2 2 0 D O C U M E N T 2 0 4 F E B R U A R Y 1 9 2 6 oscillations of the gyrocompass around the meridian.”[3] This emerges from both patent specifications with sufficient clarity. In addition, it has been very thoroughly studied by Privy Councillor Professor Dr. Sommerfeld in his opinion of 23 Febru- ary of this year, whose explanations I adopt herewith.[4] The confiscated gyrocompass of the Gesellschaft für nautische Instrumente falls under Anschütz patent no. 235 749[5] because it manifests a “flexible contact arm” that governs the turning device. This device is of fundamental importance for the correct operation of the compass, because with a rigid contact arm the torque pre- venting the correct operation of the compass would be translated from the turned part onto the compass. The fact that only contact end surfaces are shown in the illustration of the gyro- compass by the Gesellschaft für nautische Instrumente while the half-ring surfaces attached to them of Anschütz patent no. 235 749 are missing does not make the more recent design independent of the patent, but only allows it to appear as a worse or imperfect version of the protected inventive idea.[6] sig. A. Einstein 204. From Léon Brillouin [Paris,] 30 Quai de Louvre, 1st, 26 February 1926 Dear Mr. Einstein, I have to thank you very much for your kind letter.[1] I entirely agree with you that my derivation of the law of refraction is invalid, and I also wrote that in a later paper (namely, in my last paper in Journal de Physique, which I’m also enclosing) [2] but I do think that it is not very easy to introduce a single coordinate into the rotator problem. I, too, had started out that way but could not fin- ish it off because the matrix was no longer Hermitian. That is why I introduced for a real matrix full of terms , and for chose the terms and , so that both matrices were again Hermitian. As to the numerical factor α (in ), there are many rotation and vibration spectra where half-quantum numbers have been chosen in order to obtain in the end a formula . It thus seems likely to me that both values and could occur.[3] With my warmest regards, yours sincerely, L. Brillouin q1 iq2 + iq2 q1 q m m±1) , ( q2 iq1( m m , 1+ ) m m , 1– ) –iq1( ν A m α) + ( = ν Am = α 0= α 2 -- -=1