The Franck-Hertz Experiment

Ryan Rubenzahl1, ∗ and Bo Peng1, †

1
Department of Physics and Astronomy, University of Rochester, Rochester, New York, USA
(Dated: December 6, 2017)
We conduct the classic Franck-Hertz experiment in which electrons are accelerated through a
mercury vapor and inelastically collide. We find the energy required to initiate inelastic collisions
between electrons and mercury atoms to be 4.69 ± 0.05 eV, in agreement with the lowest excited
state of mercury (63 P0 ). The dependence of the Franck-Hertz curve on the tube temperature is
explored, and we find a decrease in average peak spacing with increased temperature. We find
the measurement of the lowest excited state to be stable with temperature if derived from minima
spacings, and inconsistent if derived from the maxima spacings, in agreement with previous studies.

I. INTRODUCTION claim that the minima spacing actually grows with each
successive minima. They model this effect by consider-
James Franck and Gustav Ludwig Hertz, German sci- ing the following phenomena: Once an electron reaches
entists at the University of Berlin, were the first to sufficient kinetic energy to undergo inelastic collisions
demonstrate experimentally the quantized nature of mat- (4.67 eV), it will then on average travel one mean-free-
ter in their famous experiment published in the German path before colliding with a Hg atom. In this time, the
Physical Society on April 24th, 1914 [1]. They shared electron is further accelerated, gaining more energy and
the Nobel Prize1 in 1925 for this “discovery of the laws thus possibly exciting one of the higher Hg energy levels
governing the impact of an electron upon an atom.” (63 P1 or 63 P2 ). When the voltage has been sufficiently
The original experiment consisted of measuring the increased (roughly double the first minima), electrons
current of electrons that were accelerated through a will be able to undergo two collisions before reaching the
heated tube containing mercury (Hg) gas. Once the elec- grid, causing this effect of added energy to occur twice.
trons reach a kinetic energy corresponding to the lowest The effect of this is a linear growth of the minima sepa-
excited state of Hg (the 63 P0 state, see Fig 1), inelastic ration ∆V with the minima order n:
collisions between the electrons and Hg atoms will occur, 
`

and the Hg atom will be excited to this state. Electrons ∆V (n) = 1 + (2n − 1) Va , (1)
L
accelerated to a kinetic energy corresponding to the next
excited state (63 P1 ) will also excite the Hg atoms, how- where Va is the accelerating voltage at which inelastic
ever this state is unstable, with a lifetime ∼ 105 times collisions begin to occur (4.67 V) and L is the distance
shorter than the 63 P0 state [2]. The 63 P1 state decays between the cathode and the grid. See Sec III of [3] for
almost immediately back to the ground state by the spon- a derivation of Eq 1. The slope of this line allows the
taneous emission of a photon, and is then ready to be ex- determination of the mean free path `, and the intercept
cited again. Hence while Hg atoms in the 63 P0 state will allows the determination of the lowest excitation energy
collide elastically with another incoming electron, with Ea = eVa (e = electron charge). Since the electrons will
the electron losing a negligible amount of kinetic energy follow a Maxwell-Boltzmann velocity distribution, their
in the process (because of the large mass difference), in
the same time period there will be ∼ 105 collisions ex-
citing other Hg atoms to the 63 P1 state. Franck and
Hertz’s original experiment included a window through
which the wavelength of the emitted photons could be
measured, which they found to be 2536 Å [2].
By measuring the current of electrons exiting the tube 6P 3P2
of gas as a function of their energy, Franck and Hertz
observed dips in the measured current corresponding to
the onset of inelastic collisions. While most analyses of 6P 3P1
6P 3P0
5.46 eV

the Franck-Hertz experiment assume the current minima

occur at integer multiples of the 63 P1 excitation energy
4.89 eV
4.67 eV

(4.89 eV, see Fig 1), Rapior, Sengstock, and Baev [3] 2536 Å

∗
6S 1S0
[email protected]
† [email protected]
1 https://www.nobelprize.org/nobel_prizes/physics/
FIG. 1: Energy levels of Hg relevant to this analysis
laureates/1925/ [2, 3].
 2

If an electron moving through the tube happens to

have a kinetic energy corresponding to the an energy level
of Hg, the electron may collide inelastically with a Hg
pA atom, resulting in the electron no longer having enough
energy to overcome the retarding potential. As a result of
V these collisions, at a particular value of the accelerating
Vret voltage (corresponding to electrons with kinetic energy
matching that of the Hg excited state), a measurable
+

anode drop in the current recorded by the picoammeter can be

grid
seen. Continuously increasing the accelerating voltage
Hg
will result in electrons again reaching the required kinetic
Hg
energy to inelastically collide with a Hg atom, causing a
e-
Hg second drop in the measured current. Continuing this
process results in repeated dips in the measured current
+

Hg
Vacc at regular intervals (Fig 4), with the spacing between
cathode minima corresponding to the energy required to excite
filament the Hg atoms. This is the famous result of the original
Franck-Hertz experiment, which Einstein, after hearing
a presentation of the results by Franck at a conference,
remarked to be “so lovely it makes you cry” [4].
+ The density of the Hg gas present in the tube is con-
trolled by the temperature of the oven enclosing the tube.
Vfil In our analysis, we record measurements at five oven tem-
peratures ranging between 135 ◦ C and 175 ◦ C in 10 de-
gree intervals. At each temperature, a “run” consists
FIG. 2: Schematic of the Franck-Hertz apparatus used. of recording the current measured by the picoammeter
for accelerating voltages from 3 V to 40 V, in steps of
around 0.1 V. Higher oven temperatures correspond to a
mean free path ` is given by denser Hg vapor and therefore a smaller mean free path
for the electrons, reducing the probability of an electron
1 kT kT navigating the thick Hg cloud. We therefore expect to
`= √ =√ =√ , (2)
2N σ 2pσ 2p(πR02 ) measure smaller currents at higher temperatures, as well
as observe a decrease in the slope of the line from Eq 1.
where R0 ≈ 1.5 Å is the cross-sectional radius of a Hg
atom, and the Hg vapor density N is strongly sensitive
to the temperature T of the Hg gas via the ideal gas law III. DATA COLLECTION
p = N kT . Thus we expect the mean free path of the
electrons, and hence the minima spacings, to vary with The data are recorded using the Franck-Hertz Lab-
the oven temperature as well. Specifically, the average VIEW computer program. The LabVIEW program au-
spacing between minima should decrease with increased tomates data collection after specifying a filament voltage
temperature. In this analysis we consider the dependence and start, stop and step-size values to sweep the acceler-
of the measured current minima on the oven temperature ating voltage. For all runs, we use a filament voltage of
to explore this effect. Vfil = 5 V, a retarding voltage of Vret = 1.5 V, and sweep
the accelerating voltage from 3 V to 40 V in steps of 0.1
V. This is done for oven temperatures of T = 135 ◦ C ,
II. EXPERIMENTAL APPARATUS 145 ◦ C , and 175 ◦ C. For T = 155 ◦ C and 165 ◦ C, a step
size of 0.05 V is used instead. The measurements for each
Fig 2 shows a schematic of our experimental setup. run are displayed in Fig 3. Results for two different tube
The filament voltage Vfil heats a filament resulting in temperatures are shown in Fig 4.
the emission of electrons onto the cathode. These elec- Note that we can immediately see that the minima
trons are then accelerated from the cathode to the grid locations are not equally spaced across different temper-
by the electric field created by the accelerating voltage ature runs. Fig 4 shows the smoothed data for runs at
Vacc . Electrons that reach the grid but have less energy T = 155 ◦ C and T = 175 ◦ C. From this it can be clearly
than that corresponding to the retarding potential Vret seen that the average minima spacing shrinks with in-
are halted and collected by the grid. Those with greater creased temperature, as expected. This suggests that
energy are able to overcome the retarding potential and conclusions drawn directly from measurements of the av-
make their way to the anode where they are measured as erage minima spacings are unreliable.
a current by the Keithley picoammeter (pA in Fig 2). To estimate the uncertainty in our measurement of the
 3

135 ∘ C
12.5
10.0
7.5
5.0
2.5
0.0
10∘ C
145 20 30 40 10 20 30 40
2.5
2.0
1.5
1.0
0.5
0.0
10∘ C
155 20 30 40 10 20 30 40
0.8
Anode Current [nA]

0.6

0.4

0.2

0.0
0.30 10∘ C
165 20 30 40 10 20 30 40
0.25
0.20
0.15
0.10
0.05
0.00
0.125 10∘ C
175 20 30 40 10 20 30 40
0.100
0.075
0.050
0.025
0.000
10 20 30 40 10 20 30 40
Accelerating Voltage [V]
FIG. 3: Data (left) collected for each temperature, and the smoothed and analyzed data (right). The vertical dashed
lines correspond to bin edges, and the blue curves are the individual polynomials being fit in each bin. (•) points
denote located minima, and (•) points denote located maxima.
 4

80 0.7
155 ∘ C run 12
0.6 run 13
60 run 14
0.5

Anode Current [nA]

run 15
40 run 16
0.4
Anode Current [10−2 nA]

20 0.3
0 0.2
175 C
∘
0.1
8 0.0
0.4 5 average
10 15 20 25 30 35 40
6

Residual Current
Accelerating Voltage [V]
4 0.2
2
0 0.0
5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
Accelerating Voltage [V]

FIG. 4: Franck-Hertz curves for two tube temperatures, FIG. 5: (Top) Diagnostic runs at T = 155 ◦ C with the
shifted horizontally so that the first maxima coincide. best-fit line to estimate the bias. (Bottom)
Bias-subtracted residual curves and, and the overall
average residual. There appears to be a bias present in
the amplitude as well, but this should not effect the
anode current and peak locations, we make five addi-
determination of the minima/maxima locations.
tional diagnostic runs. Each run is measured at a con-
stant temperature of T = 155 ◦ C and is recorded with a
step size of 0.5 V.
use the standard deviation of the measured locations of
each individual maxima and minima across the five tem-
perature runs as our measure of uncertainty in that peak
IV. LOCATING MAXIMA AND MINIMA
location. We estimate an overall uncertainty our mea-
surement of the peaks by computing a pooled variance
In order to measure the minima (and maxima) spac- across all located peaks. This uncertainty is expected to
ings, we first have to locate these extrema. We do this be limited by our voltage resolution (0.05 - 0.1 V), and
by first smoothing our data using a simple moving av- indeed the uncertainties in the minima and maxima lo-
erage by convolution, using a smoothing window of 1 V. cations respectively are found to be σmin = 0.09 V and
We then calculate the discrete second derivative of the σmax = 0.07 V.
smoothed data, which is searched for zero-crossings (in- The minima/maxima we locate in our data are shown
flection points). The zero-crossings define bin edges, and in Fig 3. When defining bin edges, the low S/N regions
the data-points within each bin are fit using a 2nd degree corresponding to V < 11 V are discarded. Likewise, for
polynomial, I = a + bV + cV 2 . The location of the min- the run at T = 135 ◦ C, we do not search for bins for any
imum/maximum is determined by setting the derivative voltage > 36 V. The 39 - 40 V region of each run contains
of the best-fit polynomial equal to zero, i.e. artifacts resulting from the smoothing procedure, and is
( similarly not included in the analysis.
b̂ minima ĉ > 0
V =− = (3)
2ĉ maxima ĉ < 0
V. DETERMINATION OF THE LOWEST
To estimate an uncertainty in the location of the ob- EXCITATION ENERGY OF HG ATOMS
served minima and maxima, we perform the procedure
for locating maxima and minima just described on the Following the procedure outlined in [3], we plot the
five diagnostic runs at T = 155 ◦ C. Each of these diag- spacing between each successive minima as a function of
nostic runs contains a bias in the current due to leftover the minima order n. We fit these to Eq 1 and extrap-
electrons from the previous run. The bias is removed by olate the value corresponding to n = 0.5 (Fig 6). This
subtracting the difference between a linear fit to the data value, ∆V (0.5), corresponds to the smallest energy of an
and a linear fit of the first diagnostic run. The resulting electron required to initiate inelastic collisions with a Hg
bias-subtracted runs are then searched for minima and atom, and is equal to the energy in units of electron-volts
maxima. Fig 5 shows the bias-subtraction process. We (eV). Calculating this quantity for each temperature run,
 5

145 ∘C 5.2 minima

5.3
Minima Separation [V]

Average Separation [V]

155 ∘C maxima
165 ∘C
5.1
5.1 175 ∘C

4.9 5.0

4.7 4.9

4.5 145 150 155 160 165 170 175

0 1 2 3 4 5 6 7 8
Number of minimum n Temperature [ C]

FIG. 6: Minima spacings as a function of the minima FIG. 8: Average separation of minima (•) and maxima
order n, and the best fit for each temperature to Eq 1. (•) as a function of tube temperature. The overall
The dashed vertical line corresponds to n = 0.5. averages and their 1σ uncertainty for the minima and
maxima are given by the filled bands.

5.0
Hence the value extrapolated from minima spacings will
4.5 be consistent, while the value extrapolated from maxima
V(0.5) [V]

will generally decrease with temperature as the mean free

4.0 path shortens.
Also note that the slope of the fits decreases with in-
3.5 Mean using minima creased temperature. This is reflected by the dependence
Minima of the slope on the mean free path of the electrons.
Maxima
3.0
145 150 155 160 165 170 175
Temperature [ C] VI. DEPENDENCE OF THE SPACINGS ON
THE TUBE TEMPERATURE
FIG. 7: Measured voltage corresponding to the lowest
excitation energy of Hg as a function of temperature. In this section we consider the effect of the tube tem-
Results using the minima spacings and their mean value perature on the average spacing between minima and
with a 68% central interval are shown, as well as results maxima. Plotting the average spacing as a function of
using the maxima spacings. the tube temperature yields the expected decrease, as
shown in Fig 8. This is reflected in the example Franck-
Hertz curves shown in Fig 4 as well.
we determine the mean value to be 4.69±0.05 V, in agree- Higher temperatures significantly increase the number
ment with the 4.67 eV 63 P0 state of Hg. density of the Hg vapor, which we can assume behaves
Several cuts were made on which minima were included like an ideal gas. This increased density significantly de-
in this analysis. The run at T = 135 ◦ C was excluded creases the electron mean free path, resulting in less “ex-
entirely as it has too few peaks to determine a reliable tra energy” to be gained by the electrons once reaching
trend in the spacings. The first detected minima was the lowest excitation energy. This in turn causes the
not considered if it is either not significantly pronounced spacings of the minima, which depend on this extra en-
enough or is too “messy.” Only the run at T = 165 ◦ C ergy by Eq 1, to exhibit an overall decrease with temper-
satisfies this (the T = 155 ◦ C run is is well-pronounced, ature. This is further reflected by the slope of the fits in
but odd modulations occurring in the first maximum dis- Fig 6 decreasing with temperature, which corresponds to
rupt the fit). Likewise, the final minima for T = 145 ◦ C the decrease in the electron mean free path.
was not used, as it occurs just before the minima wash The overall average across all temperatures of the av-
away. All maxima are included in the analysis. erage spacing is also shown in Fig 8 by the colored bands.
We evaluate ∆V (0.5) using both minima and maxima These values can be seen to not agree with the 4.89 V
spacings and plot the result as a function of the tube tem- spacing predicted by the lab manual [2] and as quoted by
perature, as shown in Fig 7. The minima-derived values Franck and Hertz in their original experiment [1]. This is
are consistent with temperature, and the maxima-derived not entirely unexpected, as the average spacing between
values vary with temperature. This is consistent with the peaks depends on the Hg temperature, pressure, and on
results presented in [3] and is explained the energy of the tube parameters, and is not exclusively determined by
colliding electrons, which depend partly on the tempera- the 4.89 eV transition. In fact, deviations from 4.89 V
ture, being at their highest at the observed maxima, and by a few tenths of a volt is expected and normal [5].
at their most probable value at the observed minima. This discrepancy is explained by the model described
 6

Mean free path [ m] only unmeasurable of which is the pressure p. Luckily,

the pressure of a Hg vapor in the 300 K to 500 K regime
is well-approximated by [3, 6]2
100
p = 8.7 × 109−3110/T , (4)

50 where T is given in Kelvin. From our fits shown in Fig 6,

the mean free path may be calculated using the cathode-
grid distance L = 1 cm for our setup. We plot these
0 values vs. temperature in Fig 9 and fit them to Eq 2 using
the form of the Hg pressure given above in Eq 4. From
150 160 170 this we calculate the cross section for collisions between
Temperature [ C] electrons and mercury atoms to be (1.1±0.1)×10−19 m2 .
We also calculate the radius of the Hg atom to be 1.87 ±
FIG. 9: Mean free path of electrons as a function of the 0.27 Å, in agreement with currently accepted values [2].
tube temperature.

VIII. SUMMARY
earlier and proposed in [3]. Electrons reaching 4.67 eV
during their journey through the tube will on average We conduct the traditional Frank-Hertz experiment for
travel one mean free path before colliding with a Hg four tube temperatures and determine the lowest exci-
atom. Some of these electrons may gain sufficient energy tation energy for Hg atoms to be 4.69 ± 0.05 eV. We
to excite the 4.89 V transition, and still others may gain explored the dependence of the Franck-Hertz curve on
enough to excite the 5.46 eV transition (63 P2 state). The the tube temperature and observed a decrease in average
probability of exciting these higher lines is greater when peak spacing with temperature. This phenomena is ex-
the mean free path is greater, which occurs at higher plained in terms of the mean free path of the electrons,
temperatures. This further supports the decrease in the the temperature and pressure of the Hg vapor, and the
average spacing with temperature. Therefore, the aver- properties of the tube itself. Finally, we derived the col-
age spacing of the peaks in a typical Franck-Hertz curve lisional cross section of the Hg atoms and determined the
is a complicated combination of these three excitations, radius of the Hg atom to be 1.87 ± 0.27 Å. This paper
and thus the spacing of peaks depends on which transi- is submitted in partial fulfillment of the requirements of
tions are more dominant than others. Since the mean free PHY 243W: Advanced Experimental Techniques at the
path (or equivalently the collisional cross section) is de- University of Rochester.
pendent on the tube temperature and pressure, different
tube designs will also yield varying results.
ACKNOWLEDGMENTS
VII. DETERMINATION OF THE MEAN FREE
PATH OF ELECTRONS IN HG VAPOR AND THE We thank Professor Kevin McFarland and Dr. Justin
COLLISIONAL CROSS SECTION OF HG ATOMS Winkler for their guidance in the laboratory procedure
and insightful comments that were invaluable to our anal-
Eq 1 allows the determination of the mean free path of ysis. We especially thank Professor McFarland and Mike
electrons in the Hg vapor. Eq 2 gives the expression for Culver for sustaining us with a fresh plate of cookies at
the mean free path in terms of the Hg gas parameters, the the start of every lab session.

[1] J. Franck and G. Hertz, Verh. Dtsch. Phys. Ges. 16, 457 [4] S. A. Rice and J. Jortner, James Franck, A Biographical
(1914). Memoir (National Academy of Sciences, 2010).
[2] K. McFarland, A. Melissinos, and R. Munson, PHY 243W [5] G. F. Hanne, American Journal of Physics 56, 696 (1988).
Franck-Hertz Manual, University of Rochester (2016). [6] R. Gary, Vapor pressure of the chemical elements (Else-
[3] G. Rapior, K. Sengstock, and V. Baev, American Journal vier, 1963).
of Physics 74, 423 (2006).

2 The author spent an amount of time longer than he would care after hunting down [6] in Carlson Library did he realize he had
to admit wondering if Eq 4 was incorrectly stated in [3]. Only been calculating the pressure using degrees Celsius.