# document.write (document.title)

 Math Help > Number Theory > Squares > Fourths and Squares

The sum of two fourth powers can't be a square, nor can the difference of two fourth powers.  Infinite descent proof in 19 cases (patterns) that x^r+y^s=z^t where r,s,t≠2,4.

## Fourths and Squares

### Two Fourths and a Square

This section explores various combinations of squares and fourth powers, with and without coefficients.  As it turns out, none of these patterns has a solution.

The first pattern is x4+y4 = z2, and if this has no solution, then neither does x4+y4 = z4.  In other words, we are proving Fermat's last theorem for n = 4.  Two fourth powers cannot sum to a fourth power.

We need to consider all these patterns in one go.  This is the nature of the proof.  The reasoning goes something like this.

If pattern 1 has a solution then there is a smaller triple that satisfies pattern 2.  And if pattern 2 has a solution then there is a smaller triple that satisfies pattern 1.  Thus every solution implies a smaller solution of the same type, and since positive integers cannot descend forever, there is no solution.

This approach is called the "method of infinite descent", and it is a consequence of induction.  If there is a solution there is a least solution, and yet there is always a smaller one, which is a contradiction.

What makes one triple smaller than another?  This can be a bit unclear, especially when the triples fit different patterns.  The answer is the value of the right hand side.  (If the right side is z2, the value is z2, not z.)  If one triple has a lesser "value" than another, then it is smaller.

Whenever the coprime lemma applies, we automatically assume coprime variables.  After all, this can only make the right side smaller.

### Pattern 0: x4 + y4 = z2

This pythagorean triple can be characterized as follows.

x2 u2 - v2
y2 2uv
z u2 + v2

Since 2uv is a square, and u and v are coprime, u and v are s2 and 2t2 in some order.  Therefore x2 becomes s4-4t4 or 4t4-s4.  Rewrite this as x2+4t4 = s4 or x2+s4 = 4t4.  These are patterns (1) and (2) below.

Each time we defer to another pattern, you should verify that the right side has become smaller.  In this case the right side is u2 or v2.  Either of these is smaller than z, which is smaller than z2, so we're ok.

### Pattern 1: x2 + y4 = 4z4

Odd squares cannot sum to a multiple of 4, so either x or y is even, whence the other variable is also even.  If z is even, divide x by 4 and y and z by 2, producing a smaller solution.  Thus z is odd, and x is not divisible by 4.  Divide through by 4 to get this:

(�x)2 + (�y2)2 = (z2)2

Remember that �x is odd, and so is z2.  Characterize the triple as follows.

�x u2 - v2
�y2 2uv
z2 u2 + v2

With u and v coprime, the second equation shows u and v are squares.  The third equation becomes s4+t4 = z2.  With z2 < 4z4, we have no trouble deferring to pattern (0).

### Pattern 2: x2 + 4y4 = z4

Supppose x and z are odd, and characterize the triple this way.

x u2 - v2
2y2 2uv
z2 u2 + v2

The second shows u and v are squares, and the third becomes s4+t4 = z2.  With z2 < z4, we defer to pattern (0).

Let either x or z be even, whence the other is also even.  If y is even, divide x by 4 and y and z by 2 to find a smaller solution.  Thus y is odd.  This means x = 2 mod 4, else y would be even.  Divide through by 4 to get this.

(�x)2 + (y2)2 = (�z2)2

The sum of two odd squares is an even square, and a mod 4 argument shows this is impossible.

Patterns (0), (1), and (2) all fail together, and as a corollary, the sum of fourth powers is never a fourth power.  If that's all you needed to know, you can skip the rest of this page, but if you're interested, there are many more patterns with no solutions.  Some of these act as lemmas for future theorems.

### Pattern 3: x2 + y4 = z4

If y is even, characterize the triple as y2 = 2uv and z2 = u2 + v2.  Thus s4 + 4t4 = z2, pattern (4).

If y is odd, characterize the triple as y2 = u2 - v2 and z2 = u2 + v2.  Multiply these together to show u4 - v4 = (yz)2, pattern (3).  Note that y < z, hence y2z2 < z4.

### Pattern 4: x4 + 4y4 = z2

If x and z are odd, characterize the triple this way.

x2 u2 - v2
2y2 2uv
z u2 + v2

Thus u and v are squares, and x2 = s4-t4, pattern (3).

Let x and z be even.  If z is divisible by 4 we can divide through and find a smaller solution, so let z = 2 mod 4.  This makes y odd.  Divide through by 4 and characterize as follows.

y2 u2 - v2
�x2 = 2uv
�z = u2 + v2

This implies y2 = s4-t4, pattern (3).

### Pattern 5: x4 + 2y2 = z4

The coprime lemma is in force here, hence x and z are odd.  We characterized these triples in an earlier section.

x2 (u2 - 2v2)
y 2uv
z2 u2 + 2v2

Multiply the first and third equations together.  Now (xz)2 is the difference between a fourth power and 4 times a fourth power, patterns (1) and (2).

### Pattern 6: x4 + y4 = 2z2

Refer to an earlier section for the characterization.  Note that x and y are odd, and u and v have opposite parity.

x2 u2 - 2uv - v2
y2 u2 + 2uv - v2
z u2 + v2

Multiply the first two equations together and the left side becomes (xy)2.  Add 8u2v2 two both sides and the right side becomes (u2+v2)2.  This produces the following two equations.

(xy)2 + 8u2v2 = (u2+v2)2

(xy)2 + (u2+v2)2 = 8u2v2

The second equation is impossible mod 4, so concentrate on the first.  Write 2st = 2uv, and s2+2t2 = u2+v2.

In this characterization, v might be 0, but u cannot be.  thus we can replace v with st/u.

s2+2t2 = u2 + s2t2/u2

(u2)2 - (s2+2t2)u2 + s2t2 = 0

Apply the quadratic formula.  The discriminant is s4+4t4, which must be a square.  This never happens, thanks to pattern (4).  Of course s or t could be 0; pattern (4) doesn't preclude that.  Thus v = 0, and the unique coprime solution emerges: 14 + 14 = 2�12.

### Pattern 7: x2 + y2 = 9z4, coprime terms

If x and y are divisible by 3, factor out 9 to create x2 + y2 = z4.  this has plenty of solutions, so let the terms be coprime.  Note that the coprime lemma does not apply here; we're just assuming the terms are coprime.

after characterizing the triple, we find u2+v2 = 3z2.  This is impossible.

### Pattern 7a: x2 + y4 = 9z4

The coprime lemma applies here.  There is a minimal solution that is coprime, except perhaps for 3.

If 3 does not divide x and y, there is no solution, thanks to pattern (7), so let 3 divide x and y.

If 3 divides z we can divide through and find a smaller solution, so 3 does not divide z, and x = 3 or 6 mod 9.  Let y be even and characterize the triple as follows.

x/3 u2 - v2
y2/3 2uv
z2 u2 + v2

Remember that u and v are coprime, and look at 6uv = y2.  Thus u and v are 6s2 and t2, or 3s2 and 2t2.  Substitute in the third equation.

36s4 + t4 = z2

9s4 + 4t4 = z2

Since z2 is less than 9z4, we defer to patterns (10) and (11).

Next let y be odd and find the following characterization.

y2/3 u2 - v2
x/3 2uv
z2 u2 + v2

Use the last equation to assign s2-t2 and 2st to u and v.  Then substitute in the first equation.

y2/3 = s4 - 6s2t2 + t4

The left side can be viewed as 3 times a square.  Look at this equation mod 8.  Exactly one of s and t is even, hence two of the three terms on the right drop to 0.  Now a fourth power is 3 times a square.  This is impossible mod 8, so we're done.

### Pattern 8: x2 + 9y4 = z4

The terms are coprime, except perhaps for 3.  If 9 divides x then 3 divides y, and we can divide through by 9 to find a smaller solution.

Suppose x is divisible by 3.  It must be 3 or 6 mod 9, and 3 does not divide y.  Derive the following equation, and note that it is impossible mod 3.

(x/3)2 + y4 = (z2/3)2

Thus the terms are pairwise coprime.  There are two different characterizations, depending on whether y is even or odd.  In each case z2 = u2+v2.

Let y be odd, whence u2-v2 = 3y2.  Multiply this by the equation for z2 and find 3(yz)2 = u4-v4.  Since u2 < z2, u4 < z4, and we defer to pattern (9).

If y is even then 2uv = 3y2.  Thus u and v are 6s2 and t2, or 3s2 and 2t2.  Substitute these expressions to set z2 equal to 36s4 + t4 or 4s4 + 9t4.  These are patterns (10) and (11) respectively.

### Pattern 9: 3x2 + y4 = z4

Note that y cannot be even, else 3 + 0 = 1 mod 4.  either x or z is even.

Let x be even, and z odd, and refer to the characterization of such triples.

y2 u2 - 3v2
x 2uv
z2 u2 + 3v2

Multiply the first and third together to show (yz)2 is the difference between a fourth power and 9 times a fourth power.  These are patterns (7a) and (8).

Next let z be even, whence x and y are odd.  In this case the characterization gives z2/2 = u2+3v2.  Since z2/2 is even, u and v have the same parity.  This contradicts the characterization, i.e. it makes x and y even as well.

### Pattern 10: x4 + 36y4 = z2

First assume all terms are coprime, giving 6y2 = 2uv and x2 = u2-v2.  Thus x2 is the difference between a fourth power and 9 times a fourth power, patterns (7a) and (8).

Let 2 divide x and z, while 3 does not.  For a minimal solution, z = 2 mod 4, and y is odd.  Divide through by 4 and characterize.

3y2 u2 - v2
x2/2 = 2uv
z/2 = u2 + v2

The second equation shows u and v are squares, and the first equation sets 3 times a square equal to the difference between fourth powers, which is pattern (9).

Next let 3 divide x and z while 2 does not.  For a minimal solution, z is not divisible by 9, and y is not divisible by 3.  Divide through by 9 and characterize.

x2/3 u2 - v2
2y2 2uv
z/3 u2+v2

Once again u and v are squares, and the first equation leads to pattern (9).

Finally let 6 divide x and z, but not y, so that we may have a minimal solution.

y2 u2 - v2
x2/6 2uv
z/6 u2 + v2

This time u and v are a square and 3 times a square, and the first equation sets x2 to the difference between a fourth power and 9 times a fourth power, patterns (7a) and (8).

### Pattern 11: 4x4 + 9y4 = z2

Again there are four cases to consider.  Start with coprime terms, whence 2x2 = 2uv and 3y2 = u2-v2.  The first equation makes u and v squares, and the second leads to pattern (9).

Let 2 divide y, while 3 does not divide x.  Thus 3y2/2 = 2uv, and x2 = u2-v2.  The first equation makes u and v a square and 3 times a square, and the second produces patterns (7a) and (8).

Let 3 divide x, while 2 does not divide y.  Thus 2x2/3 = 2uv, and y2 = u2-v2.  The first equation makes u and v a square and 3 times a square, and the second produces patterns (7a) and (8).

finally 3 divides x and 2 divides y.  Divide through by 36 for a coprime triple.  Thus y2/2 = 2uv, and x2/3 = u2-v2.  With u and v squares, the second equation leads to pattern (9).

This completes the circle; all patterns up to this point have no solution, save the one solution for pattern (6).  But there are more patterns to consider.

### Pattern 12: x2 + 3y4 = z4

Let y be even, whence x and z are odd.  Thus y2 = 2uv and z2 = u2+3v2.  This makes u and v a square and twice a square.  Substitute and find that z2 is either s4+12t4 or 4s4+3t4.  These are patterns (13) and (14) respectively.

If z is even then z2 is divisible by 4, which is impossible, since the characterization of z2/2 is an odd number.

### Pattern 13: x4 + 12y4 = z2

If z is even then it is 2 mod 4, and y is odd.  divide through by 4 and note that the resulting equation is impossible mod 4.

x4/4 + 3y4 = z2/4

Thus z and x are odd.  Characterize this triple and write 2y2 = 2uv, and x2 = u2-3v2.  Substitute squares into the second equation to get x2 = (s4-3t4).  This either sets a sum of squares equal to 3 times a square, which is impossible mod 4, or it creates pattern (12).

### Pattern 14: 4x4 + 3y4 = z2

If y is odd then 0+3 = 1 mod 4.  So y is even, and x is odd, and z = 2 mod 4.

Divide through by 4, and only the second term remains even.  Characterize, and write y2/2 = 2uv, and x2 = u2-3v2.  Once again x2 is the difference between a fourth power and 3 times a fourth power.  As we saw in the last pattern, this is impossible, or it creates pattern (12).

### Pattern 15: x8 + 243y8 = z2, y odd

If 3 divides x and z then 27 divides z, and 3 divides y, and 81 divides z.  We could divide z by 81 and x and y by 3 to find something smaller.  Therefore, all terms are coprime.

The following equation is equivalent.

(z+x4) � (z-x4) = 243y8.

If 3 divides both factors it divides x, and we ruled that out.  Thus 243 belongs to one of the two factors alone.

Let y be odd, whence one factor is an eighth power, and the other is 243 times an eighth power.  Subtract the two equations to get 2x4 = u8-243v8.  Reduce this mod 17.  Both u and v cannot drop to 0, as they are coprime.  An eighth power is 0 1 or -1.  In addition to these, a fourth power could be 4.  If u or v is 0 the right side is 1 or 5, while the left side is 0, 2, or 8.  This cannot be reconciled, so both u and v are nonzero.  The right side becomes 4 or 6.  The left side is still 0, 2, or 8.  We are forced to conclude that y is even.

### Pattern 16: x8 + 243y8 = z2, y even

As shown by the previous pattern, 3 cannot divide x or z, hence the terms are coprime.  Characterize as follows.

x4 u2 - 3v2
9y4 = 2uv
z = u2+3v2

To avoid squares summing to thrice a square, the first equation produces +x4.  Thus x4+3v2 = u2.

U and v are fourth powers, with coefficients of 8 and 9.  If 3 divides u it divides x, hence 9 belongs to v.  That leaves two possibilities.

Set u = s4 and v = 72t4, giving x4 + 15552t8 = s8.  Thisis pattern (13).

Setting u = 8s4 and v = 9t4 leads to the following.

x4 + 243t8 = 64s8

This is impossible mod 8, unless x and t are even.  With 64 on the right, x must be divisible by 4.  This makes s even, and we can divide through for a smaller solution.

### Pattern 17: x8 + 3888y8 = z2

Start by assuming x and z are even.  If 8 divides z then y is even, which means 16 divides z.  Reduce z by 16 and x and y by 2 to find a smaller solution.  So if z is even it is divisible by 4, and no more, and y is odd.  Divide through by 16 as follows.

16x8 + 243y8 = z2

This is impossible mod 4, hence x and z are odd.

If 3 divides x and z then 27 divides z, and 3 divides y, and 81 divides z.  We could divide z by 81 and x and y by 3 to find something smaller.  Therefore, all terms are coprime.

x4 u2 - 3v2
36y4 2uv
z u2+3v2

This time u and v are fourth powers with coefficients of 2 and 9.  If 3 divides u it divides x, hence 9 belongs to v.

Set u = s4 and v = 18t4, giving x4 + 972t8 = s8.  This is pattern (13).

Set u = 2s4 and v = 9t4 and derive x4 + 243t8 = 4s8.  This is pattern (18).

### Pattern 18: x4 + 243y8 = 4z8

If x and y are even then z has to be even as well.  This makes x divisible by 4, and we can find a smaller solution.

If x and z are divisible by 3, we can make x divisible by 9 and find a smaller solution, as we have done before.  Thus the terms are coprime.

The following equation is equivalent.

(2z4+x2) � (2z4-x2) = 243y8.

With x and z nonzero mod 3, 243 divides the first factor.  Othere than this, both factors are eighth powers.

Add the equations together to get this.

4z4 = u8+243v8

With 4z4 < 4z8, this is pattern (15).

### simpler descent proof of the sum case

From Trevor, University at Buffalo on November 3, 1996:
I want to prove, using mathematical induction, that there are no solutions to the equation x4+y4=z4, for positive values of x, y, z.
I assume that x, y, and z are supposed to be integer values; otherwise, there are plenty of solutions.

Those reading this page will likely recognize this question as the n=4 case of Fermat's Last Theorem. The general theorem (that, when n > 2, there are no positive integer solutions to the equation xn+yn=zn), was conjectured by Fermat hundreds of years ago but remained unproved until just recently, when it was proved by Andrew Wiles. The proof is highly complex and involves some deep areas of abstract mathematics.

The n=4 case, however, is relatively easy to prove and was known by Fermat. It turns out to be a little easier to prove the more general result that there is no solution in positive integers to the equation x4+y4=z2 (that is, the sum of two fourth powers cannot even be a perfect square, let alone a fourth power).

This is traditionally proven using the "method of infinite descent". The key to this method is the key lemma below (I'll explain later how this fact is proven).

Key Lemma. If x4+y4=z2 where x, y, and z are positive integers, then there would exist other positive integers u, v, and w with u4+v4=w2, and w < z.

(For those unfamiliar with the terminology: a lemma is a small theorem whose main use is in proving another, more important, theorem).

The reason this lemma implies that solutions cannot exist is as follows. Suppose solutions did exist. Among all solutions, pick one with the smallest z value (any nonempty set of postive integers has a smallest element). But now the lemma says there would be another solution with a still smaller z value, a contradiction. Therefore, solutions cannot exist.

You can phrase this using the language of mathematical induction, but it's more awkward. You would work by induction on z. When z=1 there are clearly no solutions (x and y have to be at least 1, so z2=x4+y4 has to be at least 2).

Now, if it is known that there are no solutions with z≤N, you can prove that there are no solutions with z=N+1 either: if (x,y,z) were such a solution, then the lemma implies that there is another solution (u,v,w) with w < z, so w≤N, contradicting the fact that we know no such solution exists.

The above two paragraphs form the basis and induction steps for a proof by induction. But it's a little cleaner to use the method of infinite descent.

Now for the hard part: a proof of the lemma. I won't fill in all the details (since you originally posted this in the Discussion Topics section rather than the Questions section, I assume you don't want me too), but I will outline the proof.

It depends on the theory of Pythagorean triples. A triple (a,b,c) of positive integers is called a Pythagorean triple if a2+b2=c2.  It is called a fundamental Pythagorean triple if a, b, and c have no common factor greater than 1.

For any fundamental Pythagorean triple, one of a and b is even and the other is odd. Let's use the letter a to refer to the even one, b to the odd one. Then there are positive integers m and n such that

a = 2mn
b = m2 - n2
c = m2 + n2

and m and n are relatively prime (have no common factor greater than 1).

If you don't know how to prove the above statements, feel free to post another question here.

Now, suppose x4+y4=z2.  Then (x2,y2,z) is a Pythagorean triple.

If x, y, and z have a common prime factor p > 1, then (x/p)4+(y/p)4=(z/p2)2 and we have our desired solution (u,v,w) with u = x/p, v = y/p, and w = z/p2 < z.  (You need also to explain why z/p2 is an integer; I'll leave that to you).

On the other hand, if x, y, and z do not have a common prime factor, then (x2,y2,z) is a fundamental Pythagorean triple. One of x and y is even and the other odd; let's use x to denote the even one. Then, by the theory of Pythagorean triples, there are relatively prime positive integers m and n such that

x2 = 2mn
y2 = m2 - n2
z = m2 + n2

We can rewrite the second equation as n2+y2=m2.  We know that m and n have no common factor greater than 1, so (n,y,m) is a fundamental Pythagorean triple. We know y is odd, so n must be the even one. Therefore, appealing to the theory of Pythagorean triples once again, there are relatively prime positive integers r and s such that

n = 2rs
y = r2 - s2
m = r2 + s2.

The final piece of the puzzle is the fact that if the product of two relatively prime positive integers is a perfect square, then each individually is a perfect square. From this it follows that m and n/2 are perfect squares, since the product is (m)(n/2) = (2mn)/4 = (x2)/4 = (x/2)2  (remember that x and n are even, so x/2 and n/2 are integers).

It also follows that r and s are perfect squares, since the product is rs = (2rs)/2 = n/2  which is a perfect square.

Therefore, setting r=u2, s=v2, m=w2, the equation m=r2+s2 becomes u4+v4=w2.  All that remains to complete the proof of the lemma is to show that w < z, which follows because z=m2+n2=w4+n2>w4, so w4<z, and as w is a positive integer this implies w < z as well.

If you want proofs of some of the statements I've left for you to fill in, just post a message here and we'd be glad to fill them in for you.

### Proof of the difference of two fourths isn't a square

by Ng Kwong Ming

Posted on Sunday, 14 August, 2005 - 09:48 am:

Proof that x4-y4=z^2 has no solutions.

Turning the equation around slightly,

z2+(y2)2=(x2)2, so z,y2,x2 forms a Pythagorean triple.

Assume the equation has a solution in integers.  Let x,y,z by the solution with the least x.  If GCD(x,y)=d > 1, then we can divide both sides of the equation by d4 to obtain a smaller solution, which is impossible if no smaller value than x solves the equation.  This forces x to be odd, because no primitive Pythagorean triple has an even number in the third position.

y is either even or odd.  I will consider both cases:

case 1: y is odd.

y2,z,x2 forms a primitive Pythagorean triple, with y odd.
z=2st, s>t, GCD(s,t)=1, s,t are of opposite parity.
y2=s2-t2
x2=s2+t2
s4-t4=(xy)2, so it is a solution to the equation.
0<s<sqrt(s2+t2)=x, contradicting minimal value of x

case 2: y is even.

z,y2,x2 forms a primitive Pythagorean triple, with y even.
y2=2st
z=s2-t2
x2=s2+t2.

If s is even and t odd, then from y2=2st, we have GCD(2s,t)=1, so 2s=w2, t=v2.
Since w is even, w=2u, and s=2u2
And x2=s2+t2=4u4+v4, so v2,2u2,x form a primitive Pythagorean triple with
2u2=2ab,
v2=a2-b2, and
x=a2+b2.
u2=ab implies a,b are squares, so a=c2, b=d2.
v2=a2-b2=c4-d4.  This results in a new solution c, d, v of x4-y4=z2, and what is more,
0<c=sqrt(a)<a2+b2=x, contrary to our assumption regarding x.

If t is even and s odd, then from y2=2st, we have GCD(s,2t)=1, so s=w2, 2t=v2.
Since v is even, v=2u, and t=2u2
And x2=s2+t2=w4+4u4, so w2,2u2,x form a primitive Pythagorean triple, and then the rest of it follows the same logic as the preceding paragraph.

Combining the above, the only resolution of these contradictions is that the equation cannot be satisfied in integers

### Internet references

http://www.math.toronto.edu/mathnet/questionCorner/fermat4.html -- proof that x^4+y^4=z^2 has no solution

http://www.mathreference.com/num-zext,442.html -- proof that eighteen different patterns of equations involving two fourth powers and a square have no solution by a convoluted finite state machine using infinite descent with each transition.

### Related pages in this website

The webmaster and author of this Math Help site is Graeme McRae.