You are on page 1of 22

BROWNIAN MOTION

Report Author: Lauren Honan

Report Submitted: 29 March 2020

Data Collected: 6 March 2020

Laboratory Partners: Jenna McGrath


Luis Ramos
Ervin Pilinci

Teaching Assistant: Mathew Kelley

Author’s Affiliation: Department of Chemistry and Biochemistry


University of South Carolina
Columbia, SC 29208

I pledge that my work meets the standard of the USC Honor Code.

ABSTRACT

This experiment is an approximation of Avogadro’s number using Einstein’s theories of


Brownian motion. Microscopic images of Polybead Polystyrene beads in Brownian
motion in water were taken every 10 seconds for 250 seconds. The displacements of
these beads were used to calculate Avogadro’s number using an unadjusted histogram
width and a corrected histogram width. Avogadro’s number was calculated to be
3.1847x1024[n]  1.4405x1027[o] unadjusted and 3.1807x1024[n]  1.7153x1027[o] corrected.
These approximations are associated with percent errors of 428.76% [p] and 428.09%[p]
respectively in comparison to Perrin’s value of 6.022x10 23 [1].

INTRODUCTION
This paper details the approximation of Avogadro’s number via the observation of
particles in Brownian motion. Brownian motion refers to the random movement of
particles suspended in a fluid [1]. Albert Einstein was able to quantify and explain the
phenomenon using microscopic pictures in 1905 [1]. The Brownian motion of a particle
is dependent on time, particle size, the temperature of the system, and the viscosity of
the fluid of suspension [2]. Because these are all measurable quantities, Avogadro’s
number can be determined from measuring them along with the mean-square
displacement of particles in Brownian motion [2].

My previous coursework at the University of South Carolina has prepared me for data
analysis performed in this lab. CHEM 111/112 provided me with basic chemical literacy
to understand how to use the laboratory equipment utilized in this experiment and
understand the basic chemical concepts. CHEM 332 provided me with the statistical
analysis skills required to interpret the data sets produced from this experiment.

Avogadro’s number was first determined by French physicist Jean Baptiste Perrin
multiple ways with incredible accuracy [1]. Perrin found Avogadro’s number using
Einstein’s theories on Brownian motion to be 6.022x10 23 [1].

In this experiment, Avogadro’s number was approximated by observing the Brownian


motion of Polybead Polystyrene beads suspended in water. Microscopic images of the
beads were taken every 10 seconds, and their displacements were analyzed through a
series of statistical approaches. Avogadro’s number was approximated from the data
set using an unadjusted histogram width and a corrected histogram width for particle
displacement due to Brownian motion. The unadjusted and corrected values of
Avogadro’s number were determined to be 3.1847x10 24[n]  1.4405x1027[o] and
3.1807x1024[n]  1.7153x1027[o]. These values are associated with percent errors of
428.76%[p] and 428.09%[p] respectively in comparison to Perrin’s value.

EXPERIMENTAL

A Parco BMT Series microscope (Parco Scientific. Plymouth, MI) was used in transition
mode to capture images of the movement of 0.51  0.01 m diameter Polybead
Polystyrene 0.5 Micron Microspheres (Polysciences, Inc. Warrington, PA) in solution
with water. 26 images were taken every 10 seconds (250 seconds total). FlyCap Ver
2.6.3.2 software (FLIR Systems. Wilsonville, OR) was used to display the images, and
the particle displacements were converted from pixels to microns using an Edmund
Optics USAF calibration slide (Edmund Optics. Barrington, NJ). All images were taken
at a constant temperature of 297.80  0.10 K measured using an Omega OM-DAQPro-
5300 thermocouple (Omega Scientific, Inc. Tarzana, CA).

RESULTS
The horizontal and vertical displacements of three randomly selected beads were
recorded between each image. These displacements are found in Table 1 below.

Displacement of Beads Over Time


Bead 1 Bead 2 Bead 3
Time (s) dx (pixels) dy (pixels) dx (pixels) dy (pixels) dx (pixels) dy (pixels)
10 -2 1 2 3 0 -2
20 -1 -3 6 5 1 -1
30 1 1 -4 2 10 5
40 2 1 2 3 -2 -1
50 1 -4 6 -5 -2 -2
60 1 -7 -2 -2 4 0
70 -1 2 0 -1 7 -5
80 -4 1 1 2 1 -1
90 -3 -3 0 4 2 -3
100 -1 -4 3 -4 1 3
110 2 1 -2 4 -2 2
120 3 2 4 2 1 1
130 -5 1 -1 -3 -2 -4
140 1 1 -2 2 -2 1
150 2 -4 1 4 1 -2
160 2 3 4 -2 -1 -2
170 -1 8 1 4 2 -4
180 6 -6 1 -4 -1 0
190 -5 0 2 -1 1 -1
200 -3 -1 1 1 -2 -4
210 -2 0 1 0 -1 -2
220 4 2 2 0 0 3
230 -2 -1 3 3 2 1
240 2 -1 3 -2 2 -2
250 -2 -2 0 2 -5 0
Table 1 – horizontal and vertical displacements of three beads per image.

The displacements contained in Table 1 are graphically represented in Figure 1 below.


Figure 1 organizes the displacements into a histogram that shows the most frequently
observed displacements were found between -2 and +2 in both the x and y directions.

Displacement of Beads in Horizontal and Vertical Directions


Figure 1 – histogram of x and y displacements of the beads.

DISCUSSION

Ideally, the random motion of particles in solution has a mean displacement of zero in
both the x and y direction because the distribution is symmetrical. Various factors
including convection and evaporation can cause the particles to become suspended,
thus causing the distribution to no longer behave symmetrically. Various statistical
analyses were performed to the data set to account for these deviations from ideal
behavior.

First, a t-test was performed testing the null hypothesis that the displacements in the x
and y directions have the same mean, or in other words, are centered around the same
value (ideally zero). The difference between the two means was found to be 0.76 pixels
displaced using the following equation:

1) xy = x - y [4]

The mean of x displacement was taken as 0.56  2.82 pixels and the mean of y
displacement was taken as -0.20  2.90 pixels. The variance of the difference between
the two means was then calculated to be 0.22 (pixels displaced) 2 using the equation:

1 2
2) vxy = (s + sy2) [4]
n x
The square root of this variance is thus the standard deviation of the difference between
the means. This value was found to be 0.47 pixels. Thus, the difference between the
means was 0.76  0.47 pixels. In order to accept the null hypothesis with 95%
confidence, the experimental t value must be less than the critical t value of 1.96[3]. The
experimental t value was calculated using the equation:

xy
3) texp = | | [4]
s xy

The results of this calculation gave an experimental t value of 1.63 [d]. Since the
experimental t value was found to be less than the critical t value, the null hypothesis
was accepted. Thus, the means in the x and y direction were centered around the same
value, and the displacements could be merged into one displacement data set for
further analyses. The important values of the t-test described above can be found in
Table 2.

t-Test of Mean x and y Displacements


xy 0.76 pixels[a]
vxy 0.22 pixels2[b]
sxy 0.47 pixels[c]
tcrit 1.96 [3]
texp 1.63[d]
Conclusion Accept null hypothesis
Table 2 – critical values for t-test of mean x and y displacement values.

Next, a confidence interval analysis was performed to determine if the x, y, and


combined x and y displacement sets were centered around zero. Confidence intervals
were calculated using the equation:

t sx
4) CI =  [4]
√n
The means of each data set and their corresponding confidence intervals are found in
Table 3. The critical t value was taken as 1.96 to determine 95% confidence intervals.

95% Confidence Intervals of x, y, and Combined x and y Displacements


Data Set Mean 95% Confidence Interval
x 0.56  0.64[e]
y -0.20  0.66[e]
x and y 0.18  0.46[e]
Table 3 – means and 95% confidence intervals for displacement data sets x, y, and combined x and y.

As seen in Table 3, zero is within the 95% confidence interval for each of the data sets.
Thus, it can be concluded that the ideal average displacement of zero in either direction
is within the distributions of the data sets. This further proves that the x and y sets could
be combined into one data set centered around the same value.
Next, the combined and x and y distribution was analyzed to determine if it fit a
Gaussian shape. This test for normality was based on the sample moments of the
distribution. The skew and kurtosis values were obtained from a histogram of the data
m3
using Igor. The third sample moment (√ b1= 3 /2 ¿ was used to test for skewness. A
m2
normal distribution has a √ b1 of zero. This value was calculated using the equation:

n
5) √ b1 = V_skew * √ n−1
[4]

V_skew was taken from the histogram on Igor to be 0.820. Thus, √ b1 was determined to
be 0.823[f]. For the sample size of 150, this value exceeds all critical values of √ b1
(0.322-0.519). Thus, it can be concluded that in the x direction, the data set is skewed
m4
to the right and does not fit normal distribution. The fourth sample moment (b 2 = 2 )
m2
was used to test for kurtosis. A normal distribution has a b2 of 3. This value was
calculated using the equation:

6) b2 = 3 + V_kurt [4]

V_kurt was taken from the histogram on Igor to be -0.889. Thus, b 2 was determined to
be 2.11[g]. Since this value is less than 3, it can be concluded that the distribution is
flattened in comparison to a normal Gaussian distribution. Furthermore, this value is
below all critical values of b2 for a sample size of 150 (2.24-4.34). Thus, it is not a
normal distribution. This non-normality could be a result of human bias in bead
selection. The results of the tests for skewness and kurtosis are collected in Table 4.

Test for Normality of Combined x and y Distribution


Variable Experimental Value 95% Critical Value Conclusion
√ b1 0.823[f] 0.322 Skewed to the right
b2 2.11[g] 3.65 Flattened
Table 4 – tests for skewness and kurtosis of the distribution of x and y.

The combined x and y data set was fit with a Gaussian model curve. This curvefit is
displayed in Figure 2 below.

Gaussian Fit of Combined x and y Data Set


Figure 2 – the combined x and y displacements fit with a Gaussian curve.

The results of the tests for skewness and kurtosis showed that the distribution of x and y
displacement did not behave in correspondence with a Gaussian fit. Therefore, the
width determined by a Gaussian fit histogram is greater than the actual width exhibited
by the data. Because it has been determined that the data does not display a normal
distribution, Sheppard’s Corrections were applied to the data set in order to receive a
better estimation of the width. The Sheppard Correction to the second moment is
defined by the equation in which h is the width of a group:

h2
7) m2,corr = m2,hist - [4]
12

The corrected m2 was thus 275.17 pixels2[h]. The width was then corrected from 16.60
pixels to 16.59 pixels. The standard deviation of the width was then calculated for both
the unadjusted width and the Sheppard Corrected width. To do this, first the fourth
moment was calculated using the equation:

8) m4 = (V_kurt + 3)*m22 [4]

The unadjusted fourth moment was found to be 160207.04 pixels 4[i] and the corrected
fourth moment was found to be 159819.34 pixels 4[i]. Then, the variance of the m2 values
were found using the equation:
m 4−m22
9) Vm2 = [4]
n

These values were 560.67 pixels4[j] for the corrected width and 562.03 pixels4[j] for the
unadjusted width. Then, the variance of the width was calculated using the equation:

vm 2
10)vs = [4]
4∗m2

For the unadjusted width, the variance was 0.5100 pixels 2[k] and for the corrected width,
the variance was 0.5094 pixels2[k]. The square roots of these values give the standard
deviation of the width. Thus, the unadjusted width was 16.5983  0.7141 pixels, and the
corrected width was 16.5883  0.7137 pixels. The important values in this standard
deviation calculation are provided in Table 5.

Determination of Width Standard Deviation


Unadjusted Corrected
2
m2 275.50 pixels 275.17 pixels2[h]
width 16.60 pixels 16.59 pixels
4[i]
m4 160207.04 pixels 159819.34 pixels4[i]
4[j]
vm2 562.03 pixels 560.67 pixels4[j]
vs 0.5100 pixels2[k] 0.5094 pixels2[k]
ss 0.7141 pixels 0.7137 pixels
Table 5 – the important values in determining the standard deviation of the widths.

Then, the widths were converted from pixels to microns using the conversion factor
determined from the Edmund Optics USAF calibration slide. The following equation was
used to determine the conversion factor:

1 mm 1000 um 1line pair


11)conversion factor = * *
228line pairs 1mm 27 pixels

Thus, the conversion factor was 0.1624 microns/pixel [l]. This conversion factor was used
to determine the widths and their respective standard deviations. The unadjusted width
was 2.6963  0.1160 microns[m] and the corrected width was 2.6947  0.1159
microns[m].

Next, the calculated width values were used to approximate Avogadro’s number. This
was done using the equation:

RT
12)NA = [2]
3 σ 2 ηr

In this equation, NA is Avogadro’s number, R is the gas constant 8.314 J/mol*K, T is the
temperature in Kelvin,  is the time between images in seconds,  is the calculated
width in meters divided by the square root of 2,  is the viscosity of water in kg/m*s, and
r is the radius of the beads in meters. The viscosity of water was taken as 0.891x10 -3
kg/m*s at 298 K [5]. Finally, a propagation of errors was performed on MapleSoft using
the following equation:

13)sNA = √ ¿ ¿ [3]

The approximations were compared to the known value of 6.023x10 23 using the percent
error equation:

¿
14) % error=¿ literature value −experimental value∨ literature value ∗100 % ¿ [3]

The results of these calculations are found in Table 6.

Avogadro’s Number Approximation


Avogadro’s Number Standard Deviation Percent Error (%)
24[n] 27[o]
Unadjusted 3.1847x10 1.4405x10 428.76[p]
24[n] 27[o]
Corrected 3.1807x10 1.7153x10 428.09[p]
Table 6 – approximation and standard deviation of Avogadro’s number using the unadjusted and
corrected widths.

It is evident that there are sources of error in the approximations of Avogadro’s number.
Some possible sources of error include that some of the particles that were measured
may have been stuck to the slide, and thus were not in Brownian motion. There were
several particles that had a displacement of zero or a very low number of pixels. This
could have been a source of error influencing the approximations. Additionally, air
current, evaporation, or convection may have altered the motion of the beads. This is
likely because it is evident in the Figure 2 that the histogram is not centered directly at
zero. Finally, any sort of impurity in the sample could have been mistaken for a bead
and included in the measurement even though it wouldn’t be in Brownian motion.
REFERENCES

1. Bian X, Kim C, Karniadakis GE. 111 years of Brownian motion. Soft Matter.


2016, 12(30), 6331–6346.
2. Chowdhury D. 100 years of Einstein’s Theory of Brownian Motion: from Pollen
Grains to Protein Trains. Resonance. 2005, 63-78.
3. Harvey D. Analytical Chemistry 2.1. 2016, 1045.
4. M. Kendall and A. Stuart “The Advanced Theory of Statistics, Volume 1, 4th Ed”,
Macmillan Publishing Co. (New York, 1977)
5. Crittenden J, Trussell R, Hand D, Howe K, Tchobanoglous G. MWH’s Water
Treatment: Principles and Design, Third Edition. 2012, 1861-1862.
SUPPLEMENTAL INFORMATION

Section 1: Sample Calculations


Section 2: Reference Title Pages
Section 3: Laboratory Notebook Pages

You might also like