An approximation of the acceleration of gravity using experimental methods with period, amplitude, and length dependencies

Michael Fang December 7, 2012

Introduction
Standard gravity g is used in almost every problem governed by classical mechanics. g is easily found using Newtons law of universal gravitation for a given mass and radius [1]. Instead, we will design an experimental method since Newtons law applies only to perfectly spherical bodies. We can nd a value for g through experimentation using a simple pendulum. The idea for this was rst thought of by Francis Bacon [2], who wanted to see if the force of gravity varied at dierent heights. He thought that if the pendulum had a shorter period at low elevations and a longer period at higher elevations, then the Earth must have an attractive force. For small amplitudes, we use the proportionality of the pendulums period T with the square root of its length l and g, which is given by T = 2 l/g. Despite the beautifully simple relationship, the derivation itself is slightly complicated and the relationship becomes messier for large amplitudes. Nevertheless, the derivation allows us to determine g more accurately and precisely than Newtons law of gravitation. Figure 1: A free body diagram used to derive the motion of a simple pendulum. A pendulum with mass m and length l at an angle with respect to the direction of the gravitational force on the mass (W = mg) has a tangential force (FT = ma = ml) as well. For small angles of , we can approximate sin . This gives us a simpler dierential equation. d2 g + =0 dt2 l (2)

Theory

Solving this dierential equation, we obtain the angular The motion of a simple pendulum can be formulated velocity = g/l. We can use to nd the period of by using a free body diagram, as shown in Fig. 1 [3]. the pendulum knowing that the period T is 2/. And Equating the tangential force FT to the gravitational thus, force W , we obtain l T0 = 2 (3) g d2 W = mg sin = ml 2 = FT (1) where T0 denotes the period for small angles. dt For large angles, we can solve (1) by Taylor expanding for mass m, length l, and angle as shown in Fig. 1. the sin term to the third order. We can linearize our 1

approximation by replacing our third order term with k for some k between 0 and 1 [4]. Since our dierential equation is now linear, we are able to solve (1) and nd T , which can be written with = max , the amplitude. So we now have

The amplitude of the pendulum swing was varied in 10 degree increments from 10 degrees to 80 degrees. The length and mass were held constant at 20 0.1 cm and 301.2 0.1 g respectively. The amplitude was set 2 to 3 degrees above the desired reading on the protractor to counteract the eects of string friction and air resistance. 1 2 k2 k2 T0 1 + + (4) This precaution allowed the pendulum to decay to a nal T = T0 1 6 12 amplitude 2 to 3 degrees below the initial amplitude after The value of k is found to be 3/4 for an optimal approx- 11 swings. It is reasonable to claim an uncertainty of 1 degree since negative errors cancel the eects of positive imation [4]. The result is errors. The period of the pendulum was measured for 10 trials at each amplitude with 10 swings in each trial. 1 2 T T0 1 + (5) The timer was set to start after the rst complete swing 16 and to end after the eleventh swing was complete.

The length of the pendulum string was then varied by adjusting the secondary support beams closer or further We need to vary three parameters in this experiment: away. The period of the pendulum was measured with the amplitude , length of the string l, and the mass the same procedure described previously (10 trials with m. Although mass is theoretically independent of the 10 swings) at the following lengths: 16, 20, 25, 30, 36, period, we will perform a check experiment to validate 41, 52, and 62 cm with an uncertainty of 0.1 cm. The this theory. amplitude and mass were held constant at 20 1 degrees and 301.2 0.1 g respectively. A larger angle of 20 degrees was used to minimize the factional uncertainly to < 1 % but we will have to account for the extra factor of (1 + 2 /16). A check experiment was done with 5 masses, each made of dierent materials. The mass and material for each trial consisted of: 9.2 g of cork, 22.2 g of wood, 23.7 g of aluminum, 29.6 g of brass, and 296 g of iron. The length and amplitude were held constant at 50 .5 cm and 20 1 degrees respectively. There was more uncertainty in my length since the string had to be retied as mass was varied. The secondary support beam could not be used to make precise adjustments. The previous procedure was repeated yet again. Figure 2: The set-up for this experiment consists of a mass tied a string suspended by a support bar with a protractor to measure the angle relative to vertical. A second, movable support bar is for the other end of the string, so that length can varied precisely and easily by simply moving the support system. There are many opportunities for systematical errors in this experiment. For example, our pendulum may not be restricted to only one degree of freedom. Due to Earths rotation, precession may become apparent after 11 swings. We would need to experiment at the equator to reduce such eects. Air resistance is also a factor when the surface of the mass is not smooth e.g. The setup, as shown in Fig. 2, consisted of a mass a cork will experience more drag than a marble. Thus suspended by shing wire which was axed with tape to data for a mass with a smoother surface should have a support beam. A large protractor was also attached more statistical weight in our analysis. With the use to the support beam such that we could read the an- of a hand held timer, the error can be large. However, gle between the force vector of gravity and the string at with consistent timing techniques and many trials, the its maximum angle. The opposite end of the string was standard deviation of timing between each trial was wound around an adjacent, movable, secondary support reduced to less than 0.07 seconds. beam so that the length could be varied precisely without having to retie the string. Length was determined by measuring from the pivot point of the string to the center of the mass. Mass was determined using a scale accurate to 0.01 grams and the amplitude was read o the protractor attached to the support beam. 2

Experimental Methods

Results and Analysis

Amplitude

Error
To determine error propagation in the form q = xn for our data x with our fractional uncertainty x [5], we use the following: x q = |n| (6) |q| |x| For more variables, we can generalize error propagation for q = (xn1 ) (xn2 ) . . . as the following: 1 2 q = |q| n1 x1 |x1 |
2

+ n2

x2 |x2 |

(7)

To determine a good representation of the period, the mean of the trials was determined and divided by 10 to account for each trial. To determine the precision of our data, the standard deviation of the sample represented by the 10 trials was divided by 10 to give us the standard deviation of the mean. This was done for the measured period at varying amplitude , length l, and mass m as shown in Fig. 3, 4, and 5 respectively. When amplitude and period are squared for respective plots in Fig. 3 and 4, the standard error of the mean for the relevant parameter is increased by a factor of two as shown by (6).

Figure 3: Plot of period (T) versus amplitude squared (2 ) for varying amplitudes. The amplitude was incremented 10 degrees (.1745 rad) from 10 to 80 degrees (1.396 rad). Length l and mass m were held constant at 20 0.1 cm and 301.2 0.1 g respectively. The reason for plotting our data in this way is based on (5) to obtain a linear relationship. An intercept is shown to determine the period at an amplitude very close zero. Period would be nearly impossible to measure accurately for a very small amplitudes. 1- error bars were used for period where is the standard deviation from the 10 trials. 2 ( 1) degree error bars were used for amplitude to account for error propagation (see (6))

We can use Fig. 3 determine a value for g. T0 is dened in (3) as the small angle approximation of the period. Experimentally, measuring the period for very small amplitudes would have been a challenge to do accurately. However, since our t is linear, we can extrapolate the period for an amplitude that is virtually zero. Thus we can assume that T0 is the intercept of the t line from Fig. 3 when is exactly zero (see (5)). From the linear t, the value of the intercept is determined to be 0.897 0.002 seconds. With our intercept, which shall be denoted as I, we can validate g with l = 0.20 0.001 m. So from (3) we have l I2

g = 4 2

(8)

which gives us 9.808 0.069 m/s2 . This interval contains the expected value of 9.80665 m/s2 as dened by International System of Units. 3

Length

Mass

Figure 4: Plot of period squared (T 2 ) versus length (l) for varying lengths. The lengths used are: 16, 20, 25, 30, 36, 41, 52, and 62 cm with an uncertainty of 0.1 cm. Amplitude and mass m were held constant at 20 1 degrees and 301.2 0.1 g respectively The reason for plotting our data in this way is based on (3) to obtain a linear relationship with respect to length. Our t line intersects the origin, meaning that there is minimal systematic error and T 2 = l for a constant . The error of our dependent variable, T 2 , is small so 3- error bars were used where is the standard deviation from the 10 trials. The propagation of this error was calculated with (6). Error bars were not shown for length since the fractional uncertainty was < 1%.

Figure 5: Plot of period (T ) versus mass (m) with varying mass. Each pendulum had a dierent mass and material as follows: 9.2 g of cork, 22.2 g of wood, 23.7 g of aluminum, 29.6 g of brass, and 296 g of iron. Length l and amplitude were held constant at 50 .5 cm and 20 1 degrees respectively. The resulting plot is constant, even with a very heavy mass. The plotted constant line is determined by the mean of the periods. However, there is a slight deviation from the constant line for the 9.2 g cork mass. This might be due to aerodynamic drag induced from the porous surface of the cork. 1- error bars were used where is the standard deviation from the 10 trials.

We can also use Fig. 4 to approximate the value of g. To do this, we use the fact that there is a linear relationship between the square of the period and the length of the pendulum. The linear t of Fig. 4 gives a slope of 4.116 0.032, which will be denoted . The square of the period can be expressed in the following way: 4 2 T = l = g
2

As shown in Fig. 5, the period of the pendulum is independent of mass. The smallest mass was a porous cork ball, which may explain the slight deviation from the t constant due to increased aerodynamic drag on the surface of the ball. The low density of the cork also causes the buoyant force of the surrounding air to have a greater eect of the acceleration.

2 1+ 16

Solving for g with = 20 1 , we obtain, 4 2 2 16

The theoretical period for l = 0.5 0.005 m and = 30 1 given by (5) is 1.443 0.015 seconds. The (9) mean period from the data was 1.444 0.041 seconds, which contains the theoretical value. Thus, the period is independent of mass, as the experiment shows and theory suggests. (10)

g=

1+

which gives us 9.738 0.088 m/s2 . This interval contains the expected value of 9.80665 m/s2 as dened by the International System of Units. 4

Conclusion
We have found a consistent value for g by varying two independent variables: the amplitude and length l of the pendulum. The fractional uncertainties for the approximated values of g are relatively small (< 1 %). Both intervals generated by their uncertainties also contain the value of standard gravity as dened by the International System of Units. This proves the reliability of the period of a simple pendulum to determine g. The mass independence of period has also been validated experimentally. To obtain a better approximation for g, we need a higher order approximation. Our experimental set-up needs to be dierent as well. To eliminate the eects of precession, we must conduct our experiment at the equator. However, since we are at a dierent location, we need to remember that the value of g will be inconsistent with our current experiment. To eliminate the eects of drag we need to preform our experiment in a vacuum chamber. To eliminate the eects of timing error, a laser should be used to accurately measure the period of the pendulum. With these experimental redesigns, we should be able to redetermine g with higher accuracy and precision.

Acknowledgments
The work presented has beneted from John M. Martinis, professor at the University of California, Santa Barbara, who provided invaluable feedback and advice for the work summarized here. He has pushed me further down the path to a career in physics. Zijun Jimmy Chen, graduate student at the University of California, Santa Barbara, has put in the time and eort for detailed comments that aided the cause of writing a better paper.

References
[1] Sir Issac Newton, Philosophi Naturalis Principia Mathematica (London, 1647). [2] The Works of Voltaire: A Contemporary Version (New York: Dingwall-Rock, Ltd., 1927), Vol. XIX, Part II, pp. 27-33. [3] Daniel Kleppner and Robert J. Kolenkow, An Introduction to Mechanics (Cambridge University Press, Cambridge, 2010), pp. 255-256, 276-278. [4] L.H. Cadwell and E.R. Boyko, Linearization of the simple pendulum, Am. J. Phys. 59, 979-981 (1991). [5] John R. Taylor, An Introduction to Error Analysis, (University Science Books, Sausalito, 1997), pp. 6669.