PHYS3550 General Relativity Einstein’s 1920 derivation of the Lorentz Transformations John K. Webb School of Physics, UNSW, Sydney, NSW 2052, Australia (Dated: March 16, 2015) Einstein’s 1920 derivation of the Lorentz Transformations [1] is fine up to a certain point (indicated using a red line below), but then seems to lose clarity and even perhaps permit inconsistencies (unless of course I have made a mistake. Let me know if so please!). In my lecture slides for PHYS3550 I give a different derivation. It’s the same as the first part of Einstein’s derivation but uses interval invariance to make the final part simpler. I. EINSTEIN’S DERIVATION Here I follow Einstein’s derivation, matching his equation numbering, using the convention c = 1 and using a different notation. We take the usual situation of an intertial reference frame O moving with velocity v along the positive x-axis relative to another intertial reference frame O. The goal is to find general expressions that express coordinates measured by O in terms of those measured by O (or vice versa), ensuring consistency with the Relativity Principle. A light ray traveling in the +x direction is described by O as x = t i.e. x − t = 0 . . . . . . . (1) and is described by O as x¯ − t¯ = 0 . . . . . . . (2) Equations (1) and (2) satisfy the general condition (¯ x − t¯) = λ(x − t) . . . . . . . (3) where λ is an unknown constant. Repeating the argument above for a light ray traveling in the −x direction gives (¯ x + t¯) = µ(x + t) . . . . . . . (4) where µ is another unknown constant. Now add (or subtract) equations (3) and (4) and for simplicity use a= λ+µ . . . . . . . (4a) 2 b= λ−µ . . . . . . . (4b) 2 This gives x¯ = ax − bt . . . . . . . (5a) t¯ = at − bx . . . . . . . (5b) All we have done so far is to prove that the coordinates measured by O are a linear transformation of those measured by O, which, arguably, we could have assumed as our starting point (what else could it be?). We need to find the constants a and b. Equations (5) must be consistent with all events in O. We can make use of the fact that O’s origin is permanently at x¯ = 0. Equation (5a) then gives x= bt . . . . . . . (5c) a But the velocity of O as measured by O is x/t, so v= b . . . . . . . (6) a We now invoke the Relativity Principle: The length of a rod (unit length, lying along the x-direction) at rest in O as viewed by O must be the same as the length of the same rod at rest-in O as viewed by O. In other words, each observer must measure the rod to have unit length when the rod is stationary in their own rest-frame. We can imagine verifying this in practice by taking a snapshot from O of the moving bar. Taking a snapshot “freezes” the measurements at a specific time, e.g. for convenience, at t = 0. Therefore we can put t = 0 in (5a), which gives x¯ = ax . . . . . . . (6a) 2 Since we have specified the proper length of the rod be 1, the two events corresponding to the rod-ends are thus unit distance apart when measured by O, but when measured by O, they are separated by ∆x = 1 . . . . . . . (7) a Now apply the same argument as above to the reverse situation, i.e. where the same rod is stationary in O and is viewed from O. The snapshot is then taken from O at t¯ = 0. Equation (5b) then gives at = bx, or bx . . . . . . . (7c) a t= Next, Einstein substitutes (7c) into (5a) and uses (6) to derive an expression for x¯ at t¯ = 0: x¯ = ax − bt = ax − b2 x = ax(1 − v 2 ) . . . . . . . (7d) a but since we have a unit-length bar at rest in O, the length of the bar as measured by O is ∆¯ x = a(1 − v 2 ) . . . . . . . (7a) But to be consistent with the Relativity Principle, it must be that equation (7) = equation (7a) (that is, that the proper lengths measured by both observers are the same - it’s the same rod). Therefore, we have a= √ 1 . . . . . . . (7b) 1 − v2 Note we have discarded the negative solution in (7b) as being unphysical. See the explanation given in my PHYS3550 lecture slides. We have thus determined the constant a, from which b follows, and the Lorentz Transformations follow trivially by substitution back into (5a) and (5b): x − vt . . . . . . . (8a) x¯ = √ 1 − v2 t − vx t¯ = √ . . . . . . . (8b) 1 − v2 3 II. A POSSIBLE PITFALL However, here’s a potential pitfall: Let’s restart the derivation from immediately before equation (5c): Equation (5b) then gives at = bx, or b= at . . . . . . . (5c0 ) x Next, substitute (5c0 ) into (5a) and use v = x/t (as was done in going from (5c) to (6)). We then get 1 at2 = ax 1 − 2 . . . . . . . (7d0 ) x¯ = ax − x v and recalling we have a unit-length bar in O, as before, we have ∆¯ x = a(1 − 1 ) . . . . . . . (7a0 ) v2 This is horrible! (7a0 ) can only be consistent with (7a) if v = 1. Einstein’s derivation after equation (6) seems convoluted and the method appears to permit inconsistencies. However, deriving the Lorentz Transformations is easy using interval invariance - see my PHYS3550 lecture slides. [1] Einstein, Albert. Relativity: The Special and General Theory. New York: Henry Holt, 1920, Appendix I. http://www.bartleby.com/173/a1.html or http://newt.phys.unsw.edu.au/~jkw/phys3550/Papers/AE_LT_derivations_1920.pdf 4
© Copyright 2024