'Find a Rational Between Two Reals' → Archimedean Property, Then m/N

Some problems look like magic tricks the first time you see them. "Let a and b be real numbers with a < b. Show that there exists a rational number q with a < q < b." You stare at two real numbers — possibly irrational, possibly transcendental, possibly given only as a = \pi - e and b = \pi - e + 10^{-9} — and you are asked to produce a specific ratio of integers inside a gap you cannot even see clearly. Where do you start?

The trick is not a trick. It is a fixed recipe that works every time, and the purpose of this article is to install that recipe in your head as a recognition pattern. Once you have it, problems of this shape stop being puzzles and start being exercises.

The signal

The phrase to watch for, in any of its disguises:

Given a < b in \mathbb{R}, find / show there exists / prove the existence of a rational q with a < q < b.

Variants that are the same problem wearing different clothes:

"Show that between any two distinct reals there is a rational."
"Prove that \mathbb{Q} is dense in \mathbb{R}."
"Given a < b, find p, q \in \mathbb{Z} with q > 0 and a < p/q < b."
"Approximate any real number to within \varepsilon by a rational."

All four are the same recipe. The moment you recognise the signal, you stop thinking about what to do and start thinking about the two small decisions inside the recipe.

The mental model (two moves)

Here is the full thinking process. Two moves, and nothing else.

Move (a). Pick an integer N so that \tfrac{1}{N} < b - a. This is the Archimedean property at work. Since b - a > 0 is a positive real number, the Archimedean property guarantees some positive integer N with N(b - a) > 1, which rearranges to \tfrac{1}{N} < b - a. Concretely, any N > \tfrac{1}{b - a} will do — pick the smallest such N, or round up \lceil \tfrac{1}{b-a} \rceil + 1 to be safe. Think of \tfrac{1}{N} as a ruler spacing: you have laid down a ruler on the number line whose tick marks are \tfrac{1}{N} apart, and that spacing is smaller than your gap.

Move (b). Pick an integer m with Na < m < Nb. Because you chose N so that Nb - Na > 1, the interval (Na, Nb) has length greater than 1 — and any interval of length more than 1 contains an integer. Concretely, take m = \lfloor Na \rfloor + 1: that is the smallest integer strictly greater than Na. Dividing the inequality Na < m < Nb by the positive integer N gives a < \tfrac{m}{N} < b. The rational q = \tfrac{m}{N} is your answer.

That is the whole recipe. No cleverness, no case analysis, no appeal to what a and b look like. The only fact about the real numbers you use is that the integers are dense enough on the integer lattice and sparse enough to skip no gaps of length greater than 1.

Lay a ruler with spacing $\tfrac{1}{N}$ along the number line. If the ruler is finer than the gap, at least one tick mark must fall inside $(a, b)$ — that tick is $\tfrac{m}{N}$, your rational $q$. The Archimedean property guarantees the ruler can always be made fine enough.

Walking the recipe on a concrete problem

Before the worked example, one more word about what the recipe is actually saying. The reason move (a) works is structural: the gap b - a is a specific positive real number, and the Archimedean property is exactly the axiom that says "for every positive real r, there is a positive integer N with r > \tfrac{1}{N}." You are not choosing to invoke Archimedean; you are being handed the invocation by the problem the moment you read "a < b". That is why recognition matters more than cleverness — the problem statement already contains every ingredient for the solution, and the recipe is just the order in which to reach for them.

Here is the recipe applied to specific numbers. Take a = \sqrt{2} and b = 1.42. Both are real, a is irrational, b is rational, and they are close — about 0.006 apart. Where is a rational in between?

Step 1. Measure the gap. b - a = 1.42 - \sqrt{2} \approx 1.42 - 1.41421 = 0.00579. So you need a ruler with spacing below 0.00579.

Step 2. Pick N. Any N > 1/0.00579 \approx 172.8 works. Take N = 200 for a clean number. Then \tfrac{1}{N} = 0.005 < 0.00579.

Step 3. Scale the interval by N. Na = 200\sqrt{2} \approx 282.843, and Nb = 200 \times 1.42 = 284. The interval (Na, Nb) \approx (282.843, 284) has length \approx 1.157 > 1, so it contains an integer.

Step 4. Pick m. The smallest integer strictly above Na \approx 282.843 is m = 283. Check: 282.843 < 283 < 284. Good.

Step 5. Divide back. q = \tfrac{m}{N} = \tfrac{283}{200} = 1.415.

Step 6. Sanity-check. Is \sqrt{2} < 1.415 < 1.42? Yes — 1.415^2 = 2.002225 > 2, so 1.415 > \sqrt{2}. And 1.415 < 1.42 is clear. The rational \tfrac{283}{200} lives strictly inside (\sqrt{2}, 1.42).

Notice that the whole derivation used only the Archimedean property and the floor function. Nothing about the specific nature of \sqrt{2} mattered — it could have been \pi, or e^{e}, or any other real number. The recipe does not care.

Why the two moves are forced, not clever

A natural first reaction is: "Why N > 1/(b-a)? Why not just take the midpoint \tfrac{a+b}{2}?"

The midpoint is between a and b, always. But it might not be rational. The midpoint of \sqrt{2} and \sqrt{3} is (\sqrt{2} + \sqrt{3})/2, which is irrational. The problem asks for a rational, so the midpoint is not a solution — it is only a solution when a and b are themselves rational.

The Archimedean move fixes this. By rescaling the real line by a factor of N, you convert "find a rational between a and b" into "find an integer between Na and Nb" — and integers, once you zoom in enough that the interval has length > 1, are always findable. The division by N at the end converts the integer back into a rational.

This is why move (a) has to come first. Without Archimedean, you have no guarantee that the scaled-up interval contains an integer at all. With Archimedean, you are guaranteed a scale factor that makes the interval wide enough to hold one.

Move (b) looks like a flourish, but it is actually the completeness-style fact about integers: every interval of real length > 1 contains an integer. That fact uses the floor function — itself a consequence of the Archimedean property — so the whole recipe lives inside the structure of \mathbb{R} and \mathbb{Z}, nothing deeper.

Why the recipe never fails — even for brutally narrow gaps

The first time a student meets this, the doubt that arrives is: "What if a and b are really close? Like b - a = 10^{-100}? Surely the recipe breaks." It does not, and the reason is worth stating explicitly, because it is the whole point of the Archimedean property.

Take b - a = 10^{-100}. Then the recipe wants N > 10^{100}. That is an enormous integer — far larger than the number of atoms in the observable universe — but it is still a finite positive integer. Archimedean does not care how large N needs to be; it only asserts that some finite N exists. Take N = 10^{101}. Then \tfrac{1}{N} = 10^{-101}, comfortably below the gap. The scaled interval (10^{101}a, 10^{101}b) has length 10^{101}(b - a) = 10, which is more than enough to contain an integer. The recipe produces a rational of the form \tfrac{m}{10^{101}}, and that rational — despite having a preposterous denominator — genuinely sits inside the gap.

This is the sense in which the rationals are "dense" in the reals. Not just "there are many of them" — the rationals physically infiltrate every interval, however narrow, however hidden among transcendentals. Zoom in as far as you like; you cannot zoom past them. The Archimedean property is the engine behind this infiltration. Take it away (as the hyperreal number system does, on purpose) and the recipe no longer works: you could have a gap smaller than \tfrac{1}{N} for every integer N, and no rational could fit inside.

Recognising the signal in disguise

Once the pattern is in your head, you will see it wearing unusual clothes. A few examples:

"Show that the decimal expansions of rationals are dense." Same recipe. Approximating a and b by finite decimals is exactly the N = 10^k case, and m is the integer part of 10^k a + 1.
"Given an irrational x and \varepsilon > 0, find a rational within \varepsilon of x." Set a = x - \varepsilon/2 and b = x + \varepsilon/2. Run the recipe. The rational produced is within \varepsilon of x.
"The Dirichlet approximation theorem." A deeper version of the same instinct, giving better N for irrationals with special structure. The same Archimedean bone, different flesh.
"The rationals are everywhere." Informal, but it is the density theorem talking.

When any of these appears — in a JEE paper, in a real-analysis tutorial, in a textbook proof — the mental reflex is: Archimedean, pick N, pick m, done.

A second worked example — to make the reflex automatic

One example is a demonstration. Two examples are a pattern. Here is the recipe on a second problem, deliberately chosen so that a is transcendental and b is another transcendental.

Take a = \pi and b = \tfrac{22}{7}. You know \pi \approx 3.14159265 and \tfrac{22}{7} \approx 3.14285714, so the gap is b - a \approx 0.00126. Tiny. Find a rational in between.

Move (a): pick N > 1/0.00126 \approx 793.6. Take N = 1000 for a round choice. Then \tfrac{1}{N} = 0.001 < 0.00126.

Move (b): compute Na = 1000\pi \approx 3141.59265 and Nb = 1000 \cdot \tfrac{22}{7} = \tfrac{22000}{7} \approx 3142.857. The interval (3141.59265, 3142.857) has length about 1.26, safely more than 1. Take m = \lfloor 3141.59265 \rfloor + 1 = 3142. Verify: 3141.59265 < 3142 < 3142.857. Done.

Divide back: q = \tfrac{3142}{1000} = \tfrac{1571}{500} = 3.142.

Sanity: \pi < 3.142 < \tfrac{22}{7}? Yes — \pi \approx 3.14159 < 3.142, and 3.142 = \tfrac{3142}{1000} while \tfrac{22}{7} = \tfrac{3142.857\ldots}{1000}, so 3.142 < \tfrac{22}{7}. Both checks pass.

Notice the pattern: in both worked examples you wrote down N first, wrote down Na and Nb next, took the floor to find m, and divided. Four lines of arithmetic per problem. No insight, no lateral thinking, no anxious staring at the page. The recipe is mechanical on purpose — that is what makes it a recognition pattern rather than a clever trick.

What this recognition pattern is not

It is not the same as finding an irrational between two reals. That uses density too, but with a different trick — scale an irrational like \sqrt{2} into the interval and add it to a rational inside. See Rationals Are Dense in the Reals — Animated Proof You Can Pinch for the visual version of the rational case, which shows exactly the ruler argument above.

It is also not the same as finding a rational with small denominator. The recipe gives some rational in (a, b) — with denominator N roughly 1/(b-a). If you need better (Stern–Brocot tree, continued fractions), that is a different toolbox. But for existence alone, the recipe here is complete.

The pattern in one line

"Find a rational in (a, b)" \Rightarrow Archimedean picks N > \tfrac{1}{b-a}, floor picks m = \lfloor Na \rfloor + 1, answer is \tfrac{m}{N}.

Hold that line in your head. The next time a problem asks for a rational between two reals, the recipe fires before you have finished reading the question — and you finish the problem faster than the student next to you finishes translating it.