Tutorials › Calculus II › Trigonometric Blow-Ups On Finite Intervals

Singular Integrands · Tutorial 313 of 998

Trigonometric Blow-Ups On Finite Intervals

Identify finite singularities of secant and cosecant integrands, then test each one-sided improper integral using its local power behavior.

Intermediate 10 min read

What You'll Learn

  • Locate the zeros of sine and cosine that create secant and cosecant singularities.
  • Split a finite-interval integral at every interior singularity and treat endpoints one-sidedly.
  • Relate trigonometric blow-ups to powers of the distance from a zero.
  • Apply the power-singularity threshold to test local convergence.
  • Explain why an antiderivative cannot be evaluated across an interior pole.

Why trigonometric integrals can be improper

A trigonometric function may be smooth across most of a finite interval while becoming unbounded at particular points. For example, \(\sec x=1/\cos x\) is undefined wherever \(\cos x=0\), and \(\csc x=1/\sin x\) is undefined wherever \(\sin x=0\). These points divide an integral into pieces, and convergence must be checked on each piece separately.

The key observation is that sine and cosine approach zero linearly near each of their zeros. Consequently, a reciprocal power of sine or cosine behaves locally like a reciprocal power of the distance to the singular point. The power-singularity threshold then determines convergence.

Definition: An interior singularity of an integrand on \([A,B]\) is a point \(c\) with \(A<c<B\) where the integrand is undefined or unbounded. An integral with such a point is defined by splitting at \(c\) and taking the two one-sided improper integrals separately. Both must converge for the ordinary improper integral to converge.

Locating the singular points

The zeros of cosine occur at \(x=\frac{\pi}{2}+k\pi\), and the zeros of sine occur at \(x=k\pi\), where \(k\in\mathbb{Z}\). Thus \(\sec x\) has singularities at the cosine zeros, while \(\csc x\) has singularities at the sine zeros. On any finite interval there are only finitely many such points.

Before testing convergence, list every singular point that lies in the interval. Interior singularities become split points. If a singular point is an endpoint, the integral is instead evaluated using a one-sided limit there. A singularity outside the interval is irrelevant to that integral.

Procedure: For a finite-interval trigonometric integral, first locate all zeros of the denominator in the interval. Split at each interior zero, identify any singular endpoints, and test every resulting one-sided integral. Convergence on one piece does not compensate for divergence on another.

Why the blow-up is a power of the distance

Consider a zero \(a\) of sine or cosine, and write \(x=a+t\), so \(t\) is the distance in signed form from \(a\). The standard limit \(\lim_{t\to 0}\frac{\sin t}{t}=1\) implies that, for all sufficiently small nonzero \(t\), \(\frac12\leq\left|\frac{\sin t}{t}\right|\leq\frac32\). Equivalently, \(\frac12|t|\leq|\sin t|\leq\frac32|t|\).

At a cosine zero \(a=\frac{\pi}{2}+k\pi\), the identity \(\cos(a+t)=(-1)^{k+1}\sin t\) shows that \(|\cos(a+t)|=|\sin t|\). At a sine zero \(a=k\pi\), the identity \(\sin(a+t)=(-1)^k\sin t\) gives \(|\sin(a+t)|=|\sin t|\). Therefore, near either type of zero, the absolute value of the vanishing denominator is bounded above and below by positive constant multiples of \(|t|\).

$$ \frac12|t|\leq|\sin t|\leq\frac32|t| \qquad\text{for sufficiently small }|t|. $$

For any fixed \(q>0\), taking reciprocal \(q\)-th powers reverses the bounds. Thus \(|\sec(a+t)|^q\), near a cosine zero, and \(|\csc(a+t)|^q\), near a sine zero, are each bounded above and below by positive constant multiples of \(|t|^{-q}\). The earlier power-singularity threshold states that \(\int_0^\delta t^{-q}\,dt\) converges exactly when \(q<1\), for any \(\delta>0\).

Theorem: Near any zero of the denominator, \(|\sec x|^q\) or \(|\csc x|^q\), as appropriate, has the same local convergence behavior as \(|x-a|^{-q}\). For \(q>0\), the one-sided integral converges if \(q<1\) and diverges if \(q\geq1\). At an interior singularity, this test must hold on both sides.

The comparison works on each side because the bounds hold for both positive and negative \(t\). Away from the finitely many denominator zeros, the trigonometric integrand is continuous on closed subintervals and hence bounded there. Local behavior at the zeros is therefore what decides convergence.

Worked Example: A square-root secant blow-up

Worked Example: A square-root secant blow-up

Determine whether \(\int_0^\pi|\sec x|^{1/2}\,dx\) converges.

The denominator \(\cos x\) vanishes in \([0,\pi]\) only at \(a=\frac{\pi}{2}\). Thus the integral must be split there. Let \(t=x-a\). Since \(|\cos(a+t)|=|\sin t|\), the local bounds give \(\frac12|t|\leq|\cos(a+t)|\leq\frac32|t|\). Taking inverse square roots yields \((\frac32)^{-1/2}|t|^{-1/2}\leq|\sec(a+t)|^{1/2}\leq(\frac12)^{-1/2}|t|^{-1/2}\).

On each side of \(a\), the comparison integral is finite: \(\int_0^\delta t^{-1/2}\,dt=2\sqrt{\delta}\). So the one-sided integral on the left converges, and the one-sided integral on the right converges as well. On the rest of \([0,\pi]\), away from \(a\), the integrand is continuous and bounded. Therefore the full improper integral converges.

Check: The only denominator zero is \(\frac{\pi}{2}\); both one-sided pieces have exponent \(q=\frac12<1\), and there are no other singularities. This verifies that every part of the interval has been accounted for.

Worked Example: Secant across an interior pole

Worked Example: Secant across an interior pole

Determine whether \(\int_0^\pi\sec x\,dx\) converges as an ordinary improper integral.

The denominator vanishes at \(a=\frac{\pi}{2}\), so the required pieces are \(\int_0^a\sec x\,dx\) and \(\int_a^\pi\sec x\,dx\), each defined by a one-sided limit. An antiderivative on intervals where \(\cos x\neq0\) is \(F(x)=\ln|\sec x+\tan x|\). Indeed, \(\frac{d}{dx}F(x)=\frac{\sec x\tan x+\sec^2x}{\sec x+\tan x}=\sec x\), wherever this expression is defined.

On the left, the improper integral is \(\lim_{\varepsilon\to0^+}\bigl(F(\frac{\pi}{2}-\varepsilon)-F(0)\bigr)\). Here \(F(0)=0\), while \(\sec x+\tan x\) tends to \(+\infty\) as \(x\to\frac{\pi}{2}^{-}\). Thus \(F(\frac{\pi}{2}-\varepsilon)\to+\infty\), and the left piece diverges.

On the right, the improper integral is \(\lim_{\varepsilon\to0^+}\bigl(F(\pi)-F(\frac{\pi}{2}+\varepsilon)\bigr)\). Since \(F(\pi)=0\), while \(|\sec x+\tan x|\to\infty\) as \(x\to\frac{\pi}{2}^{+}\), we have \(F(\frac{\pi}{2}+\varepsilon)\to+\infty\). The right piece therefore tends to \(-\infty\) and diverges. Since neither one-sided integral converges to a finite value, the ordinary improper integral diverges.

Check: Differentiating \(F\) gives \(\sec x\), and the limits on the two split intervals diverge separately. The conclusion follows from the definition of an interior singularity, not from combining the two limits.

Worked Example: Several cosecant singularities

Worked Example: Several cosecant singularities

Determine whether \(\int_0^{2\pi}\csc^2 x\,dx\) converges.

The denominator \(\sin x\) vanishes at \(0,\pi,2\pi\). The points \(0\) and \(2\pi\) are singular endpoints, and \(\pi\) is an interior singularity. An antiderivative wherever \(\sin x\neq0\) is \(-\cot x\), since \(\frac{d}{dx}(-\cot x)=\csc^2 x\).

At the left endpoint, the one-sided integral from \(0\) to \(\frac{\pi}{2}\) has limit \(\lim_{\varepsilon\to0^+}\int_\varepsilon^{\pi/2}\csc^2x\,dx =\lim_{\varepsilon\to0^+}\cot\varepsilon=+\infty\). At the left side of the interior singularity, \(\lim_{\varepsilon\to0^+}\int_{\pi/2}^{\pi-\varepsilon}\csc^2x\,dx =\lim_{\varepsilon\to0^+}\cot\varepsilon=+\infty\). At its right side, \(\lim_{\varepsilon\to0^+}\int_{\pi+\varepsilon}^{3\pi/2}\csc^2x\,dx =\lim_{\varepsilon\to0^+}\cot\varepsilon=+\infty\). Finally, approaching the right endpoint from the left gives \(\lim_{\varepsilon\to0^+}\int_{3\pi/2}^{2\pi-\varepsilon}\csc^2x\,dx =\lim_{\varepsilon\to0^+}\cot\varepsilon=+\infty\).

These limits use \(\cot\varepsilon\to+\infty\) as \(\varepsilon\to0^+\), and the displayed evaluations follow from the antiderivative \(-\cot x\). Each one-sided piece diverges, so the full improper integral diverges.

Check: The zero at \(\pi\) requires two separate one-sided tests, while \(0\) and \(2\pi\) each require an endpoint test. The exponent is \(q=2\), above the convergence threshold \(q<1\), consistent with every computed limit.

A common mistake: evaluating across a pole

A tempting but invalid calculation is to apply the secant antiderivative from the preceding example directly across the singularity: \(\int_0^\pi\sec x\,dx=F(\pi)-F(0)=0-0=0\). This treats the interval as if the integrand were continuous throughout it. It is not: \(\sec x\) is undefined and unbounded at \(\frac{\pi}{2}\). The fundamental theorem of calculus cannot evaluate an integral across that pole by subtracting endpoint values.

The correction is to split at \(\frac{\pi}{2}\), form the two one-sided limits, and test them separately. As shown above, they diverge in opposite signed directions, so the ordinary improper integral does not converge. An apparent cancellation between divergent pieces does not make either one-sided integral finite.

Putting the test into practice

1
Find every denominator zero.
For secant, solve \(\cos x=0\); for cosecant, solve \(\sin x=0\). Keep only points in the interval, including its endpoints.
2
Split the interval.
Use every interior singularity as a split point. Treat a singular endpoint with a one-sided limit.
3
Identify the local power.
Near a denominator zero, reciprocal power \(q\) behaves like \(|x-a|^{-q}\). Test the corresponding one-sided integrals using the threshold \(q<1\).
4
Check every piece.
The ordinary improper integral converges only if all required one-sided integrals converge. Do not combine divergent limits across a singularity.

This approach matters because a trigonometric integrand can look familiar while its finite-interval integral is undefined as an ordinary improper integral. Locating the zeros first prevents missed poles, and the local power test turns the convergence question into a precise comparison near each one.

Check Your Understanding

For each question, identify the relevant singular points and explain what must be checked.

  1. Where are the singularities of \(\sec x\) on \([0,2\pi]\)? Which are interior and which are endpoints?
  2. For \(q>0\), when does \(|\csc x|^q\) have a convergent one-sided integral near a zero of sine?
  3. Why must an integral with an interior trigonometric pole be split into two one-sided integrals?
  4. Does \(\int_0^\pi|\sec x|^{3/4}\,dx\) converge? State the local test that decides.
  5. Why is it invalid to evaluate an antiderivative at the two endpoints when a pole lies between them?

Want all 1,114 practice problems?

This tutorial covers one topic. Calc Pass covers the full Calculus AB & BC curriculum (the core of Calculus I and II): full review, every practice question, and step-by-step solutions to each one.

QR code to download the Calc Pass app

Get the Calc Pass App

Scan to download and practice on the go, with instant scoring.

Calc Pass Volume 1: 1,114 Practice Questions book cover

Vol. 1: 1,114 Questions

The complete practice question bank, organized by AB & BC topic.

View on Amazon
Calc Pass Volume 2: Solutions Manual book cover

Vol. 2: Solutions Manual

Fully worked solutions to every question in Volume 1.

View on Amazon