Real Analysis 9 | Subsequences and Accumulation Values [dark version]

Subsequences let mathematicians “thin out” a sequence without changing the order of its terms, and they preserve convergence when it already exists. That preservation matters because it turns the study of complicated behavior—especially divergence—into a search for simpler repeating patterns inside the sequence.

A subsequence is built by choosing a strictly increasing index sequence K = (K1, K2, …) and then taking the terms a_{K1}, a_{K2}, … from the original sequence a_n. The strict increase ensures the subsequence keeps the original order; it can skip infinitely many terms as long as infinitely many remain. For example, starting with a_n = 1/n and choosing K_k = 2^k produces the subsequence 1/2, 1/4, 1/8, …, which still decreases in the same order as the original. Because order is preserved, monotonicity carries over: if the original sequence visits a value like 1/2 only once and 1/4 only once (with 1/4 occurring after 1/2), the subsequence cannot swap their order.

Most importantly, subsequences do not change the limit of a convergent sequence. If a_n converges to a, then every subsequence also converges to the same a; taking K → ∞ yields the same limiting value. This fact is straightforward in the example above: both the full sequence 1/n and the subsequence 1/2^k converge to 0.

The real payoff comes when the original sequence diverges. Divergence can still hide convergent behavior inside it, and subsequences are the tool for extracting that behavior. Consider a_n = (-1)^n, which oscillates between 1 and -1 and therefore has no limit. Yet restricting to even indices gives a constant subsequence equal to 1, and restricting to odd indices gives a constant subsequence equal to -1. Those different limiting values are not random; they are the sequence’s “accumulation values.”

An accumulation value (also called a cluster point or accumulation point) is any real number a such that some subsequence converges to a. For a convergent sequence, there can be only one accumulation value—the limit itself. For a divergent sequence, multiple accumulation values can occur, reflecting the idea that the sequence keeps returning arbitrarily close to several points. A geometric picture is a sequence that keeps jumping around but repeatedly comes closer and closer to different locations; by focusing on the terms near one location, a convergent subsequence emerges.

There’s also an equivalent neighborhood-based definition: a is an accumulation value of a_n if and only if every ε > 0 interval (a − ε, a + ε) contains infinitely many terms of the sequence. In other words, the sequence does not merely pass near a; it revisits the neighborhood of a endlessly. This characterization sets up later results about when accumulation values must exist, including a forthcoming theorem attributed to Bolzano–Weierstrass.