betaFeedback

WorksPhysics and Astronomy

Looping Pendulum: Theory, Simulation, and Experiment

Abstract

The looping pendulum is a simple physical system consisting of two masses connected by a string that passes over a rod. We derive equations of motion for the looping pendulum using Newtonian mechanics, and show that these equations can be solved numerically to give a good description of the system's dynamics. The numerical solution captures complex aspects of the looping pendulum's behavior, and is in good agreement with the experimental results.

Introduction

The looping pendulum consists of two different masses, one heavy and one light, connected with a string that passes over a horizontal rod. One lifts up the heavy mass by pulling down on the light one at a given angle with the vertical axis. Releasing the light mass, it sweeps around the rod, keeping the heavy mass from falling to the ground. This mesmerizing physics demo, which we first became aware of through a video on YouTube1, appeared as one of the problems in the 2019 International Young Physicists Tournament2 and was also studied to some degree in a recent paper3, 6. Its simple implementation and spectacular dynamics make it an intriguing problem for physics students and professors alike.

In this paper we study the looping pendulum using Newtonian mechanics, simulate its motion using numerical methods, and compare the theoretical predictions with the experimental results. The experiments closely confirm the numeric simulations and reveal rich dynamics with surprising features. In some cases we notice a hint of sensitivity to small changes in the initial conditions.

The manuscript is organized as following: in Section II we derive the equations of motion, and obtain a pair of coupled, second-order, non-linear differential equations. In Section III we present the experimental setup, and in Section IV we compare the experimental results with numerical solutions of the model obtained using Mathematica4. Finally, in Section V we review our results, outline plans to improve the apparatus, and consider the possibility that, for a given range of the parameters of the model, the looping pendulum may exhibit chaotic behavior.

Equations of Motion for the Looping Pendulum

The dynamics of the looping pendulum can be described using physics that is typically covered in first-year and second-year undergraduate courses in United States universities. Our setup is shown in Figure 1. A heavy mass M2M_2 is suspended vertically by a string that passes over a rigid rod of radius RR. The distance from M2M_2 to the point where the string makes contact with the rod is hh. The string winds around the rod for an angle θ\theta, then extends a distance ℓ\ell from the rod to a lighter mass M1M_1 which is held at rest. The lighter mass is released, and the system is allowed to move freely. During this motion the quantities

Image not extracted

Figure 1. This cross-section view of the looping pendulum shows masses M1M_1 and M2M_2 connected by a string that passes over a rod with radius RR. The radius of the rod is much smaller than the initial values of ℓ\ell or hh in our experiments, and is exaggerated here for clarity.

hh, θ\theta, and ℓ\ell will change, but the total length of the string remains fixed at L=ℓ+Rθ+hL = \ell+ R\theta+ h.

The motion of mass M2M_2 is best described using a fixed coordinate system x^\hat{x} and y^\hat{y} with the origin taken at the center of the rod’s circular cross section. Mass M2M_2 is assumed to move straight up and down with no horizontal component to its motion. Its position is r⃗2=Rx^−hy^\vec{r}_2 = R\hat{x} - h\hat{y} and its velocity is v⃗2=−h˙y^\vec{v}_2 = -\dot{h}\hat{y}, where a dot indicates the derivative with respect to time. Since the length LL of the string is fixed, L˙=ℓ˙+Rθ˙+h˙=0\dot{L} = \dot{\ell} + R\dot{\theta} + \dot{h} = 0 and the velocity and acceleration of M2M_2 can be written as:

v⃗2=(ℓ˙+Rθ˙)y^,a⃗2=(ℓ¨+Rθ¨)y^.(1)\vec{v}_2 = (\dot{\ell} + R\dot{\theta})\hat{y}, \qquad\vec{a}_2 = (\ddot{\ell} + R\ddot{\theta})\hat{y}. \tag*{(1)}

The lighter mass M1M_1 follows a more complicated path. Its position is given by

r⃗1=(Rcos⁡θ−ℓsin⁡θ)x^+(Rsin⁡θ+ℓcos⁡θ)y^.(2)\vec{r}_1 = (R\cos\theta- \ell\sin\theta)\hat{x} + (R\sin\theta+ \ell\cos\theta)\hat{y}. \tag*{(2)}

However, it is convenient to describe its motion using a co-moving system of coordinates that point parallel and perpendicular to the string connecting M1M_1 to the rod. We define the orthonormal unit vectors e^1\hat{e}_1 and t^1\hat{t}_1 (shown in Figure 1) as

e^1=sin⁡θx^−cos⁡θy^,(3)\hat{e}_1 = \sin\theta\hat{x} - \cos\theta\hat{y}, \tag*{(3)}
t^1=−cos⁡θx^−sin⁡θy^.(4)\hat{t}_1 = -\cos\theta\hat{x} - \sin\theta\hat{y}. \tag*{(4)}

Then the position of M1M_1 can be written as

r⃗1=−ℓe^1−Rt^1.(5)\vec{r}_1=-\ell\hat{e}_1-R\hat{t}_1 . \tag*{(5)}

Because these new unit vectors depend on θ\theta, they will change as the string loops around the rod. Their first and second derivatives with respect to time are

e^˙1=−θ˙t^1,t^˙1=θ˙e^1,(6)\dot{\hat{e}}_1=-\dot{\theta}\hat{t}_1,\qquad\dot{\hat{t}}_1=\dot{\theta}\hat{e}_1, \tag*{(6)}
e^¨1=−θ¨t^1−θ˙2e^1,t^¨1=θ¨e^1−θ˙2t^1.(7)\ddot{\hat{e}}_1=-\ddot{\theta}\hat{t}_1-\dot{\theta}^{2}\hat{e}_1,\qquad\ddot{\hat{t}}_1=\ddot{\theta}\hat{e}_1-\dot{\theta}^{2}\hat{t}_1 . \tag*{(7)}

Using these results, we obtain for the velocity and acceleration of M1M_1

v⃗1=−(ℓ˙+Rθ˙)e^1+ℓθ˙t^1,(8)\vec{v}_1=-(\dot{\ell}+R\dot{\theta})\hat{e}_1+\ell\dot{\theta}\hat{t}_1, \tag*{(8)}
a⃗1=(ℓθ˙2−ℓ¨−Rθ¨)e^1+(2ℓ˙θ˙+ℓθ¨+Rθ˙2)t^1.(9)\vec{a}_1=(\ell\dot{\theta}^{2}-\ddot{\ell}-R\ddot{\theta})\hat{e}_1+(2\dot{\ell}\dot{\theta}+\ell\ddot{\theta}+R\dot{\theta}^{2})\hat{t}_1 . \tag*{(9)}

As a check, notice that the e^1\hat{e}_1 component of v⃗1\vec{v}_1 – the rate at which the string is moving onto or off of the rod – is minus the rate at which M2M_2 is moving downwards or upwards in Eq. (1).

Now that we have expressions for the accelerations of each mass, we need to identify the forces acting on them. For mass M2M_2 there is an upward force due to a tension F⃗T,2\vec{F}_{T,2}, a downward force F⃗g,2\vec{F}_{g,2} due to gravity, and a small but non-negligible amount of air resistance F⃗air,2\vec{F}_{\mathrm{air},2} in the direction opposite to v⃗2\vec{v}_2. These forces are

F⃗T,2=T2y^,F⃗g,2=−M2gy^,F⃗air,2=−γ2∣v⃗2∣v⃗2,(10)\vec{F}_{T,2}=T_2\hat{y},\qquad\vec{F}_{g,2}=-M_2g\hat{y},\qquad\vec{F}_{\mathrm{air},2}=-\gamma_2\lvert\vec{v}_2\rvert\vec{v}_2, \tag*{(10)}

where the coefficient γ2\gamma_2, discussed in more detail in § 0.4, encodes the dependence of F⃗air,2\vec{F}_{\mathrm{air},2} on the drag coefficient of M2M_2 and the area it presents in a plane perpendicular to v⃗2\vec{v}_2. Using Eq. (1), the net force on mass M2M_2 is

F⃗net,2=(T2−M2g−γ2∣ℓ˙+Rθ˙∣(ℓ˙+Rθ˙))y^.(11)\vec{F}_{\mathrm{net},2}=\left(T_2-M_2g-\gamma_2\lvert\dot{\ell}+R\dot{\theta}\rvert(\dot{\ell}+R\dot{\theta})\right)\hat{y}. \tag*{(11)}

Notice that the force due to air resistance changes sign based on the direction of v⃗2\vec{v}_2. For mass M1M_1 there are four forces: a tension F⃗T,1\vec{F}_{T,1} pulling along the direction e^1\hat{e}_1 of the string, gravity pulling downwards, air resistance in the direction opposite to v⃗1\vec{v}_1, and an effective force associated with air resistance experienced by the string. The first three forces have the form

F⃗T,1=T1e^1,F⃗g,1=−M1gy^,F⃗air,1=−γ1∣v⃗1∣v⃗1.(12)\vec{F}_{T,1}=T_1\hat{e}_1,\qquad\vec{F}_{g,1}=-M_1g\hat{y},\qquad\vec{F}_{\mathrm{air},1}=-\gamma_1\lvert\vec{v}_1\rvert\vec{v}_1 . \tag*{(12)}
Image not extracted

Figure 2. The forces acting on M1M_1 and M2M_2 include gravity, tension in the string, and air resistance.

The final force acting on M1M_1 comes from drag on the string connecting it to the rod. Because drag forces are proportional to area, and the diameter of the string’s cross section is much smaller than its length, we approximate this force as being due entirely to the area of the string in a plane perpendicular to t^1\hat{t}_1 and ignore its motion in the e^1\hat{e}_1 direction. Therefore,

F⃗air,str≃−γstr∣v⃗1⋅t^1∣(v⃗1⋅t^1)t^1.(13)\vec{F}_{\mathrm{air,str}} \simeq-\gamma_{\mathrm{str}} \lvert\vec{v}_1 \cdot\hat{t}_1\rvert(\vec{v}_1 \cdot\hat{t}_1)\hat{t}_1 . \tag*{(13)}

The area of the string factoring into γstr\gamma_{\mathrm{str}} is proportional to its length ℓ\ell, which changes as the masses move. To account for this, we write γstr=ℓαstr\gamma_{\mathrm{str}} = \ell\alpha_{\mathrm{str}}. Using Eqs. (3)–(4) and Eq. (8), the net force on M1M_1 is then

F⃗net,1=(T1+M1gcos⁡θ+γ1∣v⃗1∣(ℓ˙+Rθ˙))e^1+(M1gsin⁡θ−γ1∣v⃗1∣ℓθ˙−αstr∣ℓθ˙∣ℓ2θ˙)t^1.(14)\begin{aligned} \vec{F}_{\mathrm{net},1} ={}& \left(T_1 + M_1 g \cos\theta+ \gamma_1 \lvert\vec{v}_1\rvert\left(\dot{\ell} + R\dot{\theta}\right)\right)\hat{e}_1 \\ &+ \left(M_1 g \sin\theta- \gamma_1 \lvert\vec{v}_1\rvert\ell\dot{\theta} - \alpha_{\mathrm{str}}\lvert\ell\dot{\theta}\rvert\ell^2\dot{\theta}\right)\hat{t}_1 . \tag*{(14)} \end{aligned}

The coefficients γ1\gamma_1 and αstr\alpha_{\mathrm{str}} characterizing the air-resistance forces will be discussed in more detail in § 0.4.

The tensions T1T_1 and T2T_2 will differ because of friction between the string and the rod. Figure 3 shows an infinitesimal segment of string covering an angle dθ\mathrm{d}\theta along the rod. We assume here that the string is sliding across the surface of the rod in the clockwise direction. This produces a small amount of friction dFf=μ dN\mathrm{d}F_f = \mu\,\mathrm{d}N in the counterclockwise direction, where μ\mu is the coefficient of friction between the string and the rod and dN\mathrm{d}N is the normal force acting on the string segment. Since dθ≪1\mathrm{d}\theta\ll1, we use cos⁡dθ2≃1\cos\frac{\mathrm{d}\theta}{2} \simeq1 and sin⁡dθ2≃dθ2\sin\frac{\mathrm{d}\theta}{2} \simeq\frac{\mathrm{d}\theta}{2} to write the net force on the segment of string as

Image not extracted

Figure 3. Friction between the string and the rod determines the difference between the tensions T1T_1 and T2T_2.

F⃗net=(dT−μ dN)x^+(dN−T dθ)y^.(15)\vec{F}_{\mathrm{net}} = (\mathrm{d}T - \mu\,\mathrm{d}N)\hat{x} + (\mathrm{d}N - T\,\mathrm{d}\theta)\hat{y} . \tag*{(15)}

Requiring the net force to vanish gives

dT=Tμ dθ.(16)dT = T\mu\,d\theta. \tag*{(16)}

Integrating dTdT over the interval between the two points where the string makes contact with the rod, as illustrated in Figure 1, gives the relationship between T1T_1 and T2T_2:

T2=T1e−sign⁡(v2)μθ.(17)T_2 = T_1 e^{-\operatorname{sign}(v_2)\mu\theta}. \tag*{(17)}

The factor of −sign⁡(v2)-\operatorname{sign}(v_2) in the exponent accounts for the fact that the string may slide along the rod in either direction, depending on whether M2M_2 is moving downwards or upwards. The clockwise sliding shown in Figure 3 corresponds to M2M_2 moving downwards, in which case sign⁡(v2)<0\operatorname{sign}(v_2)<0 and T2>T1T_2>T_1.

We are now in position to write the equations of motion for the looping pendulum. For mass M1M_1, the e^1\hat{e}_1 and t^1\hat{t}_1 components of Newton’s second law give

M1(ℓθ˙2−ℓ¨−Rθ¨)=T1+M1gcos⁡θ+γ1(ℓ˙+Rθ˙)2+(ℓθ˙)2(ℓ˙+Rθ˙),(18)M_1\left(\ell\dot{\theta}^{2}-\ddot{\ell}-R\ddot{\theta}\right) = T_1+M_1g\cos\theta +\gamma_1\sqrt{\left(\dot{\ell}+R\dot{\theta}\right)^2+\left(\ell\dot{\theta}\right)^2}\left(\dot{\ell}+R\dot{\theta}\right), \tag*{(18)}
M1(2ℓ˙θ˙+ℓθ¨+Rθ˙2)=M1gsin⁡θ−γ1(ℓ˙+Rθ˙)2+(ℓθ˙)2ℓθ˙−αstr∣ℓθ˙∣ℓ2θ˙.(19)M_1\left(2\dot{\ell}\dot{\theta}+\ell\ddot{\theta}+R\dot{\theta}^{2}\right) = M_1g\sin\theta -\gamma_1\sqrt{\left(\dot{\ell}+R\dot{\theta}\right)^2+\left(\ell\dot{\theta}\right)^2}\ell\dot{\theta} -\alpha_{\mathrm{str}}\left|\ell\dot{\theta}\right|\ell^2\dot{\theta}. \tag*{(19)}

For mass M2M_2 the equation of motion is

M2(ℓ¨+Rθ¨)=T2−M2g−γ2∣ℓ˙+Rθ˙∣(ℓ˙+Rθ˙).(20)M_2\left(\ddot{\ell}+R\ddot{\theta}\right) = T_2-M_2g-\gamma_2\left|\dot{\ell}+R\dot{\theta}\right|\left(\dot{\ell}+R\dot{\theta}\right). \tag*{(20)}

Solving Eq. (18) for T1T_1 and using Eq. (17) leads to the following pair of coupled, second-order, non-linear differential equations for ℓ(t)\ell(t) and θ(t)\theta(t):

0=2ℓ˙θ˙+ℓθ¨+Rθ˙2−gsin⁡θ+γ1M1(ℓ˙+Rθ˙)2+(ℓθ˙)2 ℓ˙+αstr∣ℓθ˙∣ℓ2θ˙,(21)0 = 2\dot{\ell}\dot{\theta}+\ell\ddot{\theta}+R\dot{\theta}^{2}-g\sin\theta+\frac{\gamma_{1}}{M_{1}}\sqrt{(\dot{\ell}+R\dot{\theta})^{2}+(\ell\dot{\theta})^{2}}\,\dot{\ell}+\alpha_{\mathrm{str}}|\ell\dot{\theta}|\ell^{2}\dot{\theta}, \tag*{(21)}
0=ℓ¨+Rθ¨+g+γ2M2∣ℓ˙+Rθ˙∣(ℓ˙+Rθ˙)−e−sign⁡(v2)μθ(M1M2(ℓθ˙2−ℓ¨−Rθ¨−gcos⁡θ)−γ1M2(ℓ˙+Rθ˙)2+(ℓθ˙)2(ℓ˙+Rθ˙)).(22)\begin{aligned} 0 ={}& \ddot{\ell}+R\ddot{\theta}+g+\frac{\gamma_{2}}{M_{2}}|\dot{\ell}+R\dot{\theta}|(\dot{\ell}+R\dot{\theta})\\ &-e^{-\operatorname{sign}(v_{2})\mu\theta}\left(\frac{M_{1}}{M_{2}}\left(\ell\dot{\theta}^{2}-\ddot{\ell}-R\ddot{\theta}-g\cos\theta\right)-\frac{\gamma_{1}}{M_{2}}\sqrt{(\dot{\ell}+R\dot{\theta})^{2}+(\ell\dot{\theta})^{2}}(\dot{\ell}+R\dot{\theta})\right). \tag*{(22)} \end{aligned}

These equations cannot be solved in closed form, so in this paper we use Mathematica to solve them numerically. The numerical solution is compared with our experimental results in § 0.4.

The phenomenon that first drew our attention to the looping pendulum is the tendency of mass M2M_{2} to fall some distance from its initial position and then stop moving. In some cases its motion will stop temporarily before falling again, or it may even briefly move upwards. But once M2M_{2} stops completely – a behavior we refer to as halting – the subsequent motion of M1M_{1} simplifies somewhat. First, the shape of its trajectory (assuming it is moving fast enough to maintain tension in the string) is entirely determined by the action of the string wrapping around the rod. Second, if air resistance is ignored then the time dependence of ℓ\ell and θ\theta are well-approximated by a simple function. These observations, which aren’t essential in what follows, are discussed in more detail in Appendix § 0.8.

Experimental Setup

To test the model derived in § 0.2 we constructed several looping pendulums and recorded the motion of masses M1M_{1} and M2M_{2} for a range of initial conditions. The xx and yy positions of each mass were then extracted from a slow-motion video using the Tracker Video Analysis and Modeling Tool1.

A schematic of our apparatus is shown in Figure 4. The frame consisted of three aluminum rods. Two vertical rods were mounted on lab stands, with clamps connected to each rod supporting a third, horizontal rod. The third rod, which was aligned using a level, appears in cross section in Figure 1. The two masses were connected by a string which was placed over this rod. The lighter mass was then released from rest at different initial angles. The subsequent motion of the masses was recorded in a 240 fps video using an iPhone placed in front of the apparatus. A large white panel with a grid drawn on it was mounted behind the apparatus to improve alignment of the camera and visibility of the masses.

Image not extracted

Figure 4. The arrangement of the rods, masses, string, and camera is shown in this schematic of the apparatus used in our experiments.

The materials used in the apparatus were selected because they were on hand at the beginning of the project. With the exception of a set of cylindrical masses used in later trials, all materials were taken from one of the undergraduate project rooms in the Physics Department at Loyola University Chicago. Multiple pendulums were built following the setup described above, but there was some variation in materials. A number of different aluminum rods were used for the horizontal rod. While all the rods had a length of approximately 1 m and a diameter between 1.2-1.3 cm, rough spots and variations in surface finish gave different values for the coefficient of friction μ\mu. The string connecting the masses was taken from a spool of generic kite string and was not selected for any particular properties other than its negligible mass. The masses themselves initially comprised bundles of washers with individual masses in the range of 7.5-7.7 g. The irregular shape of the bundles made it difficult to model air resistance so they were eventually replaced with a set of cylindrical masses. Later versions of the apparatus also corrected problems like alignment of the camera, using a leveled tripod to mount the iPhone rather than a standard lab clamp.

There were several dimensionless parameters that could be varied when running the experiment: the ratio M2/M1M_2/M_1 of the masses; various ratios of the lengths LL, ℓ\ell, and RR; and the initial angle θi\theta_i for the point of contact between the string and the rod. We observed the behavior of the masses for several values of M2/M1M_2/M_1, multiple values of the initial angle θi\theta_i for each value of the mass ratio, and fixed2 values of LL, ℓ\ell, and RR.

A run of the experiment began by selecting two masses that gave a specific value of M2/M1M_2/M_1. The masses were connected by a string, a spot was marked on the string where it would make contact with the rod, and the total length of the string LL and the length ℓ\ell from the point of contact to M1M_1 were measured. The mass M1M_1 was then positioned at different initial angles, held at rest but with tension in the string, and released. The participant recording the run would scrutinize the video for signs that the masses had made contact with each other or with the string. If there was any indication of contact the run would be repeated. This happened frequently, so getting a “clean” run with no collisions typically required several attempts. This process was then repeated for a new initial angle, with multiple initial angles tested for each mass ratio.

Once a run was successfully completed, the 240 fps video was trimmed to begin at the frame where the masses were released and loaded into the Tracker Video Analysis and Modeling Tool. The length scale was set by identifying the distance between two points in the first frame with one of the measured lengths (typically ℓ\ell), and then verified by comparing the computed and measured values of the other lengths (LL, RR, and hh). The initial angle that the string connected to M1M_1 made with the horizontal direction was measured and converted into the initial angle θ\theta shown in Figure 1. Then the positions of each mass were recorded frame-by-frame until mass M1M_1 reached the rod – typically resulting in between 200 and 400 data points. The positions were marked on screen at the points where the string attached to the masses. Once all the positions had been marked, the data was exported to CSV files containing the (t,x,y)(t,x,y) data points for each mass. The data was then imported into Mathematica and compared with the numerical solution of our model.

There were aspects of the apparatus which could not be incorporated into the model of § 0.2. For example, the model assumes that the trajectory of the masses lies in a plane. However, it was difficult to realize this in practice. When the masses experienced coplanar motion they often collided with each other or with the string. To avoid this, the trajectory of mass M1M_1 had to have a small component in the direction parallel to the rod, which is not included in our simulation. Another problem arose from the string looping over itself as it wound around the rod. This changed the value of the coefficient of friction in a way that we could not account for in the model. Fortunately, these effects could be minimized: a careful release of the mass reduced motion parallel to the rod, and a trial could be repeated if the video showed the string looping over itself and subsequently slipping. More significant sources of error which can be addressed in future experiments with different materials or equipment are discussed in § 0.5.

Experimental Results and Comparison with Simulation

To compare our experimental results to the model of § 0.2 we solve Eqs. (21)-(22) numerically. The equations depend on several parameters which must be measured or estimated. The quantities RR, LL, ℓ\ell, and θi\theta_i were measured directly and/or extracted from video using the Tracker Video Analysis and Modeling Tool, while the coefficient of friction μ\mu was determined using a procedure described below. The drag coefficients could not be measured and were instead estimated based on the shape and dimensions of the masses and string.

The coefficient of friction μ\mu between the string and rod was determined by tracking the motion of masses M1M_1 and M2M_2 after they were suspended vertically and released from rest. This arrangement is shown in Figure 5. In that case the masses move up or down with constant acceleration. The motion of the masses was recorded, the position of M2M_2 as a function of time was extracted from the video using Tracker, and μ\mu was determined by fitting this data to the trajectory

Image not extracted

Figure 5. The masses are suspended vertically and released from rest. Mass M2M_2 moves downwards with a constant acceleration (ignoring air resistance) that depends on gg, μ\mu, and the ratio M2/M1M_2/M_1.

y2(t)=y2(0)−12gM2−M1eμπM2+M1eμπt2(23)y_2(t) = y_2(0) - \frac{1}{2} g \frac{M_2 - M_1 e^{\mu\pi}}{M_2 + M_1 e^{\mu\pi}} t^2 \tag*{(23)}

An example of a best-fit for the case M2/M1=10M_2/M_1 = 10, yielding the value μ=0.34\mu= 0.34, is shown in Figure 6.

Image not extracted

Figure 6. For the case M2/M1=10M_2/M_1 = 10, the green curve shows the trajectory with μ=0\mu= 0, the purple circles show the positions extracted from video of the falling mass, and the blue curve represents the best-fit value μ=0.34\mu= 0.34 for Eq. (23).

Air resistance of the masses and string was more difficult to quantify. In Eqs. (21)-(22) the three coefficients γ1\gamma_1, γ2\gamma_2, and αstr=γstr/ℓ(t)\alpha_{\mathrm{str}} = \gamma_{\mathrm{str}}/\ell(t) determine the magnitude of the drag forces. These coefficients are generally expected to take the form

γ=12CρairA(24)\gamma= \frac{1}{2} C \rho_{\mathrm{air}} A \tag*{(24)}

where CC is a drag coefficient of order 1, ρair≃1.2 kg/m3\rho_{\mathrm{air}} \simeq1.2\ \mathrm{kg/m^3} is the density of air, and AA is the area that the object presents in a plane perpendicular to its direction of motion. In early runs of the experiment the masses comprised bundles of washers. The irregular shape of the bundles made estimates of AA difficult, especially as M1M_1 changed orientation during its motion, so the washers were later replaced with a set of cylindrical masses. Assuming a drag coefficient CC of order 1, and treating the area AA as that of a rectangle with dimensions given by the height and diameter of the cylinder, the coefficients γ1\gamma_1 and γ2\gamma_2 were estimated to both be of order 10−3 kg/m10^{-3}\,\mathrm{kg}/\mathrm{m}. For the string connecting M1M_1 to the rod, the area in the plane perpendicular to t^1\hat{\mathbf{t}}_1 is equal to its length ℓ(t)\ell(t) times its diameter. The string used in the experiment has a diameter of approximately 2 mm, so the ratio αstr=γstr/ℓ(t)\alpha_{\mathrm{str}}=\gamma_{\mathrm{str}}/\ell(t) is estimated to be αstr≃10−3 kg/m2\alpha_{\mathrm{str}}\simeq10^{-3}\,\mathrm{kg}/\mathrm{m}^2.

Data for a typical trial with mass ratio M2/M1=10M_2/M_1=10 (M1=10M_1=10 g, M2=100M_2=100 g) is compared with the numerical solution of our model in Figure 7. The first subfigure, Figure 7a, shows the path followed by M1M_1 as it loops around the rod. The simulation accurately tracks the measured position of M1M_1 from release until the point where M2M_2 halts (the solid black dot), and then precisely replicates the spiral produced by the string wrapping around the rod (see also Appendix § 0.8). The primary source of error affecting these plots is uncertainty in accurately judging the position of M1M_1, caused by the blurring of the mass in the video.

Image not extracted

Figure 7. These plots show experimental data for M1M_1 (red circles) and M2M_2 (green circles), along with the numerical solution of our model (blue curve), for M2/M1=10M_2/M_1=10. The vertical purple line in Figure 7b, Figure 7c, and Figure 7d indicates the time when M2M_2 halts.

Figure 7b compares the observed and simulated dynamics of M2M_2 as it falls vertically. The model predicts that the mass drops a distance Δy2,num=−0.37 m\Delta y_{2,\mathrm{num}} = -0.37\,\mathrm{m} before halting, and the observed value was Δy2,exp=−0.37 m\Delta y_{2,\mathrm{exp}} = -0.37\,\mathrm{m}. The numerical solution predicts that M2M_2 halts 0.38 s0.38\,\mathrm{s} after M1M_1 is released, which is again in good agreement with the observed halt time of 0.40 s0.40\,\mathrm{s}. However, in Figure 7b the mass can be seen to pass directly through its final y2y_2 value closer to the predicted halt time. It briefly dips below its final y2y_2 value and then returns to the same point at a slightly later time. We suspect that this behavior is caused by the slight elasticity of the string, which was not accounted for in the model.

Finally, Figure 7c and Figure 7d show the xx and yy positions of mass M1M_1 as a function of time. As with M2M_2, the numerical solution precisely captures the dynamics of M1M_1 as it loops around the rod. The numerical solution predicts that M1M_1 reaches the rod (ℓ=0\ell= 0) at tend=1.1 st_{\mathrm{end}} = 1.1\,\mathrm{s}. This prediction can’t be directly compared to experiment, since the finite size of M1M_1 in our apparatus causes it to collide with the rod before the length ℓ\ell reaches zero. However, we can compare the times at which x1x_1 and y1y_1 reach local extrema where v1,x=0v_{1,x} = 0 or v1,y=0v_{1,y} = 0. These times are shown in Table I of Appendix A and the predictions of the numerical solution are in excellent agreement with the observed values. We should emphasize that the effects of air resistance are essential for the accuracy of the model. Without air resistance, the model’s prediction for when M1M_1 reaches the rod is typically 5−10%5 - 10\% earlier than with air resistance included. Although this might amount to less than a tenth of a second, it is enough for the late time behavior of the simulation to move completely out of phase with the data. To faithfully track the dynamics of M1M_1, air resistance must be included in the model.

v1,x=0v_{1,x}=0v1,x=0v_{1,x}=0v1,y=0v_{1,y}=0v1,y=0v_{1,y}=0
texpt_{\mathrm{exp}}tnumt_{\mathrm{num}}texpt_{\mathrm{exp}}tnumt_{\mathrm{num}}
10.38 s0.38 s0.25 s0.24 s
20.52 s0.52 s0.45 s0.45 s
30.63 s0.63 s0.58 s0.58 s
40.75 s0.75 s0.70 s0.69 s
50.84 s0.84 s0.79 s0.79 s
60.91 s0.92 s0.87 s0.88 s
70.98 s0.99 s0.95 s0.96 s

Table I. Experiment and numerical solution values for the times when x1x_1 and y1y_1 experience local extrema in Figure 7c and Figure 7d.

Our experiments show that the behavior of the looping pendulum for large values of M2/M1M_2/M_1 usually involves a quick halt, followed by a regular spiral like the one in Figure 7a.8 However, the looping pendulum exhibits more complicated behavior at smaller values of M2/M1M_2/M_1, which is also well accounted for in our model.

An example is shown in Figure 8 for the case M2/M1=2M_2/M_1 = 2 (M1=50 gM_1 = 50\,\mathrm{g}, M2=100 gM_2 = 100\,\mathrm{g}). The mass M1M_1 follows a path with a qualitatively different shape than the trajectory shown in Figure 7a. Nevertheless, it is well described by the numerical solution. The motion of M2M_2 is also very different than what is seen in Figure 7b. It is observed to drop, briefly stop at time tAt_A, then move upwards before stopping again at time tBt_B, and then resume falling at time tCt_C before finally halting. The same behavior appears in the numerical solution of our model. Table II in Appendix § 0.7 summarizes the observed and predicted values for the times and positions of these events. The dynamics match up well for the first few events, with small differences between the predicted and observed values for tAt_A and tBt_B. There is a larger disagreement between the observed and predicted values for the time tCt_C at which M2M_2 begins its final downward plunge. But the model’s halting time is once again in good agreement with the data, and it successfully reproduces the final distance that M2M_2 drops. (There appears to be another small “bounce” in the data as M2M_2 abruptly stops, likely due to stretching of the string.) Despite slightly larger errors for some of the intermediate features in M2M_2’s trajectory, the model gives a compelling description of what was observed in the experiment. In fact, the M2/M1=2M_2/M_1 = 2 trial was carried out because the simulation predicted this interesting behavior for M2M_2.

Image not extracted

Figure 8. The numerical solution captures nontrivial features of the trajectory and dynamics of the masses for the mass ratio M2/M1=2M_2/M_1 = 2.

QuantityExperimentNumericalPercent Error
tAt_A0.31 s0.32 s4.6
Δy2(tA)\Delta y_2(t_A)−0.21-0.21 m−0.23-0.23 m8.5
tBt_B0.46 s0.48 s3.0
Δy2(tB)\Delta y_2(t_B)−0.20-0.20 m−0.21-0.21 m8.6
tCt_C0.71 s0.62 s12.7
thaltt_{\mathrm{halt}}1.0 s0.95 s4.6
Δy2(thalt)\Delta y_2(t_{\mathrm{halt}})−0.44-0.44 m−0.44-0.44 m0.7

Table II. Observed and predicted values for the times and changes in position at different points along the trajectory of M2M_2 as illustrated in Figure 8b.

Discussion

In this paper we developed a model of the looping pendulum using Newton’s Laws. The result is a system of two coupled, nonlinear differential equations which we solved numerically. The numerical solution accurately describes the trajectory and dynamics of the masses, and reproduces complex behaviors observed for smaller values of the ratio M2/M1M_2/M_1.

The apparatus described in Section III was built with materials that were on hand. Replacing some of the components would reduce uncertainty in data collection, eliminate the effects of forces we did not account for, and improve our ability to estimate parameters that appear in the model.

The single largest source of uncertainty when collecting position data was blurring of the fast-moving masses (especially M1M_1) in the video. Future versions of the experiment will rely on a dedicated high speed camera instead of an iPhone, which will reduce blurring and allow us to more accurately identify the positions of the masses in each frame.

Another source of error is the slight elasticity of the string. The string connecting the masses will be replaced with a material that is less susceptible to stretching. We expect that this will reduce the small bounce seen in M2M_2’s position as it halts, resulting in more reliable measurements of the halting time.

A final challenge is factoring in air resistance, which proved important for reproducing the dynamics of the pendulum. The orientations of the masses, and hence the area they present in a plane perpendicular to their velocity, change throughout their motion. This alters the drag coefficients, which are already difficult to estimate. Replacing bundles of washers with cylindrical masses was an improvement, but spherical masses that present the same area throughout their motion would better address this problem.3

Using an improved apparatus, we will explore a larger range of the looping pendulum’s parameter space. These first investigations sampled a relatively small number of initial angles for mass ratios 2≤M2/M1≤102 \le M_2/M_1 \le10. We hope to extend this to higher and lower values of M2/M1M_2/M_1, and to explore a much larger set of initial angles. For lower values of M2/M1M_2/M_1 the system only seems to halt for isolated values of the initial angle, so it would be interesting to see if there is a nontrivial value of M2/M1M_2/M_1 below which halting does not occur at all. Also, the presence of air resistance in the model means that the individual masses are important, as opposed to just their ratio. It would also be interesting to explore the behavior of the system as M1M_1 and M2M_2 are increased (or decreased) with M2/M1M_2/M_1 held fixed. The ratios γ1/M1\gamma_1/M_1 and γ2/M2\gamma_2/M_2 appear in the equations of motion, so the area scaling of γ\gamma and volume scaling of MM should make air resistance less important for larger masses.

Finally, at smaller values of M2/M1M_2/M_1 our model seems to show sensitivity to small changes in initial conditions. This is reminiscent of simple systems that display chaotic behavior, like the double-pendulum. For larger values of M2/M1M_2/M_1 the numerical solutions are robust in the sense that small changes in the initial angle (or other parameters of the model) lead to small changes in the trajectory and quantities like thaltt_{\mathrm{halt}}. But as we decrease M2/M1M_2/M_1 these “nearby” solutions often describe qualitatively different behaviors. Figure 9 shows an example of

Image not extracted

Figure 9. The simulated behavior of y2y_2 and ℓ\ell for M2/M1=2.5M_2/M_1 = 2.5 and initial angles 75∘75^\circ, 75.1∘75.1^\circ, 75.2∘75.2^\circ, and 75.3∘75.3^\circ suggests sensitive dependence on the initial conditions.

this for M2/M1=2.5M_2/M_1 = 2.5 (M2=38M_2 = 38 g, M1=15.2M_1 = 15.2 g). Beginning at an initial angle of 75∘75^\circ and incrementing in steps of 0.1∘0.1^\circ, the solutions exhibit the same initial behavior before branching out into a wide range of halting times. It is not clear yet whether this is a genuine feature of the looping pendulum or an artifact of the numerical methods used to solve the equations of motion. But we suspect that the looping pendulum may exhibit chaotic behavior, and we are currently investigating this possibility.

Acknowledgments

We wish to thank Loyola University Chicago for facilitating lab access during the COVID-19 pandemic shutdown, and the Department of Physics for providing the materials and work space used in the experiments.

A Data

Table I shows the observed and predicted times for the local extrema of x1x_1 and y1y_1 in Figure 7c and Figure 7d, respectively. Times in the texpt_{\mathrm{exp}} columns were found by identifying points with the largest local values (positive or negative) of x1x_1 and y1y_1 in the data. The values in the tnumt_{\mathrm{num}} columns were obtained by evaluating the derivatives x˙1\dot{x}_1 and y˙1\dot{y}_1 for the numerical solution and finding their roots. As explained in the text, the numerical solution is in excellent agreement with the data. Table II records the times and positions for various events experienced by mass M2M_2 as shown in Figure 8b.

Motion of M1M_1 After Halting

Once the heavier mass halts, the shape of the path followed by the lighter mass is completely determined by the string wrapping around the rod. The shape of this spiral can be described using the polar coordinates (ρ,ψ)(\rho,\psi) shown in Figure 10, where ρ=R2+ℓ2\rho= \sqrt{R^2+\ell^2} is the distance from the center of the rod to M1M_1 and ψ=θ+arctan⁡(ℓ/R)\psi= \theta+ \arctan(\ell/R) is its angular position. Let ℓhalt\ell_{\mathrm{halt}} and θhalt\theta_{\mathrm{halt}} be the values of ℓ\ell and θ\theta, respectively, when M2M_2 halts. Since hh is now

Image not extracted

Figure 10. After M2M_2 has halted, the subsequent path of M1M_1 is determined by the fixed length of string ℓhalt\ell_{\mathrm{halt}} looping around the rod. The polar coordinates (ρ,ψ)(\rho,\psi) shown here are used to describe this path.

constant, L−h=ℓ+Rθ=ℓhalt+RθhaltL-h=\ell+R\theta=\ell_{\mathrm{halt}}+R\theta_{\mathrm{halt}} and the shape of the curve is

ρ(ψ)=R2+(ℓhalt+Rθhalt−Rθ(ψ))2,(25)\rho(\psi)=\sqrt{R^2+\left(\ell_{\mathrm{halt}}+R\theta_{\mathrm{halt}}-R\theta(\psi)\right)^2}, \tag*{(25)}

where θ(ψ)\theta(\psi) is defined implicitly by the transcendental equation

ψ=θ(ψ)+arctan⁡(ℓhaltR+θhalt−θ(ψ)).(26)\psi=\theta(\psi)+\arctan\left(\frac{\ell_{\mathrm{halt}}}{R}+\theta_{\mathrm{halt}}-\theta(\psi)\right). \tag*{(26)}

The value of θ\theta when M1M_1 reaches the rod (ℓ=0\ell=0) is θhalt+ℓhalt/R\theta_{\mathrm{halt}}+\ell_{\mathrm{halt}}/R, so ψ\psi takes values in the range

θhalt+arctan⁡(ℓhaltR)≤ψ≤θhalt+ℓhaltR.(27)\theta_{\mathrm{halt}}+\arctan\left(\frac{\ell_{\mathrm{halt}}}{R}\right)\leq\psi\leq\theta_{\mathrm{halt}}+\frac{\ell_{\mathrm{halt}}}{R}. \tag*{(27)}

The values ℓhalt\ell_{\mathrm{halt}} and θhalt\theta_{\mathrm{halt}} depend on the dynamics of the system, and for some initial conditions halting may not occur at all. But once M2M_2 has completely stopped moving, the shape of the curve traced out by M1M_1 is geometrically determined by the string looping around the rod. As an example, Figure 11 reproduces this path for M1M_1 in the M2/M1=10M_2/M_1 = 10 trial shown in Figure 7a.

Figure 11. After mass M2M_2 halts in Figure 7a, mass M1M_1 follows the spiral described by Eqs. (25)–(26) with ℓhalt=0.196 m\ell_{\mathrm{halt}} = 0.196\ \mathrm{m} and θhalt=4.738\theta_{\mathrm{halt}} = 4.738.

Once M2M_2 has completely halted we can also give an approximate solution for ℓ(t)\ell(t) and θ(t)\theta(t) in the case where air resistance is negligible. As above, ℓ+Rθ=ℓhalt+Rθhalt\ell+ R\theta= \ell_{\mathrm{halt}} + R\theta_{\mathrm{halt}} is constant and therefore ℓ˙+Rθ˙=ℓ¨+Rθ¨=0\dot{\ell} + R\dot{\theta} = \ddot{\ell} + R\ddot{\theta} = 0. Then the t^1\hat{\mathbf{t}}_1 component of the equation of motion for M1M_1 becomes

ℓ˙θ˙+ℓθ¨=gsin⁡θ,(28)\dot{\ell}\dot{\theta}+\ell\ddot{\theta}=g\sin\theta, \tag*{(28)}

which can be rewritten as

ℓ˙2+ℓℓ¨=−gRsin⁡(θhalt+ℓhaltR−ℓR).(29)\dot{\ell}^{2}+\ell\ddot{\ell}=-gR\sin\left(\theta_{\mathrm{halt}}+\frac{\ell_{\mathrm{halt}}}{R}-\frac{\ell}{R}\right). \tag*{(29)}

If we ignore the right-hand side of the equation then the solution from the time thaltt_{\mathrm{halt}} when M2M_2 halts until the time tendt_{\mathrm{end}} when M1M_1 reaches the rod is

ℓ(t)=ℓhalttend−ttend−thalt(30)\ell(t)=\ell_{\mathrm{halt}}\sqrt{\frac{t_{\mathrm{end}}-t}{t_{\mathrm{end}}-t_{\mathrm{halt}}}} \tag*{(30)}
θ(t)=θhalt+ℓhaltR−ℓhaltRtend−ttend−thalt.(31)\theta(t)=\theta_{\mathrm{halt}}+\frac{\ell_{\mathrm{halt}}}{R}-\frac{\ell_{\mathrm{halt}}}{R}\sqrt{\frac{t_{\mathrm{end}}-t}{t_{\mathrm{end}}-t_{\mathrm{halt}}}.} \tag*{(31)}

For the cases we have considered these expressions differ from the numerical solution of the system by no more than a few percent. This suggests that an analytic solution of the full equation Eq. (29) can be obtained perturbatively. However, our experiments show that the effects of air resistance are important for capturing the dynamics of the system, so we will not pursue this further.

References

References

No references were extracted for this paper.

Paper details

vlnc:W_h7q233ex
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents