We've talked about limits as a value a sequence 'approaches' or 'gets closer to'. This is a great starting point, but in mathematics, we need to be more precise. What does 'getting closer' actually mean? How close is close enough?

To make this idea rigorous, mathematicians developed a formal definition. Think of it as a game of challenge and response. Someone challenges you with a tiny margin of error, and you have to prove that your sequence eventually stays within that margin forever.

The formal definition of a convergent sequence uses two key variables: epsilon ($ε$) and N. Epsilon represents the 'challenge'—an arbitrarily small positive number that defines an error tolerance around the limit. N is our 'response'—a point in the sequence after which every single term is guaranteed to be within that epsilon tolerance of the limit.

This turns our intuitive notion into a concrete, testable statement. A sequence $a_n$ converges to a limit $L$ if, for any positive epsilon you can imagine, you can find a corresponding N that satisfies the condition.

Here is the formal statement in the language of mathematics:

The expression $|a_n - L| < \varepsilon$ is just a formal way of saying the distance between the sequence term and the limit is smaller than our tolerance. It can be rewritten as $L - \varepsilon < a_n < L + \varepsilon$. This creates a small horizontal band, or 'strip', around the limit $L$. The definition says that for any strip you draw, no matter how narrow, the sequence must eventually enter it and never leave.

Let's prove that the sequence $a_n = \frac{1}{n}$ converges to $L = 0$. We need to show that for any given $\varepsilon > 0$, we can find an $N$ that works.

Our condition is $|a_n - L| < \varepsilon$, which becomes |rac{1}{n} - 0| < \varepsilon. Since $n$ is a positive integer, this simplifies to rac{1}{n} < \varepsilon.

Now, we solve for $n$. Rearranging the inequality gives us n > rac{1}{\varepsilon}.

This is our rule for finding $N$. If someone challenges us with an $\varepsilon$, we can respond by choosing any integer $N$ that is greater than rac{1}{\varepsilon}. For any term $n$ after that $N$, the condition will hold. For example, if the challenge is $\varepsilon = 0.01$, then we need $n > \frac{1}{0.01}$, or $n > 100$. So we can choose $N=100$. Every term from $a_{101}$ onward will be less than 0.01 away from zero. This work was pioneered by mathematicians like Augustin-Louis Cauchy.

Since we can find a valid $N$ for any positive $\varepsilon$, we have formally proven that the limit of $\frac{1}{n}$ as $n$ approaches infinity is 0.

The definition is also powerful because it gives us a way to prove a sequence diverges. A sequence diverges if it does not converge. To show this, we negate the definition of convergence.

Negating the statement 'For all $\varepsilon$, there exists an $N$...' becomes 'There exists an $\varepsilon$ such that for all $N$...'.

Consider the sequence $a_n = (-1)^n$, which alternates between -1 and 1. Let's propose it converges to $L=1$. Can we find a 'bad' epsilon? Let's pick $\varepsilon = 0.5$. The epsilon-strip is $(0.5, 1.5)$. No matter how large an $N$ you choose, you can always find an odd $n > N$ where $a_n = -1$, which is outside the strip. Therefore, the sequence does not converge to 1. A similar argument shows it doesn't converge to -1 or any other value.

An important consequence of this rigorous definition is that a limit, if it exists, must be unique. A sequence cannot 'get close' to two different numbers at the same time. If it did, you could just choose an epsilon small enough to draw two separate, non-overlapping strips around the two proposed limits, and the sequence can't eventually be in both.