Equation xy = yx explained

xy=yx

has solutions, such as

x=2,y=4.

History

The equation

xy=yx

is mentioned in a letter of Bernoulli to Goldbach (29 June 1728). The letter contains a statement that when

x\ney,

the only solutions in natural numbers are

(2,4)

and

(4,2),

although there are infinitely many solutions in rational numbers, such as

(\tfrac{27}{8},\tfrac{9}{4})

and

(\tfrac{9}{4},\tfrac{27}{8})

.The reply by Goldbach (31 January 1729) contains a general solution of the equation, obtained by substituting

y=vx.

A similar solution was found by Euler.

J. van Hengel pointed out that if

r,n

are positive integers with

r\geq3

, then

rr+n>(r+n)r;

therefore it is enough to consider possibilities

x=1

and

x=2

in order to find solutions in natural numbers.

The problem was discussed in a number of publications. In 1960, the equation was among the questions on the William Lowell Putnam Competition,[1] which prompted Alvin Hausner to extend results to algebraic number fields.[2]

Positive real solutions

Main source:

Explicit form

An infinite set of trivial solutions in positive real numbers is given by

x=y.

Nontrivial solutions can be written explicitly using the Lambert W function. The idea is to write the equation as

aeb=c

and try to match

a

and

b

by multiplying and raising both sides by the same value. Then apply the definition of the Lambert W function

a'ea'=c'a'=W(c')

to isolate the desired variable.

\begin{align} yx&=xy=\exp\left(ylnx\right)&\\ yx\exp\left(-ylnx\right)&=1&\left(multiplyby\exp\left(-ylnx\right)\right)\\ y\exp\left(-y

lnx
x

\right)&=1&\left(raiseby1/x\right)\\ -y

lnx\exp\left(-y
x
lnx
x

\right)&=

-lnx
x

&\left(multiplyby

-lnx
x

\right) \end{align}

⇒ -y

lnx
x

=W\left(

-lnx
x

\right)

y=

-x
lnx

W\left(

-lnx
x

\right)=\exp\left(-W\left(

-lnx
x

\right)\right)

W(x)/x=\exp(-W(x))

.

Here we split the solution into the two branches of the Lambert W function and focus on each interval of interest, applying the identities:

\begin{align} W
0\left(-lnx
x

\right)&=-lnx&for&0<x\lee,\\ W-1\left(

-lnx
x

\right)&=-lnx&for&x\gee. \end{align}

0<x\le1

:

-lnx
x

\ge0

\begin{align}y&=

\exp\left(-W
0\left(-lnx
x

\right)\right)\\ &=\exp\left(-(-lnx)\right)\\ &=x\end{align}

1<x<e

:

-1
e

<

-lnx
x

<0

y=

\begin{cases} \exp\left(-W
0\left(-lnx
x

\right)\right)=x\\ \exp\left(-W-1\left(

-lnx
x

\right)\right) \end{cases}

x=e

-lnx
x

=

-1
e

y=

\begin{cases} \exp\left(-W
0\left(-lnx
x

\right)\right)=x\\ \exp\left(-W-1\left(

-lnx
x

\right)\right)=x \end{cases}

x>e

:

-1
e

<

-lnx
x

<0

y=

\begin{cases} \exp\left(-W
0\left(-lnx
x

\right)\right)\\ \exp\left(-W-1\left(

-lnx
x

\right)\right)=x \end{cases}

Hence the non-trivial solutions are:

Parametric form

Nontrivial solutions can be more easily found by assuming

x\ney

and letting

y=vx.

Then

(vx)x=xvx=(xv)x.

Raising both sides to the power

\tfrac{1}{x}

and dividing by

x

, we get

v=xv-1.

Then nontrivial solutions in positive real numbers are expressed as the parametric equation

The full solution thus is

(y=x)\cup\left(v1/(v-1),vv/(v-1)\right)forv>0,v1.

Based on the above solution, the derivative

dy/dx

is

1

for the

(x,y)

pairs on the line

y=x,

and for the other

(x,y)

pairs can be found by

(dy/dv)/(dx/dv),

which straightforward calculus gives as:
dy
dx

=

2\left(v-1-lnv
v-1-vlnv
v

\right)

for

v>0

and

v1.

Setting

v=2

or

v=\tfrac{1}{2}

generates the nontrivial solution in positive integers,

42=24.

Other pairs consisting of algebraic numbers exist, such as

\sqrt3

and

3\sqrt3

, as well as

\sqrt[3]4

and

4\sqrt[3]4

.

The parameterization above leads to a geometric property of this curve. It can be shown that

xy=yx

describes the isocline curve where power functions of the form

xv

have slope

v2

for some positive real choice of

v1

. For example,

x8=y

has a slope of

82

at

(\sqrt[7]{8},\sqrt[7]{8}8),

which is also a point on the curve

xy=yx.

The trivial and non-trivial solutions intersect when

v=1

. The equations above cannot be evaluated directly at

v=1

, but we can take the limit as

v\to1

. This is most conveniently done by substituting

v=1+1/n

and letting

n\toinfty

, so

x=\limv\tov1/(v-1)=\limn\toinfty\left(1+

1n\right)
n

=e.

Thus, the line

y=x

and the curve for

xy-yx=0,y\nex

intersect at .

As

x\toinfty

, the nontrivial solution asymptotes to the line

y=1

. A more complete asymptotic form is

y=1+

lnx
x

+

3
2
(lnx)2
x2

+.

Other real solutions

An infinite set of discrete real solutions with at least one of

x

and

y

negative also exist. These are provided by the above parameterization when the values generated are real. For example,
x=1
\sqrt[3]{-2
},
y=-2
\sqrt[3]{-2
} is a solution (using the real cube root of

-2

). Similarly an infinite set of discrete solutions is given by the trivial solution

y=x

for

x<0

when

xx

is real; for example

x=y=-1

.

Similar graphs

Equation

The equation

\sqrt[x]y=\sqrt[y]x

produces a graph where the line and curve intersect at

1/e

. The curve also terminates at (0, 1) and (1, 0), instead of continuing on to infinity.

The curved section can be written explicitly as

x))
W
0(ln(x
y=e

for0<x<1/e,

W-1(ln(xx))
y=e

for1/e<x<1.

This equation describes the isocline curve where power functions have slope 1, analogous to the geometric property of

xy=yx

described above.

The equation is equivalent to

yy=xx,

as can be seen by raising both sides to the power

xy.

Equivalently, this can also be shown to demonstrate that the equation

\sqrt[y]{y}=\sqrt[x]{x}

is equivalent to

xy=yx

.

Equation

The equation

logx(y)=logy(x)

produces a graph where the curve and line intersect at (1, 1). The curve becomes asymptotic to 0, as opposed to 1; it is, in fact, the positive section of y = 1/x.

External links

Notes and References

  1. Web site: 21st Putnam 1960. Problem B1 . 20 Oct 1999 . bot: unknown . https://web.archive.org/web/20080330183949/http://www.kalva.demon.co.uk/putnam/putn60.html . 2008-03-30 .
  2. Hausner . Alvin . November 1961 . Algebraic Number Fields and the Diophantine Equation mn = nm . . 68 . 9 . 856–861 . 10.1080/00029890.1961.11989781 . 0002-9890.