Why does the Ratio Test become inconclusive at the endpoints?

At an endpoint x = a ± R, the limit L in the Ratio Test equals 1 exactly. The test only distinguishes convergence when L 1. At L = 1, the test gives no information because the ratio of successive terms does not shrink fast enough to guarantee convergence, nor does it grow to guarantee divergence. You must use another test to decide.

Can both endpoints be included in the interval of convergence?

Yes. For Σ x n /n 2 , R = 1, and both endpoints converge: x = 1 gives the p-series with p = 2 (converges), and x = −1 gives the alternating version (converges absolutely). The interval is [−1, 1].

Can neither endpoint be included?

Yes. For Σ x n , R = 1, but the series diverges at both x = 1 (geometric with r = 1) and x = −1 (geometric with r = −1, oscillates and diverges). The interval is (−1, 1).

What is the most common mistake students make with endpoints?

Assuming that the interval of convergence is simply (a − R, a + R) without testing the endpoints. This is the failure mode listed in the research: forgetting to check endpoint convergence for power series. Always test x = a + R and x = a − R separately.

Which tests do I use for endpoint testing?

The p-series test works when the endpoint series becomes Σ 1/n p . The Alternating Series Test works for alternating endpoints. The comparison test and limit comparison test work for positive-term series. The divergence test (nth-term test) can sometimes show divergence quickly if the terms do not approach 0.

Endpoint Tests After Finding a Radius of Convergence

The Radius Settles The Interior, Not The Endpoints

You find the radius of convergence R by applying the Ratio Test (or Root Test) to the power series Σ c_n (x − a)ⁿ. The test gives L = lim |c_n+1/c_n| · |x − a|, then you set L < 1 to get |x − a| < R. That inequality defines an open interval (a−R, a+R) where the series converges absolutely. Every x in that interior satisfies L < 1, so the Ratio Test is decisive there. What the test cannot tell you is what happens at x = a + R or x = a − R. At those two boundary points, L = 1 exactly, and the Ratio Test becomes inconclusive. That is why you must run separate endpoint tests after finding a radius of convergence: the radius settles the interior, but it tells you nothing about the two points where the test mechanism itself breaks down.

Why The Ratio Test Stops At L = 1

The Ratio Test works by comparing the size of successive terms. When L < 1, terms shrink fast enough to force convergence. When L > 1, terms eventually grow and force divergence. At L = 1 the test cannot distinguish between a series whose terms shrink slowly (like Σ 1/n) and one whose terms shrink just a little faster (like Σ 1/n²). The first diverges; the second converges. The Ratio Test sees the same limit for both and returns a blank verdict. This is not a bug in the test; it is the boundary of what a ratio limit can measure. The radius R is the distance from the center a to that boundary. Inside that boundary, the ratio condition is strict; on the boundary, it is not. That is why you need endpoint tests after finding a radius of convergence: the radius of convergence calculator gives you R, but you must bring other convergence tests to examine the boundary itself.

Endpoint Tests After Finding A Radius Of Convergence Example: Σ (x−2)<sup>n</sup> / (n·3<sup>n</sup>)

Finding The Radius

For the power series Σ (x−2)ⁿ / (n·3ⁿ), apply the Ratio Test. Write a_n = (x−2)ⁿ / (n·3ⁿ). Then |a_n+1/a_n| = |x−2|/3 · n/(n+1). The limit as n → ∞ is |x−2|/3. Set |x−2|/3 < 1, giving |x−2| < 3. So R = 3. The open interior is (2−3, 2+3) = (−1, 5). The Ratio Test guarantees absolute convergence for every x inside that interval.

Testing The Left Endpoint: x = −1

Substitute x = −1 into the series. The numerator becomes (−3)ⁿ. The series becomes Σ (−3)ⁿ / (n·3ⁿ) = Σ (−1)ⁿ / n. This is the alternating harmonic series. Apply the Alternating Series Test: the terms b_n = 1/n decrease to 0. The test shows conditional convergence. So x = −1 is included in the interval of convergence.

Testing The Right Endpoint: x = 5

Substitute x = 5. The numerator becomes 3ⁿ. The series becomes Σ 3ⁿ / (n·3ⁿ) = Σ 1/n. This is the harmonic series, a p-series with p = 1. The p-series test says it diverges. So x = 5 is not included in the interval of convergence.

The Final Interval

The interval of convergence is [−1, 5). This is a failure case for anyone who assumes both endpoints behave the same way after finding R = 3. At x = 5, the series diverges. The radius R gave the correct interior, but you needed separate endpoint tests to settle each boundary.

Calculation Steps

Compare With Σ (x−2)ⁿ / n²·3ⁿ

The Ratio Test still gives R = 3 for this variant. The interior is again (−1, 5). But at the endpoints, the behavior differs. At x = 5, the series becomes Σ 1/n², a p-series with p = 2, which converges. At x = −1, the series becomes Σ (−1)ⁿ/n², which converges absolutely. The interval of convergence is [−1, 5] in this case. The radius R is identical, but the endpoint results are different. That is why endpoint testing is never optional: the radius alone cannot distinguish between the harmonic series (p = 1) and the p = 2 series. You must test each boundary point using the p-series test, the Alternating Series Test, or the comparison test, depending on the series that appears.

Why You Cannot Skip This Step

A common mistake is to assume that because R = 1 for Σ xⁿ/n, both endpoints behave the same way. At x = 1, the series is Σ 1/n (harmonic, diverges). At x = −1, the series is Σ (−1)ⁿ/n (alternating harmonic, converges conditionally). One endpoint is included, one is not. The radius gave you the open interval (−1, 1), but to write the interval of convergence you had to test each endpoint individually. That is exactly why the research calls this a failure mode: forgetting to check endpoint convergence for power series.

Why the Method Works

The Ratio and Root Tests determine convergence by checking whether the ratio of successive terms stays below 1. At the endpoints x = a ± R, that ratio equals 1 exactly, and the tests can say nothing more. The behavior at those points depends on the specific coefficients c_n and must be decided by other convergence tests. That is the full explanation of why the radius settles the interior but not the two boundary points.