T HE D IFFRACTION OF L IGHT

A P REPRINT

Clayton Williams and Dustin Shipp

Department of Physics
Utah Valley University

July 18, 2019

A BSTRACT

A draft for the Diffraction chapter for PHYS 3115 Introduction to Experimental Physics I. This
experiment introduces light diffraction and interference as measurement tools and some basic error
analysis techniques. Throughout the experiment, students should consider what else these tools could
be used for.

c 2019 Utah Valley University

1 Background

1.1 History

The beginning of the 19th century saw a dramatic conclusion to a debate dating from Newton’s time on the nature of
light when the English physicist Thomas Young established with his double-slit experiment that light can behave in a
wave-like manner. Since Christiaan Huygen’s work on waves in the 17th century it was well established that waves
diffract around obstacles which impede their motion, and that waves undergo interference when they combine. Most
interestingly to scientists of the time, it was found that the degree of diffraction depends on the wavelength of the
incident light and that light often travels in a seemingly particle-like way due to the very short wavelength of visible
light. It was also discovered by Jacques Babinet that complementary apertures produce diffraction patterns which are
identical except for their phase.
Until Young’s time most optical experiments used prisms, lenses, and mirrors to refract and redirect light; however, his
results on diffraction established the use of another optical instrument in the laboratory, that of the diffraction grating.
This instrument proved very useful in the hands of Gustave Kirchhoff, who combined diffraction gratings with prisms
to separate closely spaced spectral lines and helped establish the science of elemental spectrometry. Fundamental to
Kirchhoff’s results was the use of a Bunsen burner, invented by Kirchhoff’s friend Robert Bunsen [1], which isolated
sources of background light. Kirchhoff also attempted to mathematically derive Fesnel’s results, with mixed success:
while Kirchhoff’s methods and approximations are insufficient to apply to the study of electromagnetism in general,
they work exceptionally well for visible light and are the basis for most studies in diffraction [2]. Indeed, Babinet’s
result on complimentary apertures can be derived immediately from Kirchhoff’s theory.
Later, in the 20th century, it was discovered by pairing wave-particle duality in quantum theory with the wave function of
Erwin Schrödinger that repeating Young’s double-slit experiment with single photons or even electrons, fired individually
and spaced over long periods, can yield diffraction patterns identical to those of a continuous beam of light provided
information about the initial conditions of the system remains sufficiently obscure. Conventional interpretations posit
that in these experiments it is the wavefunction of the particle which is diffracted, and indeed treating the system in this
way yields results which are mathematically and experimentally consistent. This is particularly interesting because of
the collapse of the diffraction pattern when certain information about the system is acquired, and the diffraction pattern
can even be restored when this information is destroyed. What these results mean is still mysterious and epistemological
debates over determinism and the nature of quantum mechanics ignited by these experiments continue to this day.
1.2 Physical Principles

1.2.1 Huygen’s Principle

Naturally you are already somewhat acquainted with the principles of interference, that when two waves meet the
resultant wave is the sum of the amplitude of the component waves at each point, resulting in both constructive and
destructive interference. An understanding of this principle, combined with basic ray-tracing techniques, allows one to
accurately predict diffraction patterns without much of a complicated mathematical apparatus. Indeed, as J.D. Jackson
notes in his Classical Electrodynamics, diffraction is a second order approximation for electromagnetic waves, with ray
tracing being the first order. Calculations within the diffraction framework are achieved by the same techniques used
in geometric optics save one considers points where ‘waves’ or rays meet with path lengths differing by odd whole
number multiples of λ/2 (e.g. λ/2, 3λ/2, etc. with λ being the wavelength) as intensity minimums and points with path
lengths differing by multiples of λ as intensity maximums. When combined with Huygen’s Principle, that all points on
a spherical wavefront may be considered to be the source of a spherical wave, this second order approximation becomes
a powerful tool. Let’s look at this in more detail.

1.2.2 Single-Slit Diffraction

Suppose we have a beam of light of wavelength λ incident on a slit of width a, and we are attempting to calculate
the location of intensity maxima on a screen. Let the height of the maximum above the center of the slit be y and the
distance between the slit and the screen be D. Furthermore, we desire to use the small angle approximation for the angle
θ between the perpendicular from the slit to the screen and y be small. See figure 1. By Huygen’s Principle we must
treat each point on the wavefront as being the source of a spherical wave; in particular, we will focus the two points
on the edges of the slit 1 . We can consider only these points because our slit is rather narrow, and by treating these 2
most distant points we can characterize the interference pattern without much loss in accuracy. Then the difference in
path length between these two points is δ = d − d0 . Note in particular the angle from the central length to the point y,
namely θ, is nearly θ ≈ tan θ = y/D.

Figure 1: Single-Slit Diffraction Setup

Now because of our small angle approximation we see that the rays from our two points on the slit are nearly parallel, at
least near the slit. Then the triangle formed by projecting the shorter ray onto the longer one is a right triangle with the
interior angle θ0 being approximately θ, so θ0 ≈ θ 2 . From our small angle approximation then tan θ ≈ θ, tan θ0 ≈ θ0 ,
and hence y/D = δ/a so y = Dδ/a. Now since y is a maximum the difference in path lengths is a whole number of

1
Justify this. Points near the center of the slit combine destructively with wavelets generated on the opposite side of the central
line. Let the height of a point above the center of the slit be x. Now use a parallel ray approximation to show that | x |≤ d/4 there
is a point opposite the central line a distance x and theese points generate wavefronts which combine destructively, so that the
difference in their path lengths to y is mλ2d
when the conditions for equation 1 hold.
2
Prove this using the theorems for parallel lines crossed by a transversal (hint: you’ll be constructing 2 right triangles).

2
wavelengths, so rays originating from the top and bottom of the slit combine constructively 3 . Hence δ = d − d0 = mλ
for some m a natural number, and

mλD
y= , m ∈ N. (1)
a

If we plot light intensity as a function of position on the screen, then positions of minima and maxima become quite
apparent, as shown in the experimentally obtained data in figure 2. Note that the greatest intensity is at the m = 0 order
maxima, in the center of the diffraction pattern.

Figure 2: Single-Slit Diffraction

Now if it seems that this calculation required an excessive amount of approximation, recall that the key assumption (and
essentially our only assumption) was the small angle approximation from which all our calculations followed. The
approximation sin θ ≈ θ is accurate to less than 5% error up to 30◦ = 0.53 rad, and hence we can expect that for low
order maxima the main source of error in our experiment will not be that due to the small angle approximation 4 . If you
still aren’t convinced, complete the diffraction experiment to determine exactly how far our theoretical calculation is
from reality! 5 .

1.2.3 Double-Slit Diffraction

The setup for double-slit diffraction is quite similar to that for single-slit diffraction. The main difference in our
presentation is the use of the variable d to refer to the displacement between slits and variables r, r1 , and r2 to refer to
lengths between the slit and screen.
3
Draw a diagram of wave interference convincing yourself that if the path length differs by a multiple of a whole wavelength
the result is constructive interference, while if the path length differs by a multiple of a half wavelength the result is destructive
interference.
4
To motivate the small-angle approximations, note the Taylor series for f (x) = sin x = x − 61 x3 + ... ≈ x, for small values of
x. Indeed, x3 → 0 more quickly than x → 0. A similar case is true for h(x) = tan x. Plot x, sin x, and tan x on the same graph
near x = 0 to visualize this. It is worth noting that the small angle approximation tan θ ≈ θ is less accurate than the approximation
sin θ ≈ θ. Make a table demonstrating the relative errors of these approximations for small angles to convince yourself of their
accuracy and usefulness, and to gain an understanding of when these approximations can reliably be used.
5
Perhaps add a note on the difficulty of evaluating Kirchoff’s Integral.

3
 Figure 3: Double-Slit Diffraction Setup

The derivation of the location of maxima is somewhat more involved than for the single-slit case and left as an exercise
6
. Without deriving the relation, and simply knowing that for double-slit diffraction

d sin θ = mλ, m ∈ N, (2)

we can analyze double-slit diffraction in general. First, note that information about the location y of maxima can be
obtained from sin θ when D is known. Now recognize that two closely spaced slits each generate their own single-slit
patterns superimposed on the double-slit pattern. This superposition becomes apparent when viewing the double-slit
diffraction pattern on a screen (see figure 4). In the figure the large-scale structure of the pattern is a single-slit diffraction
pattern, with the finer details resulting from the double-slit interference. Indeed, double-slit diffraction patterns are
contained within single-slit envelopes, which you can confirm during the course of your experiment 7 . This means you
can use equation 2 to determine the separation of a double slit and equation 1 to determine the width of the slits, once
you determine the location of a single-slit maxima.

6
To derive the location of the maxima, use the small angle approximation for θ as before, noting this makes r1 and r2
approximately parallel and equal to r near the slits. Reference the setup in figure 3 and use the law of cosines on the triangles there,
along with the relation r2 − r1 = δ, to determine δ and note that for constructive interference δ = mλ.
7
In general (n + 1)-slit diffraction patterns are contained in n-slit diffraction patterns.

4
 Figure 4: Double-Slit Diffraction

1.3 Why do we care?

Interference and diffraction are most apparent when light propagates through apertures or around objects (Babinet’s
Principle says these are equivalent cases) that are very small (usually measured in micrometers). It becomes much
easier to measure the fringe patterns (millimeters in size) than to measure the aperture or object itself. This becomes a
powerful tool in many applications of microscopy and nanoscience. What applications have you heard of?

1.4 Error Propagation

Naturally an essential skill when obtaining, analyzing, and publishing data is the proper handling of uncertainty. Despite
our conventionally negative attitude towards the term ‘error’, it is inevitable in experimentation and, when sufficiently
minimized, is essentially synonymous with uncertainty. Error or uncertainty in one measurement may have an effect
on the overall experiment which is greater than the uncertainty in just that measurement. This is error propagation.
To demonstrate error propagation, consider the uncertainty in the measurement of the circumference of a circle, say
of circumference C and radius r. By taking a differential (implicit differentiation) on the relation C = 2πr, we have
∆C = 2π∆r, and hence the error in r is approximately ∆r = ∆C/2π. Now for the area of the circle, however,
A = πr2 results in an error of ∆A = 2πr∆r = 2πCδC, and scales with the value of r (and hence your measurement
of C).
This process is rather simple when measuring simple properties of circles, but not as intuitive in other experiments.
For example, what happens when the differential yields a negative value? Or when there are multiple sources of error
in a single measurement? To resolve this conundrum it is standard to add errors in quadrature. If an experimenter is
determining a value y which is a function of variables x1 , x2 , ..., xn the uncertainty in y(x1 , x2 , ..., xn ), called σy , can
be approximated by
 2  2  2
∂y ∂y ∂y
σy2 ≤ σx21 + σx22 + ... + σx2n , (3)
∂x1 ∂x2 ∂xn
where σxi is the uncertainty in the measurement xi and the function is evaluated at the mean for each of xi .
In practice it can be somewhat tricky to determine what uncertainty to use for each σx . Ideally each piece of equipment
will have an attached statement with a sensitivity rating; however, this is often not the case. It will often be up to your

5
personal discretion to determine what value to use for σx , however, common sense and good practice dictates that
oftentimes 1/2 the value of the most accurate decimal place on your measurement is a wise choice 8 . For example,
with a ruler that measures up to millimeters the uncertainty would be ±0.5 mm. And, it is important to note that when
confronted with two reasonably valid choices for the uncertainty it is best practice to pick the larger value.
Now, when designing and conducting experiments it is important to be aware of two kinds of errors: systemic errors
and random errors. A systemic error is one which is due to the experimenter or system performing the experiment,
and is consistent through each trial. Being aware of these kinds of error and minimizing them is a critical part of
experimentation. An example of systemic error may be a difference in delays between sensors receiving a signal and
the recording of those signals by a computer. Oftentimes systemic errors can be dealt with by careful examination of
your apparatus; however, systemic errors will always persist in any experiment and it is important to note possible
sources of systemic error and their magnitude when reporting your results.
The other kind of error mentioned, random error, can be controlled for by simply increasing the number of trials run
during your experiment. These errors result from inconsistencies in setup and execution which cannot be controlled. For
example, while Hooke’s law for springs F~ = −k~x is incredibly accurate and predictable across an appropriate ranges
for x, multiple trials of the same experiment measuring the spring constant k will result in a predictably randomly
distributed distribution even for the same spring, which distribution is centered on the mean value for k.
The formula for the standard error of the mean (standard deviation) is presented in equation 4
v
uN
uX xi − x̄
σ=t , (4)
i=1
N −1

which describes the spread of data from the mean x̄ due to random error and is optimized when σ is small, tells us much
about general ways to improve experiments 9 . Most importantly, it tells us that the influence of random errors decreases
with a larger number of samples N, and that the standard deviation is undefined for a sample size of 1. Indeed, this
indicates to us that when we run an experiment a single time we have no useful information about the size of the random
error and we may conclude that any result obtained from such a shallow experiment is invalid. But, by increasing the
number of trials the standard deviation decreases, and with the correspondingly smaller influence of random error on
our our data the results are more reliable.

1.4.1 Reporting Your Results

When CERN reported the landmark discovery of the Higgs boson in 2012 they did so with the exemplary caution that
characterizes the scientific community when publishing results of any kind. This can be seen in their report, Observation
of a new particle in the search for the Standard Model Higgs boson with the ATLAS detector at the LHC [3]. It is
remarkable that nowhere in the article do they state that they have definitively discovered the Higgs boson, and even
in the title the most they claim is that they had discovered a new kind of particle. While the particle they discovered
seemed to have many of the characteristics of the Higgs boson, CERN researchers waited for consensus of the scientific
community that the particle observed is indeed almost certainly the Higgs before declaring their discovery in stronger
terms.
Diving more deeply into the CERN article we can learn much about the proper reporting of scientific results, applicable
both in landmark cases and in situations more modest. The report includes multiple Systemic Uncertainties sections, one
for each type of particle event described, in which we find an in-depth analysis of their experiment including background
events and possible complications confounding their results. This analysis is not just qualitative and includes a numeric
uncertainty associated with every claim, as well as a possible source of that uncertainty. For example, the article states
that:
“The dominant experimental uncertainty on the signal yield (±8% and ±11% [for the 7 TeV and 8 TeV data respec-
tively]) comes from the photon reconstruction and identification efficiency, which is estimated with data using electrons
from Z decays and photons from Z → l+ + l− + γ events.”
8
In some cases this method will result in an underreporting of error, especially if you’re using the default setting on your
measuring apparatus. For example, in some experiments the main source of error may be in human reaction time, which may find
it difficult to match the accuracy of a stopwatch with 0.1 s accuracy. In this case it may be best to state the uncertainty in your
measurement as 0.5 s instead of 0.05 s.
9
Note this σ, the standard deviation, is dimensionless and related to but not the same as the theoretical uncertainty in your
measurement σy in equation 3. It is unfortunate that these two are so closely related and use similar symbols, but this is the standard
usage.

6
Without going into the science behind the CERN experiment, we can still be inspired by their methods for accurately
reporting data. From this statement not only is their result clear (the 7 TeV and 8 TeV data) but also the uncertainty of
that data (±8% and ±11% respectively), potential sources of that uncertainty (photon reconstruction and identification
efficiency), and the method used to calculate that uncertainty (data from Z decays ). This is a statement not only of
their results but also of their confidence in those results.
When reporting your own results it is important to include as much relevant detail as necessary 10 . Oftentimes you will
be reporting the mean result from your data, and stating with that mean the standard deviation, number of trials, and
theoretical uncertainty from equation 3. If you do so you will be reasonably covered in case of controversy, so long
as you do not claim any discoveries which don’t follow from your data. Also, it is common for a calculator to return
many more decimal points than are appropriate for your experiment. Report only up to the accuracy given by your
theoretical uncertainty 11 . In general, it is best to simply restate what it is you actually measured as well as the method
of that measurement, in which case no one can take the results of your experiment away from you. As an example, the
measurement of a spring constant in the laboratory may be stated in the following manner:
“We measured the stiffness k of a spring by measuring the displacement of the spring from equilibrium when the spring
was hung vertically with one end fixed and various masses hung from the free end. The spring constant k was found to
be k = 0.50 ± .04 N m−1 (or k = 0.50(4) N m−1 ) with a standard deviation of σ = 0.043 using 10 differing masses
ranging from 0.1 kg to 5 kg Possible sources of error include inaccuracies in the masses used, which were measured by
a digital scale with 0.05 kg accuracy; or inaccuracies in measuring the displacement perhaps caused by imperceptible
gentle bobbing of the spring or by measuring along an angle not directly in line with the vertical of the spring-mass
system, which displacement was measured by ruler sticks with 0.5 mm accuracy.”.

2 The Experiment

2.1 The Apparatus

The main part of the experiment consists of a track on which various optical instruments are placed. In particular, these
instruments include a red 650 nm laser, which shines on apertures of various dimensions. The resulting diffraction
patterns are recorded by a light sensor, which is placed on a track of its own (the linear translator) which is perpendicular
to the path of the laser. Measurements along the linear translator are correlated with measurements of light intensity as
read by the light sensor, which data is interpreted by a Pasco Scientific 550 Universal Interface and recorded by the
software Pasco Capstone. Pasco Capstone allows the student to locate maxima and minima, and in doing so the student
will be able to corroborate or refute the results of equations 1 and 2.

10
The key word is relevant. Some reports, even professional ones, will include superfluous and irrelevant information such as the
temperature in the room, levels of ambient light, elevation above sea level, etc. Depending on the experiment this may be pertinent
information and it will be up to your judgement to determine what to include. In general include all detail which would be practically
useful to some attempting to replicate your experiment.
11
So, for example, instead of reporting 13.54325 ± .05 eV report 13.54 ± .05eV.

7
 Figure 5: Diffraction Apparatus

2.2 Performing the Experiment

We begin with assembling the apparatus.

1. Insert the nut under the frame of the linear translator into the slot on the track, slide the linear translator into a
desirable position, and tighten.
2. Insert the laser and aperture selector into the optical track.
3. Connect the laser and the Pasco 550 Universal Interface to power.
4. Connect the Pasco 550 Universal Interface to the computer via USB and the Rotary Motion Sensor (attached
to the linear translator) to the 1 and 2 ports on the Pasco 550 Universal Interface. Connect the light sensor to
the A port on the Pasco 550 Universal Interface.

Figure 6: Hardware Setup

8
Now to set up Pasco Capstone:

1. Open Pasco Capstone on the computer.

2. Click "Hardware Setup".
3. Click the 1 and 2 ports on the image of the Pasco 550 Universal Interface. Scroll down to select the rotary
motion sensor.
4. Click on the settings for the Rotary Motion sensor and make sure that the "Medium Pulley" is selected to
ensure correct calibration of rotation to distance.
5. Click the A port on the image of the Pasco 550 Universal Interface and scroll down to select the light sensor.
6. Exit hardware setup and open the graph. On the graph’s vertical axis select measurements by clicking on the
axis and selecting "Light Intensity".
7. Similarly, select "Position" for the horizontal axis.

To make measurements select the desired aperture on the aperture selector and press the "Record" button. Then slowly
drag the sensor from one side of the linear translator to the other. It may take several trial runs to learn how to smoothly
drag the sensor from one side to the other. Zoom and scale your graph as appropriate.

2.3 Analysis

Now that we have empirical data on our diffraction setup we can compare our results to those predicted by equations 1
and 2. Knowing the wavelength of the laser light is 650 nm and by measuring the distance from the aperture to the light
sensor, we can the locations of maxima y to measure the slit width a and distance between slits d.

1. For a single-slit aperture, use your plot of light intensity to determine the distance from the central maximum
to the first order maximum. What is the distance between these points?
2. Use equation 1 to calculate the slit width a. Does this seem reasonable?
3. Perform an error analysis for the single-slit experiment to calculate your predicted uncertainty in the slit width.
What contributes to this uncertainty and how can you combine these effects in your calculations?
4. See if your theoretical prediction falls within the error bounds of your experiment. How does this affect your
confidence in your measurement?
5. Think of ways that you could reduce the uncertainty in your measurements. Try them out now if you can!
6. Repeat this procedure for various aperture widths and repeat the entire experiment for the double slit experiment.
Can you confirm the double-slit diffraction pattern is bounded by the single-slit diffraction pattern of the same
aperture width? Are there any variables unaccounted for which affect your experiment?

Carefully collect your findings and record them in a dedicated laboratory notebook. Use this notebook to write a report
on diffraction summarizing your findings, including proper error analysis. Answer the following questions:

1. What do your data mean? Why?

2. How can this experiment or the analysis be improved?
3. What predictions can you make and verify in future experiments?

3 The Next Step

What else could you measure using light diffraction? Are there things that this procedure could measure that would be
difficult or impossible using other methods?
With these types of measurements in mind, describe another experiment you could perform. What questions could you
ask about the things you are measuring? What other equipment would you need? What measurements would you need
to make? Could any of the other experiments you have done also be useful?
If this plan seems feasible, it can be the beginnings of the proposal you write for this class. Your designed experiment
(measurements and analysis) should take you about 10 class periods. Do some background research to predict what
results you would get. Adapt your plan as necessary. Contact your instructor if you need equipment that is currently
unavailable. Funding has been obtained for student-designed experiments in the class in past semesters (prior results do
not guarantee a similar outcome).

9
References
[1] Isaac Asimov. Asimov’s Biographical Encyclopedia of Science and Technology. Doubleday and Company, Inc.,
Garden City, New York, 1982.
[2] John David Jackson. Classical electrodynamics. AAPT, 2 edition, 1999.
[3] Georges Aad, Tatevik Abajyan, B Abbott, J Abdallah, S Abdel Khalek, Ahmed Ali Abdelalim, O Abdinov, R Aben,
B Abi, M Abolins, et al. Observation of a new particle in the search for the standard model higgs boson with the
atlas detector at the lhc. Physics Letters B, 716(1):1–29, 2012.

10