INSTITUTE OF PHYSICS PUBLISHING EUROPEAN JOURNAL OF PHYSICS

Eur. J. Phys. 23 (2002) 533–548 PII: S0143-0807(02)35947-6

Critical potentials of mercury with a

Franck–Hertz tube
Peter Nicoletopoulos1
Faculté des Sciences, Université Libre de Bruxelles, Brussels, Belgium

E-mail: [email protected] and [email protected]

Received 17 April 2002, in final form 16 July 2002

Published 6 September 2002
Online at stacks.iop.org/EJP/23/533

Abstract
A classic experiment, where a four-electrode Franck–Hertz tube is used to show
various energy states (critical potentials) to which a neutral mercury atom can
be raised by electron impact, is upgraded and elucidated thanks to a novel
arrangement of applied inter-electrode potentials. By that means, a suitably
located plasma region, the working field-free scattering chamber, can be created
dynamically and maintained thereafter under appropriate conditions of cathode
emission current and mercury vapour pressure. Concurrently, a more efficient
mechanism for displaying excitation peaks becomes operant in the adjoining
electron collection cell, by means of the window for slow-electron transmission
that is available in mercury below the resonance peak in the cross section at about
0.4 eV. The resultant enhanced resolution in current–voltage curves permits the
identification and measurement within 0.05 eV of the leading features in the
known electron impact excitation spectrum of mercury.

1. Introduction

This paper is about an experiment that can lift itself by its own bootstraps: the attempted
setup, inefficient at the outset, is spontaneously redesigned during operation by a conspiracy
of dynamical phenomena whose actuator is the very element that is being subjected to
investigation.
Suppose that you want to demonstrate—indeed, measure—atomic critical potentials. Take
a vacuum diode. Interpose two grids distant from each other near the electrodes, say grid 1
close to the cathode and grid 2 close to the anode.
(a) Short the grids so that their interspace will be equipotential. Insert a pure gas of atoms,
say mercury.
(b) Adjust the pressure so that the electronic excitation free path is longer than the gap between
the cathode and grid 1.
Thus, when a voltage is applied across this gap, preaccelerated electrons traversing the vapour
between the grids will conserve their energy unless they suffer inelastic collisions.
1 Address for correspondence: Rue Joseph Cuylits 16, 1180 Brussels, Belgium.

0143-0807/02/050533+16$30.00 © 2002 IOP Publishing Ltd Printed in the UK 533

534 P Nicoletopoulos

(c) Provide a small retarding potential between grid 2 and the anode in order to prevent slow
electrons from reaching the anode.
Now vary the voltage of grid 1 and simultaneously monitor the electron current between
grid 2 and the anode. You expect to see downturns in this current whenever the accelerating
voltage passes through a critical potential of the gas.
At first sight this looks simple enough for an undergraduate laboratory. The question is,
how well does it work? Usually not so well, as attested by the literature: the data commonly
obtained are so elusive that textbooks of laboratory practice proposing this method have been
loath to show sample results [1, 2].
In 1979 I introduced this experiment in our undergraduate laboratory because it is an
immediately available extension of the standard Franck–Hertz experiment, thanks to a Franck–
Hertz tube of the type described in [2] which incorporates grid 1 as a luxury (see section 2).
It took years to grope a way toward a reasonably consistent interpretation of the electron
excitation spectrum of mercury obtained in that way, and some time more to find a theoretical
basis for the gaseous discharge phenomena that are involved. The main outcome of this search
is presented here. The technical elaboration of the transport-theoretical and plasma-physical
aspects will be presented elsewhere [3, 4].
Although this experiment dates from the days of the early quantum theory, only scarce
data exist in the literature with which to compare the results on mercury. Three papers contain
current–voltage curves in the range between zero and about 25 V. The authors were:
(1) Mohler et al [5] in 1920 at NBS,
(2) Einsporn [6], a student of James Franck, in 1921 in Berlin, and
(3) Liu [7] in 1987 in China.
These plots were explained by matching the peaks with suitable combinations of two or
three of the lowest critical potentials of mercury, multiple inelastic scattering being favoured at
the pressure employed. Einsporn interpreted his or her plot in terms of the 3 P1 and 1 P1 states.
Mohler et al, who presented curves at more than one value of pressure, concluded likewise but
noted that one or two additional critical potentials might be involved. Liu, whose features are
sharper and more numerous, was forced to invoke a third critical potential, the 3 P2 state, and
yet four features remained outside this scheme.
Curves resembling one or another of the ones published are readily obtained with a Franck–
Hertz tube of the type described in [2]. At times, however, additional peaks show up among
the ones usually present and continually frustrate all attempts to interpret the spectrum.
The problem is that the pressure required for producing strong peaks and valleys in the
anode current is so large that positive ions are produced very early by repeated electron
impact on atoms excited to the metastable 3 P0 and 3 P2 states. Thus, the electrostatic potential
distribution between the grids evoked by condition (a) is strongly modified by the dynamics
of the interaction of positive and negative space charge. In addition to that, the effectiveness
of condition (c) becomes questionable because the anode current is bipolar.
It was realized that a maximum of structure is obtainable if condition (c) is abandoned or
reversed and a sufficiently large decelerating voltage is applied across the grids. This structure
is identifiable as the superposition of two spectra (see section 3). The inference is that the space
potential between the grids has been distorted into a steplike configuration of two field-free
plasma regions at different potentials joined by a narrow space charge layer. This phenomenon
is well known in plasma physics.
Varying the intergrid potential can move the narrow zone dividing the two drift spaces
inwards or outwards. In this way one can eliminate the intermixing of spectra. Particularly
good results are obtained under conditions that privilege the display of the spectrum excited
at the lower potential, namely in the field-free plasma adjacent to the outer grid.
The reason is that we are then enabled to shift from the intended beam-depletion experiment
to its complementary version in which critical potentials are displayed as peaks in the diffusion
current of slow electrons crossing the ‘detection chamber’ between grid 2 and the anode.
Critical potentials of mercury with a Franck–Hertz tube 535

Table 1. Observed critical potentials and their interpretation.

Measured Accepted
No value (eV) Attribution value (eV)

1 0.4 (5d10 6s2 6p)2 P state of Hg− 0.42a

2 4.9 (5d10 6s6p)3 P1 4.89
3 5.5 (5d10 6s6p)3 P2 5.47
4 6.7 (5d10 6s6p)1 P1 6.70
5 7.8 Unresolved cluster of (5d10 6s7s)3 S1 7.73
and (5d10 6s7s)1 S0 7.92
6 8.8 (5d9 6s2 6p
)3 D3 8.80
7 9.8 Twofold 6 3 P1 9.78
or (5d9 6s2 6p
)1 P 9.77
8 10.4 6 3 P1 + 6 3 P2 10.36
or (5d9 6s2 6p
)(3/2 1/2)2 10.35
9 11.0 Twofold 6 3 P2 10.94
or (5d9 6s2 6p
)3 P 11.00
a Peak in the elastic cross section.

Such electrons, of the very kind that we had initially planned to hold back, are subject to
preferential collection near their thresholds of production due to an effect intrinsic to elastic
e− –Hg scattering: a low-energy shape resonance (see sections 4 and 5.1).
The features displayed by this contrivance are stronger and much sharper than those
obtained ordinarily and can be measured within 0.05 eV. Thus it becomes possible to interpret
the structure in terms of the known electron impact excitation spectrum of mercury (see table 1).

The coarser curves of [5–7] are spectra obtained under conditions that privilege the display
of the spectrum excited in the high-potential plasma. The consistency of these data with the
series in table 1 is established in sections 5.2 and 6. A detailed consideration is necessary
because a further difficulty caused by space charge must be faced.
The phenomena underlying the changes of the space potential are quite complicated and
their details are outside the scope of this paper. The main points are outlined in section 7.
The dependence of optimal operation on the pressure of mercury vapour and on the
emission current density and velocity distribution of the electron beam is given in section 8.

2. Apparatus

The tube used in this experiment is a commercially available Franck–Hertz tube (Leybold–
Heraeus 55580) an earlier version of which (55080) was described in [2]. Although a triode is
sufficient for the standard Franck–Hertz experiment, this tube contains an extra grid g1 close to
the oxide-coated cathode K. This grid, designed to act merely as a ‘current control’ electrode,
is promoted here to accelerating grid.
All electrodes are cylindrical except for g1 which is a flat, 0.2 cm wide helix evenly coiled
around two upright supports siding the 0.05 cm diameter cathode. A sort of shield, close to
and shorted with g1 , obstructs the vertical ends of this assembly. The open central part, 0.7 cm
tall, contains about 40 turns of thin (0.005 cm?) wire.
The outer grid g2 is a cylindrical net made of 0.02 cm wire (0.1 cm square mesh); it is
3 cm tall. The height of the anode is 2 cm. The distances d1 , d2 , d3 , between K and g1 , g1 and
g2 , g2 and A (the anode), respectively, are: d1 = 0.02, d2 = 0.8, d3 = 0.2 cm.
The experimental arrangement is shown in figure 1. The tube lies vertically inside a small
homemade electric oven, essentially a 50 Hz ac coil. The temperature was measured with an
ordinary mercury thermometer inserted between the coil’s support and the tube through the
536 P Nicoletopoulos

Figure 1. Schematic diagram of tube and circuit.

lid. Temperature readings T (◦ C) vary somewhat along the axis of the oven. The readings of T
given in the text were taken with the tip of the thermometer positioned about midway between
the drop of mercury at the bottom of the tube and the electrodes.
M1 and M2 are ordinary vacuum tube meters. The role of M1 is essentially to provide
a standard, called Ic , of the emission capability of the cathode. The quantity Ic is defined
arbitrarily as the reading of M1 (operated in current mode) at T = room temperature,
Va = 8.6 V, V f = 3.5 V, where Va is the voltage between the cathode and g1 , and V f is
the effective value of the ac filament voltage. As will be discussed in section 8, the value of
Ic depends on the state of activation of the cathode, which is an important factor.
The output of M2 is fed to the Y terminals of an X–Y recorder whose X-axis is driven
by the accelerating voltage Va . The scale of X was set at 1 V cm−1 . Va is varied very slowly
by hand via a 5 k
helipot. Smoothness of recording was enhanced by adding the condenser
C (640 µF). The curves shown in this paper are representative of hundreds of similar plots
obtained in the period 1985–7.
Preparations must also be made for small constant voltages of either sign to be applicable
between g1 and g2 and between g2 and A. The algebraic values (in volts) of these potentials
are called g and a, respectively; positive values accelerate outward-bound electrons.
The magnitude of g, accurately controlled by a finely adjustable constant-voltage
generator, is a crucial element in this experiment.

3. Untangling the energy scale: the main step

Figure 2 shows a plot of anode current versus Va obtained with g = 0 and a = 0. A
discussion of how different portions of such curves are influenced by mercury vapour pressure,
cathode state of activation and filament voltage is deferred until section 8; meanwhile, it will be
implicit that these factors have been adjusted for optimal visibility of a maximum of structure
throughout the energy range.
The excitation curve of figure 2 is considerably more complex than the original curves
of Einsporn [6] and Mohler et al [5]. It differs, too, from the modern curve of Liu [7]. The
central purpose of this section is to show that the increased quantity of structure obtained here
is due to the fortuitous presence of a property intrinsic to well-seasoned samples of our type
of tube.
Before we can measure critical potentials, we need a reference for calibrating the energy
scale. To all appearances, the first large peak, feature a, should correspond to a poorly
Critical potentials of mercury with a Franck–Hertz tube 537

Figure 2. Excitation curve with g = 0, showing a mixture of ‘standing’ and ‘moving’ peaks.

resolved clump of the lowest multiplet of excited states of mercury, the 6 3 P triplet around 5 V,
which is the state involved in the standard Franck–Hertz experiment. Despite the differences
between figure 2 and the excitation curves of previous authors, a similar peak is displayed in
all occasions, and is commonly placed at 4.9 V to calibrate the energy scale.
On the other hand, it is clear that there is a point Va = V0 ∼
= 2.5 on the voltage axis, below
which there is no anode current. This must mean that the energy scale has been shifted to the
right by a retarding (negative) contact potential difference (cpd) of 2.5 V between K and the
grids, a fact that would be consistent with a low-work-function cathode surface.
If the abrupt rise of current is the zero of absolute energy, then feature 1 (also shown in
amplified form) is compatible with the peak at 0.4 eV in the elastic cross section due to a shape
resonance (see section 4). Feature a, however, which we would like to attribute to the 6 3 P
triplet, appears at about 3 V on that scale instead of 5 V. We are therefore confronted with a
serious inconsistency.
A clue for the solution of this problem is obtainable from a more elaborate exploration
of the origin of the retarding cpd: let Ug1 and Ug2 be the contact potentials between K and g1
and between K and g2 . If Wc , Wg1 , Wg2 are the work functions of these three electrodes, then
eUg1 = Wc − Wg1 and eUg2 = Wc − Wg2 . The true electrostatic voltages Ta and Tg2 felt by
cathode electrons traversing g1 and g2 , respectively, are Ta = Va + Ug1 and Tg2 = Va + Ug2 .
Therefore Tg2 − Ta = Ug2 − Ug1 = (Wg1 − Wg2 )/e. This means that the electrostatic potential
drop between shorted grids is zero only when Ug2 = Ug1 or Wg1 = Wg2 .
Now V0 = Va −Tg2 = Va −Ta −U , where U = Ug2 −Ug1 . But the location of peak a on
the Va axis indicates that Ta is close to Va . And if the current to g1 is plotted against the voltage
Va , its onset potential is found to be very close to Va = 0, which means that the cpd between
K and g1 must be negligible so that indeed Ta = Va . Thus V0 = −U = (Wg1 − Wg2 )/e. In
other words the 2.5 V delay of the onset of anode current must be due to a negative contact
potential between g1 and g2 .
538 P Nicoletopoulos

Figure 3. Mixed excitation curve with g = −0.5; feature b has come into view, f and g have
been overtaken by 6 and 7.

If this is true, then upon addition of an external negative voltage g, V0 should advance
to V0 + |g|. Curves obtained for g = −0.5, −0.8, −1.6, are displayed in figures 3–5. It
is seen that in each case the onset of anode current has indeed been delayed by an additional
|g| volts. The threshold of g1 current is not affected. This confirms that the cpd is effective
between g1 and g2 .
Unexpectedly, however, the changes of intergrid potential have also affected the overall
aspect of the excitation curves, certain portions of which have been significantly modified.
The key to solving the puzzle of the energy scale is recognizing that the source of the
alterations in the excitation curves is dual. Firstly, certain peaks (those labelled by numbers)
are shifted to the right by |g| volts concurrently with the shift of the threshold of anode
current. Secondly, there is a class of peaks (labelled by low-case letters) whose location is
independent of the value of g. Thus, as g is made more negative, a standing peak may be
overtaken and partly or totally masked by a prominent moving one. For instance, in figure 4,
feature 4 has crept forward to almost overtake peak d. The unmasking of a previously invisible
standing feature can also occur, as is the case for peak b in figures 3 and 4.
The standing structure (except of course for feature a) is overtaken entirely at g ∼ = −1.2.
For g < −1.2, namely after this mergence, the standing peaks do not re-emerge. The moving
spectrum—now unblemished—improves toward its optimal form, attained at g ∼ = −1.5.
Curves recorded in the range of g between −1.5 and −1.8 offer the strongest display of
structure. Beyond g ∼ = −1.8, the peaks become narrower and their amplitude decreases.
Because the location of moving peaks is governed by the potential Tg2 of g2 with respect
to the cathode, it is natural to suspect that these features show excitations effected by electrons
with energy equivalent to this potential, these electrons somehow reaching the vicinity of g2
(decelerated by V = g + U ) without suffering previous inelastic collisions. Impressive
support for this conjecture is obtained if the pure moving spectrum of figure 5 is measured
from Tg2 = 0, namely with respect to the onset of anode current. As discussed in section 4, the
Critical potentials of mercury with a Franck–Hertz tube 539

Figure 4. Mixed excitation curve with g = −0.8; feature 4 has almost overtaken peak d.

Figure 5. Excitation curve with g = −1.6, showing a pure ‘moving’ spectrum plus feature a.
540 P Nicoletopoulos

Figure 6. Excitation curve with g = +1.6, showing a pure ‘standing’ spectrum plus feature 1.

values obtained by this procedure correspond to all the prominent features commonly observed
in the electron impact spectrum of mercury.
What about the origin of the standing peaks? By their very name they must be due to
excitations occurring in the vicinity of the first grid, by electrons with energy equivalent to the
true potential Ta (∼
= Va ) of g1 with respect to the cathode. We can now explain the inconsistency
encountered with regard to the position of feature a: since this peak is stationary, it must be
measured on the g1 voltage scale and on this basis its location is not incompatible with its
interpretation in terms of the 6 3 P state.
Needless to say, the existence of two scales set off from each other, means that different
peaks in the same curve can show the same energy level excited in different regions in the tube.
For instance, features 2 and 3 correspond to the two highest members of the 6 3 P triplet excited
at g2 potential. Clearly, much better resolution is attainable amongst the moving structure.
But the very existence of the moving structure is critically dependent on the presence of a
retarding potential of sufficient magnitude between the two grids. It is easily verified that far
less satisfactory results would have been achieved had we used a tube with no cpd between g1
and g2 (U = 0) in order to really accomplish the arrangement desired originally: indeed,
the effect of compensating the negative value of U by making g increasingly positive is
to blunt the exposed structure by favouring the standing peaks. Quite soon, at g ∼ = +0.4,
namely for V still in the retarding direction but inferior in magnitude to |g + U | ∼ = 2.1,
all the moving peaks are washed out. Thereafter, the remaining structure becomes broader
until ultimately a stable form is attained at g ∼ = +1.5. This type of curve, shown in figure 6,
is quite similar to the plots obtained in [5, 6].
The fact that even for g = 0 a retarding value of V is operant in the present tube is
doubtless due to deposition of oxide material from the cathode coating onto the surface of the
close-lying first grid. As a result, the cpd initially present between the cathode and a clean
g1 surface has been neutralized, but a large retarding potential has become effective between
Critical potentials of mercury with a Franck–Hertz tube 541

Figure 7. Intergrid space potential, established for g < 0.4 and a strong beam, showing two
plasma regions and a space charge double layer (DL) with the high-potential plasma at (a) applied
value of grid 1 voltage and (b) raised value operant beyond feature A. Observed values of δ range
from 1.6 to 2.7.

the first and second grid. Ostensibly, the cathode activation procedure described in section 8.2
enhances this effect by inducing impact evaporation from the cathode by positive ions.
Let us take stock: our initial intention of creating field-free drift space within the tube
by shorting the grids was thwarted by the presence of a retarding potential between these
electrodes. Far from being a handicap, this mishap is a bonus, and enhancing its effect upgrades
the experiment. Thus, it is found that in the most effective arrangement for showing critical
potentials of mercury with the present four-electrode tube, the second grid must be maintained
negative with respect to the accelerating grid by 4–4.3 V, including the cpd. The structure in
excitation curves recorded with retarding values of g is interpretable in terms of inelastic
events occurring in regions of the tube where the electron beam energy is equivalent to either
the potential of g1 , or that of g2 , with respect to the cathode.
If it is correct, that interpretation implies that only an insignificant fraction of the inelastic
events can take place at intermediate energies. Therefore, it would require the presence, in the
interspace between grids, of a stepwise decreasing potential configuration consisting of two
equipotential regions joined by a precipitous fall where the beam is abruptly decelerated from
g1 to g2 potential, as shown in figure 7(a). The natural unit here is the electronic excitation
free path λe . So, expressed in this scale, the widths p1 and p2 of the high- and low-potential
regions and the width s of the sheath joining them should be in accordance with the relations:
p1  1, p2  1, s ∼ = 1.
Potential configurations of this type can indeed arise from the global interaction of
positive and negative space charges. In fact, we have come across an instance of a nonlinear
phenomenon which is of current interest in plasma physics: the formation of a narrow space
charge sheath between two weakly ionized plasma regions at different potentials (see section 7).
The foregoing discussion indicates that for values of g exceeding +0.4, the sheath
advances outward through the intergrid volume so that p2 ∼ = 1, since only the standing spectrum
is observed. For g < −1.2 the converse phenomenon takes place.
A closer look at the structure of the stationary spectrum beyond feature a (see section 6)
will bring to light a further complication: the shift of scale |V |, namely the potential difference
between the two field-free regions, jumps from |g + U | to a higher value |g + U | + δ in
542 P Nicoletopoulos

the course of operation. Thus, the final shape of the high-potential region is the one shown in
figure 7(b). The formation of a hump in the space potential is another topic of current research.

4. Measuring critical potentials

A striking trait of figure 5 is that the pattern comprising peaks 2–9 is met repeatedly, with
somewhat reduced amplitude but no appreciable change in shape. Two complete cycles
(unprimed and primed numbers) and part of a third (doubleprimed numbers) are included
in our energy range.
The energy interval between similar peaks of adjacent periods is 6.7 eV. This value agrees
precisely with the energy of the 1 P1 singlet state of mercury. Primed and doubleprimed peaks
are therefore interpreted as multiple step excitations involving the corresponding unprimed
level plus one and two 1 P1 transitions, respectively. The exclusive involvement of the 1 P1 state
in these periodic multiple-impact transitions is in accordance with the fact that the excitation
cross section for this state does not drop significantly after reaching a peak but extends virtually
undiminished well beyond its maximum, so that above 12 eV it exceeds the excitation functions
of other prominent levels by several orders of magnitude [8, 9].
Peak 2 is located nearly 5 V beyond the onset of anode current. Therefore, the eight peaks
within the first period should reflect the structure of the e− –Hg excitation spectrum in a range
of 6.7 eV past the first excitation threshold. Paired excitations involving the lowest lying levels
(the 6 3 P triplet) are energetically possible only beyond feature 6 so that we must be observing
at least five single-impact critical transitions.
Feature 1 is special and will be discussed individually in the later part of this section.
The energy scale whose zero is the steep onset of anode current is calibrated by placing
peak 4
, the second occurrence of the strongest feature, at 13.4 V. This procedure was found
to give better overall agreement than placing peak 4 at 6.7 V, because peak 4
always appears
at the same distance from the dip at feature 1, whereas peak 4, and the two preceding it, are
sometimes shifted (by as much as 0.2 V) farther from feature 1 and closer to their primed
counterparts. This effect is more noticeable for less negative values of g. Apparently a
slight scale distortion can exist during an interval of about 2 V following the creation of the
low-potential drift space. Beyond feature 4, complete linearity is established since the energies
obtained for peaks 5–9, 2
and 3
by measuring from peak 4
, are precisely 6.7 eV below those
of their primed counterparts 5
–9
, 2

and 3

.
The values found for the nine basic features of figure 5 are given in table 1. Remarkably,
this is an exhaustive list of the electron impact spectrum of mercury obtainable with a resolution
of 0.2 eV or so in the range 0 < E
12 eV (see the spectra in [10, 11]).
Peaks 2–5 correspond respectively to the 6 3 P1 , 6 3 P2 , 6 1 P1 , and to a cluster of the 7 3 S1
and 7 1 S0 states of the 5d10 6s configuration of mercury converging to the well known 2 S1/2
ionization potential at 10.4 V. The lowest member of the 6 3 P triplet, 6 3 P0 at 4.7 eV, is not
observed.
Prominent structures commonly displayed in e− –Hg excitation spectra above 8 eV are
attributed nowadays to states of the 5d9 6s2 configuration converging to the 2 D5/2 and 2 D3/2
ionization potentials at 14.8 and 16.7 V respectively. Therefore, feature 6 at 8.8 V is assigned
to the 6p
3 D3 level in that configuration, a metastable state as discussed in [12].
Peaks 7–9 correspond to the three pairwise combinations of features 2 and 3 and would
surely be identified with those were it not that single-impact events at the same energies are
seen in e− –Hg spectra obtained by one method or another.
Single-impact candidates for features 7 and 9 are the 6p
1 P and 6p
3 P states at 9.77 and
11.00 eV, both of which appear strongly in the energy loss spectra of [9]. As to feature 8, a
case has been made of late [13] for a new (metastable) state close to 10.5 eV which is claimed
to be related to a pronounced resonance structure observed [13, 14] at about 10.35 eV. This
neutral state is best described in j – j coupling notation (see table 1).
Critical potentials of mercury with a Franck–Hertz tube 543

Next we turn to feature 1 which is of a different nature. As mentioned earlier, it is

natural to attribute this structure to the ‘shape’ resonance identified in [15] and confirmed
elsewhere [16, 17]. Remarkably, this feature had already been discovered 50 years earlier
with mercury and cadmium in precisely our type of apparatus (with a > 0) and correctly
recognized as being due simply to a variation with energy of the elastic mean free path [18].
A shape resonance of this type is an unstable compound state (negative ion) formed
when an impacting electron is temporarily trapped in a neutral atom by a combination of
polarization and centrifugal barrier potentials [19]. This effect produces an anomaly—in this
case a pronounced asymmetric peak of considerable width—in the total elastic scattering cross
section σT (E), near the energy of the compound state. Comprehensive reviews of experimental
and theoretical studies of negative ion resonances are provided in [20, 21].
How is the shape resonance observed here? Suppose that electrons of energy E are injected
into a field-free region of gas of width x. It is known [22, 23] that at moderate pressure and
in the absence of inelastic events, the current traversing this region is proportional to the total
elastic free path λT (E) and thus inversely proportional to σT (E). Therefore, in a plot of current
versus E, the resonance will appear as a dip and the peak in the cross section should be near
the minimum.
With our calibration of the energy scale, the dip at feature 1 is found to lie near 0.4 eV, in
agreement with the total cross section results of [17]. Thus, arbitrary as it may have seemed,
the procedure of identifying peaks at features 2–9 with thresholds of excited states is consistent
with the determination of the energy of feature 1 at the dip. The absence of feature 1 in the
plots of [5–7] is understandable since a retarding potential (a < 0) was applied.

5. Two complementary detection mechanisms

5.1. The low-potential spectrum

Obviously, the mechanism responsible for observing feature 1 should also operate at thresholds
of excited states whereupon electrons of very low energy, stemming from inelastic collisions
in the low-potential drift space, diffuse through the vapour in the adjacent detection chamber:
a sharp peak should appear near each threshold due to the narrow window for slow-electron
transmission formed below the energy of the resonance. Thus, the entire low-potential spectrum
is formed by pre-eminent collection rather than rejection of the slowest electrons.
The simple formulae of [22, 23] for the diffusion-dominated anode current are only
applicable to a field-free g2 –A region. A generalization of these results for a = 0 was
presented by McMahon [24]. This extended treatment is used in [3] in order to model the
effect of the elastic cross section on the formation of our moving peaks at various pressures
and to demonstrate that a must be zero or slightly positive for optimal operation.

5.2. The stationary spectrum

Regardless of the secondary current of slow electrons, the primary beam (what is left of it) is not
prevented from reaching the anode and its proper variations compose added structure amongst
the slow-electron peaks, as in figures 2–4. But this structure is due to inelastic collisions
occurring in the high-potential drift space as evidenced by their ‘standing’ nature.
A second experiment is being performed simultaneously, but a retarding potential is now
interposed between the scattering chamber and the detector. As a result, critical potentials
are displayed by the complementary detection mechanism as conceived originally, namely the
removal of the slower electrons.
A pure standing spectrum is shown in figure 6. Although the structure is much coarser
than in figure 5, it is clear that the pattern formed by d, e, f , g , . . . is a loose sketch of the
sequence 4
, 5
, 6
, 7
, . . . of figure 5, with successive periods differing by 6.7 V. Finer standing
spectra are obtainable with lower values of the filament voltage V f .
544 P Nicoletopoulos

We also see that a dip labelled z , z
, z

, better displayed in figure 2, shows up in the
standing spectrum. Its position at 0.5 eV or so below d, d
, d

suggests that z may be due
to the structure at E ∼= 12.8–13 eV exhibited in [9, 13, 14]. Analogues of z and z
in the
pure moving spectrum would be situated, inconveniently, inside the pronounced pits preceding
peaks 4
and 4

. On occasion, they have indeed been seen to stand out in curves like figure 5.

6. Untangling the energy scale: the final step

There is still a serious difficulty concerning the absolute location of the pattern of structure in
figure 6 following feature a. If peak d is the analogue of feature 4
, then it should be close
to 13.4 V on the scale of the g1 potential. Instead, we find that it occurs at Va = 10.8 V in
figures 2–4 and slightly higher (11 V) in figure 6.
We are forced to infer that a sudden leap of 2.4–2.6 V in the Va scale has taken place
between a and d so that the space potential has taken the form shown at (b) in figure 7. A
similar instance of abrupt vitiation of the energy scale due to space charge, also with mercury,
has been observed elsewhere with an entirely different apparatus [25]. Remarkably, the jump
of potential reported there (2.7 V) is very close to our values.
The signature for the occurrence of a leap in g1 potential is structure A, the clear
discontinuity in anode current at Va ∼ = 9.5 V of figures 2 and 3, which is weaker but still
discernible in figures 4–6. The higher the cathode emission, the stronger is feature A and the
closer to 2.7 V is the jump in accelerating potential. Conversely, it was observed that if the
cathode is insufficiently activated, feature A is smoothed out, and the entire pattern comprising
b, c, d . . . moves to the right. Thus d advances past 10.8 V. Even so, this peak never reaches
13.4 V, which shows that a significant rise of potential always takes place. Observed values of
δ range from 1.6 to 2.7 V.
Evidently, the uphill region is subject to large potential variations between a and A so
that no standing peaks are seen in this interval. The sudden reappearance of standing peaks
after feature A indicates that a sudden transformation takes place at that point which creates
the field-free plateau (b) of figure 7. A detailed scenario describing this phenomenon is given
in [4].
The essence of this is that the distance between peak a and the subsequent structure of
the high-potential spectrum is affected by the magnitude of the emission current. No wonder
then that the identification of features has been haphazard in the past: of all the peaks that
could have been chosen by earlier authors as the reference for reading the spectrum, peak a is
actually the most unsuitable!
Consider, for instance, the excitation curve of Liu [7]. Starting from peak number ‘2’, the
entire spectrum of [7], including the four unidentified peaks, is found to agree perfectly with
the structure of figure 5, if Liu’s feature ‘4’ is taken as the counterpart of our peak 4
(13.4 eV).
The location of the first maximum (peak ‘1’) in Liu’s curve indicates that a negative cpd of
1.65 V exists between the cathode and g1 . It follows that Liu’s peak ‘4’ (located at 13.25 V)
is really at 11.6 V (13.25–1.65) on the g1 -potential scale. This shows that a jump of potential
of 1.8 V has taken place somewhere between the first and the second peak.
Summing up, it seems that we can safely infer that:
(1) onwards from peak b, the sequence of maxima—or should we say downturns?—of the
stationary spectrum is identical to the sequence of peaks of the moving spectrum, and
(2) peak a is unsuitable for calibrating the g1 energy scale because a jump of space potential
inevitably occurs before the emergence of the remaining standing structure.

7. Space charge phenomena

Potential configurations resembling feature (a) of figure 7 are well known in plasma physics.
A narrow zone encompassing most of the potential drop between wide field-free regions at
Critical potentials of mercury with a Franck–Hertz tube 545

different potentials is created by the strife among the various components of space charge to
attain stress balance. Such zones are composed of two adjacent sheaths of opposite space
charge excess and are thus called ‘double layers’ or simply DLs. Several kinds of DLs have
been studied with various devices in laboratory experiments, computer simulations and in the
formation of auroras [26, 27].
A bulging high-potential region similar to that of (b) in figure 7 is often seen to accompany
DL formation in gaseous discharge experiments (see for instance [27], p 65). At the pressure
used in this experiment, this phenomenon always takes place. As a result, the high-potential
plasma is inevitably above the potential of grid 2, regardless of the value of g. Hence the
drift space producing the standing spectrum is always followed by a decelerating space charge
sheath.
In contrast, no significant jump analogous to δ (namely a uniform depression below
g2 potential followed by an ascending sheath) takes place on the low-potential side. This
asymmetry is due to a fundamental difference in the dynamics of formation of the high- and
low-potential plasmas [4].
The superior detection mechanism involving the low-energy shape resonance can only
operate if the drift space preceding the detection chamber is very close to g2 potential.
Therefore, the success of this experiment hinges upon the creation and subsistence of the
low-potential plasma. The presence of this plasma requires (a) an adequately large beam
current density and (b) a sufficiently negative intergrid potential.
The current density requirement is understandable: the cathode beam produces plasma at
the exit of the electron gun and must remain strong enough after traversing this region to create
a second plasma close to the outer grid. The initial formation of this plasma occurs just below
feature 2. At that point the beam energy at g2 exceeds 4.7 eV and thus positive ions become
available in that neighbourhood by cumulative ionization thanks to the metastable nature of
the 6 3 P0 state.
The problem is that shortly afterwards, at feature A, a strong transformation takes place on
the high-potential side which produces the field-free plateau of figure 7(b). The charge density
in this region increases considerably because single-impact ionization becomes energetically
possible. As a result, the DL is pushed outwards and the useful low-potential plasma tends to
disappear. This is where the requirement on g comes in.
One can estimate the lowest value of the DL amplitude ψ = |U + g| + δ that will
sustain the display of moving peaks. As we saw in section 3, for both plasmas to be present
we need g < 0.4. Assuming that δ ∼ = 2.6 and U ∼ = −2.5 we find that, past feature A,
ψ > 4.7.
The remarkable closeness of this lower bound to the excitation potential of the 6 3 P0 state
suggests that the values of |V | sustaining the presence of a DL inside the intergrid volume
should exceed the threshold of production of metastable atoms. In that case an additional
internal discharge is initiated: a fraction of the slow electrons produced by inelastic scattering
in the low-potential region are accelerated backwards towards the high-potential plasma by
the DL. When the amplitude of the DL exceeds 4.7 V, the threshold for cumulative ionization,
the backflowing beam can confine the high-potential plasma by modifying the composition of
space charge at the expanding front.

8. The influence of gas pressure and electron emission current

Irrespective of the value of g, the intensity and definition of the structure displayed in our
curves are conditioned by the pressure of mercury vapour and by the state of activation and
temperature of the cathode. This section is a description of the changes induced by varying
each of these three parameters assuming that a value of g in the optimal range (−1.5 to −1.8)
is operant.
546 P Nicoletopoulos

8.1. Mercury vapour pressure

At room temperature, at most some frail structure corresponding to peaks 2–4 is discernible
in the anode current, feebly modulating a smoothly rising curve.
As the oven temperature is raised, the average anode current is attenuated, but the overall
form of the plots is not affected. An end to this ‘smooth’ regime takes place abruptly at
about 110◦ when the sudden drop at feature A appears in the curve. Structure is now clearly
observable in the portion of the curve past the break A. Near 120◦ the features become
narrower and more numerous. At 130◦ the peaks are clear enough for the curve to look like
a weak replica of figure 5 but the lowest excitation peak displayed is peak 4. The curve then
evolves into its optimal form, attained, as in figure 5, around 160◦ . Above this temperature
no significant changes are observed until, past 180◦ , the peaks become narrower and their
amplitude diminishes perceptibly.
Feature 1 is brought to view as a peak followed by a clear dip at about 120◦ . Below this
temperature it looks more like an inflection and is very sensitive to a.

8.2. Cathode state of activation

At any value of filament voltage, the emission characteristic of the cathode is dependent on
its state of activation; a measure for the latter, denoted Ic , was defined in section 2. If a
tube remains idle for several weeks, Ic dwindles, sometimes by as much as 50-fold. Such
‘poisoning’ [28] of the cathode prevents the occurrence of the moving spectrum despite the
application of a negative value of g.
Fortunately, a process for reactivating a poisoned cathode was discovered in the course
of the early tryouts. In an attempt to determine the effect of higher values of filament voltage
V f , it was observed that for V f ∼ = 7 V, as the voltage Vhel of the accelerating helipot is
cranked upwards, a critical value, say Vosc , of the voltage Va across g1 is reached whereupon
an oscillatory behaviour sets in: instead of growing past this value on the graph, Va drops
suddenly by a few tenths of volt, rises less rapidly back to Vosc , drops again abruptly and so on.
When V f is subsequently lowered, stable behaviour is restored, but upon cooling the tube, Ic
is found to have grown at least tenfold, an obvious indication that substantial activation of the
cathode has taken place. Of course, raising V f can increase the current by orders of magnitude
and might thus appear to be hazardous for the life of the tube; all the same, this procedure has
been employed time and again with no visible damage, whenever reactivation was desirable.
No useful activation has ever been achieved unless oscillations actually set in. The
procedure is most effective when T and V f exceed certain limits, which seem to depend
on the degree of poisoning; usually T > 170◦ works best and V f ∼ = 7 V will suffice, but
occasionally higher values of V f (up to 9 V) have been necessary.
The minimal value of Va at which a poisoned tube will oscillate is usually the voltage of
feature A (about 9.5 V) but it can be as high as 11–12 V in cases of severe poisoning. To
reactivate, Vhel should be raised until oscillations set in and then be held fixed when the rate of
oscillations has reached about 1 Hz. If reactivation is becoming effective, then upon bringing
Vhel back to zero and gradually increasing it again, it will be found that the value of Vosc has
retracted to 9.5 V. When good activation is finally attained, the oscillating regime may even be
triggered at peak a.
For our purposes, activation is optimal when Ic is between 1000 and 1500 µA, but values
as low as 600 µA have been seen to be acceptable.
Activation is not always a rapid procedure. In one of our two model-55580 tubes, called
tube I, a dozen or so oscillations at feature A have commonly been sufficient, even when
poisoning is severe. Tube II, however, usually had to be left oscillating for minutes or even
hours for Ic to recover adequately. Two older-model tubes (Leybold–Heraeus 55080) which
had remained inactive for several years could not be activated beyond Ic ∼ = 200 µA, which
was inadequate for producing the moving spectrum with any tube at any value of g.
Critical potentials of mercury with a Franck–Hertz tube 547

Once a tube’s cathode has been sufficiently activated, Ic stays remarkably constant for
hours on end. And although Ic diminishes somewhat, usually no reactivation is required for
several months if the experiment functions frequently. When a tube was left inactive for
months, poisoning invariably took place.
An oxide cathode is a metal core, usually nickel, coated with a mixture of the alkaline earth
oxides CaO, SrO, BaO. Activation consists in the formation in the coating, by ion impact [28],
of excess atoms of Ca, Sr and Ba. The sequence of events underlying the phenomenon of
relaxation oscillations probably breaks up space charge limitation of the ion current flowing
to the cathode and produces puffs of increased ion bombardment which induce activation in
the coating [4].

8.3. Filament voltage

When electrons escape from a thermionic cathode, their velocity distribution is half-
Maxwellian. Hence its width is proportional to the temperature of the cathode and we would
expect the energy resolution to improve with decreasing filament voltage V f . This effect is
indeed perceptible: it was verified that the peaks in the curve of figure 5 are conspicuously
sharper than they would have been, had we used a higher value of V f . On that basis it would
appear that lowering V f would be profitable. In reality, this is only partly true. As V f is
reduced below 3.5 V, the peaks past feature 4 in figure 5 are indeed sharpened. On the other
hand, however, in the portion of the curve below peak 4, the resolution declines.
Insofar as it is desirable to display a maximum of structure throughout the energy range
with good resolution, the setting of V f should be the best compromise between the narrowing
of the energy distribution and the smearing of the early moving peaks. In both our tubes the
best range was between 3.3 and 3.6 V.

9. Conclusion

This mode of medium-pressure electron impact spectroscopy was introduced in the literature
more than 80 years ago. It is surprising, therefore, that despite the progress of physics, modern
accounts [2, 7] of the experiment do not differ substantially from those of the original authors.
Perhaps the reason for this is that while the scope of the experiment is in atomic physics and
quantum mechanics, its conception really belongs to the plasma field of gaseous discharges, a
complex and less well known subject.
As a result, it is not easy to make this a routine experiment for undergraduate university
laboratories, unless an apparatus is developed that provides completely straightforward access
to plasma formation at the second grid. Among the types of Leybold tubes, only the 55580,
duly conditioned, is guaranteed to produce the moving spectrum. Older tubes did not, as I
pointed out. A newer-model tube (55585) acquired in 1987 was briefly tried with no success.
Since then the author has not carried out experimental work on this matter.
Readers whose tube is only able to produce the standing excitation spectrum are at least
supplied, for the first time, with the interpretation of the observed critical potentials, and with
the correct way to calibrate the energy scale. In addition, direct observation of the resonance
at feature 1 should always be possible with non-negative a and negative g.

Acknowledgments

I am grateful to Paul Burrow for several motivating overseas conversations in the early stages of
this venture and for sending me an unpublished electron transmission spectrum of mercury. As
to plasma double layers, I am indebted to Jean Wallenborn for introducing me to the literature
and to Hans Schamel for teaching me about them and for his constant concern. Of course
without Martine this work would never have been completed.
548 P Nicoletopoulos

References
[1] Harnwell G P and Livingood J J 1933 Experimental Atomic Physics (New York: McGraw-Hill) pp 317–18
[2] Melissinos A C 1966 Experiments in Modern Physics (New York: Academic) pp 8–17
[3] Nicoletopoulos P The 0.4 eV shape resonance of electron scattering from mercury in a Franck–Hertz tube,
to be submitted
[4] Nicoletopoulos P Theory of a strong double layer in a Franck–Hertz tube, to be submitted
[5] Mohler F L, Foote P D and Meggers W F 1920 J. Opt. Soc. Am. 4 364
[6] Einsporn E 1921 Z. Phys. 5 208
[7] Liu F H 1987 Am. J. Phys. 55 366
[8] Moiseiwitsch B L and Smith S J 1968 Rev. Mod. Phys. 40 267
[9] Skerbele A, Ross K J and Lassettre E N 1969 J. Chem. Phys. 50 4486
[10] Schulz G J 1958 Phys. Rev. 112 150
[11] Borst W L 1969 Phys. Rev. 181 257
[12] Borst W L 1977 J. Chem. Phys. 67 2920
[13] Newman D S, Zubek M and King G C 1985 J. Phys. B: At. Mol. Phys. 18 985
Zubek M and King G C 1990 J. Phys. B: At. Mol. Opt. Phys. 23 561
[14] Burrow P D 1980 private communication, unpublished
[15] Burrow P D, Michejda J A and Comer J 1976 J. Phys. B: At. Mol. Phys. B 9 3225
[16] Johnston A R and Burrow P D 1982 J. Phys. B: At. Mol. Phys. 15 L745
[17] Jost K and Ohnemus B 1979 Phys. Rev. A 19 641
[18] Minkowski R 1923 Z. Phys. 18 258
[19] Biondi M A, Herzenberg A and Kuyatt C E 1979 Phys. Today 32 44
[20] Schulz G J 1973 Rev. Mod. Phys. 45 378–422
[21] Buckman S J and Clark C W 1994 Rev. Mod. Phys. 66 539–655
[22] Langmuir I 1931 Phys. Rev. 38 1656
[23] Chen F F 1965 Plasma Diagnostic Techniques ed R H Huddleston and S L Leonard (New York: Academic)
pp 152–5
[24] McMahon D R A 1983 Aust. J. Phys. 36 27
[25] Jongerius H M 1962 Philips Res. Rep. (Suppl.)
[26] Schamel H 1986 Phys. Rep. 140 161–91
[27] Raadu M A 1989 Phys. Rep. 178 25–97
[28] Herrmann G and Wagener S 1951 The Oxide-Coated Cathode vol 2 (London: Chapman and Hall) ch 6