Perhaps not a very bright question (and more of a personal after-thought), but how is one able to interpret or figure out the boost's magnitude (i.e. speed), direction, and rotation (about the z-axis) after a matrix multiplication?
In this "follow-up" thread, I provided below an example given by a helpful expert showing how two perpendicular boosts equal to a rotation after a boost.
Here, in the answer (shown in the image below), I noticed that the author specifically separated his solved answer with two multiplying matrices. While I'm fairly certain that this helps with finding the magnitude, direction, and rotation of the new boost, I am unsure how was the author able to accomplish this and what mathematical tools helped him determine that these two matrices setup will ultimately contribute to the solution process.
Thus, to boil it down simply, here are my two questions:
How can I tell what the new boost's magnitude, direction, and rotation (about the z-axis) are when I have my matrix solution?
How can I appropriately separate a - usually - one matrix solution into two to properly find for the information (i.e. magnitude, direction, and rotation) regarding the new boost?
Thank you for reading through this question and I will sincerely appreciate any amount of assistance to help me better understand how to derive the information for the new boost.
Source: Special Relativity - Perpendicular Boosts Equaling to a Rotation after a Boost
Answer
In my previous answer, I mentioned that the form of a general boost to speed βc in the direction of the unit vector ˆn is
B(β,ˆn)=(γ−γβnx−γβny−γβnz−γβnx1+(γ−1)n2x(γ−1)nxny(γ−1)nxnz−γβny(γ−1)nynx1+(γ−1)n2y(γ−1)nynz−γβnz(γ−1)nznx(γ−1)nzny1+(γ−1)n2z)
where
γ=1√1−β2.
A general rotation by angle θ around an axis determined by the unit vector ˆu (with rotational direction given by the right-hand-rule) is
R(θ,ˆu)= (10000cosθ+u2x(1−cosθ)uxuy(1−cosθ)−uzsinθuxuz(1−cosθ)+uysinθ0uyux(1−cosθ)+uzsinθcosθ+u2y(1−cosθ)uyuz(1−cosθ)−uxsinθ0uzux(1−cosθ)−uysinθuzuy(1−cosθ)+uxsinθcosθ+u2z(1−cosθ))
By multiplying these matrices, you can see that a general boost followed by a general rotation has the form
R(θ,ˆu)B(β,ˆn)=(γ−γβnx−γβny−γβnz−−−−−−−−−−−−)
and a general rotation followed by a general boost has the form
B(β,ˆn)R(θ,ˆu)=(γ−−−−γβnx−−−−γβny−−−−γβnz−−−)
Here the matrix elements indicated by a dash are complicated expressions involving both the boost parameters and the rotation parameters. But the first row (in the RB case) or the first column (in the BR case) is simple: It depends only on the boost parameters!
This means that if we want to break down a general Lorentz transformation matrix into a boost and a rotation, we can just extract the boost parameters from the first row or column.
Let's see how this works in the example I gave of composing two boosts, B(12,ˆx) followed by B(12,ˆy). Multiplying the two boost matrices gives
B(12,ˆy)B(12,ˆx)=(2√30−1√300100−1√302√300001)(2√3−1√300−1√32√30000100001)=(43−23−1√30−1√32√300−23132√300001)
The matrix on the right (the composition of the two boosts) is clearly not a boost, because it isn't symmetric. And it's clearly not a rotation, because it mixes space and time coordinates. It is a more general Lorentz transformation that is neither a boost nor a rotation but a combination of them.
If we want to write it in RB (boost-then-rotation) form, we extract the boost parameters from the first row. From the top-left element we find that
γ=43
and thus the boost speed is
β=√1−1γ2=√74
and the product γβ is
γβ=√73.
From the other three elements in the top row, we can easily find the boost direction ˆn. We have
−γβnx=−23
and thus
ˆn=(2√7,√3√7,0).
Once you have extracted the parameters β and ˆn for the boost part of a general Lorentz transformation Λ=RB from the first row of Λ, you can use the general boost formula to compute B. In our case, it is
B=B(√74,(2√7,√3√7,0))=(43−23−1√30−23252127√30−1√327√38700001)
One can then invert it and find the rotation factor R=ΛB−1.
If you prefer to write a general transformation in the form Λ=BR, then extract the boot parameters from the first column of Λ, compute B, and then compute R=B−1Λ.
In either case, computing the inverse boost factor B−1 is easy: Rather than actually inverting a 4×4 matrix, you can just reverse the direction of the boost parameter ˆn. So in our example,
B−1=[B(√74,(2√7,√3√7,0))]−1=B(√74,(−2√7,−√3√7,0))=(43231√3023252127√301√327√38700001)
The rotation matrix R satifying Λ=RB is then
R=ΛB−1=(43−23−1√30−1√32√300−23132√300001)(43231√3023252127√301√327√38700001)=(100004√37−1700174√3700001)
This is recognizable as a rotation around the z-axis because it mixes only x and y together. Rotations by θ around the z-axis look like
(10000cosθ−sinθ00sinθcosθ00001)
so we immediately see that the rotation angle is such that
cosθ=4√37
and
sinθ=17.
But this raises the question of how to extract the rotation angle and rotation direction if R doesn't have such a simple form as in this example.
By taking the trace of the general rotation matrix, we see that
TrR(θ,ˆu)=2+2cosθ
so the rotation angle of a general rotation R is
θ=cos−1TrR−22.
And, since the unit vector ˆu along the rotation axis doesn't get rotated by the rotation,
R(1uxuyuz)=(1uxuyuz)
This says that ˆu is the spatial part of an eigenvector of R with eigenvalue 1.
Let's see how this works in an example where R is a more complicated rotation. Consider the Lorentz transformation
Λ=(83√3−43√3−23−1√3−1√32√300−23132√30−43√323√3132√3)
which I constructed by composing three boosts: the first by c/2 along ˆx, the second by c/2 along ˆy, and the third by c/2 along ˆz.
Extracting the boost parameters as before from the first row, we find
β=√378
and
ˆn=(4,2√3,3)√37.
After working out B and B−1, one finds the rotation to be
R=(10000237(6+7√3)237(3√3−8)137(9−8√3)0137(8√3−9)237(6+7√3)237(3√3−8)0−237(3√3−8)137(8√3−9)237(6+7√3))
The trace formula gives the rotation angle as
θ=cos−142√3−174≈14.18∘.
An unnormalized spatial eigenvector with eigenvalue 1 is (0,1,−1,1). (I suggest using a computer algebra system to compute the eigenvalues and eigenvectors! There is also a trivial temporal eigenvector (1,0,0,0) also with eigenvalue 1, and two complex eigenvectors with complex eigenvalues eiθ and e−iθ. The sum of the eigenvalues equals the trace value, 2+2cosθ, as it should.)
Therefore the rotation axis is the normalized spatial part of this eigenvector,
ˆu=(1,−1,1)√3.
No comments:
Post a Comment