From: mpalenik on
On Feb 25, 2:57 am, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
>
> The statement A.B = INVARIANT means  A.B = A'.B' which necessarily
> means B is a vector as it can be transformed to B'.
>

I just want to emphasize this one more time, the way the problem in
Jackson is phrased, the existence of B' does not on its own mean that
B is a 4-vector. B' is constructed by frame dependent measurements,
in this case (w, kx, ky, kz). In this case B *is* a 4-vector but it
could easily have been something like (E, x, y, m) which does not
transform as a 4-vector (i.e. (E', x', y', m) is not (E, x, y, m)
under a coordinate transformation). We are using the invariance of
A*B to PROVE that B transforms as a 4-vector.

I agree that this is completely valid but the OP's question was "is
the only possible way that A*B can be invariant if B is a 4-vector."
The answer is yes--but although A*B DOES give an invariant when B is a
4-vector, it is not clear by this definition that the ONLY way for A*B
to be invariant is for B to be a 4-vector.
From: mpalenik on
On Feb 25, 8:28 am, mpalenik <markpale...(a)gmail.com> wrote:
> On Feb 25, 2:57 am, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
>
>
>
> > On Feb 25, 3:32 pm, mpalenik <markpale...(a)gmail.com> wrote:
>
> > > BTW, I only skimmed your post--but thanks for the extensive, yet
> > > totally unnecessarry, description of Jacobian matrices.  My guess is
> > > that Jackson E&M is meant to be used in the context of flat, Minkowski
> > > spacetime, and the Jacobian matrices that you deal with are the
> > > standard Lorentz boost matrices.
>
> > I didn't reference any specific geometry or coordinate transformations
> > at all.
>
> > > And once again, to rehash, you don't need to repeatedly measure B' in
> > > every possible frame, as the problem *GIVES* you the information that
> > > A*B is invariant and it *GIVES* you the information that A is a 4-
> > > vector.
>
> > [1] If your question is:
> > "If A . B = invariant and A is a vector, prove B a vector"
>
> > Proof: The dot product is DEFINED for vectors so B is necessarily a
> > vector.
>
> > [2] If your question is:
>
> > "If A_x B_x + A_y B_y + ... = invariant and A is a vector, prove the
> > set of B forms a vector"
>
> Yes, this is the OP's question, not my question.  I wasn't asking a
> question.
>
>
>
> > Proof:
> >  - Overview: I define "invariant" as a "rank-0 tensor" so A and B are
> > necessarily invariant vectors (i.e. either contravariant or covariant
> > rank-1 tensors).
>
> Yes, I don't disagree with this.  My point is, it's not obvious from
> this statement that the only way to get an invariant is by taking the
> dot product of two vectors.

And just so we're clear, before you tell me that's not the ONLY way to
get a scalar, I'm not discounting contracting a tensor, i.e. T^mu_mu
as a method of obtaining an invariant, or other such tensor
operations.

The question is whether or not there are frame dependent quantities
that *don't* transform as a 4-vector but will give an invariant when
contracted with a 4-vector. And once again, I AGREE with you that the
answer is no, there are not. And you can prove this, but your first
answer didn't really address this.
From: Schoenfeld on
On Feb 25, 11:39 pm, mpalenik <markpale...(a)gmail.com> wrote:
> On Feb 25, 2:57 am, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
>
>
>
> > The statement A.B = INVARIANT means  A.B = A'.B' which necessarily
> > means B is a vector as it can be transformed to B'.
>
> I just want to emphasize this one more time, the way the problem in
> Jackson is phrased, the existence of B' does not on its own mean that
> B is a 4-vector.  B' is constructed by frame dependent measurements,
> in this case (w, kx, ky, kz).  In this case B *is* a 4-vector but it
> could easily have been something like (E, x, y, m) which does not
> transform as a 4-vector (i.e. (E', x', y', m) is not (E, x, y, m)
> under a coordinate transformation).  We are using the invariance of
> A*B to PROVE that B transforms as a 4-vector.
>
> I agree that this is completely valid but the OP's question was "is
> the only possible way that A*B can be invariant if B is a 4-vector."
> The answer is yes--but although A*B DOES give an invariant when B is a
> 4-vector, it is not clear by this definition that the ONLY way for A*B
> to be invariant is for B to be a 4-vector.

Okay. So reading all that, I look at question as:

QUESTION
"If A.B = INVARIANT SCALAR and A is an invariant vector, prove B MUST
be an invariant vector". By invariant vector, we mean covariant or
contravariant which is acceptable usage of word invariant.

PROOF

We know that A.B = A'.B', so in tensor form

[1] A'^r B'_r = A^r B_r

We chose A as contravariant, so now we must show that given [1], B is
covariant for all B.

So we know that A' is contravariant a priori

[2] A'^r = A^j (@x'^r/@x^j)

Substitute [2] into LHS of [1]

[3] A'^r B'_r = A^j (@x'^r/@x^j) B'_r

Substituting covariant B'_r = B_j (@x^j/@x'r) in RHS of [3]

[4] A'^r B'_r = A^j (@x'^r/@x^j) B_j (@x^j/@x'r)

Simplifying

[5] A'^r B'_r = A^j B_j KroneckaDelta^r_r

Thus recover equation [1]

[6] A'^r B'_r = A^j B_j


[7] So in step 4 we assumed B was covariant we recovered equation [1]
as the jacobian elements cancelled out to a contraction of the
KroneckaDelta tensor.

The question is now was the ONLY POSSIBLE way to recover [1] the
subsitution made in [4]? And yes, it is based on the uniqueness of the
inverse.

For example

A*B = A*B*10*C implies that C = 1/10 and only 1/10

So observe that

A^j B_j = A^j B_j (@x'^r/@x^j) C^j_r

Uniquely implies 'C' cancel out the previous term such that

A'^r B'_r = A^j B_j (@x'^r/@x^j) (@x^j/@x'r) = A^j B_j

But rather than cancelling it, just rewrite it as

A'^r B'_r = A^j (@x'^r/@x^j) B_j (@x^j/@x'r)

and you can clearly see that given [1], B' is UNIQUELY

B'_r = B_j (@x^j/@x'r)

which is covariant.

QED

If you chose A_r as covariant then the above would show B is uniquely
contravariant.



From: mpalenik on
On Feb 25, 8:26 pm, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
> On Feb 25, 11:39 pm, mpalenik <markpale...(a)gmail.com> wrote:
>
>
>
>
>
> > On Feb 25, 2:57 am, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
>
> > > The statement A.B = INVARIANT means  A.B = A'.B' which necessarily
> > > means B is a vector as it can be transformed to B'.
>
> > I just want to emphasize this one more time, the way the problem in
> > Jackson is phrased, the existence of B' does not on its own mean that
> > B is a 4-vector.  B' is constructed by frame dependent measurements,
> > in this case (w, kx, ky, kz).  In this case B *is* a 4-vector but it
> > could easily have been something like (E, x, y, m) which does not
> > transform as a 4-vector (i.e. (E', x', y', m) is not (E, x, y, m)
> > under a coordinate transformation).  We are using the invariance of
> > A*B to PROVE that B transforms as a 4-vector.
>
> > I agree that this is completely valid but the OP's question was "is
> > the only possible way that A*B can be invariant if B is a 4-vector."
> > The answer is yes--but although A*B DOES give an invariant when B is a
> > 4-vector, it is not clear by this definition that the ONLY way for A*B
> > to be invariant is for B to be a 4-vector.
>
> Okay. So reading all that, I look at question as:
>
> QUESTION
> "If A.B = INVARIANT SCALAR and A is an invariant vector, prove B MUST
> be an invariant vector". By invariant vector, we mean covariant or
> contravariant which is acceptable usage of word invariant.
>
> PROOF
>
> We know that A.B = A'.B', so in tensor form
>
> [1] A'^r B'_r = A^r B_r
>
> We chose A as contravariant, so now we must show that given [1], B is
> covariant for all B.
>
> So we know that A' is contravariant a priori
>
> [2] A'^r = A^j (@x'^r/@x^j)
>
> Substitute [2] into LHS of [1]
>
> [3] A'^r B'_r = A^j (@x'^r/@x^j) B'_r
>
> Substituting covariant B'_r = B_j (@x^j/@x'r) in RHS of [3]
>
> [4] A'^r B'_r = A^j (@x'^r/@x^j) B_j (@x^j/@x'r)
>
> Simplifying
>
> [5] A'^r B'_r = A^j B_j KroneckaDelta^r_r
>
> Thus recover equation [1]
>
> [6] A'^r B'_r = A^j B_j
>
> [7] So in step 4 we assumed B was covariant we recovered equation [1]
> as the jacobian elements cancelled out to a contraction of the
> KroneckaDelta tensor.
>
> The question is now was the ONLY POSSIBLE way to recover [1] the
> subsitution made in [4]? And yes, it is based on the uniqueness of the
> inverse.
>
> For example
>
>    A*B = A*B*10*C   implies that C = 1/10 and only 1/10
>
> So observe that
>
>  A^j B_j = A^j B_j (@x'^r/@x^j) C^j_r
>
> Uniquely implies 'C' cancel out the previous term such that
>
>  A'^r B'_r = A^j B_j (@x'^r/@x^j) (@x^j/@x'r) = A^j B_j
>
> But rather than cancelling it, just rewrite it as
>
>  A'^r B'_r = A^j (@x'^r/@x^j)  B_j (@x^j/@x'r)
>
> and you can clearly see that given [1], B' is UNIQUELY
>
>  B'_r =  B_j (@x^j/@x'r)
>
> which is covariant.
>
> QED
>
> If you chose A_r as covariant then the above would show B is uniquely
> contravariant.- Hide quoted text -
>
> - Show quoted text -

And hence we have a more general version of the exact same thing I
wrote several posts ago, where I used the Lorentz matrices (written as
L) instead of general Jacobian matrices, since the problem dealt
specifically with flat, Minkowski spacetime.
From: mpalenik on
On Feb 25, 8:26 pm, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
> On Feb 25, 11:39 pm, mpalenik <markpale...(a)gmail.com> wrote:
>
>
>
>
>
> > On Feb 25, 2:57 am, Schoenfeld <schoenfeld.fore...(a)gmail.com> wrote:
>
> > > The statement A.B = INVARIANT means  A.B = A'.B' which necessarily
> > > means B is a vector as it can be transformed to B'.
>
> > I just want to emphasize this one more time, the way the problem in
> > Jackson is phrased, the existence of B' does not on its own mean that
> > B is a 4-vector.  B' is constructed by frame dependent measurements,
> > in this case (w, kx, ky, kz).  In this case B *is* a 4-vector but it
> > could easily have been something like (E, x, y, m) which does not
> > transform as a 4-vector (i.e. (E', x', y', m) is not (E, x, y, m)
> > under a coordinate transformation).  We are using the invariance of
> > A*B to PROVE that B transforms as a 4-vector.
>
> > I agree that this is completely valid but the OP's question was "is
> > the only possible way that A*B can be invariant if B is a 4-vector."
> > The answer is yes--but although A*B DOES give an invariant when B is a
> > 4-vector, it is not clear by this definition that the ONLY way for A*B
> > to be invariant is for B to be a 4-vector.
>
> Okay. So reading all that, I look at question as:
>
> QUESTION
> "If A.B = INVARIANT SCALAR and A is an invariant vector, prove B MUST
> be an invariant vector". By invariant vector, we mean covariant or
> contravariant which is acceptable usage of word invariant.
>
> PROOF
>
> We know that A.B = A'.B', so in tensor form
>
> [1] A'^r B'_r = A^r B_r
>
> We chose A as contravariant, so now we must show that given [1], B is
> covariant for all B.
>
> So we know that A' is contravariant a priori
>
> [2] A'^r = A^j (@x'^r/@x^j)
>
> Substitute [2] into LHS of [1]
>
> [3] A'^r B'_r = A^j (@x'^r/@x^j) B'_r
>
> Substituting covariant B'_r = B_j (@x^j/@x'r) in RHS of [3]
>
> [4] A'^r B'_r = A^j (@x'^r/@x^j) B_j (@x^j/@x'r)
>
> Simplifying
>
> [5] A'^r B'_r = A^j B_j KroneckaDelta^r_r
>
> Thus recover equation [1]
>
> [6] A'^r B'_r = A^j B_j
>
> [7] So in step 4 we assumed B was covariant we recovered equation [1]
> as the jacobian elements cancelled out to a contraction of the
> KroneckaDelta tensor.
>
> The question is now was the ONLY POSSIBLE way to recover [1] the
> subsitution made in [4]? And yes, it is based on the uniqueness of the
> inverse.
>
> For example
>
>    A*B = A*B*10*C   implies that C = 1/10 and only 1/10
>
> So observe that
>
>  A^j B_j = A^j B_j (@x'^r/@x^j) C^j_r
>
> Uniquely implies 'C' cancel out the previous term such that
>
>  A'^r B'_r = A^j B_j (@x'^r/@x^j) (@x^j/@x'r) = A^j B_j
>
> But rather than cancelling it, just rewrite it as
>
>  A'^r B'_r = A^j (@x'^r/@x^j)  B_j (@x^j/@x'r)
>
> and you can clearly see that given [1], B' is UNIQUELY
>
>  B'_r =  B_j (@x^j/@x'r)
>
> which is covariant.
>
> QED
>
> If you chose A_r as covariant then the above would show B is uniquely
> contravariant.- Hide quoted text -
>
> - Show quoted text -

And hence we have a more general version of the exact same thing I
posted a little while back in the thread, where I used the Lorentz
matrices (written as L) instead of general Jacobian matrices.

I hope this doesn't appear twice. My last reply didn't show up for
some reason.