Is this true?? (1 Viewer)

no_arg

Member
Joined
Mar 25, 2004
Messages
67
Gender
Undisclosed
HSC
N/A
Prove that if z ranges over a fixed circle in the complex plane (not containing 0)
then the locus of 1/z is also a circle
 

darkliight

I ponder, weak and weary
Joined
Feb 13, 2006
Messages
341
Location
Central Coast, NSW
Gender
Male
HSC
N/A
I'm not sure what you're asking ... say your fixed circle has radius r, do you mean ||z|| = r or ||z|| <= r ?

If the former, it's true, if the later, it's false.
 

no_arg

Member
Joined
Mar 25, 2004
Messages
67
Gender
Undisclosed
HSC
N/A
More precisely

suppose that |z-a|=r (z lies on a circle of radius r centred at a) and |a| is not equal to r (circle does not contain 0)

prove that

the locus of 1/z is also a circle.

Its trivial of course when a=0

I'm after a variety of different proofs (no insane determinant arguments please!)
 
Last edited:
Joined
Jul 7, 2002
Messages
722
Gender
Undisclosed
HSC
N/A
The determinant method is an easy way to do it. Here it is:

 

Mill

Member
Joined
Feb 13, 2003
Messages
256
Gender
Male
HSC
2002
If I recall correctly, the question is equally as interesting (and perhaps just as difficult) if you instead create the constraint that the circle must pass through the origin.

From memory, in this case, the locus of 1/z will actually be a straight line.



Disclaimer: I may not recall correctly. :)
 

no_arg

Member
Joined
Mar 25, 2004
Messages
67
Gender
Undisclosed
HSC
N/A
Anyone have a proof within the syllabus?
What is the radius and centre of the new circle?
 
Last edited:

Riviet

.
Joined
Oct 11, 2005
Messages
5,593
Gender
Undisclosed
HSC
N/A
This is not a proof, but just a way to confirm it. Let z=x+iy, then 1/z = x-iy, which is the conjugate of z, ie the reflection of z in the real axis. Therefore all the points representing z that lie on that fixed circle will be reflected in the real axis to form another circle.
 
Joined
Jul 7, 2002
Messages
722
Gender
Undisclosed
HSC
N/A
no_arg said:
What is the radius and centre of the new circle?
Well if the original circle is |z-a|=r, I got (again using determinants) that the new one has centre conj(a)/(|a|<sup>2</sup>-r<sup>2</sup>) and radius r/||a|<sup>2</sup>-r<sup>2</sup>| - but it is not necessary to find these things in order to answer your original question.

It can be generalised to linear fractional transformation f(z)=(az+b)/(cz+d) whereby circles not through the origin are mapped to circles, and if they pass through the origin, they are mapped to lines:
http://mathworld.wolfram.com/LinearFractionalTransformation.html
 
Last edited:
Joined
Jul 7, 2002
Messages
722
Gender
Undisclosed
HSC
N/A
OK. Well here's one, but I prefer my determinants method above. It's faster and more elegant.
 
Last edited:
Joined
Jul 7, 2002
Messages
722
Gender
Undisclosed
HSC
N/A
Lets look at an easier problem.

Suppose 2x<sup>3</sup>-5x<sup>2</sup>+4x+6=0 has roots &alpha;, &beta;, &gamma;. Find an equation with roots 1/&alpha;, 1/&beta;, 1/&gamma;.

Solution.

Replace x with 1/x.

&there4; 2/x<sup>3</sup>-5/x<sup>2</sup>+4/x+6=0.

&there4; 6x<sup>3</sup>+4x<sup>2</sup>-5x+2=0.

x is a dummy variable. That's why we can say "replace x with 1/x". It's the same with locus.

But if you don't like saying "replace x with 1/x", you can instead do it like this:

Let y=1/x.

&there4; 2/y<sup>3</sup>-5/y<sup>2</sup>+4/y+6=0.

&there4; 6y<sup>3</sup>+4y<sup>2</sup>-5y+2=0.

The result is the same, with dummy variable y instead of x.

It's the same in locus problems. In my syllabus solution, instead of saying "replace z with 1/z", you can say "Let w=1/z" and get the same result, i.e, the same circle with the same centre and same radius, but with a dummy variable w instead of z in the equation.

Hope that helps.
 
Last edited:

KFunk

Psychic refugee
Joined
Sep 19, 2004
Messages
3,323
Location
Sydney
Gender
Male
HSC
2005
Hey, I don't know if this suffices as a proof or something which is sufficiently within the syllabus, but a possible approach uses the fact that reix = r(cosx + isinx) = rcis(x) (which comes from Euler's formula: http://en.wikipedia.org/wiki/Euler's_formula ... If you look up the taylor series stuff it's understandable using what is within the syllabus and the rest follows).

So that it makes more sense, first consider the simple case where the circle is centered around the origin:

Let f(x) and F(x) be functions where f(x) = reix = rcis[x] = z (and 0 <= x <= 2pi), so f(x) is the anticlockwise circle from 0 to 2pi of radius r in the complex plane, and F(x) = z-1.

F(x) = F(x) = z-1 = (1/r)e-ix which is a circle of radius (1/r) in the complex plane going clockwise from 0 to -2pi.

...

If the circle can be anywhere then let f(x) = reix + z2 = z1 + z2 = z (z2 is any complex number x2 + iy2).

F(x) = z-1

= 1/(z1 + z2)

= conj.(z1 + z2)/(|z1 + z2|)

= (conj.z1 + conj.z2)/k ... (k since the modulus of the denominator will take a real value)

= (r/k)e-ix + (1/k)conj.z2

which is a clockwise cirlce with a radius r/k... note that when z2 = 0 then k = r2 which is the first result. Anyhow, there is my BoS cameo... sorry for any mistakes, I'm quite out of practice.
 
Last edited:
Joined
Jul 7, 2002
Messages
722
Gender
Undisclosed
HSC
N/A
KFunk said:
1/(z1 + z2)

= conj.(z1 + z2)/(|z1 + z2|)
conj.(z<sub>1</sub> + z<sub>2</sub>)/(|z<sub>1</sub> + z<sub>2</sub>|)

should be

conj.(z<sub>1</sub> + z<sub>2</sub>)/(|z<sub>1</sub> + z<sub>2</sub>|<sup>2</sup>)

Also, I reckon KFunk'd better write something about the circle not going through (0,0) because otherwise, you get a line, not a circle for the locus of 1/z.

dreampianist needent worry. The syllabus states that linear fractional transformations of circles are not required for the course.

My determinants method proved the locus is a circle without having to find the centre and radius and is the most elegant one. When no_arg prodded me to stick to the syllabus I constructed a more complicated argument finding the centre and radius using methods only in the syllabus. But this is an embrassment. The syllabus methods are grotesque compared to the beauty of the determinants method. So the syllabus committee thought it best to exclude linear fractional transformations of circles altogether. But since when did that stop boredofstudies.org members?
 
Last edited:

KFunk

Psychic refugee
Joined
Sep 19, 2004
Messages
3,323
Location
Sydney
Gender
Male
HSC
2005
Unfortunately my real mistake, which I noticed having looked back over it, was in assuming that |reix + z2|2 takes a constant value for all values of x... which is significantly easier to do if you use a z1 substitute like I did. Disregard the second part, the first part for circles centred at (0,0) still holds.
 
Joined
Jul 7, 2002
Messages
722
Gender
Undisclosed
HSC
N/A
I can see why you prefer a polar form instead of my Cartesian form. But you'd have to show more working and show your centre is conj(z<sub>2</sub>)/(|z<sub>2</sub>|<sup>2</sup>-r<sup>2</sup>) and radius r/||z<sub>2</sub>|<sup>2</sup>-r<sup>2</sup>|.
 

Users Who Are Viewing This Thread (Users: 0, Guests: 1)

Top