Hydrogenated Diamond Surfaces studied by Atomic and Kelvin Force Microscopy
B. Rezek1, C. E. Nebel, and M. Stutzmann
Walter Schottky Institut, Technische Universität München, Am Coulombwall, 85748 Garching,
Germany
Abstract
Atomic force (AFM) and Kelvin force microscopy (KFM) are employed to characterize and modify
both morphologic and electronic properties of (100) hydrogenated diamond surfaces with high
lateral resolution. Contrasts in AFM and KFM images are discussed and values of the local surface
work function and Fermi level are deduced using gold and aluminum pads. A shift of Fermi level is
observed by KFM during sample illumination, in agreement with an increase in conductivity. Using
negatively biased cantilevers in AFM, surface hydrogen termination of diamond is locally removed
a with resolution of 10 nm. Thus insulating nanoscale patterns are scribed into the diamond surface.
Conductivity through such patterns is orders of magnitude lower then for H-terminated diamond.
Keywords: diamond crystal, surface microscopy, surface electronic properties, oxidation
PACS: 81.05.Uw, 81.16.Rf, 68.37.Ps, 73.20.At, 73.30.+y
Introduction
Since high surface conductivity has been detected by Landstrass and Ravi [1] on as grown intrinsic
diamond layers it has gained a lot of interest. The high surface conductivity has been found to be
generated by hydrogen termination of the diamond surface. Besides p-type surface conductivity,
hydrogen induces hydrophobic surface properties and a negative electron affinity, which is due to
the formation of a high density of C-H dipoles at the surface.
Some theories have been proposed to explain the nature and origin of the p-type surface
conductivity, varying from hydrogen induced surface acceptors [2] to an acceptor subsurface region
created by incorporation of hydrogen [3], or even to a buried acceptor layer [4]. The surface
1
On leave from: Institute of Physics ASCR, Cukrovarnická 10, 162 53 Prague, Czech Republic
Present address: Nanotechnology Group, ETHZ, Tannenstr.3, 8092 Zürich, Switzerland
PREPRINT, to appear in Diamond and Related Materials
1
conductivity was shown to be also strongly affected by atmospheric adsorbates on the surface of
hydrogen-terminated diamond [5]. It has been proposed that adsorbates act as a sink for electrons
transfered out of the diamond. Thus, a hole accumulation region in the valence band is generated.
This model is also favoured in our recently published work [6]. Up to now, however, little is known
about the details of this phenomenon.
In contrast to the high conductivity of hydrogenated diamond surfaces, surfaces terminated by
oxygen are strongly resistive. Thus, in-plane electronic devices can be realized by a control of
surface termination. The type of surface termination can be modified on a microscopic scale by use
of atomic force microscopy (AFM). A negatively biased gold coated AFM cantilever can be
applied to induce anodic oxidation which resulted in the formation of insulating patterns on
diamond surfaces [7,8]. Compared to other techniques, AFM patterning has the advantage that no
lithographic masks are required and no radiation damage occurs. By combination of AFM
patterning with electron beam lithography, in-plane transistor structures could be produced and
characterized [9].
In this paper, we summarize and discuss our results from Kelvin force microscopy (KFM) to obtain
information about the electronic properties of diamond surfaces and to develop a surface
conductivity model for hydrogen-terminated diamond. AFM is further used to prepare microscopic
insulating patterns on hydrogen-terminated surfaces, and the electronic properties of these patterns
are characterized. We show that the negatively biased cantilevers not only modify the electronic
properties but also can etch nanoscale patterns into the diamond surfaces. The effect of contact and
non-contact AFM on the evaluation of these patterns is discussed.
Experimental
Three types of samples were used in this study. Sample I was a diamond layer of 300 nm thickness
grown epitaxially on a (100) oriented Ib natural diamond substrate. For the growth a hot filament
process was used, which generates a hydrogen-terminated surface (1.2% CH4, 2200°C tungsten
filament temperature, 690°C sample temperature, 6 mm filament-sample distance, pressure 50
mbar, gas flow 100 sccm). Sample II was a (100) IIa natural diamond, which was hydrogenated in
a microwave plasma (30 min, 40 Torr, 800 W, 850°C). Sample III was a (100) Ib natural diamond
overgrown homoepitaxially by chemical vapour deposition and subsequently hydrogenated in a
microwave plasma.
PREPRINT, to appear in Diamond and Related Materials
2
To explore surface Fermi energies, isolated hydrogen-terminated patterns have been realized on the
diamond surface by oxygen plasma treatment through photolithographic masks with the shape of
bars or cross-like patterns. Gold (for ohmic contacts) and aluminum (for Schottky contacts) were
thermally evaporated on hydrogen-terminated regions of the surface. The gold layer was also used
for calibration of KFM.
Ambient AFM was employed to measure surface morphologies and to prepare microscopic
insulating patterns on diamond surfaces. The AFM was operated in contact mode, except for few
instances discussed below. Square patterns were scribed by a sequence of controlled cantilever
movements with a bias voltage of –10 V applied to the AFM tip. During this procedure, the AFM
feedback maintained the cantilever tip in contact with the grounded sample surface.
Images of contact potential difference (CPD) were acquired using the lift-mode technique of KFM
[10]. The principle of KFM is following: An a.c. voltage is applied between between the tip and the
sample. The resulting electrostatic force is detected by a lock-in amplifier and a feedback adjusts a
d.c. voltage on the tip Vdc until the CPD is compensated. Potential at the sample surface than equals
-Vdc. In this work, the compensation voltage is used to render the images. Thus for the silicon AFM
tips with work function lower than that of sample the signal is negative (see Fig.1). And vice versa
for metallized tips with higher work function. In both cases, brighter regions in CPD images
correspond to relatively lower work function.
For samples I and III, n-type doped silicon cantilevers were used, with spring constants in the range
20-100 N/m and conductivities of ≈ 100 (Ωcm)-1. Forces applied through the cantilever tips during
contact mode were in the range of 0.1-1 µN. For comparison, PtIr5-coated cantilevers were used for
sample II.
Results
The AFM morphology of sample I is shown in Fig.2(a). Hillocks on the diamond surface are a
result of epitaxial regrowth. With AFM, no difference between hydrogen-terminated and oxidized
diamond can be detected. The edge of the gold contact is revealed along the left border. The white
line indicates the scan position of the spatial profile shown in Fig.2(b). On the other hand, the
PREPRINT, to appear in Diamond and Related Materials
3
hydrogen-terminated part of the diamond surface is clearly resolved in the image and spatial profile
of the CPD at the same place (displayed in Figs.1(c) and (d)). The bright hydrogenated region is
similar to the gold layer. The oxidized diamond parts are dark. The contrast between oxygen and
hydrogen-terminated surface was independently corroborated by scanning electron microscope [6].
Note that on insulating surfaces (such as on oxidized diamond) the CPD data cannot be interpreted
in terms of absolute work functions or Fermi levels. Yet characterization of relative variations, e.g.
due to local surface charging, is possible there.
The AFM morphology of sample II including aluminum and gold contacts is shown in Fig. 3(a).
Straight lines indicate AFM line profiles across aluminum (1) and gold (2) which are shown in Fig.
3(b). Corresponding image and spatial profiles of CPD at the same place are displayed in Figs. 3(c)
and (d). Aluminum exhibits a potential difference of 0.6 eV compared to gold. Thus even on highly
conductive equipotential surfaces KFM can measure the CPD locally.
After prolonged exposure of diamond to a hydrogen plasma, a few nanometer thin soft layer can be
detected on the surface. As shown by the pattern in Fig. 4(a), the layer is located on the
hydrogenated bar, while outside of the bar the layer was mostly removed by oxygen plasma
treatment. The layer could be swept aside by scanning in contact mode AFM using forces in the
range of 0.1-1 µN. A non-contact image of such a swept area is shown in Fig. 4(b). The ”dirt” layer
is effectively removed in the area which has been previously scanned in contact mode. The “dirt”
may be an amorphous carbon layer generated by etching effects of hydrogen plasma, this still has
to be proved though. Fig. 4(c) shows that this layer has a pronounced effect on CPD. Whereas the
contrast appears dark on the original surface, the hydrogenated bar exhibits bright contrast in the
clean region. This indicates a lower work function there. In order to characterize the clean
hydrogen-terminated diamond surface, CPD measurements were performed generally after AFM
contact mode cleaning.
The influence of illumination on the CPD between the cantilever and the hydrogen-terminated
surface is illustrated by Fig.5. In the course of CPD scanning, the sample was at first in the dark
(left). After a 1 µm scan the sample was exposed to ambient light (middle). This results in a
decrease of the CPD signal by 75 mV. Additional illumination by a halogen light results in a further
decrease by 135 mV (right).
PREPRINT, to appear in Diamond and Related Materials
4
The typical surface morphology of a pattern prepared by AFM oxidation is shown in Fig.6(a). The
image was obtained by AFM in contact mode. A trench scribed into the flat and smooth diamond
surface is detected. The line profile in Fig.6(b) shows that the depth of the trench is 0.6 nm. Along
the trench, the depth varies between 0.1-0.6 nm. The width of the trench at half maximum is about
60 nm. In general, the width can be controlled by adjustment of the bias voltage [7,9]. In a special
case, the inscribed pattern was as deep as 3.4 nm at the same bias voltage [8]. The difference in
depth is most likely related to the shape and quality of the specific AFM tip used.
Changing the AFM mode of operation to non-contact mode (the cantilever resonance frequency
was about 320 kHz), the pattern looks as if it was deposited onto the surface with a thickness of
about 0.2 nm. This is illustrated in Fig.7 by a comparison of the contact and non-contact AFM
images of the same pattern. In the non-contact mode, the transition from a height to a depth profile
could be observed as the cantilever was brought closer to the surface by adjustment of set-point.
The question of what the real structure is will be discussed in the next section.
The AFM scribing into diamond affects not only the morphology but also electronic properties at
the surface. Fig.8 shows I-V characteristics measured by use of an AFM tip as a local contact. The
measurement setup is schematically illustrated by the inset in Fig.8. When the tip is inside the
square, the I-V characteristics is controlled by electronic trasport through the oxidized barrier. Note
that currents at negative bias are smaller compared to positive tip voltages as the onset of oxidation
around –4 V affects the measurement. I-V characteristics measured outside the insulating pattern
are ohmic and show a by more than three orders of magnitude larger conductivity, in agreement
with macroscopic measurements on hydrogenated diamond surfaces.
Discussion
For samples I and II the CPD spatial profiles in Fig.2(b) and Fig.3(b) show that work functions of
gold contacts and of hydrogenated diamond surfaces are identical on both samples in spite of
different CPD values due to use of silicon or metal coated cantilevers. The work function of the
gold contacts was found to be 4.9±0.1 eV by photoemission experiments. This value is in a
reasonable agreement with 5.0 eV reported in literature for polycrystalline gold [11]. Thus the
Fermi energy in the hydrogen-terminated diamond surface is deduced about 4.9 eV below the
vacuum level.
PREPRINT, to appear in Diamond and Related Materials
5
The accuracy of this number is mostly determined by the reference work function of the gold
contact. Although due to a possible inhomogeneity of the tip coating the CPD can also be a
function of the tip-sample distance in KFM [12], the results obtained with coated and uncoated
cantilevers are in agreement in our case. Small fluctuations in the spatial CPD spatial profiles are in
the range of ±0.02 eV. With respect to 4.9 eV this error is negligible.
Taking into account the reported negative electron affinity of χ = -1.3 eV [13] and an energy gap of
Eg = 5.5 eV, the Fermi level lies 0.7 eV below the valence band maximum (VBM) of the hydrogenterminated surface. This is consistent with the large accumulation of holes in the subsurface region.
On the oxidized surface, the Fermi level cannot be deduced from relative CPD measurements due
to the high resistivity of both the diamond surface and the bulk.
Note that the above-mentioned negative electron affinity of -1.3 eV was obtained for a clean,
hydrogen-terminated diamond surface in vacuum. Exposure to air may influence this value. This is
a matter of ongoing research [14]. Studies of dipole-oriented water layers on mica substrates
reported a change of the surface potential by 0.4 eV at room temperature and ambient humidity
[15]. Assuming a similar value for water on C-H surfaces, the Fermi level would lie 0.3 eV below
VBM. Theoretical calculations using self-consistent solutions of the Schrödinger and Poisson
Equations have shown that the band bending at the interface of hydrogen-terminated diamond and
water should be in the range 0.3 to 0.9 eV. Thus, despite the uncertainty introduced by ambient
conditions, both Kelvin probe experiments and theoretical calculations support the conclusion that
the Fermi level is rather deep in the valence band. This corroborates the hole accumulation layer
model of surface conductivity introduced by Maier et al. [5].
The CPD between the cantilever and the hydrogen-terminated surface was also susceptible to
illumination as demonstrated in Fig.5. This effect is referred to as surface photovoltage (SPV)
[16,17]. Here, the reduction of CPD with respect to a metallized cantilever corresponds to an
increase in diamond work function and thus to a shift of the Fermi level further into the valence
band (by as much as 0.2 eV).
This effect may appear counter-intuitive since SPV is usually expected to reduce surface band
bending. However, this is not the only possible mechanism (see Ref. [16]). The observed effect
may be explained as follows: In contrast to e.g. silicon, where the band bending is a result of charge
PREPRINT, to appear in Diamond and Related Materials
6
equilibrium between bulk and surface states, there are no surface states on hydrogenated diamond.
But there is a negative electron affinity (NEA). In the model adopted in this paper the NEA gives
rise to transfer of electrons from diamond into an air-adsorbate layer and hence to hole
accumulation and increase in conductivity at the diamond surface. The thin hole accumulation
region can be considered as a two-dimensional potential well. The photogenerated holes cannot
escape the surface potential well, they increase the hole density there and thus surface conductivity
(until they recombine). To accommodate more holes the surface Fermi level has to move deeper
into a valence band, thereby increasing the work function. Due to the NEA, photo-electrons are
transferred into the adsorbate layer.
While the surface potential of a hydrogen-terminated diamond surface is the same as for gold,
aluminum exhibits a difference of 0.6 eV. Using gold as a reference, this results in an aluminum
work function of 4.3 eV, in agreement with literature values for polycrystalline aluminum [18].
Considering the aluminum work function of 4.3 eV and the negative electron affinity of
hydrogenated diamond (-1.3 eV), the Fermi level at the aluminum-diamond interface is coincident
or only slightly below the VBM (accuracy ±0.1 eV). In this case no hole accumulation would be
generated below the aluminum layer. Electronic properties of aluminum Schottky junctions on
hydrogenated diamond have also been characterized by Garrido et al. [19]. Next to the aluminum
contacts, capacitance-voltage characteristics indicated a lateral depletion region width of 0.5 - 100
nm.
When KFM was applied to measure the depletion profile and width across Al Schottky junctions,
CPD decayed on diamond across several micrometers in lateral direction from the contact edge.
We have already discussed that this decay is mostly dominated by stray capacitance effects [9], as
the potential profile is affected not only by band bending in the depletion region but also by the
edge transfer function [10]. In addition, the spatial resolution is limited by the AFM tip shape and
by the height of the aluminum contact [9]. Thus, unambiguous interpretations of the CPD profiles
are not possible.
The modification of hydrogen-terminated diamond surfaces by negatively biased AFM cantilevers
gave rise to insulating patterns which appeared to be scribed into the surface when measured by
AFM in contact mode. This observation seems to be in contradiction to previously reported growth
accompanying the oxidation process on diamond [7]. However, it is well known that plasma
PREPRINT, to appear in Diamond and Related Materials
7
oxidation of diamond very effectively removes carbon. A comparable effect may take place in a
strong electric field around the negatively biased tip. Details will be discussed elsewhere.
As a matter of fact, in the non-contact AFM the patterns looked as deposited onto the surface, but a
shift from height to depth profile was observed as the cantilever was brought closer to the surface
as shown in Fig.7. Such an effect can be attributed to switching between attractive and repulsive
interaction controlling the measurement in the tapping mode [20,21]. The real surface morphology
is revealed in contact or tapping mode where the tip is touching the surface and stronger repulsive
forces are dominating the interaction. In non-contact mode the interaction between the tip and the
sample is small and the cantilever is very sensitive to long range forces. We shall discuss two
factors which could explain the different contrast.
Cantilever oscillations can be influenced by a different electrostatic potential above the oxidized
pattern and the surrounding hydrogenated surface, because the oxidation changes the electron
affinity and increases the total work function of diamond [13]. A larger work function difference
between the oxidized surface and the cantilever gives rise to a larger electrostatic interaction, which
can result in apparent height contrast in the morphology image [22].
Operation in ambient atmosphere can also influence the non-contact mode measurement. Air
humidity is expected to preferentially adsorb at the oxidized patterns, because of the hydrophilic
properties of the oxidized surface compared to the hydrophobic properties of the hydrogenated
surface [23]. This also can result in a variation of surface height. In ambient humidity (33-37%), a
typicla water layer thickness is in the order of 0.1 nm [24], which agrees well with the observed
height difference on the oxidized (hydrophilic) patterns. While in our opinion this effect is
dominating, conclusive interpretation requires further research.
Summary
To summarize, on microscopic hydrogen-terminated patterns prepared by oxygen plasma treatment
on hydrogen-terminated (100) diamond surfaces, images of contact potential difference clearly
revealed hydrogenated as well as oxidized regions. From the potential images, the Fermi level was
deduced to 0.7±0.1 eV below the valence band maximum at the hydrogen-terminated surface, thus
giving rise to a hole accumulation layer. Illumination of the sample resulted in an additional shift of
the surface Fermi level into the valence band by as much as 0.2 eV. This was corroborated by an
PREPRINT, to appear in Diamond and Related Materials
8
increase in surface conductivity. Evaluation of CPD across aluminum Schottky contacts indicated
that no accumulation layer is generated below the contacts, in agreement with previous
observations in the literature. To directly control the termination of hydrogenated diamond
surfaces, AFM was applied using negatively biased silicon cantilevers. By this technique patterns
were inscribed into the diamond surface, with a lateral resolution of ~ 10 nm and a depth of ~ 3 nm.
The conductivity through these patterns was low. This nanoscale patterning can be an important
feature for the preparation of in-plane electronic devices of diamond.
The authors gratefully appreciate cooperation with Prof. H. Okushi from the Diamond Research
Center in Tsukuba, Dr. P. Bergonzo and E. Snidero from LIST(CEA) in Saclay, and Dr. J. Ristein
and Prof. L. Ley from the University of Erlangen-Nürnberg. Financial support by the European
Community contract HPRN-CT-1999-00139, Deutsche Forschungsgemeinschaft contract NE524-2,
and the Procope French-German program is gratefully acknowledged.
PREPRINT, to appear in Diamond and Related Materials
9
References
[1] M. I. Landstrass, K. V. Ravi, Appl. Phys. Lett. 55 (1989) 975.
[2] J. Shirafuji, T. Sugino, Diam. Relat. Mater. 5 (1996) 706.
[3] K. Hayashi, S. Yamanaka, H. Watanabe, T. Sekiguchi, H. Okushi, K. Kajimura, J. Appl. Phys. 81 (1997)
744.
[4] A. Denisenko, A. Aleksov, A. Pribil, P. Gluche, W. Ebert, E. Kohn, Diam. Relat. Mater. 9 (2000) 1138.
[5] F. Maier, M. Riedel, B. Mantel, J. Ristein, L. Ley, Phys. Rev. Lett. 85 (2000) 3472.
[6] B. Rezek, C. Sauerer, C. E. Nebel, M. Stutzmann, J. Ristein, L. Ley, E. Snidero, P. Bergonzo, Appl.
Phys. Lett. 82 (2003) 2266.
[7] M. Tachiki, T. Fukuda, K. Sugata, H. Seo, H. Umezawa, H. Kawarada, Jap. J. Appl. Phys. 39 (2000)
4631.
[8] B. Rezek, C. Sauerer, J. Garrido, C. E. Nebel, M. Stutzmann, E. Snidero, P. Bergonzo, Appl. Phys. Lett.
82 (2003) 3336.
[9] B. Rezek, J. Garrido, C. E. Nebel, M. Stutzmann, E. Snidero, P. Bergonzo, phys. stat. sol. (a) 193 (2002)
523.
[10] H. O. Jacobs, H. F. Knapp, A. Stemmer, Rev. Sci. Instrum. 70 (1999) 1756.
[11] S. M. Sze, Physics of semiconductor devices, 2nd Edition, John Wiley & Sons, New York, 1981.
[12] C. Sommerhalter, T. W. Matthes, T. Glatzel, A. Jäger-Waldau, M. C. Lux-Steiner, Appl. Surf. Sci. 157
(2000) 263.
[13] J. B. Cui, J. Ristein, L. Ley, Phys. Rev. Lett. 81 (1998) 429.
[14] G. Piantanida, A. Breskin, R. Chechik, O. Katz, A. Laikhtman, A. Hoffman, C. Coluzza, J. Appl. Phys.
89 (2001) 8259.
[15] H. Bluhm, T. Inoue, M. Salmeron, Surf. Sci. 462 (2000) L599.
[16] L. Kronik, Y. Shapira, Surf. Interface Anal. 31 (2001) 954.
[17] V. Švrček, I. Pelant, P. Fojtík, J. Kočka, A. Fejfar, J. Toušek, M. Kondo, A. Matsuda, J. Appl. Phys. 92
(2002) 2323.
[18] G. Chiarotti (Ed.), Landolt-Börnstein New series III/24b, Springer, Berlin, 1994.
[19] J. A. Garrido, C. E. Nebel, M. Stutzmann, E. Snidero, P. Bergonzo, Appl. Phys. Lett. 81 (2002) 637.
[20] A. Kühle, A. H. Sørensen, J. B. Zandbergen, J. Bohr, Appl. Phys. A 66 (1998) S329.
[21] R. García, A. S. Paulo, Phys. Rev. B 60 (1999) 4961.
[22] T. Shiota, K. Nakayama, Jpn. J. Appl. Phys. 40 (2001) L986.
[23] K. Larsson, H. Bjorkman, K. Hjort, J. Appl. Phys. 90 (2001) 1096.
[24] M. Luna, J. Colchero, A. M. Baró, Appl. Phys. Lett. 72 (1998) 3461.
PREPRINT, to appear in Diamond and Related Materials
10
Figures
Fig. 1: Schematic energy band diagram of CPD measurement on hydrogenated diamond.
Fig. 2: AFM morphology and CPD images of the epitaxially grown diamond layer (sample I) with
an oxygen plasma generated pattern and a gold contact.
PREPRINT, to appear in Diamond and Related Materials
11
Fig. 3: AFM morphology and CPD images of the plasma hydrogenated (100) diamond (sample II)
with an oxygen plasma generated pattern and with aluminum (1) and gold (2) contacts.
Fig. 4: Non-contact AFM images of a surface layer on a hydrogenated diamond (a) prior to and (b)
after scanning in contact AFM. (c) CPD image after contact AFM scanning.
Fig. 5: Contact potential difference on a hydrogen-terminated surface (sample II) as a function of
illumination.
PREPRINT, to appear in Diamond and Related Materials
12
Fig. 6: (a) Typical AFM surface morphology image and (b) depth profile across a square inscribed
by AFM into hydrogenated diamond surface (sample III).
Fig. 7: (a) Contact mode and (b) non-contact mode AFM image of the same pattern. In non-contact
image the shift from height to depth profile can be observed.
Fig. 8: Current-voltage characteristics measured by AFM tip from inside the pattern across the
oxidized line (squares) and on the hydrogenated surface outside the pattern (circles). Schematic
setup is shown in the inset drawing.
PREPRINT, to appear in Diamond and Related Materials
13