), with \(dy/dx=-x/\sqrt{1-x^2}\) on \(0\le x<1\). The radius is \(y\), so
The inverse description is \(x=\sqrt{1-y^2}\), with \(dx/dy=-y/\sqrt{1-y^2}\) for \(0\le y<1\). Here the radius to the \(x\)-axis is \(y\), so
The inverse derivative becomes unbounded as \(y\) approaches \(1\), but the full surface-area integrand is integrable. The calculation is justified as an improper integral: evaluate up to \(y=r<1\), then take the limit as \(r\) approaches \(1\). The limit is \(2\pi\), agreeing with the \(dx\) calculation.
Common Setup Mistakes
A frequent error is to switch to \(dy\) by replacing \(dy/dx\) with its reciprocal but leaving the old radius or the old limits unchanged. For example, for the line \(y=2x+1\) revolved about the \(y\)-axis, a wrong setup is
This mixes the \(dx\)-based radius \(x\) and \(x\)-limits with a \(dy\)-based derivative, and the differential does not match the limits. The correction is to use the inverse \(x=(y-1)/2\), the \(y\)-limits \(1\) to \(5\), and \(dx/dy=1/2\) throughout:
Another mistake is assuming that strict monotonicity alone makes the inverse derivative finite at every point. For example, \(f(x)=x^3\) is strictly increasing, but \(f'(0)=0\), and its inverse does not have a finite derivative at \(0\). Check the derivative condition before using a pointwise inverse-derivative formula; where an endpoint is singular, check the limit of the complete integrand instead.
Choosing an Orientation Reliably
Confirm that both proposed setups describe the same curve segment and rotation axis.
Use distance to the axis: height for the \(x\)-axis, horizontal distance for the \(y\)-axis.
For \(dx\), use \(dy/dx\) and \(x\)-limits; for \(dy\), use \(dx/dy\) and \(y\)-limits.
Combine radius and arc-length factor before deciding which integral is simpler.
Use an inverse only on a single-valued branch, and justify any unbounded derivative by examining the full integrand and its limiting integral.
Why the Choice Matters
Choosing the orientation is a small but important modeling decision. A good choice can turn an awkward derivative or radius into a simple expression, make the limits natural, and expose an integrable endpoint limit. A poor choice can obscure the same geometry behind unnecessary algebra.
The choice does not change the surface. It changes only how the profile’s arc length and radius are expressed in the integral. Keep the curve segment, axis, radius, derivative, and integration limits consistent, and either valid orientation will produce the same area.
Check Your Understanding
For each question, identify which parts of the setup change when the graph orientation changes.
- For a profile revolved about the \(y\)-axis, what is the radius in a \(dx\)-based formula, and what is it in a \(dy\)-based formula?
- What derivative and limits belong in a \(dy\)-based surface-area integral?
- Why does strict increase alone not ensure that the inverse derivative is finite everywhere?
- For \(y=x^2\), \(0\le x\le1\), revolved about the \(y\)-axis, which orientation simplifies the combined integrand, and what is the resulting exact area?
- When an inverse derivative is unbounded at an endpoint, what should be checked before accepting the integral?
