Physica B 444 (2014) 44–48

Contents lists available at ScienceDirect

Physica B
journal homepage: www.elsevier.com/locate/physb

More precise determination of work function based on Fermi–Dirac

distribution and Fowler formula
Liu Changshi
Nan Hu College, Jiaxing University, Zhejiang 314001, PR China

art ic l e i nf o a b s t r a c t

Article history: More precise numerical method to simulate current–voltage of metal at fixed temperature is presented
Received 24 January 2014 in this paper. The new algorithm for the simulation has been developed via Fermi–Dirac distribution step
Received in revised form by step. These calculated characteristics are shown to remain in excellent agreement with the experi-
15 March 2014
mental ones, taken for a range of different metals, which strong supports the validity of the model. It is
Accepted 18 March 2014
Available online 26 March 2014
also shown that based on the Fowler formula, higher precise work function can be determined.
& 2014 Elsevier B.V. All rights reserved.
Keywords:
Current
Voltage
Thermionic current
Simulation
Fowler formula.

1. Introduction behavior in the electrode region, which would contribute to deve-

Although the purely thermal electron emission was not origin-
ally developed for the sake of interpreting the work function of
metal [1–3], the work functions of many metals have been
detected via thermal electron emission [4–12]. Thermionic emis-
sion is important in the context of development of cathodes [13]
and has been investigated extensively for variety of applications
such as X-ray tubes, high power microwave device, cathode ray
tubes, and fluorescent lamps [14,15]. Historically, there are many
applications of electric arcs in industrial processing of materials,
for example in welding, cutting, thermal spraying, heating, surface
treatment and chemical vapor deposition (CVD) [16,17]. In general,
thermonic electron emission from the cathode surface is a func-
tion of both the cathode temperature and the work function of the
electrode materials, which is today known as the Richardson–
Dushman equation. Since thermonic emission requires a very high
surface temperature close to the melting point of metal, energy
transfer from the free-burning arc to the cathode electrode is
necessary for maintaining this high surface temperature. For
example, ions accelerated by the electric field between anode
and cathode fall zone.
More precise measuring the work function of metals would
be extremely helpful for a better understanding of the physical
E-mail address: liucs4976@sohu.com
lopments of advanced control and evaluation system for materials
processing by the arcs through the prediction of properties of the
total arc processing system [18]. The more precise thermonic
current can be obtained, the more precise work function of metals
may be calculated, and the behavior of the emitted current
depends on the strength of the electric field applied to remove
and detect the emitted electrons. At low field strengths which
refer to voltage is p lower
ffiffiffiffiffiffiffiffiffi than 25 V, experimental plot IðV; TÞ
against Ið0; TÞexpðp1 V=T Þ cannot explain the experimental curve
constructed via I and V since the electrons are not removed rapidly
enough and their repulsive charge distribution slows the rate at
which electrons can actually reach the detector. When the applied
electric field strength is large which means voltage is higher than
25 V, the work function of the material is decreased, and the
measured emitted current is related to the actual current by the
following relation:
pffiffiffiffi!
p1 V
IðV; TÞ ¼ Ið0; TÞexp
T
The Richardson–Dushman equation is a convenience; in practice,
however, there is still a lack of practical method to determine the
work function more precise than that of by Richardson–Dushman.
Hence, a model is needed which would allow prediction current of
metal at one temperature by voltage at the highest precise level.
Such a model is developed and verified experimentally in this
paper. In this article the main features of the Fowler formula on

http://dx.doi.org/10.1016/j.physb.2014.03.037
0921-4526/& 2014 Elsevier B.V. All rights reserved.
 L. Changshi / Physica B 444 (2014) 44–48 45

calculation work function are also concerned using the most Experimental results of field emission tell that as the electric
precise results of thermal electron emission. potential of cathode approaches low enough relative to anode, the
anode current, I(V), goes asymptotically to a constant value know
as the forward saturation current, Imax:
2. Mechanism between current and voltage
IðVÞjV-1 ¼ I max ð9Þ
In the past, electron transport over the work function from the
surface of metal is described by thermionic emission theory (TE) Boundary condition (9) predicts that the field emission current
based on the Schottky effect with the expression: and voltage should be related as
pffiffiffiffi!
p V I2
IðV; TÞ ¼ Ið0; TÞexp 1 ð1Þ IðVÞ ¼ I 1 þ   ð10Þ
T 1 þ exp V 
V0
Vc

where I(V,T) is the current generated by both electronic field and

heating, I(0,T) is thermionic current, in other words, thermionic Combining Eqs. (8), (9) and (10) yields
current is electron beam emission only by heating. T is absolute
temperature of filament, p1 is constant controlled by the structure I 1 ¼ I max ð11Þ
of both anode and cathode, the V is the voltage between anode
and cathode. One notable is that the electric potential of cathode
Substitution of Eq. (11) into Eq. (10) leads to following expres-
must be lower than or equal to that of anode in function (1). For
sions:
thermalp electron
ffiffiffiffi emission, however, the nonlinearity in LnðIðV; TÞÞ
against V cannot be interpreted via Eq. (1). I2
IðVÞ ¼ I max þ ð12Þ
Because electron is fermions, the Fermi–Dirac distribution 1 þ expððV  V c Þ=V 0 Þ
should apply to electron. [3,19] Fermi–Dirac distribution is usually
where Imax is the forward saturation current predicted via Eq. (12),
written:
I2 is the current constant expected by Eq. (12), Vc is the inflection
1 point voltage, when the voltage is less than Vc, the I–V curve is
f ðEÞ ¼ ð2Þ
1 þ expððE  Ef Þ=kTÞ concave, while, as long as the voltage is higher than Vc, the I–V
curve is large convex, meanwhile, the average value of the field
where f(E) is the probability that a particle will have energy E, Ef is
emission current is fixed at Vc, the V0 is constant of voltage. Imax, I2,
Fermi energy, T is Absolute temperature, k is Boltzmann constant
Vc and V0 will be optimized.
and its value is k ¼ 1:38  10  23 in SI unit.
In according to the definition of thermionic current, when
The kinetic energy of thermion electron is eV  eφ  eδ after it
there is no any accelerating field which means that the voltage
has been liberated from the metal and enter into the external
between anode and cathode is zero in Eq. (12), the mechanism of
electronic field with V. Where e¼ electron charge¼ 1.602  10–19,
the thermionic current can be read:
eφ is the work function of the metal. eφ is the energy you must
give to an electron at the Fermi level to kick it out of the metal and I2
I 0 ðTÞ ¼ Iðo; TÞ ¼ I max þ ð13Þ
turn it into a free election. An electron below the Fermi level needs 1 þ expð  V c =V 0 Þ
more than eφ to escape, δ is an additional contact potential
because the surfaces of the anode and cathode are different. An
electron now has kinetic energy eV  eφ  eδ, thus the probability
that an electron will have the kinetic energy eV  eφ  eδ is 3. Application

1 1
f ðEÞ ¼ ¼   ð3Þ The anode current against voltage curves was collected at each
1 þ expððE  Ef Þ=kTÞ 1 þ exp ðV  δ  ϕÞ U kT
e
different temperature of tungsten as filament (cathode). Use the
let commercial measurement system for work function typed ZN17-
SLWY-III to sweep the anode voltage from 16 V to 122 V. I–V plots
Vc ¼ δþφ ð4Þ of the experimental data for tungsten (W) are shown in Fig. 1.
In this paper, the field current curves of tungsten were analyzed
kT
V0 ¼ ð5Þ at different temperatures in accordance with the above Eq. (12).
e Then through experimental data employed in the method of the
regression analysis, the experimental curves for W at various
Function (3) become temperatures have been simulated using the experimental data as
1 1 shown in Fig. 1. Optimized parameters employed to simulate the
f ðEÞ ¼  ¼   ð6Þ component curves are also listed in Table 1. The modeled verifica-
1 þ exp ðV  δ  ϕÞ U kT
e
1 þ exp V  V c
V0 tion is also carried out by a comparison of modeled field current
obtained by numerical integration of field current to measured field
One natural conclusion is that current liberated by electric field at
temperature of T, I(V), is proportional to the probability, f(E), that a current in Fig. 1.
Although the current of Tungsten has been calculated via
electron will have energy eV  eφ  eδ, in mathematical language:
voltage at several temperatures, this does not give enough idea
I2 about how accuracy of the methodology suggested in this paper.
IðVÞ ¼   ð7Þ
V  Vc To test for the determination of the field current of any kind of
1 þ exp V0
metal using experimental data of field current, Experimental I–V
data of Potassium is taken from published paper and displayed in
But mathematical result shown that: Fig. 2 [20]. At the same time, all of the data in Fig. 2 was fitted to
1 the functional (12) and illustrated as cross in Fig. 2. The best
 jV -1 ¼ 0 ð8Þ parameters to obtain the best results of calculation are also listed
V  Vc
1 þ exp V0 in Table 1.
46 L. Changshi / Physica B 444 (2014) 44–48

Fig. 1. Experimental and theoretical voltage depends on current at several temperatures for tungsten. The symbols ○ refers to experimental results, þ and  stand for
theoretical values by Eqs. (12) and (1).

Table 1 I im is measured data, and I is stands for the value obtained by

Parameters used in the simulation to obtain the thermion current of potassium simulation. The maximum value of ARE is 7.03%. As is shown in
(K) and tungsten (W) as a function of the voltage at various temperatures (T). Table 1, Figs. 1 and 2, it can be stated that a satisfactory agreement
Metal T (K) Function form (10  3A) R of ARE of R of ARE
between the measured and the modeled current is achieved by the
Eq. Eq. Eq. of Eq. functions given in Table 1.
(12) (12) (1) (1)% To compare the precise between the results obtained from Eq.
(1) and Eq. (12), all of the data in Figs. 1 and 2 was also fitted to the
W 1880 IðV Þ ¼ 40:68 1 þ expððV 12:0 0.998 0.28 0.990 0.64
22:54Þ=31:81Þ functional of Schottky Equation (1). As shown in Table 1 and
1912 IðV Þ ¼ 70:74 1 þ expððV þ134:55 0.999 0.48 0.992 0.57
218:61Þ=105:47Þ Figs. 1 and 2, one can then find that the relationship between field
1944 IðV Þ ¼ 107:52 1 þ expððV 297:56
þ 152:18Þ=57:84Þ
0.996 0.32 0.991 0.50 current and voltage given by function (12) is more precise than
1976 IðV Þ ¼ 172:34 1 þ expððV 328:63
þ 149:05Þ=67:09Þ
0.996 0.19 0.996 0.34 expressed by function (1).
2008 IðV Þ ¼ 256:75 1 þ expððV2537:26
þ 123:39Þ=31:84Þ
0.986 0.61 0.956 1.10 In way labeled (13) the intercepts of thermionic current for
2040 IðV Þ ¼ 406:64 1 þ expððV861:32
þ 145:29Þ=63:0Þ
0.999 0.17 0.995 0.38 Tungsten and Potassium in Figs. 3 and 4 were all determined
2072 IðV Þ ¼ 598:57 1 þ expððV1713:76
þ 160:51Þ=60:46Þ
0.998 0.23 0.994 0.42 precisely. Since the thermionic current, obtained by Eq. (13) was,
K 455 IðV Þ ¼ 3:87 1 þ expððV þ30:14
368:46Þ=187:46Þ
0.994 4.42 0.966 20.78 therefore, measured experimental, the effective work function
463 IðV Þ ¼ 7:56 1 þ expððV þ96:26 0.998 4.42 0.984 15.52 during operation should be estimated by fitting with Fowler`s
697:31Þ=276:48Þ
482 IðV Þ ¼ 9:51 1 þ expððV þ272:87 0.994 6.73 0.97 22.16 theoretical function, given by
549:081Þ=159:65Þ
8  

490 IðV Þ ¼ 15:85 1 þ expððV þ284:46 0.994 7.03 0.977 14.82 >
>
N ð 1Þn  1 hν eχ hν  eχ
551:39Þ=184:89Þ > AT 2
pffiffiffiffiffiffiffiffiffiffiffiffiffi !0
496 IðV Þ ¼ 23:70 1 þ expððV þ147:25 0.998 2.43 0.994 4.22 < eχ 0  hνn∑
>
¼1 n 2
exp n
kT kT
411:85Þ=215:78Þ
Iðν; TÞ ¼  2 N ð  1Þn  1  

>
>
> AT 2 π 1 hν  e χ hν eχ hν eχ
p ffiffiffiffiffiffiffiffiffiffiffiffiffi
2
>
: ½ 6 þ 2 kT ∑ exp n Z0
R: correlation coefficient, ARE: average relative error. eχ 0  hν n n2 kT kT

In order to verify the accuracy of the simulation, the correlation
coefficient between the measured and the simulated data is given
in Table 1; the minimum magnitude of the correlation coefficient
is 0.986. The average relative error (ARE) 1n∑ii ¼ n jI im  I is j
¼ 1 I im
evaluating the simulation results is also shown in Table 1, where
ð14Þ
where A is a coefficient incorporating constants and instrumental
parameters, eχ is the work function, eχ 0 is written as eχ 0 ¼ eχ þ Ef ,
 100% T is absolute temperature; k is Boltzmann constant and hν is the
energy of photon [3].
 L. Changshi / Physica B 444 (2014) 44–48 47

Fig. 2. Comparisons between experimental and calculated voltage–current curve of potassium at several temperatures. The symbols ○ refers to experimental results data
taken from Young, þ and  stand for theoretical values by Eqs. (12) and (1).

Fig. 3. Geometry used to determine the work function of tungsten by Fowler`s Fig. 4. Numerical results of the thermion current of potassium as a function of the
theoretical function. The symbols þ is simulation via Eq. (15), ○ refers to the temperature.
experimental data obtained by Eq. (13).

At finite temperatures, because the cathode was not irradiated A plot of I(0,T) again T is known as a ‘‘Fowler plot,’’ and by fitting
by any light, the work function (in eV) can then be extracted from data to the universal curve I(0,T) using function (15) one obtains
the Fowler as the work function. Figs. 3 and 4 containing experimental I and T
were used to fit to the Fowler`s law functional form (15). Figs. 3 and 4
N ð  1Þn  1  eχ  eχ 
Ið0; TÞ ¼ BT 2 ∑ exp  n n !0 ð15Þ represent fits with Fowler theoretical function of Eq. (15), and the
n n2 kT kT effective work functions of Tungsten and Potassium are determined
48 L. Changshi / Physica B 444 (2014) 44–48

Table 2 state that the Fowler equation (15) is very much better. The
Results and justification of simulation to obtain the work function of potassium (K), experimental results of thermionic current for Platinum are taken
tungsten (W) and platinum (P) via Fowler equation.
from paper published and shown in Fig. 5 [21], the conclusion of
Metal n P2 by I0 from (13) χ(eV) R P2 by I0 from (1) φ(eV) R Platinum is reached successively by the Fowler equation (15) and
given by Table 2 and Fig. 5. Again, sufficiently precise value of the
K 1 13696.91 1.18 0.947 10439.25 0.90 0.977 effective work functions for Platinum could be found out as long as
2 13698.03 1.18 0.947 10439.91 0.90 0.977 n is equal to or greater than 3.
3 13698.02 1.18 0.947 10439.91 0.90 0.977
4 13698.02 1.18 0.947 1043.91 0.90 0.977
W 1 25209.73 2.17 0.905 43843.98 3.78 0.996
2 42523.39 3.66 0.993 47717.77 4.11 0.999 4. Conclusions
3 45155.78 3.89 0.996 48517.78 4.18 0.999
4 45155.78 3.89 0.996 48517.78 4.18 0.999 It appears from the experimental and theoretical curves shown
P 1 20322.60 1.75 0.903
above that the current–voltage for metal at fixed temperature is in
2 22719.34 1.96 0.916
3 22715.77 1.96 0.916 complete quantitative agreement with experiment, whether the
4 22715.77 1.96 0.916 current–voltage for metal at fixed temperature be analyzed by
current–voltage equation suggested in this paper from Fermi–
Dirac distribution, or the most precise value of thermionic current
can be obtained via application the current–voltage equation
suggested in this paper. The theory suggested in this paper
therefore makes a step forward in our understanding of electric
field and thermionic electron emission processes. It is particularly
fortunate that Fowler has presented his theory in a form which is
both understandable and useful to experimental physicist in
determining work function, and it may be expected to play an
important role in future experimental study.

References

[1] O.W. Richardson, Proc. R. Soc. A. 9119 (15) 524.

[2] Saul Dushman, Rev. Mod. Phys. 2 (4) (1930) 381.
[3] R.H. Fowler, Phys. Rev. 38 (1931) 45.
[4] S.K. Kolekar, S.P. Patole, P.S. Alegaonkar, J.B. Yoo, C.V. Dharmadhikari, Appl.
Surf. Sci. 257 (23) (2011) 10306.
[5] Franz A.M. Koeck, Robert J. Nemanich, Yasodhaadevi Balasubramaniam,
Ken Haenen, Jeff Sharp, Diam. Relat. Mater. 20 (2011) 1229.
[6] W. Schlager, M. Neiger, J. Phys. D: Appl. Phys. 33 (2000) 3083.
Fig. 5. Scatter plots of both the thermion current of Platinum expected by the [7] X. Zhou, J. Heberlein, J. Phys. D: Apply. Phys. 31 (1998) 2577.
[8] R. Liyoshi, H. Shimoyama, S. Maruse, J. Electron Microsc. 45 (1996) 514.
Fowler formula against the temperature and the experimental results given by Eq.
[9] J. Haidar, A.J. D. Farmer, J. Phys. D: Appl. Phys 28 (1995) 2089.
(13).
[10] J. Haidar, A.J. D. Farmer, Rev. Sci. Instrum. 64 (1993) 542.
[11] M. Ushio, A.A. Sadek, F. Matsuda, Plasma Chem. Plasma Process. 11 (1991) 81.
[12] M I Boulos, P Fauchais, E Pfender, Thermal Plasmas, Plenum, New York, 1994
from these fits and listed in Table 2 corresponding n equate to 1, 2, 3 (Michaelson H. B).
[13] J F Lancast (Ed.), The Physics of Welding, PergamonOxford, 1984.
4. It is easy for one to know that when the n is 1, the Fowler equation [14] Herbert B. Michaelson, J. Appl. Phys. 48 (1977) 4729.
(15) becomes Richardson–Dushman equation. It will be seen from [15] Fomenko V. S., Naukova Dumka, Kiev. Emission Properties of Materials, 1970
Table 2 that the effective work functions of Potassium is in perfect (in Russian).
[16] C. Herring, M.H. Nichols, Rev. Mod. Phys. 21 (1949) 185.
alignment whether it was calculated by the Fowler equation (15) or
[17] A.L. Reimann, Proc. Phys. Soc. Lond. 501 (1938) 496.
by Richardson–Dushman, but the effective work functions of Tung- [18] Manabu Tanaka, Masao Ushio, Mitsuru Ikeuch, Yoshiro Kagebeyashi, J. Phys. D:
sten can be obtained most precise until n equal to 3. Appl. Phys. 38 (1) (2005) 29.
Perhaps it is still too early to assert with absolute confidence [19] C. Kittel, Introduction to Solid State Physics, seventh ed., John Wiley and Sons
(Asia), Singapore (1996) 146.
the general and exact validity of the Fowler equation. Never- [20] A.F.A. Young, Proc. R. Soc. A 104 (1923) 611.
theless, it must be admitted that more application is needed to [21] Saul Dushman, Phys. Rev. 21 (1923) 623.