2
$\begingroup$

Hi all:

I'm wondering if there is a simple formula for this.

Simple Example:

x=cos(pt);
y=cos(qt);
where p,q are integers. 

Question: How many intersection points are there?

0) only need to consider (p,q) are relatively prime.

1) Firstly, I thought it was just basic counting:

Let N(p,q) be the number of crossing.

If p is odd, q is odd: N(p,q) = p(q-1) + (p-1)q = 2pq - p - q. 
If p is even, q is odd: N(p,q) = (p-1)(q-1)/2. 

But later I found some special cases p=2,q=5,where the intersection becomes unusual and the formula falls down.

Thanks

$\endgroup$
5
  • $\begingroup$ FWIW, all the references I have that I checked (e.g. Shikin's "Handbook and Atlas of Curves") give the very unhelpful "the number of nodes is finite and depends on the value of m and n." On the other hand, they all consider the more general case that m and n are rational (for irrational values, the curve has infinitely many nodes and does not close on itself.). $\endgroup$ Sep 9, 2010 at 0:48
  • $\begingroup$ If $p=2, q=5$, your second formula applies, $N(2,5)=(2-1)(5-1)/2=2$, which when you draw out $x=cos(2t)$ and $y=cos(5t)$, shows 2 intersections which is what your formula predicts. Also, you changed variables from $m,n$ above to $p,q$ below; perhaps change that to consistently use the same variables if that is what you meant. $\endgroup$ Sep 9, 2010 at 0:53
  • $\begingroup$ Consider that for the node to self-intersect, $x(t+\tau_i)=x(t)$ and $y(t+\tau_i)=y(t)$. Consider this for $gcd(m,n)=1$. $\endgroup$ Sep 9, 2010 at 0:55
  • 1
    $\begingroup$ @Eminemya, the web site you point to uses a parametrized Lissajous curve defined by y=sin(mt),x=sin(nt) while your question uses a different definition. The site also uses a constant you can add in, but it's easier if you stay consistent with your definitions. Your question, however, poses the question with an even cosine function, y=cos(mt),x=cos(nt) which yields a different number of intersections. Could you clarify and edit your question? Also, you change variables from m,n at first, to p,q later. Could you please fix that also? Thanks. $\endgroup$ Sep 9, 2010 at 1:46
  • $\begingroup$ sorry about my slopiness...Thanks for everyone's considerations:) $\endgroup$
    – Eminemya
    Sep 13, 2010 at 4:34

3 Answers 3

1
$\begingroup$

EDIT: with both functions switched to cosine I get $$ \frac{(p-1)(q-1)}{2} $$ for both odd-odd and for even-odd.

ORIGINAL, both functins sine: For $p$ even and $q$ odd I get $$ 2 p q - p - q $$ which is the same as you have when both are odd.

Examples, I had to count over a few times, (p,q,count) :

(2,1,1); (2,3,7); (2,5,13); (2,7,19);

(4,1,3); (4,3,17); (4,5,31); (4,7,45);

(6,1,5); (6,5,49); (6,7,71); (6,11,115);

Then for both odd I get $$ \frac{(p-1)(q-1)}{2} $$

(1,q,0);

(3,q, q-1);

(5,q, 2 q - 2);

(7,q, 3 q - 3);

However for both odd the figure is traced twice, there is no way around that, look on the website you supplied and slowly vary the maximum of $\Theta.$ So it is also reasonable to claim that both segments are doubled and what appears to be a single intersection counts as four as far as the parameter values are concerned. If you prefer that, the formula for both odd becomes $$ 2 p q - 2 p - 2 q + 2. $$

$\endgroup$
6
  • 1
    $\begingroup$ @Will Jagy, your $(p,q,count)$ triplets are valid for a parametrized Lissajous curve defines as $y=sin(mt), x=cos(nt)$, which is the parametric definition used on the web site referred to in the question. The question itself, however, poses the question with an even $cosine$ function, $y=cos(mt), x=cos(nt)$ which yields a different number of intersections. $\endgroup$ Sep 9, 2010 at 1:40
  • $\begingroup$ obviously, I meant "defined" rather than "defines" in my comment above. $\endgroup$ Sep 9, 2010 at 1:47
  • $\begingroup$ sleepless, there is a saying from the old country that covers this situation: Well, that blows. $\endgroup$
    – Will Jagy
    Sep 9, 2010 at 1:49
  • $\begingroup$ @Will Jagy, that's pretty colourful language in that old country of yours. :) I think the questioner is just unclear on what they're asking. At least your conclusions are valid for the question which you answered. $\endgroup$ Sep 9, 2010 at 1:54
  • $\begingroup$ also, obviously, I meant $y=sin(mt), x=sin(nt)$ in my first comment to Will. These typos must be fully explained by my name here, alas. For I am truly... $\endgroup$ Sep 9, 2010 at 1:59
1
$\begingroup$

Whether you use $sin$ or $cos$ matters, as sine is an odd function because $sin(-\theta)=-sin(\theta)$, whereas cosine is an even function because $cos(-\theta)= cos(\theta)$.

If you draw a parametrized lissajous curve with $x=cos(nt), y=cos(mt)$, with n=2, m=5, you will get two intersections as I remarked in the comments above. If you draw a parametric lissajous curve defined as $x=sin(nt), y=sin(mt)$ with n=2, m=5, you will get the 13 points of intersection that Will Jagy got above.

Your question posits a parametric Lissajous curve with an even $cosine$ function. The website you point to draws out the example curves with an odd $sine$ function. You should clarify exactly what you mean in your question.

Otherwise, the points I made in the comments above should help.

Consider the intersections occuring at the simultaneous constraints for multiple values of $\tau_k$, $k$ varying from $1$ to the number of intersection points:

$cos(nt+\tau_k)=cos(t)$

$cos(mt+\tau_k)=cos(t)$

I believe that if you look at the correct parametric form of the lissajous curve, your equation will be correct.

$\endgroup$
1
$\begingroup$

First let us think of generic curves: We say that a plane curve $C$ is generic (or non-degenerate) if $C$ is an immersed curve in the plane having only transverse dounble points. A Lissajous curve with frequencies $p,q$ and phase $c$ is given by the parametrization $$x(t)=cos (pt+c), y(t)=cos(qt).$$ Except for some finitely many values $c$, the curve is generic. Let $L(p,q)$ denote such a generic Lissajous curve, indeed, its topological type does not depend on the generic value $c$. A good reference is [V. F. Jones et al, Lissajous knots, Jour. Knot Thoery and Its Ramification (1994) 121--140].

By the way, V. I. Arnold introduced three basic invariants of topological types of generic plane curves, denoted by $J^+$, $J^-$ and $St$: [V. I. Arnol'd, Topological Invariants of Plane Curves and Caustics, Univ. Lect. Ser. Vol. 5, Amer. Math. Soc. (1994)]. The number $d$ of double points of a generic curve is always equal to $J^+-J^-$.

For generic $L(p,q)$ with coprime $p$ and $q$, Arnold invariants are computed as follows: $$J^+(L(p,q))=(p-1)(q-1), J^-(L(p,q))=-pq+1, St(L(p,q))=0,$$ (Sunao Kan-mura, 2002). In particular, $d=2pq-p-q$. A degenerate Lissojous curve $L_0$, e.g. in case of $c=0$ as in the question, is doubly covered and has two singular `end' points. By slightly perturbing $c$, we obtain a generic Lissajous curve $L_c$. Observe how double points of $L_c$ are created, then we see the number of the intersection points of $L_0$ is $(p-1)(q-1)/2$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.