Method stop · When and why do we use it?

Common Methods for Diophantine Equations

In the Chinese book: 不定方程的常用解法

Factoring, estimation and congruences each cut infinitely many cases to a few.

Before this stop, visit:Linear Diophantine Equations (and Systems) (Skill)

One idea, one pictureChoose an equation at the top and a method below. Each of the four equations has one method that works best. Which is it for each?

Loading figure…

In one sentence

What makes integer solutions hard is that there are “infinitely many possibilities”. The three common methods all do the same thing: turn infinitely many into finitely many.

  • Factoring: rewrite the equation as a constant. An integer has only finitely many divisors.
  • Estimation (bounding): use inequalities to lock the unknowns into a finite range, then check each case.
  • Congruences: pick a modulus for which the two sides can never leave the same remainder, and the possibilities drop straight to zero.

Each of the three methods in action

Step 1 / 6
Loading figure…
  1. Factoring: first build a product

    . The left side looks like the first three terms of expanded; it is only missing the constant . Add to both sides.

Stop 1 · Factoring

Spot it: an term together with linear terms in and , or a difference of squares. Do it: add a constant to build a product, then list the divisors: each split gives a small system of equations to solve.

Predict first

becomes . There are ways to split into two integer factors (counting negative factors and order). For how many of them do you get integer and ?

Choose an answer first; then you can see the result.

Stop 2 · Estimation

Spot it: sums of squares, sums of reciprocals, or one side that clearly “grows faster” than the other. Do it: put the unknowns in order of size or complete the square to get an inequality, which limits the smallest (or largest) unknown to just a few values.

One problem, several solutions

Find positive integers with .

Estimation

, so . Also , so , which gives , that is, .

You only need to check , working out :

  • : , ✓
  • : , ✓
  • : , is not an integer ✗
  • : , ✓

CostNo clever rewriting needed, but you must check 4 candidates.

Factoring

Clear the denominators: , that is, . Add to both sides:

and both are greater than , so are positive divisors: , , .

This gives .

CostYou must think of adding 16 after clearing denominators; after that it is one step.

Both methods find the same three solutions. When factoring is possible it is usually faster; when it is not (for example, a sum of reciprocals of three unknowns), estimation still works.

Stop 3 · Congruences

Spot it: the problem asks you to “prove there are no integer solutions”, or the equation has squares or cubes. Do it: choose a modulus that makes some terms vanish and leaves the powers as few remainders as possible.

Loading figure…
. Drag the modulus from to : which moduli leave no cell at all? Which ones get you nowhere?

Mod and mod each settle it in one blow (and so, of course, do the multiples and ); mod all leave surviving cells. Good first choices:

If the equation hasTry the modulusBecause
squares, , leaves only (mod , ) or (mod )
cubes, leaves only
a coefficient that is a multiple of that term simply vanishes

How to choose

There is no method that always works. Try every one of the four equations with each of the three methods on the workbench above and you get this table:

FactoringEstimationCongruences
✓ four solutionscan't fence it inonly narrows it down
✓ four solutionscan't fence it inonly narrows it down
doesn't factor✓ check 121 cellsonly narrows it down
doesn't factorcan't fence it in✓ no solutions

Practice

Exercise 1Basic

How many positive integer solutions does have?

Exercise 2Basic

Find all positive integers for which there is a positive integer with .

Exercise 3Advanced

Positive integers satisfy . How many such triples are there?

Exercise 4Challenge

Prove that has no integer solutions.

Mistake Museum · Exhibits from this stop

Before you leave, look at the traps other people fell into here.See every exhibit in the museum →

Exhibit

Forgetting negative divisors when splitting

The wrong step: , so or : only two integer solutions.
What went wrong? Think first, then open

Diagnosis The question asks for integer solutions, and are valid splits too. They give and .

The right step: For integer solutions, list both positive and negative divisors: solutions in all. Only when the problem asks for “positive integer solutions” may you drop the negative ones because of the range.

“For integer solutions, divisors come with both signs”

Exhibit

No contradiction mod m, so the equation has a solution

The wrong step: Mod , the two sides of can match (with and both odd), so the equation has integer solutions.
What went wrong? Think first, then open

Diagnosis If one modulus finds no problem, that only shows this modulus is not good enough. Switch to mod : the left side is , the right side leaves , and there is an immediate contradiction.

The right step: The congruence method can only prove there are no solutions. To prove there is a solution, you have to actually produce one.

“One modulus letting it through does not mean all of them do”

Exhibit

Finding the range, then not checking each case

The wrong step: From we get , so the solutions are the lattice points in this range.
What went wrong? Think first, then open

Diagnosis The range is a necessary condition: every solution is inside the box, but most points in the box are not solutions. Of the cells, only are solutions.

The right step: Once you have a finite range, patiently substitute the cases one by one (symmetry can save some of the work).

“Fencing in the range is only the first half”