HAL Id: jpa-00249078
https://hal.archives-ouvertes.fr/jpa-00249078
Submitted on 1 Jan 1993
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Ballistic electron emission microscopy of the Au-Si (100) interface
Roland Coratger, François Ajustron, J. Beauvillain
To cite this version:
Roland Coratger, François Ajustron, J. Beauvillain. Ballistic electron emission microscopy of the Au-Si (100) interface. Journal de Physique III, EDP Sciences, 1993, 3 (12), pp.2211-2220.
�10.1051/jp3:1993270�. �jpa-00249078�
J. Phys. III France 3 (1993) 2211-2220 DECEMBER 1993, PAGE 221
Classification Physic-s Abstiacts
61.16P 68.20 73.40N
Ballistic electron emission microscopy of the Au-Si (100)
interface
R. Coratger, F. Ajustron and J. Beauvillain
CEMES-LOE/CNRS. 29 rue J. Marvig, B-P. 4347, 31055 Toulouse, France
lReceii,ed 24 May 1993, revised 2 Septembei1993, accepted 24 September 1993)
Abstract. In this paper, experiments performed on Au-Si ( loo) junctions by Ballistic Electron Emission Microscopy (BEEM), a method for studying the metal-semiconductor interface, are presented. BEEM spectra show a characteristic threshold that yields the Schottky barrier height at the interface. The average barrier height is found to be lower than the usual values of Au-Si
contacts. BEEM images result from the ~patial variations of the current in the semiconductor and
present unusual local contrasts. These phenomena are analysed in terms of interface quality.
Introduction.
Over the last decade, Scanning Tunneling Microscopy (STM) [II and Atomic Force
Microscopy [2] (AFM) have opened up new fields of investigation in surface science. New
applications in superconductivity, biology, as well as molecular chemistry have shown the
possibilities of these novel techniques for observations at the atomic scale [3]. In solid state
physics and surface science, the images are interpreted in terms of surface roughness, surface
structures or surface reconstructions for semiconductors and metals. The information only
concerns the surface itself or the first layer below this surface contrary to techniques such as
Transmission Electron Microscopy where information is averaged on a certain thickness.
In 1988, Kaiser et al. used Ballistic Electron Emission Microscopy (BEEM), a new
technique derived from STM, to study the interface between a metal and a semiconductor [4].
For the first time, a near field microscope yielded depth information on a material. This type of microscopy relies on the following principle electrons ballistically injected in a thin metal
layer may overcome the Schottky barrier (eV~) between this metal and a semiconductor, provided the electrons have a sufficient incident energy and the thickness of the metal film is not too large compared to the electron mean free path in the bulk material. The kinetic incident
energy of the electrons is given by the voltage between the tip and the metal surface and the Fermi level of the electrodes. Current in the semiconductor (I~) is therefore zero when the
voltage is less than V~ and increases when V exceeds V~. In the I~ vs. V spectrum, the threshold voltage allows direct determination of V~ (BEES : Ballistic Emission Electron Spectroscopy).
By applying a constant bias voltage above the Schottky barrier height, it is possible to detect the spatial variations of the collector current in the semiconductor while scanning the tip on the
metal surface. A BEEM image consists of a grey scale representation of these local variations.
In this paper, BEEM experiments carried out on thin gold films evaporated on Si (100) are presented. The spectroscopic measurements show that the barrier height is relatively uniform.
The Kaiser and Bell model is used to fit the experimental curves. The average Schottky barrier
height deduced from these experiments is found to be lower than the values usually yielded by
BEEM on Au-Si junctions. Our BEEM images also demonstrate the weak role of surface
roughness in the electrical characteristics of these Schottky junctions. They also show unusual
large domains mainly attributed to interface defects.
Experimental.
BEEM experiments have been performed on a home-made STM using a tripod of piezoelectric ceramics already described elsewhere [5]. Nevertheless, the electronic setup has been slightly
modified to detect the collector current (typically a few tens of picoamperes) and to provide a
variable bias voltage between tip and sample. The sign, frequency as well as range of this
voltage (from 0 to 10 V) can be modified by the user. A frequency of about fifteen spectra per
minute appears to be the most suitable. This setup has therefore been used during the
spectroscopic measurements and a spectra sequence was stored in a computer for possible
further processing. Each spectrum results from the collector current variations as a function of
sample bias voltage and at a constant tunneling current between tip and sample. BEEM images
consist of the variations of the collector current while scanning the surface at a constant voltage above the Schottky barrier height (typically from I to 1.5 V) and at a constant tunneling current
(from I to 5 nA).
A Si loo) substrate is HF cleaned before introduction into the preparation chamber. This
substrate is of the n type with a doping of 2 x 10~~ cm~~ The
vacuum in the preparation
chamber is kept in the 10~ ~-10~~
torr range. An Ausb ohmic contact is achieved on the back of the wafer to allow collector current detection. The Au films are then deposited on 0.8 mm
diameter disks with a thickness ranging from 5 to l~.snm. Each diode is therefore a
I x I mm~ of Si loo) with a circular Au film in its middle (Fig. I). An electrical contact is
made before each BEEM observation by bonding a small 50 ~Lm diameter copper wire on a
corner of the circular gold film.
,,,,,,,,,,,,,,,
'jjjjqjpjjji'
lit,~,~,ii' 'lxxx'
'xxi' ~t
'jj t
Au fihn ,
~w n~si(lW)
Onfldc contact
l'
Fig. I. Schema of the Au-Si (100j Schottky diode used in the experiments. The tunneling current flows between the tip and the metal surface. Collector current is detected in the semiconductor through an
ohmic contact achieved on the back of the silicon substrate.
N° 12 BEEM OF THE Au~si loo) INTERFACE 2213
Results.
SPECTROSCOPY. Because of the potential barrier at the metal-semiconductor interface, the SEEM spectra should present a characteristic threshold and a strong increase of the collector
current beyond this value. Figure 2 shows the average over fifty spectra taken with a tunneling
current of 5 nA and a mean bias voltage of 300 mV (Au thickness on the silicon : lo nm). For
clarity, the spectrum has been represented for a minimum voltage of 0.5 V. The collector
current is equal to the noise of the diode and the electronic setup below 0.7 V (~ l pA).
Beyond this value, I~ slowly increases at first, more strongly afterwards and is about 400 pA at 1.3 V. This corresponds to 8 §6 of the injected electrons. In this case, a gold tip was employed
but similar results were obtained with Ptlr or W tips.
. Exzfimen&lda~
Calcohwdcwve
5o
10oo
. It
= 7.5 nA
. It= 5nA
.
' It = 2.5 nA .
° It = I nA
~
c~-t
j
.
~-
,
mm
.
~. .
. ."
.." ;."
.:.
;.
°
500100 7l©
Bias voltage
a)
N° 12 BEEM OF THE Au-Si loo) INTERFACE 2215
. Eq&men&lda~
Calcdawdcwve
4tXl 8l© 12W l6tU 2W© 24l©
Bias voltage (mV) a)
Derivative (a.u.)
4
.~
3 "
m~
a
w
m
I . .
m o
1 00
main reasons. On the one hand, the Schottky barrier height may locally vary at the interface.
On the other, the film and interface properties may alter the probability of transmission of the electrons and then the current in the semiconductor. Previous studies have demonstrated that these Au-Si interfaces are relatively uniform and that they give constant Schottky barrier
heights [4, 6]. By successive acquisitions, it is possible to detect and analyse these variations.
Our experiments show that the eV~ variations are small, at least less than the accuracy of the BEEM method (I.e.
m 0.05 eV). The average value deduced from all these measurements is
eV~
=
0.77 ± 0.05 eV, a value quite different from those usually obtained on such Metal- Semiconductor contacts (generally above 0.8 eV). The slope of the curves above the threshold voltage is found to be the only varying parameter during the slow tip displacement from one place to another. This peculiar characteristic obviously leads to the image contrast in BEEM
experiments.
BEEM IMAGES. These images yield the variations of the collected current while the bias
voltage and the tunneling current are kept constant. As for the spectra, the contrast may
originate from the different Schottky barrier height or from the different transmission coefficient in the metal layer or at the interface. Other aspects such as the injection angle of the electrons have also been discussed. Spectrum analysis shows the variations of the collector
current to be due only to local modifications in the injection yield. It is therefore considered
that this property accounts for most of the images obtained by this technique.
Figure 5 is an example of a BEEM image (Au thickness 12.5 nm). Bias voltage and
Fig. 5. STM topography (left) and BEEM image (right) of a Schottky diode with a 12.5 nm Au thickness. The grey level gives either the roughness of the sample (5 nm) or the ballistic current in the silicon (from I to 50pA).
N° 12 BEEM OF THE Au-Si ( loo) INTERFACE 2217
tunneling current are 1.2 V and 3 nA respectively. The image scale is 175 x 175 nm~. The left image gives the topography of the sample surface and the right one represents local variations of I~ bright levels correspond to high injected currents while the lower collector currents give
the darker grey levels.
The topography shows the characteristic arrangement of gold hillocks on the diode surface
with a maximum roughness of 7.5 nm. These hillocks are often observed by STM as soon as
the growth rate of gold becomes too important. Their size is found to depend on the substrate temperature as well [7]. The BEEM image looks like the topography because the local variations of I~ follow the grain boundaries. Nevertheless, two similar grains may yield two different rates. On the other hand, different grains may give the same collector current. This appears in the center of the ballistic image where a large domain with a low injection yield is
evidenced.
This current is found to be independent of the injection angle because the grains are uniform
in the BEEM images while the injection angle is certainly different near the grain edges.
Another striking feature is the low role of metal thickness. Some images show that the thicker
grains may have the higher yields for ballistic electron injection. This is not surpri~ing if one considers the usual electron mean free path in gold (» lo nm).
As previous studies on Au-Si loo) and Au-Si (I II) junctions do not show any contrast in the BEEM images, the most interesting result remains the unusual presence of large areas with
an uniform ballistic current and the high reproducibility of this phenomenon whatever the Au thickness of the metal layer.
Discussion.
The main advantage of the Ballistic Electron Emission Microscopy is its high resolution [4].
Indeed, the ballistic electrons are injected in the metal base with a certain angle relative to the interface normal. However, limitations on energy and transverse momentum conservation lead to reflection conditions. Electrons with a large kj component (where kjj is the component parallel to the interface) are then reflected in the metal layer and do not participate in the
collector current. This critical angle of injection H~ is very small and depends mainly on the m~/m ratio, where mj is the electron effective mass parallel to the interface within the semiconductor and m the free electron mass. For silicon, in which m~ is equal to 0. I m, we
obtain a critical angle of 5° 50 (for V =1.5 V) while in the same conditions in GaAs
(mj = 0.065 ml, H~ is equal to 4° 20. This focusing explains that the expected lateral resolution is less than I nm at the interface for a metal thickness of lo nm. Hence, each spectrum is a
local characterization of the metal-semiconductor contact.
In earlier studies, spectra have been interpreted by transport calculations in a plane electrode geometry, as introduced by Simmons in 1963 [8]. The collector current results from the carrier transport through the potential barrier between tip and surface with specific conditions about the energy components parallel (E~) and perpendicular (E,) to the interface. The following
expression may be written for the collector current
I~
=
RC ~~ D(E,)
~~"
f(E) dE~ dE, with C =
~ ~ )"~~ (l)
E~~~ o h
A is the area, R is a constant and f(E) is the Fermi function. D(E,) is the well-known
expression of the transmission probability for a square barrier in the case of the planar tunneling theory. E~~~ and E~~, are given by the energy and momentum matching conditions at the interface.
Using the same formalism, it is possible to calculate the tunneling current I~. The collector
current in terms of tunneling current then becomes
cc E~~~
D (E, f (E) dE~ dE,
~ o
~C " ~~t
cc
~~~
cc
~~~
D (E,) f(E) f(E + eV )] dEj dE,
o o
The injected current therefore depends on parameters R and V~ where R gives the attenuation due to scattering in the metal layer.
Kaiser et al. have demonstrated that the iormer relation gives a square law relation above the threshold V~ in the I~ vs. V spectra, I.e., close to the behavior experimentally observed [9]. We have also used this formalism to fit our experimental data in the case of low bias voltages
w 0.5 V above the threshold). The solid line in figure 2 gives the results of the calculations for R
= 0.94 eV~ and V~ = 0.78 V using (2).
Near the threshold, the model perfectly fits the experimental results. But for higher voltages,
the calculated spectra increase more strongly than the BEEM data. This discrepancy appears
above I.I V in iigure 2. The tendency to reach saturation in the experimental results has been
extensively studied by Prietsch and Ludeke [lo, 11]. To take this effect into account, they introduce in the I~ calculation a function that exponentially depends on the mean free path of the electrons in the base layer. This assumption leads to calculated spectra which present a
saturation at « high » bias voltages mainly because the mean free path A~_~, due to electron- electron scattering, decreases when the electron incident energy increases [12]. The solid line in figure 4a has been calculated after introduction of inelastic scattering in the metal layer as
described in [11]. Our observations are therefore compatible with these theoretical predictions.
However, we have to introduce in the fit function an unusual low value of r~ which is the volume per electron in Au (see [I I]). This traduces a local
« desorder » encountered by the electrons during transport through the metal layer and at the interface.
Our observations also confirm that the Schottky barrier height is uniform for Au~si contacts.
However, the average value (0.77 V) is smaller than the values usually obtained by BEEM (0.82 in [9] and 0.88 [6]) or by others techniques (such as J(V) and C (V) measurements)
which give a threshold generally above 0.8eV [13]. Hence, the main variations of
I~ arise from the fluctuations of the R parameter.
The BEEM images are in good agreement with the precedent result the variations of the
injected current are due to scattering in the base or defects at the interface that reduce the number of electrons in the semiconductor rather than to barrier height modifications. The same
threshold has been obtained whatever the (x, y tip position on the sample (by threshold, we mean the lowest tunneling voltage giving the first electrons in the BEEM image). Contrary to
previous works on Au-Si junctions in which the BEEM images are very uniform [6, 9, 13], the collector current intensity at a fixed bias voltage is found to be strongly dependent of the lateral
tip position and large domains appear on the BEEM images. In addition, these domains to not exhibit the granular structure of the surface except for their edges which follow some grain
boundaries (Fig. 5).
Comparatively to previous works on the same Metabsemiconductor contact, the following discrepancies are therefore revealed relative low barrier height, more important disorder
which appears in the spectra at high voltage and large domains in the BEEM images.
Two phenomena may be invoked to explain these observations : scattering in the base layer and interface defects. For scattering in the gold layer, one should obviously admit that the
granular structure of the metal film plays the most important role. However, this granular
structure has been already observed in BEEM experiments (lee the STM topography in [9])
N° 12 BEEM OF THE Au-Si (100) INTERFACE 2219
and the BEEM images only reveal smooth interfaces : domains have never been reported. As the BEEM contrast is also independent of the Au thickness, scattering events in the base layer
may be discarded. Interface defects therefore dominate the transport properties.
Among these defects, we think that the presence of Si oxides could explain the lowering of the Schottky barrier height. Indeed, previous studies have demonstrated that oxidized or etched
silicon surfaces may give lower barrier heights [14]. Hallen et al. have also reported that no
BEEM current is observed if the oxide layer is too thick [13]. In this case, darker areas in the
BEEM image could correspond to interface domains with thicker oxide layers. In our
experiments, oxides on the Si substrate are certainly present before gold deposition, I.e. during
achievement of the Ausb ohmic contact.
The local disorder in the high voltage spectra is assumed to result from the presence of the oxide layer and from intermigration [15] at the interface that considerably decrease the electron
mean free path at high kinetic energies. Interface contaminants and intermigrations are
therefore assumed to play the most important role in the contrast observed and in the electronic characteristics of this Au-Si (100) contact.
Further experiments are being done but these first results already suggest that BEEM may be used for the characterization of imperfect interfaces. These results also underline the role of substrate cleaness and first metal layers in the interface formation. More generally, this
technique appears to be fundamental for Schottky junctions optimization.
Summary.
Ballistic Electron Emission Microscopy experiments have been carried out on Au-Si (100)
interfaces. The spectra show that the Schottky barrier height is constant whatever the Au
thickness on the silicon and are interpreted through the Kaiser and Bell formalism. The
experiments performed with different tips show the threshold to be independent of the tip metal (Ptlr, W or Au). A lowering of the Schottky barrier height is also observed. BEEM images
present large characteristic domains whose edges correspond to the grain boundaries of the metal film. Interface defects are invoked to account for these observations.
Acknowledgments.
We would like to thank F. Rossel for sample preparation and G. Peuch for technical assistance.
References
[1] Binnig G., Rohrer H., Gerber Ch., Weibel E., Appt. Phys. Lent 40 (1982) 178.
[2] Binnig G., Quate C. F., Gerber Ch., Phys. Rev. Lent. 56 (1986) 930.
[3] See for example the proceedings of the 6th conference on STM. P. Descouts and H. Siegenthaler Eds., part A and B (North Holland 1992),
[4] Bell L. D., Kaiser W. J., Scanning Microsc. 2 (1988) 1231.
[51 Coratger R., Beauvillain J., Ajustron F., Lacaze J. C., Trdmolli6resC., Rev. Sci. Instrum. 62
(1991) 830.
[6] Palm H., Arbes M., Schulz M., Appt. Phys. A 56 (1993j 1.
[7] DeRose J. A., Thundat T., Nagahara L. A., Lindsay S. M.. Surf Sci. 256 (199I) 102.
[81 Simmons J. G., 34 (1963) 1793.
[9] Bell L. D., Kaiser W. J., Phys. Rev. Lent. 61(1988) 2368.
[lo] Ludeke R., Prietsch M., Samsavar A., J. Vac. Sci. Technol. B 9 (1991) 2342.
[I11 Ludeke R., Prietsch M., J. Vac. Sci. Technol. A 9 (1991) 885.
[121 Quinn J. J,, Phys. Rev 126 (1962) 1453,
[13] Hallen H, D., Femandez A., Huang T,, Buhrman R, A., Silcox J,, J. Vac. Sci. Technol. B 9 (1991) 585.
[14] Metal-Semiconductor Contacts, E, H. Rhoderick, R. H. Williams Eds. (Clarendon Press Oxford, 1988j.
[15] Cro~ A.. Derrien J., Salvan F., Surf. Sci. l10 (1981j 471.
