An equation with infinitely many solutions
The equation asks for integers whose squares differ by one after the indicated scaling. Direct substitution finds and , but a list of examples does not explain whether there are infinitely many or whether some have been missed. Multiplication in the set of numbers produces solutions; a descent argument will show that, in this particular equation, it produces all of them.
We first classify solutions with and . Changing either sign preserves the equation, so this also determines all integer solutions after the signs are restored. The argument below concerns specifically; it does not prove existence or classification for every nonsquare coefficient.
Multiplication preserves the equation
Definition 7.1.1
(The norm of )
Write . Direct multiplication shows .
Let , so and .
Theorem 7.1.2
(All nonnegative solutions)
The solutions with and are exactly
The exponent is unique.
Every nonnegative integer power has integer coefficients, norm one, positive and nonnegative . To prove that no other solutions exist, we need a way to make any nontrivial solution smaller.
Proof
(The inverse step descends)
Suppose , and . If , then , impossible for an integer. Thus and , so .
Multiply by :
The coefficients are integers and . Since and its conjugate is , the identities
show and . Also because , so . The nonnegative integer strictly decreases. Repeating the inverse step must reach , where . Reversing the steps expresses the original solution as a power of . Since , distinct nonnegative powers are distinct, proving uniqueness.
Common mistake
(The descent stops at )
The inverse step is applied only while . Applied to the trivial solution it gives , which leaves the sign range: the bound used , and that fails for . The process therefore ends exactly at , which is why the count of steps is the exponent .
Recurrence and an example
Multiplying by gives the integer recurrence
Worked example 7.1.3
(Three steps without searching a grid)
The first nontrivial pairs are , and . For example, . The recurrence gives further solutions, while the descent proof—not the recurrence alone—shows that every solution in the stated sign range appears.
Worked example 7.1.4
(Descending from )
Apply , repeatedly:
Each intermediate pair satisfies the equation, for instance , and falls . Three steps were needed, so . Expanding and multiplying once more by gives , confirming the exponent.
Worked example 7.1.5
(Restoring signs: the solution )
The pair solves the equation, since , but lies outside the range . Its absolute values give . Because , we have , and therefore
The same reasoning writes every integer solution as with . The edge case gives the two solutions , the only ones with .
Counterexample mode
A congruence is not a full solution
Every solution has odd and even : reducing the equation modulo two makes odd, and then reducing modulo four forces even. But meets these parity conditions and gives , not . Necessary congruence conditions can reject candidates without classifying all solutions.
Rational approximations
For any solution with and , the equation implies . Factoring the difference gives
Thus approaches from above along these positive-sign solutions as grows. The recurrence makes strictly increase after the first nontrivial pair, so there are arbitrarily large denominators. This estimate is a consequence of this equation; it is not a claim that every good rational approximation must arise from it.
Exercises
Quick check
(Starting from , use the recurrence to obtain the next solution and verify its equation.)
Substitute into the recurrence.
Solution · Answer
The recurrence gives , and .
Quick check
(Show why cannot occur in the stated sign range.)
Substitute into the equation.
Solution · Answer
It would require , which is not an integer square.
Quick check
(If solve the equation with , show that is odd and is even.)
Reduce the equation modulo , then modulo .
Solution · Answer
Modulo two, , so is odd. An odd square is one modulo four; the equation then gives , so is even.
Quick check
(Use the factorization above to bound by an explicit positive multiple of .)
Use to bound from below.
Solution · Answer
By the factorization, , which is positive. Since (because ), we have , so the difference is less than .