2

Can someone please show me how to prove that the intersection of the plane

$$x+y+z=0$$ and the sphere

$$x^2+y^2+z^2=1$$

can be expressed as

$$x(t)=\frac{\cos t-\sqrt3 \cdot\sin t}{\sqrt6}$$

$$y(t)=\frac{\cos t+\sqrt3 \cdot\sin t}{\sqrt6}$$

$$z(t)=\frac{-2\cos t}{\sqrt6}$$

Ps: Also, why am I not getting the correct notation? I am using a macbook pro (safari) if that is a concern?

vadim123
  • 82,796
amanda
  • 425
  • 1
    Notation should not depend on the machine you're using, since it's MathJax. You did have some / where you should have . – vadim123 May 16 '13 at 03:57
  • 2
    I find it strange that the intersection of two objects in $\mathbb R^3$ would be a path in $\mathbb R$. In other words, are we looking for $y$ and $z$ of $t$ as well? – Brady Trainor May 16 '13 at 03:59
  • @Milind: I'm afraid you have edited it wrong, check out the original version. – Inceptio May 16 '13 at 04:01
  • 1
    I know how to find the intersection by simply equating these equations which gives me $$x^2+y^2+xy=\frac{1}{2}$$ but i don't know what the next step is?... @BradyTrainor – amanda May 16 '13 at 04:02
  • Thankyou @Inceptio , the correct equation is now up :) – amanda May 16 '13 at 04:03
  • @BradyTrainor Sorry, the full question is now up, i didnt think that y and z mattered in the question but i realised it was giving the wrong question, apologies guys. – amanda May 16 '13 at 04:07
  • @amanda, I ran into that $xy$ term as well. Perhaps if I had studied conics a bit more... I smelled an ellipse, but it looks like vadim123 has spotted the nice substitution below. – Brady Trainor May 16 '13 at 04:10
  • @amanda, do you know some linear algebra? If so, rather than messing with these ellipses, it's easier to rotate the coordinate system so that in the new $x'y'z'$ coordinates, the plane $x+y+z=0$ becomes $z'=0$. – Ted Shifrin May 16 '13 at 12:51

2 Answers2

4

You can solve for $z=-x-y$ and plug in $x^2+y^2+(-x-y)^2=1$, which rearranges to $2x^2+2xy+2y^2=1$. Hence for any fixed $z$ you have a rotated ellipse.

Let $u=(x+y/2)$. Then $u^2=x^2+xy+y^2/4$, so the equation becomes $2u^2+\frac{3}{2}y^2=1$. In standard form this is: $$\frac{u^2}{(\frac{1}{\sqrt{2}})^2}+\frac{y^2}{(\frac{\sqrt{2}}{\sqrt{3}})^2}=1$$

We can solve this parametrically as $u=\frac{1}{\sqrt{2}}\sin t$, $$y=\sqrt{\frac{2}{3}}\cos t$$ We can now find $$x=u-y/2=\frac{1}{\sqrt{2}}\sin t-\frac{1}{\sqrt{6}}\cos t$$ and $$z=-x-y=\frac{-1}{\sqrt{2}}\sin t +(-\frac{1}{\sqrt{6}}-\sqrt{\frac{2}{3}})\cos t$$

Alas, you posted your revisions after I'd made my choices so my solution will not agree with yours. There are three choices at the first step (solve for $z,y,x$), and two choices at the second step (replace $x$ or $y$). One of those six choices might give the result you have. :-)

Followup: A cross term ($xy$) is a rotation and can always be eliminated by a change of variables, pointing in the directions of the major and minor axes. There is a systematic way to do this; see here or here, or you can do it by the seat of your pants, like I did. I didn't directly rotate, instead I sheared, which made the computations a bit simpler.

Edit: fix sign error, thanks @Mhenni.

Edit 2: add general method

vadim123
  • 82,796
  • Wow, Thankyou so much @vadim123 . i understand everything but I am little confused as to how you got $$u=\frac{(x-y)}{2}$$ ? Also, I think that should be a plus? But what I dont understand is why you let $$u=\frac{(x-y)}{2}$$ ? – amanda May 16 '13 at 04:20
  • It should be $ 2x^2+2xy+2y^2=1 $. The sign of $xy$. – Mhenni Benghorbal May 16 '13 at 04:20
  • (In the first sentence) – Brady Trainor May 16 '13 at 04:21
  • @MhenniBenghorbal I cannot tag him so I am unsure if vadim123 can see my comment but I am unsure as to why Vadim let $$u=x-\frac{y}{2}$$ ? – amanda May 16 '13 at 04:26
  • That substitution is a change of variables that undoes the rotation of the ellipse. There are many others, such as $u=y-\frac{x}{2}$, and also the symmetric one @Brady offers. – vadim123 May 16 '13 at 04:28
  • @vadim123 Ok, I am sort of understanding it. Maybe answering this question will help me understand it, so why didn't we let $$u=x^2 + y^2$$ ? – amanda May 16 '13 at 04:40
  • Yeah my substitution isn't as slick as I was hoping. Perhaps more insight can be gathered if one take the matrix point of view on the transformation of conics. I couldn't find a nice treatment on the internet. – Brady Trainor May 16 '13 at 04:40
  • @amanda, the goal in the substitution is to still have an ellipse, but to no longer have a cross term. If we let $u=x^2+y^2$, the equation becomes $2u+2xy=1$. Not only is this not an ellipse, but you haven't eliminated anything (there are now 3 variables). – vadim123 May 16 '13 at 04:43
  • @BradyTrainor, see my links (especially 2nd). – vadim123 May 16 '13 at 04:44
  • Not to sound old or anything, but rotation of conic sections to kill the cross term was a standard part of the course when I took calculus 25 years ago. Every once in a blue moon it comes in handy. :-) – vadim123 May 16 '13 at 04:46
  • Thankyou so much Vadim! :) It took me a while to respond because I was trying to understand it and now I do. Thankyou so much for your patience and effort :) @vadim123 – amanda May 16 '13 at 05:21
1

(Originally I had $u,v=x+y,x-y$, but after going through half the problem, I considered that $2u,2v=x+y,x-y$ would be a little more convenient.)

amanda, after your substitution, we consider $x^2+xy+y^2={1\over2}$.

Here are is a nice choice for the next substitution,

\begin{align} 2u&=x+y,\\ 2v&=x-y. \end{align}

We next consider

\begin{align} 4u^2&=x^2+y^2+2xy,\\ 4v^2&=x^2+y^2-2xy. \end{align}

To get $x^2+y^2+xy$, we need $2u^2+2v^2+(u^2-v^2)$. Thus our problem is transformed to

$$6u^2+2v^2=1.$$

Then we have parametric solutions

$$u=\sqrt{1\over6}\cos t,\ v=\sqrt{1\over2}\sin t.$$

From there, we find for instance

$$x=u+v={\cos t+\sqrt3\sin t\over\sqrt6},$$

and the rest should be straightforward.

To get the sign to match what is shown in your equations, I can think of several ways to go.

  • You could simply note the symmetry in $x$ and $y$ and switch them (which suggests that if we had made the substitution $2v=y-x$ instead of $x-y$, that would have worked also).
  • From $6u^2+2v^2=1$, we could have chosen $v=-\sqrt{1\over2}\sin t$.
  • If we make the substitution $t=-s$, then we have $x={\cos (-s)+\sqrt3\sin (-s)\over\sqrt6}={\cos s-\sqrt3\sin s\over\sqrt6}$. Now, call it $t$ again.
  • ohhhh I absolutely understand this now! Thankyou so much Brady, I couldn't of understood it without the help of you and Vadim. It feels so good understand it, thankyou so much :) – amanda May 16 '13 at 05:20
  • Glad to help. I think the $xy$ term motivated a little "creativity". – Brady Trainor May 16 '13 at 05:28
  • Sorry to bother you again Brady but how did you get $2u^2+2v^2-(u^2-v^2)$. simplifying to

    $$6u^2+2v^2=1.$$ ?

    – amanda May 16 '13 at 06:21
  • also, shouldn't it be $2u^2+2v^2+(u^2-v^2)$ – amanda May 16 '13 at 06:58
  • @amanda, yes, you are right about the sign error. That gives coefficients of 3 and 1, which after clearing the fractions in the original equation, becomes the coefficients of 6 and 2. – Brady Trainor May 16 '13 at 07:06
  • Sorry i am little lost now. What do you mean by "clearing the fractions in the original equation"? How does that bring about 6 and 2? – amanda May 16 '13 at 07:28
  • oh right, after substituting it in lol silly question by me. Thankyou :)) – amanda May 16 '13 at 07:35
  • Lastly, what do you mean by substitute s=-t? Is it possible if you could edit the example if it's not too much to ask? After that step, I think i have mastered any difficulty given by this question :) @BradyTrainor – amanda May 16 '13 at 08:59
  • @amanda, I have added some ideas to the end. – Brady Trainor May 16 '13 at 15:25
  • Thankyou Brady, I really needed that part :) – amanda May 16 '13 at 17:22