The Photoelectric Effect: Determination of Planck’s Constant

Javier M. G. Duarte
Massachusetts Institute of Technology, MA 02142∗
(Dated: September 25, 2008)
We test Einstein’s theory of the photoelectric effect, which predicts a linear relation between the
energy of light incident on a metal surface and the energy of the emitted electrons. We determine the
electron stopping voltage as a function of frequency of incident radiation by illuminating a potassium
cathode and constraining the wavelength of incident radiation using several optical band-pass filters.
From this, we calculate the constant of proportionality h/e as well as the constant offset, known as
the work function. Our analysis included two different selection criteria to determine the stopping
voltage, Vcutoff , each of which led to a different estimate of Planck’s constant and the effective
work function, φeff . Our final results for both methods are h = (7.12 ± 0.88) × 10−34 J·s and
φeff = 2.20 ± 0.32 eV and h = (4.08 ± 0.88) × 10−34 J·s and φeff = 1.28 ± 0.32 eV .

INTRODUCTION given that the incident light has sufficient intensity.

On the contrary, there exists a longest wavelength,
In 1887, Hertz was the first to discover that a metallic specific to the metal, such that light of longer wave-
surface, when illuminated by light of sufficient frequency, length cannot liberate electrons from the metallic
may emit electricity [1]. This ‘photoelectric effect’ was surface.
unexplained until Einstein connected this experimental (3) Classically, an electron would not be emitted from
curiosity with Planck’s idea that radiation comes in small the surface until the incident light transferred
packets, or quanta [2]. He proposed that the energy of enough of its energy to the illuminated metal. This
the ejected electrons is proportional to the energy of the would require a delay from when the light first hit
incident light with a constant offset that is unique to the surface to when the first electron was ejected.
the metal, referred to as the work function [3]. This But, experimentally it is seen that the emission of
phenomenon was a crucial precursor to the formulation of electrons occurs very shortly after the arrival of the
quantum mechanics as it was one of the first to show the radiation.
wave-particle duality of light. We will examine this effect,
test the hypothesized linear relation, and extract values Motivated by these, Einstein produced one of his semi-
for Planck’s constant and the effective work function. nal papers in 1905, “A heuristic point of view concerning
the production and transformation of light,” which put
forth his law of the photoelectric effect. He extended
THEORY Planck’s notion that the energy of radiation comes in
chunks of the frequency multiplied by a constant, hν, to
There are several main features of the photoelectric light. In effect, he treated light as if it were composed of
effect that cannot be explained by the classical wave de- particles, or photons, each carrying an energy hν.
scription of light [4]: To explain the problem of (2) above, Einstein reasoned
that is a work threshold required for an electron to es-
(1) According to Maxwell’s equations, the magnitude cape the metal. The particular amount of work is a com-
of the electric field vector of a light wave goes plicated combination of the energy required for optical
like the√square root of the intensity of the light, excitation of a valence band electron into the conduction
~ ∝ I. Thus, as the intensity increases , the
|E| band as well as the energy required in the diffusion of the
electric field magnitude increases. Since the force photo-excited electron through the solid. Thus, the work
on an electron is proportional to the electric field is related to the electronic structure and lattice configu-
vector, it is expected that the kinetic energy of a ration of the metal among other things. As a result, this
photoelectron should increase with the intensity of work function is extremely dependent on the metal.
incident light. However, it is observed that the Further, he derived a beautiful relation from conserva-
maximum kinetic energy of the photoelectrons does tion of energy. In short, an ejected electron can have at
not depend on intensity. most the incident energy minus the total energy needed
to escape the metal.
(2) Classical wave theory predicts that photoemission
should occur at all wavelengths of incident light, Kmax = hν − φmetal (1)
In the actual experiment, we an expect the ejected
electrons to have energy less than this due to the ran-
∗ Electronic address: [email protected] dom processes involved in the photoemission: thermal
 2

vibrations, diffusion, multiple photon interactions, and current reading on the ammeter. Next, we incremented
scatterings. the applied voltage starting from 0.00 V over an appro-
priate range and recorded the current to achieve at least
seven data point pairs. We repeated this procedure five
EXPERIMENTAL SETUP times for each filter, for a total of 25 runs. All the data
was taken on the same day over a two hour period.
Design

DATA AND ANALYSIS

Anode Cathode (K)
Photon Source (Hg) Filter e-
Systematic Errors

Photons Two sources of systematic error that may confound the

e- data and affect our analysis include the spread in frequen-
Reverse Current cies permitted to pass through the filter and the negative
current. The former is a result of the reported error on
1.25 V 0.00 pA
the central wavelength of the band-pass filter: ±2.0nm.
Auxiliary Filters Voltage Supply Ammeter Originally, we considered this as a main contributer in
the nonlinearity of our data. However, further investi-
FIG. 1: The experimental setup including the high power gation, revealed that the spectral output of the mercury
Mercury lamp lamp is far too peaked for this to be a significant source
of error.
Our actual experimental setup had many of the same Over time, potassium residue builds up on the anode
features as the schematic shown in figure 1. For the first and light reflected from the cathode back onto the anode
component, the photon source, we used a high power may cause photoemission at the anode, resulting in elec-
mercury lamp with well documented transition lines [5]. trons being accelerated toward the cathode [6]. In effect,
We aimed this source through an optical narrow band- this means that a zero current reading doesn’t correspond
pass filter, which isolates a specific emission line. The to no photoelectrons reaching the anode from the cath-
light continues through an aperture and then an anode ode, but it indicates where the photocurrents from the
ring until finally striking the potassium coated cathode, cathode and from the anode cancel out. Therefore the
which acts as a photoemission surface. If electrons are true stopping voltage at which electrons with maximum
ejected, this may complete the electrical circuit and al- kinetic energy are being stopped is detected as a negative
lows us to detect a current with an ultra-sensitive amme- current. This complicates our analysis greatly, as we now
ter. A voltage power supply was used to apply a known need a Vcutoff selection method that takes this effect into
potential between the anode and the cathode, which may account.
manipulate the trajectory of the ejected electrons from Since the reverse current decreases monotonically with
the cathode. increasing positive retarding voltage, we expect that the
By adjusting the applied retarding voltage and record- data should have a characteristic linear tail.
ing the measured current, we may determine the poten-
tial difference required to stop the photoelectrons from
crossing the gap. This potential corresponds to the max- Data Reduction
imum kinetic energy of the electrons divided by the el-
ementary charge, displayed in equation ??. Utilizing Data reduction and analysis was done using Matlab
five different filters that select distinct mercury transition and linear fit scrips provided by the Junior Lab staff.
lines, we can take several data point pairs of (ν, Vcutoff ) Raw data of current versus voltage looked similar to that
and test the linear relation among incident photon energy of figure 2, with a characteristic linear tail below the
and maximum electron kinetic energy. zero current level as well as a departure from linearity.
However, as seen in figure 3, the 577.0 nm data never
actually crosses the zero current mark. In addition, the
Execution departure from linearity is much less pronounced in this
data set (keeping in mind the relative scales). Finally,
The apparatus was grounded properly and the path of the cutoff voltage, as determined by method II because
light was protected from ambient light background. For method I does not apply (explained later in this section),
each band-pass filter, i.e. wavelength of incident light, is completely off the linear structure that the other four
we first recorded the voltage corresponding to a zero net data sets follow so closely.
 3

FIG. 2: Method I: the simple selection criteria for the stop- FIG. 4: Method II: the double linear fit selection criteria for
ping voltage. The stopping voltages chosen in this way will the stopping voltage. The stopping voltages chosen in this
I II
be denoted Vcutoff . way will be denoted Vcutoff

I II
λ med-errorbar Vcutoff Vcutoff
365.0 nm ± 14.4 pA 0.24 ± 0.04 V 0.12 ± 0.04 V
404.7 nm ± 6.54 pA 0.80 ± 0.10 V 0.50 ± 0.10 V
435.8 nm ± 9.21 pA 1.20 ± 0.20 V 0.60 ± 0.20 V
546.1 nm ± 1.35 pA 1.50 ± 0.25 V 0.75 ± 0.25 V
577.0 nm ± 0.11 pA N/A N/A

TABLE I: These are the results of the data, the med-errorbar

is the median error bar for each plot of current versus retard-
ing voltage, taken at the 95 % confidence level.

compare this to figure 2. We know that the asymptotic

behavior of the curved data on each side of the range
should approximate a line. So we are merely, ‘flattening
FIG. 3: The data for the 577.0 nm transition is not in agree- out’ the curve to find where the departure from linearity
ment with the characteristic photoemission line. takes place.
For each wavelength, the mean and standard deviation
was calculated for each voltage bin; error bars were taken
We determined two distinct Vcutoff selection methods
at 95 % confidence levels, or 2σ. Furthermore, for both
based on the systematic errors we were trying to over-
Vcutoff selection methods, the standard error associated
come as well as the photoelectric hypothesis. The first,
with the cut off voltage was approximated to be the reso-
hereafter referred to as method I, consisted of recalling
lution of the current vs. voltage plot for that wavelength.
the voltage which gave a zero net current reading during
So, for example, for the 435.8 nm data, we used ±0.20
the experiment and selecting the first voltage bin to the
V. This data is summarized in table I.
right of it (the next greatest voltage measured at regular
interval). The method is portrayed in figure 2. We jus-
tify this method by again noting that the true stopping
voltage is actually detected as a negative current due to
the reverse current tail.
For the second selection technique, method II, we fit
the first three bins and the last three bins of data to Finally, the magnitude of the cut off voltage points and
separate lines by a χ2 minimizing procedure, and ex- their associated error bars were plotted against frequency,
trapolated. Then, we chose the voltage bin closest to figure 5, for both methods, and fitted to a line by least
the x-value (Voltage) of the intersection of the two lines. squares. The intercept, slope and respective error bars
Figure 4 displays method II on 435 nm data. You can correspond to φeff /e and h/e, respectively.
 4

criteria and they give you a wide range of final results.

Method I not only provides a better approximation to
Planck’s constant, but to the effective work function as
well. Assuming that our analysis of the 577.0 nm data is
correct, it is safe to assume that no photoemission took
place during the trials with this wavelength. We note
that these photons have an energy of E = hc/λ = 2.15
eV. We see that within error, method I yields a work
function greater than the energy of the incident photons
in this case, thus explaining the lack of photoemission.
However, method II does not yield parameters that agree
with this interpretation within standard error.
We conclude that these issues might have been resolved
had we backed out the anode, which could have highly
reduced the reverse current and limited the ambiguity in
FIG. 5: The linear fits to the cut off voltage points for each
frequency for both methods of selection. The chi-square val- selecting a cut off voltage. Also, we could have improved
ues of the fit to the data from method I and method II was the accuracy of our measurement by taking more filter
χ2I /dof = 0.6523/2 and χ2II /dof = 0.1325/2, respectively. cycles, with better resolution, and out to further voltages.
However, time constraints prevented us from doing so.
The properties of the photoelectric effect that we
Results observed reflect the wave-particle duality of light as
we see that there is a defined minimum packet, or
Assuming the accepted value for the elementary charge quanta, of light energy, which is proportional to fre-
(e = 1.602 × 10−19 C), we find for method I: quency. This seeming contradiction within the frame-
work of Maxwellian theory is resolved in the much richer
h = (7.12 ± 0.88) × 10−34 J·s context of quantum mechanics, a branch of physics which
this experimental observation helped to develop and
eventually flourish.
φeff = 2.20 ± 0.32 eV

and for method II:

h = (4.08 ± 0.88) × 10−34 J·s

[1] Harnell and Livingood. Experimental Atomic Physics.
McGraw-Hill, 1933.
[2] MIT Department of Physics. The photoelectric effect. Lab
φeff = 1.28 ± 0.32 eV
Guide, 2007.
[3] A. Einstein. Annalen der Physik, 132, 1905.
We see that only one of the methods seems to agree [4] R. Eisberg and R. Resnick. Quantum Physics of Atoms,
with the accepted value of h = 6.626 × 10−34 J·s. Molecules, Solids, Nuclei, and Particles. John Wiley &
Sons, Inc., 1985.
[5] Thermo-Oriel. Oriel high power mercury lamp specifica-
CONCLUSIONS tions. Specs.
[6] Y. Kraftmakher. Am. J. Phys., 2, 2006.
Only one of the methods managed to be in agreement The author acknowledges Sara Campbell’s equal part-
with nature. This may be a result of the arbitrary na- nership in the preparation and execution of this experi-
ture of the criteria: it seems there are many reasonable ment.