3. Sequences

Convergence of sequences via ε-N arguments: monotone and Cauchy sequences, limsup and liminf.

• The ε–N Definition — Narrow the ε-band as much as you like: from some moment on, the whole tail sits inside it. This scene watches that moment arrive. Converging means closing in on the target value without end. A narrower band only delays the moment the tail fits inside. As long as that moment always comes, the sequence converges. The exact promise is the formula below. For example, a ⚽ ball rolled toward a line stops closer and closer to it. Demand it land within 10 cm, then within 1 cm — from some roll on, it always stays inside.
• Standard Limits: x^n and x^{1/n} — Powers of a number below 1 in absolute value, and roots of a positive number, take the stage. The scene checks where each of them gathers. Keep multiplying a number below 1 and it sinks toward 0. Take roots of a positive number again and again and it gathers at 1. At −1 the sign just flips back and forth. Beyond 1 in size, powers grow without end. Bernoulli's inequality handles every case. For example, eat half of the remaining 🍰 cake each day. What is left goes one half, one quarter, one eighth — nearly vanishing.
• Sequence and Function Spaces — Whole sequences get collected into a single space here. Each sequence becomes one point of the space. Sequences can be added and scaled, so the space is a vector space. The distance between two sequences can also be measured. The measuring rule decides which space a sequence fits into. A finite sum of absolute values fits the smallest space. A finite sum of squares fits the middle one. Merely being bounded fits the largest one. So each smaller space sits inside the bigger ones. The norm behind the distance bridges the two worlds. That these spaces are complete is stated as a known theorem only. For example, picture three nested 📦 boxes. Anything inside the small box also lies in the middle and the large one. Sequence spaces nest in exactly the same way.
• Cauchy–Schwarz Inequality — How large can the inner product of two vectors become? It can never exceed the product of the two norms. The inner product measures how much one vector overlaps the other's direction. The overlap is largest when the two vectors point the same way. Tilt them apart and the overlap falls below the product of the norms. The proof starts from one fact: a length is never negative. Reading that fact as a discriminant drops the inequality out. Equality holds only when the two vectors are parallel. For example, shine a light on a ✏️ pencil and its shadow is never longer than the pencil. Only when the pencil lines up with the light does the shadow match its full length.
• Hölder’s Inequality — This scene widens Cauchy–Schwarz to general exponents. The key is that the two exponents come as a matched pair. Multiply two sequences term by term and add them up: even that total has a ceiling. The ceiling is the product of the two sequences' norms. With both exponents equal to 2, it is exactly Cauchy–Schwarz. The key step is Young's inequality, which trades a product for shares of a sum. The precise formulas are below. For example, at a 🏪 shop you multiply each item's price by its quantity and add everything up. That total never exceeds the product of the price list's size and the quantity list's size.
• The Parallelogram Law — An identity hides between the two diagonals and the two sides of a parallelogram. This scene slides the exponent $p$ and checks when that identity survives. Every norm that comes from an inner product satisfies the identity. Computing with $e_1$ and $e_2$ makes the left side $2 \cdot 2^{2/p}$ and the right side $4$. The two agree only at $p = 2$. So an $\ell^p$ norm with $p$ different from 2 cannot come from any inner product. For example, a 📐 parallelogram drawn with a ruler stays exactly the same, but the measuring ruler is swapped. If swapping the ruler alone breaks the diagonal identity, that ruler never came from an inner-product blueprint.
• Why ℝ Has No Holes — A sequence built only from rationals creeps toward a place that is not rational. The scene shows the hole in the rational world. The iteration halves the error of each guess. Every term stays rational. Yet the destination is not a rational number. So rationals alone cannot fill that spot. The full irrationality proof is skipped; only the idea is shown. For example, measure the diagonal of a square with a 📏 ruler. Try 1.4, then 1.41, then 1.414 — no fraction ever fits it exactly. That spot is the hole in the rationals.
• Construction of the Real Numbers — The hole in the rationals gets filled and a real number is actually built. One real number is defined as a bundle of rational sequences. Lengthen the decimal expansion and the rational terms march into the hole. Sequences whose difference vanishes count as the same number. That bundle itself is one real number. Built this way, the number line ends up with no holes. Completeness is exactly what this construction buys. For example, climb a 🪜 staircase one step at a time toward the top. The numbers 1, 1.4, 1.41, 1.414 are footprints heading upward. The whole parade of footprints is the destination itself: the real number.
• A Convergent Subsequence Always Hides Inside — Every bounded sequence hides a convergent subsequence inside it. This scene finds that subsequence by halving the interval. The trick is to keep halving the interval. Always follow the half that holds infinitely many terms. Each halving makes the interval narrower. The points trapped inside pile onto a single value. Stringing those points together gives a convergent subsequence. For example, thirty 🧑 classmates scatter across a playground. Split the playground in half and one side still holds a crowd of them. Keep halving that side and you corner a single spot.
• Monotone and Cauchy Sequences — An increasing sequence blocked from above climbs to the very top. The scene also watches the tail squeeze into a single point. Bounded above, the sequence has nowhere to go but its highest allowed spot. That spot is the supremum, and the sequence converges to it. Cut to the tail and the remaining terms spread less and less. A sequence whose spread dies to 0 is a Cauchy sequence. Every convergent sequence is Cauchy. For example, drop coins into a 🐷 piggy bank: half the goal, then half of what remains, and so on. The total creeps toward the goal without ever jumping to it. The coins added later also differ from each other less and less.
• Lim sup and Lim inf — Watch only the tails of a sequence and track their highest and lowest values. The scene checks where the two of them settle. The supremum of the tails keeps stepping down, and the infimum keeps stepping up. The upper settling point is the lim sup, the lower one the lim inf. If the two meet, the limit exists. An oscillating sequence keeps them apart forever. The precise definitions are the formulas below. For example, record each day's high and low temperature with a 🌡️ thermometer. As days pass, the highs gather near one value and the lows near another. If the two values coincide, the weather has settled on a single temperature.
• Stolz–Cesàro Theorem — When the full ratio is stubborn, this scene detours through the ratio of differences. It is the sequence version of l'Hôpital's rule. Compare the growth of single steps, and that ratio settles first. The full ratio follows slowly from behind. The denominator must increase strictly and grow without bound. Both ratios share the same destination. The headline example's limit is the second formula below. For example, on a 🚗 car trip the speed over the last minute settles at 60 km/h. Then the average speed over the whole trip gets dragged toward 60 km/h as well.
• Double Sequences and Iterated Limits — This scene is about double sequences. Watch the values $a_{m,n}$ on a grid, and see how the row-wise and column-wise limits split apart. The double limit lets both indices grow together. An iterated limit sends one index first, then the other. Swapping the order can change the answer. And even when both iterated limits agree, the double limit may not exist. For example, reading a 📋 table row by row is not the same as reading it column by column. Change the reading order, and you can land on a completely different conclusion.