Question 9.5: Which trajectory of the mass of a mathematical pendulum yiel...
Which trajectory of the mass of a mathematical pendulum yields a pendulum period that is independent of the amplitude?
Learn more on how we answer questions.
We consider Fig. 9.4. From energy conservation, we have
\frac{m}{2} \dot{s}^{2}(y)+gmy = mgh (9.20)
or
\dot{s}(y) = \sqrt{2g(h−y)}. (9.21)
From this, one can calculate the period by separation of the variables:
\frac{1}{4} T = \int\limits_{0}^{T/4}{dt} =\int\limits_{0}^{s(h)}{\frac{ ds}{\sqrt{2g(h−y)}}}=\int\limits_{0}^{h}{\frac{(ds/dy)dy}{\sqrt{2g(h−y)}}}. (9.22)
Using the variable u = y/h, (9.22) changes to
\frac{T}{4} = \int\limits_{0}^{1}{\frac{(ds/dy) \sqrt{h}du}{\sqrt{2g(1− u)}}}. (9.23)
We now require that T be independent of the maximum height h:
\frac{dT}{dh}= 0 for all h. (9.24)
Thus, we get from (9.23) (s^{′} ≡ ds/dy)
\frac{d}{dh} \int\limits_{0}^{1}{\frac{s^{′}\sqrt{h}du}{\sqrt{2g(1 −u)}}}= \int\limits_{0}^{1}{\frac{du}{\sqrt{2g(1− u)}}} \left(\frac{1}{2}h^{−1/2}s^{′}+\sqrt{h} \frac{ds^{′}}{dh}\right)= 0 for all h. (9.25)
With the condition that we keep the dimensionless variable u = y/h constant, we can rewrite the derivative with respect to h as a derivative with respect to y,
\frac{ds^{′}}{dh}=\frac{uds^{′}}{d(uh)}= u \frac{ds^{′}}{dy}= us^{″}, (9.26)
and thus, we can transform (9.25) into
\int\limits_{0}^{1}{\frac{du}{\sqrt{8g(1−u)}}(s^{′} + 2ys^{″}) \frac{1} {\sqrt{h}}}= 0 for all h. (9.27)
Any periodic function f (u) satisfying \int_{0}^{1}{ f (u)du} = 0 can generally be expanded into
a Fourier series:
f (u) =\sum\limits_{m=1}^{∞}{[a_{m} sin(2πmu) + b_{m} cos(2πmu)]} . (9.28)
Therefore, from (9.27) it follows that
s^{″} + \frac{1}{2y}s^{′} = \frac{ \sqrt{8gh(1 −u)}}{2y} \sum\limits_{m=1}^{∞}{a_{m} sin(2πmu) +b_{m} cos(2πmu)}
=\frac{ \sqrt{8gh(1 −u)}}{2y} \sum\limits_{m=1}^{∞}{\left(a_{m} sin\left(2πm \frac{y}{h}\right) +b_{m} cos \left(2πm \frac{y}{h}\right)\right)}. (9.29)
This holds for all values of h. The left-hand side of (9.29) does not contain h; therefore, the right-hand side must be independent of h too. This holds only for a_{m} = b_{m} = 0 (for all m), as we shall prove now.
To have the right-hand side of (9.29) independent of h, we must have
\sum\limits_{m=1}^{∞}{ \left[a_{m} sin \left(2πm \frac{y}{h}\right) + b_{m} cos \left(2πm \frac{y}{h}\right) \right]}= \frac{constant · (y/h)h^{1/2}}{\sqrt{8g(1 −y/h)}} (9.30)
or
\sum\limits_{m=1}^{∞}{ \left[a_{m} sin (2πmu) +b_{m} cos (2πmu)\right]} = \frac{u}{\sqrt{1−u}} \frac{h^{1/2}}{\sqrt{8g}}C. (9.31)
By integrating (9.31) from 0 to 1, we obtain
0 = \frac{h^{1/2}}{\sqrt{8g}}C \int\limits_{0}^{1}{\frac{u}{\sqrt{1− u}}du}= \frac{4}{3} \frac{h^{1/2}}{\sqrt{8g}}C, (9.32)
thus, C = 0. (This reflects the fact that u/\sqrt{1− u} cannot be expanded into a Fourier series à la (9.31).)
Inserting this result C = 0 again into (9.30), we have a_{m} = b_{m} = 0 ∀m, and thus, from (9.29)
s^{″} + \frac{s^{′}}{2y}= 0. (9.33)
From this, one finds by integrating once
\frac{s^{″} }{s^{′}}=− \frac{1}{2y} ⇒ s^{′} ≡ \frac{ds}{dy} = \widetilde{C} e^{−(1/2) ln y} =\frac{\widetilde{C}}{\sqrt{y}}. (9.34)
The constant is usually denoted by
\widetilde{C}=\sqrt{\frac{l}{2}}, (9.35)
so that we have to solve
\frac{ds}{dy}=\sqrt{\frac{l}{2}} \frac{1}{\sqrt{y}}. (9.36)
This is the differential equation of a cycloid.²
² See W. Greiner: Classical Mechanics: Point Particles and Relativity, 1st ed., Springer, Berlin (2004), Problem 24.4.
