# Thread: Lorentz invariance of certain zero angles

1. Originally Posted by Pete
No, $\vec{v_1}$ and $\vec{v_2}$ are not unit vectors, they are of arbitrary length.

The S observer is considering the velocity vectors:
\begin{align}
\vec{v_1} &= (0,v_1) \\
\vec{v_2} &= (0,v_2)
\end{align}
The units vectors of $\vec{v_1}$ and $\vec{v_2}$ are identical, so the S observer concludes that $\vec{v_1}$ and $\vec{v_2}$ are parallel.

Is that correct?
No, it isn't correct. You have a class of (parallel) vectors, $(0,v_i), 1. None of these vectors is a unit vector. The representative of this class is the unit vector$(0,1)$. When judging vector parallelism, you need to stop using the members of the class and start using the class representative. This way, the notion of parallelism is independent of vector norm (length).

2. Originally Posted by Tach
No, it isn't correct.
Which bit?
You have a class of (parallel) vectors, $(0,v_i), 1.
So you agree with the conclusion, that v1 and v2 are parallel.
None of these vectors is a unit vector. The representative of this class is the unit vector$(0,1)$. When judging vector parallelism, you need to stop using the members of the class and start using the class representative.
And you agree with how the conclusion was reached.
So what's the problem?

3. Originally Posted by Pete
Which bit?

So you agree with the conclusion, that v1 and v2 are parallel.

And you agree with how the conclusion was reached.
So what's the problem?
The problem is that you insist on using unnormalized vectors in your reasoning.
Normalized vectors retain their parallelism under Lorentz transformations, unnormalized ones don't (when the lengths of the vectors differ). The geometric notion of parallelism has nothing to do with vector length.

4. Originally Posted by Tach
The problem is that you insist on using unnormalized vectors in your reasoning.
Why is that a problem?
Normalized vectors retain their parallelism under Lorentz transformations
Trivially true, because normalised parallel vectors are identical.
unnormalized ones don't (when the lengths of the vectors differ).
I agree.
The geometric notion of parallelism has nothing to do with vector length.
I agree.

Conclusion: Zero angles between velocity vectors are not lorentz invariant, except for the trivial special case of identical vectors.

5. Originally Posted by Pete

Conclusion: Zero angles between velocity vectors are not lorentz invariant, except for the trivial special case of identical vectors.
Vectors that have the same unit vector aren't identical.

6. They are when they're normalized.
If they're not, then (as you said in your previous post) they don't retain their parallelism under Lorentz transformations.

7. Originally Posted by Pete
They are when they're normalized.
If they're not, then (as you said in your previous post) they don't retain their parallelism under Lorentz transformations.
Well, I see that we are not seeing eye to eye in terms of what vector parallelism means in SR. This is not surprising since there is no definition for what an angle is in SR (otherwise we wouldn't have the dispute on the Lorentz invariance of zero angles). Let me try to explain the same part of my document we have been discussing a different way. The velocity of one of the ion beams models $\vec{v_P(0)}$ and the speed of the ions is , accordingly, $v_P(0)$. The velocity of the second ion beam models $\vec{T_1}$. Since $\vec{T_1}$ is a vector of any length, one can choose any value for its norm $T_1$, so there is nothing stopping the choice $T_1=v_P(0)=U$ in the writeup.

8. Originally Posted by Tach
Well, I see that we are not seeing eye to eye in terms of what vector parallelism means in SR.
Only because you are using your incorrect conclusion that parallelism is lorentz invariant as an assumption, which means you can't develop a self-consistent argument.

If you held to the obvious and simple meaning of parallelism that you were happy with before you found that it contradicted your side of the debate, there would be no problem.

This is not surprising since there is no definition for what an angle is in SR
Sure there is. You've used it yourself in your document. If two vectors form the same angle with the x-axis, they are parallel.
Or, use the dot product if you like.

The only problems arise when questions arise over exactly what vectors are being compared, but this is not an issue for velocity vectors because there is no confusion about how they transform between frames.

Let me try to explain the same part of my document we have been discussing a different way. The velocity of one of the ion beams models $\vec{v_P(0)}$ and the speed of the ions is , accordingly, $v_P(0)$. The velocity of the second ion beam models $\vec{T_1}$. Since $\vec{T_1}$ is a vector of any length, one can choose any value for its norm $T_1$, so there is nothing stopping the choice $T_1=v_P(0)=U$ in the writeup.
As we established earlier, the velocity of the $v_p(0)$ modelling ions must be equal to $v_p(0)$ (ie u=0).
So, the ions must have the same speed as the guns, making the guns redundant. The ions are actually a simple light source of known frequency attached to P, just as I suggested back in post 49.

It is true that you can choose the S' angle of the ion velocity within certain limits by setting u, the ion velocity relative to the gun.

Now you just have to find what value of u (if any) makes the T1 ion velocity tangent to the wheel in S'.

9. Originally Posted by Pete
Only because you are using your incorrect conclusion that parallelism is lorentz invariant as an assumption, which means you can't develop a self-consistent argument.
You have agreed earlier with me that the notion of parallelism should not be dependent on vector norm (length). You have agreed that the parallelism of the unit vectors is frame-invariant. Yet, you insist in bringing back vector length in the argument....

If you held to the obvious and simple meaning of parallelism that you were happy with before you found that it contradicted your side of the debate, there would be no problem.
...applied to unit vectors, such that it is independent of vector length, as you agreed earlier.

The only problems arise when questions arise over exactly what vectors are being compared, but this is not an issue for velocity vectors because there is no confusion about how they transform between frames.
This is not the issue. The issue is what parallelism should be defined as.
We both agree that it should be defined as the identity of unit vectors.
You apply the above definition as "transform the vectors, normalize them and calculate their relative angle".
I apply the above definition as "transform the unit vectors and calculate their relative angle".

As we established earlier, the velocity of the $v_p(0)$ modelling ions must be equal to $v_p(0)$ (ie u=0).
So, the ions must have the same speed as the guns, making the guns redundant. The ions are actually a simple light source of known frequency attached to P, just as I suggested back in post 49.
Yes, so what? Adding an extra degree of liberty (u) and finding out that it has to have a specific value (i.e. u=0) does not affect the problem in a negative way.

It is true that you can choose the S' angle of the ion velocity within certain limits by setting u, the ion velocity relative to the gun.

Now you just have to find what value of u (if any) makes the T1 ion velocity tangent to the wheel in S'.
Err, no, you are missing the point explained in the previous post: I can chose the velocity of the ions such that the vector modeling $\vec{T_1}$ equals $\vec{v_P(0)}$ by choosing to make their lengths equal. This way, the two velocity vectors transform into identical vectors in S'.

10. Originally Posted by Tach
This is just your prejudice on the subject. You have agreed earlier with me that the notion of parallelism should not be dependent on vector norm (length).
Correct.
In any given reference frame, the length of two vectors does not affect whether they are parallel - ie the notion of parallelism does not depend on vector length.

However, that has no bearing on whether parallelism is frame-invariant.

You have agreed that the parallelism of the unit vectors is frame-invariant.
Correct.
Yet, you insist in bringing back vector length in the argument.
Yes, because you have agreed that the parallelism of vectors that have different lengths is not lorentz invariant.

...applied to unit vectors, such that it is independent of vector length, as you agreed earlier.
No, you correctly applied the obvious and simple meaning of parallelism to arbitrary vectors. Parallelism doesn't depend on length, as we agreed, so there is simply no need to normalize.

This is not the issue. The issue is what parallelism should be defined as.
W both agree that it should be defined as the identity of unit vectors.
No, I said that it could be defined as the identity of unit vectors, not should.
You apply the above definition as "transform the vectors, normalize them and calculate their relative angle".
No, I apply the definition as "normalize the vectors $\vec{v_1}'$ and $\vec{v_2}'$, and calculate their relative angle".
I apply the above definition as "transform the unit vectors and calculate their relative angle".
Which tells you nothing about $\vec{v_1}'$ and $\vec{v_2}'$.

Yes, so what? Adding an extra degree of liberty (u) and finding out that it has to have a specific value (i.e. u=0) does not affect the problem in a negative way.
So, you wasted a lot of time faffing about with first lasers, then ion beams, instead of considering my suggestions. It would be appropriate for you to apologize.

Err, no, you are missing the point explained in the previous post: I can chose the velocity of the ions such that the vector modeling $\vec{T_1}$ equals $\vec{v_P(0)}$ by choosing to make their lengths equal. This way, the two velocity vectors transform into identical vectors in S'.
There's only one tangent orientation. You need to model it correctly.

11. Originally Posted by Pete

Yes, because you have agreed that the parallelism of vectors that have different lengths is not lorentz invariant.
First off, I never agreed to that.
Second off, I have repeatedly pointed out that , in my opinion, the notion of parallelism has to be defined independent of vector length. So, you are not only contradicting yourself, you are also putting words in my mouth.

No, I apply the definition as "normalize the vectors $\vec{v_1}'$ and $\vec{v_2}'$, and calculate their relative angle".
...which is precisely what I said you were doing. By doing so, you create the paradox that the transformed vectors are no longer parallel , since the unit vectors, as computed by your approach are no longer parallel because the result of your approach DEPENDS on the original length of the vectors.

So, you wasted a lot of time faffing about with first lasers, then ion beams, instead of considering my suggestions. It would be appropriate for you to apologize.
You are not only wrong, you are also being rude.

There's only one tangent orientation. You need to model it correctly.
...which is precisely what I have done.

12. Originally Posted by Tach
Firs off, I never agreed to that.
Yes you did, just a couple of posts ago:
Originally Posted by Tach
Normalized vectors retain their parallelism under Lorentz transformations, unnormalized ones don't (when the lengths of the vectors differ).
Second off, I have repeatedly pointed out that , in my opinion, the notion of parallelism has to be defined independent of vector length.
And I have repeatedly agreed. The problem is that you are confusing parallelism with transforms.
Parallelism is independent of length.
Transforming between frames is not.

...which is precisely what I said you were doing.
Wrong.
When considering vectors in frame S', I do not start from S.
How do you decide if $\vec{v_1}'$ and $\vec{v_2}'$ are parallel?
Not by looking at $\vec{v_1}$ and $\vec{v_2}$, obviously.

By doing so, you create the paradox that the transformed vectors are no longer parallel
Finding a contradicting to an assumption means the assumption is wrong. It's not a paradox unless you can't let it go.

the result of your approach DEPENDS on the original length of the vectors.
You're treating frame S as privileged.
There's nothing "original" about frame S.

You are not only wrong, you are being rude.
No Tach, you are wrong.
I politely pointed out very early in the thread that your lasers and ion beams were a waste of time, but you ignored me. Now you've discovered I was right, but you don't acknowledge it. I think you should apologize.

...which is precisely what I have done.
You have a vector with adjustable direction, and you think that no matter how you adjust it will always be tangent to the wheel?
Prove it.

13. Originally Posted by Pete
Parallelism is independent of length.
Not the way you are trying to apply it, that much is clear.

Transforming between frames is not.
This is only because you are miss-applying the notion of parallelism by taking through the transformation two vectors of different lengths. Then you say "see, they are no longer parallel".

You have a vector with adjustable direction, and you think that no matter how you adjust it will always be tangent to the wheel?
Prove it.
You are again putting words in my mouth, we aren't talking about vectors that "no matter how you adjust...", we are talking about vectors tangent to the wheel (at least in frame S). The fact that you see them no longer tangent in any other frame is your problem. See part 1 of my writeup. The part that you failed to understand earlier. The part that you keep insisting on making $dt=0$ without any justification.

14. The debate has stalled.

15. Originally Posted by Pete

Transform to S'...

\begin{align}
t &= \gamma(t' + vx'/c^2) \\
x' &= \gamma(x - vt) \\
&= \gamma r\cos(\omega\gamma t' + \omega\gamma vx'/c^2 + \theta) - v\gamma^2t' - v^2\gamma^2x'/c^2 \\

\end{align}

Differentiate...

\begin{align}

\frac{\partial x'}{\partial \theta} &= -r\gamma(\frac{\gamma\omega v}{c^2}\frac{\partial x'}{\partial \theta} + 1)\sin(\omega \gamma t' + \omega \gamma vx'/c^2 + \theta) - \frac{v^2\gamma^2}{c^2}\frac{\partial x'}{\partial \theta} \\

\end{align}
I think that you may have learned by now how to calculate the partial derivatives, if not for multivariate functions, at least for univariate ones. Anyways, the error in the above stems from missing terms:

Since :
$t'=\gamma(t-vx/c^2)$ implies that you are missing the term:

$\frac{\partial t'}{\partial \theta}=-\gamma v/c^2 \frac {\partial x}{\partial \theta}=-\gamma vr/c^2 sin (\omega t +\theta)$

In other words, you are missing:

$\gamma^3 v^2r/c^2 sin (\omega t +\theta)$ from your failed attempt at calculating $\frac{\partial x'}{\partial \theta}$

BTW, this is not the only part you got wrong, the bit with $+1$ in $(\frac{\gamma\omega v}{c^2}\frac{\partial x'}{\partial \theta} + 1)$ is also wrong. Same mistake, you are missing $\frac{\partial t'}{\partial \theta}$

16. I'm satisfied with the correctness of my answer, as I'm sure you are with yours.

As the debate has stalled, no more posts may be made to this thread until we have prepared our summaries, in accordance with the agreed debate rules.

17. Originally Posted by Pete
The debate has stalled.
You don't decide unilaterally what happens next and most definitely you don't get to order me around. I will give you one more formal proof, then we will see where we go, by mutual agreement as the rules state. In the meanwhile, I pointed out where you went wrong in your calculations, perhaps you can take some time to look at it while I transcribe my third (and final proof). This should give you some time to correct your errors in the calculations of the partial derivatives.

18. My summary is complete. When yours is ready, we'll post them up and the thread can be closed.

19. The declared intent of the thread was to debate and learn. Here is another explanation. Before you attempt to shut down the conversation, may I suggest that you read it.

20. Sorry Tach, I'm done wasting time with you.