Research Article
Received: 24 August 2009
Revised: 20 October 2009
Accepted: 21 October 2009
Published online in Wiley Interscience: 1 December 2009
(www.interscience.wiley.com) DOI 10.1002/sia.3138
Compact plasma source for removal
of hydrocarbons for surface analysis
Catherine A. Dukes∗ and Raúl A. Baragiola
A low–energy, constricted-anode Anders-type plasma source was built and tested for the chemical removal of adventitious
carbon on surfaces. Oxygen plasma, generated in the source, extends to the sample surface through an aperture in the anode.
This plasma reacts with surface hydrocarbons and removes them in less than a minute without influencing the intrinsic surface
stoichiometry of nonoxidizing samples such as minerals, glasses, and metal oxides. Adventitious carbon removal is critical
for accurate binding energy determination and quantitative measurements in XPS and AES, particularly in multicomponent
materials. We measure the plasma to be composed primarily of O+ and O2 + , with minor H+ , H2 + , and O++ components. Ion
energy distributions were measured for O+ and O++ and show all emitted ions have energies less than 50 eV, confirming
chemical desorption as the primary removal mechanism. The plasma source, easily built ‘in house’, is compact and can be
c 2009 John Wiley & Sons, Ltd.
mounted on a 2.75-in flange for in situ specimen cleaning prior to surface analysis. Copyright Keywords: adventitious carbon; plasma; plasma cleaning; plasma source; hydrocarbons; sample cleaning
Introduction
40
Adventitious carbon is a ubiquitous feature of surface analytical
spectra – appearing in XPS, AES, and SIMS. These carbonaceous
deposits exist on all surfaces exposed to the atmosphere and are
thought to be a type of hydrocarbon contamination, based on the
binding energy of the carbon 1-s photoelectron peak, ∼285.0 eV,
measured on conductive surfaces. This layer of surface carbon,
which can be a fraction of a monolayer to several monolayers
in thickness, forms in islands on sample surfaces, attenuating
signal from the specimen.[1] The adventitious carbon layer occurs
always on the outermost surface of materials. While adventitious
carbon does not generally bond chemically to the surface, it adds
error to the quantitative calculation of concentration for intrinsic
surface elements and obscures signal due to trace elements and
impurities.[2]
However, the ubiquitous nature of adventitious carbon has
also made it an attractive standard to use as a binding energy
reference for nonconductive samples; researchers fix the spectral
binding energy scale by shifting the entire spectrum such that
the primary carbon peak sits at a given binding energy value,
typically 284.8 ± 0.2 eV. The spread in binding energy is known to
vary with the amount of carbon on the surface and the substrate
material to which it adheres, limiting the accuracy of the energy
scale.[3] Adventitious carbon is recommended as a binding energy
standard only in cases where no better standard is available or the
absolute binding energy measurement is unimportant.[4]
A number of techniques have been employed to remove
adventitious carbon from samples: solvent baths, chemical
etching, sputter cleaning, heating, and ex situ plasma etching.[5]
Each of these removal techniques is helpful under specific
conditions, but ex situ carbon removal does not eliminate carbon
redeposited between the cleaning process and insertion into the
analysis vacuum chamber, and chemical etching often introduces
surface impurities from etchant residue. The most common in situ
technique, sputtering, is typically done in an analysis or preparation
chamber with keV Ar ions; however, sputtering produces atomic
Surf. Interface Anal. 2010, 42, 40–44
layer mixing and preferential element removal. Heating to
temperatures required for hydrocarbon removal (∼400 ◦ C) is not
an option for polymers or many samples with mobile or thermal
segregating elements. Thus, a technique to remove adventitious
carbon that can be mounted on the analysis chamber and produces
no damage or impurities would be useful, as well as unique.
Description of Constricted-Anode Plasma
Source
Low-pressure plasmas are nearly neutral mixtures of neutral atoms
and molecules, excited neutrals, low-energy ions and electrons.
They are formed in a gas when a sufficiently high voltage is
applied between two metal electrodes, cathode and anode. While
the peak of the electron energy spectrum lies below the ionization
or dissociation energies of gas-phase molecules, the tail of the
distribution lies above. It is these higher energy (>50 eV) electrons
that generate the plasma as they collide, ionizing and exciting
atoms as well as forming radicals. The lower energy electrons
are contained within the plasma by the plasma sheath, which
forms between the walls of the cathode and the higher potential
plasma. Plasma sources are defined by their method of generation,
pressure, ion and electron energy, composition, and density.[6]
To remove the adventitious carbon from mineral samples,
we constructed a constricted-anode plasma source based on
a design developed by Anders at Lawrence Berkeley National
Laboratory.[7 – 9] Anders’ design features a d.c. glow discharge
tube with a small aperture in the anode, which serves as a
nozzle through which the plasma flows outward. The unique
∗
Correspondence to: Catherine A. Dukes, Laboratory for Atomic and Surface
Physics, Engineering Physics, University of Virginia, Charlottesville, VA
22904, USA. E-mail: [email protected]
Laboratory for Atomic and Surface Physics, Engineering Physics, University of
Virginia, Charlottesville, VA 22904, USA
c 2009 John Wiley & Sons, Ltd.
Copyright Compact plasma source for removal of hydrocarbons for surface analysis
Figure 1. Cutaway schematic of the constricted-anode plasma source of
cylindrical geometry. The cathode is held at high potential (−350 V) with
respect to the anode at ground, which is separated by a thin BN spacer.
Surf. Interface Anal. 2010, 42, 40–44
Figure 2. Oxygen plasma discharge. The ion-constricted anode source fits
on a 2.75-in conflat flange and is mounted in the sample introduction
chamber to the surface analysis instrument.
the emission current. The discharge voltage is measured directly
between the cathode and the grounded anode.
To ignite the source, we flow O2 gas through the quartz capillary
tube, and its pressure is controlled with a leak valve between the
gas bottle and the capillary tube. At a pressure of ∼5 Torr and
an applied voltage of −750 V, a glow discharge ignites and the
cathode voltage drops to about a half. The green discharge can
also be observed with the unaided eye (Fig. 2). Once the plasma
starts, both the pressure and the power supply voltage can be
reduced, resulting in a stable discharge current of ∼20 mA. Our
standard running discharge voltage was approximately −350 V at
a pressure ∼2 Torr measured on the capillary tube. However, the
source will operate in a stable manner at pressures between 200
mTorr and >5 Torr which has a strong influence on the discharge
parameters and output current.
Plasma Characterization
The plasma source was characterized using a Hiden Analytical Inc.
EQS300 with energy analyzer to determine the ionic composition
of the plasma, as well as the ion energy distribution. The source
was located in line about 30 cm from the entrance aperture of the
mass/energy analyzer. The ion mass spectrum was taken at four ion
energies: 5, 10, 25, and 50 eV (Fig. 3), using an electrostatic energy
filter at the entrance to the mass spectrometer. The oxygen plasma
contains roughly equivalent amounts of O+ and O2 + ions, with
O++ , H+ , and H2 + as minor (<7%) components. No unexpected
impurities are apparent, ensuring no spurious material is deposited
on the sample surface during plasma etching. The hydrogen in
the plasma is likely from absorbed water in the BN insulating
disc which, according to manufacturer specifications, contains
0.5–3.0% water.
The total plasma current was measured on a copper collector
plate at ∼10 cm in front of the source anode. We determined the
positive component (ions) by deflecting the electrons away from
the plate with a magnet and measuring the positive plate current
with a Keithley 610B electrometer. We measured an ion current of
0.7 mA at the source conditions specified above.
It is important to identify the energy of the ions to ascertain
the nature (physical or chemical) of the hydrocarbon removal.
c 2009 John Wiley & Sons, Ltd.
Copyright www.interscience.wiley.com/journal/sia
41
quality of this plasma source is that the majority of the ions
have low energies, at or below the sputtering threshold. This low
energy occurs when an insulator with a very small aperture is
used to separate the anode from the cathode, thus effectively
reducing the size of the anode and increasing the current density
at the ‘reduced’ anode. A low-voltage, double-layer virtual anode
forms on the cathode side of the insulating aperture.[7 – 11] Ions
emitted from this high-density region have a kinetic energy
similar to the potential of the virtual anode. In addition, recent
experiments have shown that an even lower energy spectrum
(<20 eV) may be obtained by floating, rather than grounding, the
anode electrode.[9]
Our constricted-anode plasma source was designed specifically
to provide low-energy oxygen ions to a sample with superficial
hydrocarbon contamination. The oxygen ions react chemically
with the hydrogen and carbon on the surface and selectively
remove surface hydrocarbon and organic contamination. This
occurs without disturbing the native specimen, except in oxidizing
samples such as pure metals and metal alloys where a thin metal
oxide grows in response to the oxygen gas in the plasma source. In
addition, we required a source that could be easily mountable on
a ultra-high vacuum (UHV) vacuum chamber, for in situ cleaning,
and fit on a standard 2.75-in conflat flange through a stainless
steel tube (inner diameter (ID) 1.35 ).
The plasma source, sketched in Fig. 1, consists of a copper
support cup mounted by two aluminum legs attached to
the conflat flange. The flange has three colinear liquid/gas
feedthroughs welded through it: two for water cooling and a
single center tube for gas admission. The source is cooled by water
flowing through a copper tube wound around the copper cup
and braised in place for maximal heat transfer. The water lines are
fixed to the two water feedthroughs in the mounting flange using
VCR fittings. Inside the copper cup sits a second cup machined
from BN, which electrically isolates the cathode from the grounded
copper cup, and then an aluminum cup that serves as a cathode.
A Kapton-coated copper wire, inserted through an aligned hole
in the copper and insulating cups, delivers high voltage to the
cathode and connects to the cathode cup via a small set screw.
This screw is accessed through an aperture in the two outer cups.
The cathode wire is connected to an electrical feedthrough on an
auxiliary flange. Oxygen gas is fed from outside vacuum through a
quartz capillary (ID ∼1 mm) to the cathode chamber. The cathode
is then capped by an insulating disc (BN) with an aperture of 2 mm
in it and then an aluminum disc (anode) with a 1-mm orifice which
serves as an exit for the plasma.
A single power supply (1 kV; 30 mA) is required to provide
voltage to the cathode. This supply is used in series with a
current limiting 18 k power resistor and an ammeter to quantify
C. A. Dukes and R. A. Baragiola
Figure 3. Typical mass spectra of ions within the oxygen plasma at energies
5, 10, 25, and 50 eV.
In the standard method of cleaning by sputtering, ion energies
are typically between 500 eV and 3 keV to maximize the physical
removal of atoms from the surface. Each ion–material combination
requires a threshold impact energy of ∼50 eV or more for knock-on
(physical) sputtering. Above this threshold, ion bombardment can
change the intrinsic surface stoichiometry of a multicomponent
sample by preferentially removing one species with respect to
another. Below this threshold, removal has to occur by a chemical
mechanism, where the projectiles react chemically with atoms in
the solid, making a volatile compound that subsequently desorbs.
To determine the primary cleaning mechanism for this ion
source, we measured the energy spectrum for the most abundant
ions: atomic and molecular oxygen. From the results in Fig. 4, we
see that 95% of the oxygen ions have energies of less than 40 eV,
below typical knock-on sputtering thresholds, with a small tail
that extends to 55 eV. In fact, the peak of the distribution lies at
about 3 eV for atomic oxygen (10 eV for molecular oxygen), an
energy which lies well below the sputtering threshold for most
materials. Adjusting the capillary pressure and cathode supply
voltage changes the energy distribution of ions and their absolute
intensity. Thus, running parameters can be chosen to optimize
ion species, energy, or intensity. Lower pressure and reduced
cathode voltage diminish the higher energy tails, minimize
the peak energy, and increase the O+ : O2 + ratio. However,
reduction of these parameters also decreases the overall plasma
current. Thus, an operator may choose to optimize the plasma
source maximizing the ion current or minimizing the electron
energy.
Performance of Plasma Source on Carbon
Removal
42
We measure the performance of the plasma source with respect
to hydrocarbon removal by XPS, acquiring spectra before and
after exposure to the plasma source. By monitoring the area
of the carbon photoelectron peak relative to other elemental
peaks intrinsic to the material, we determine the efficacy of the
oxygen plasma produced by the constricted-anode source. All
XPS spectra were acquired with a Physical Electronics Inc. 560
XPS/SAM instrument, equipped with a double-pass cylindrical
www.interscience.wiley.com/journal/sia
mirror analyzer, dual-anode X-ray source, and a low-energy
electron flood gun. XPS survey spectra were taken to assess relative
elemental changes (spectrometer energy resolution: 3.2 eV); high
resolution spectra were taken to quantify changes in surface
chemistry (spectrometer energy resolution: 0.6 eV). The analysis
chamber base pressure was maintained at ∼5 × 10−10 Torr with a
combination 200 l/s ion pump and Ti sublimation pump. The
plasma source was mounted in the introduction/preparation
chamber, perpendicular to the sample transfer rod. Samples
were mounted onto the transfer rod in the load lock chamber
which was then evacuated with a turbo pump backed by a
scroll pump. The sample could then be admitted directly to the
analysis chamber. After taking initial XPS spectra, the sample
was transferred back to the load lock chamber and exposed to
the oxygen plasma for a period of time, usually less than 1 min.
After exposure, the plasma source gas is turned down, quenching
the plasma. After pumping the oxygen from the load chamber
(<5 min), the sample was readmitted to the analysis chamber.
For these studies, we used several materials: gold (conductive
metal), silicon (semiconductor), and forsterite (a nonconductive
nesosilicate mineral) to demonstrate the effectiveness of the
plasma source on hydrocarbon removal.
A copper sample with a thin gold deposit, 100 nm thick, was
exposed to the oxygen plasma for varying times and analyzed
before and after plasma treatment; the results are shown in Fig. 5.
As exposure increases from 0 to 600 s, the atomic concentration of
C decreases from 45 to 10%. Concurrent with the reduction of the
carbon overlayer, there is an increase in the intrinsic Au and Cu
peaks. However, the O : Au ratio does grow with increasing oxygen
plasma exposure due to oxidation of copper. Continued plasma
exposure beyond 5 min does not increase the surface oxygen
further. Plasma etching allows minor constituents and impurities
to be measured, rather than being attenuated or obscured by the
surface hydrocarbons.
The ability of the constricted-anode plasma source to remove
adventitious carbon from the surface of a silicon wafer is shown
in Fig. 6. The silicon wafer, stored in atmosphere, was analyzed
by XPS and found to bear thin oxide with 15.8% (atomic percent)
surface carbon. A subsequent exposure to the oxide plasma for 5 s
removes the carbon to a level of 1.8%. Additional exposure to the
oxide plasma reduces this C only slightly further, never completely
eliminating it. We attribute this to redeposition of hydrocarbons
desorbed from the walls of our unbaked preparation chamber.
High resolution spectra (Fig. 6, inset) show in detail the increase
in silicon peak intensity relative to the corresponding decrease in
surface carbon. No change in the surface chemistry is apparent,
nor is there a change in the Si : O ratio.
Insulating surfaces, such as a minerals or glasses, can be cleaned
with an oxygen plasma, before charging electrostatically to the
extent that low-energy ions are repelled. While the plasma itself
must be electrically neutral, the impinging atoms, ions, and
electrons incident on an insulating material may eject electrons,
causing a sample to charge positively. The effect of surface
charging can be seen as a ‘dark’ region in front of the surface
of insulating samples, of initial width ∼1 mm and expanding with
exposure time. Charging, however, does not preclude the ability
of the ions and atoms to clean an insulator surface effectively.
An example of carbon removal from forsterite, a dielectric silicate
mineral (Fig. 7), shows the effect of 30-s oxygen plasma exposure.
The C is reduced to 10%, while the Mg Auger peaks and Si, Mg, and
O photoelectron peaks all intensify. There is no increase of oxygen
c 2009 John Wiley & Sons, Ltd.
Copyright Surf. Interface Anal. 2010, 42, 40–44
Compact plasma source for removal of hydrocarbons for surface analysis
Figure 6. Effect of 5-s exposure to O2 plasma from constricted-anode
source on adventitious carbon-coated silicon wafer (with native oxide
surface). The carbon is removed to the 1.5% level. Inset shows the effect of
30-s plasma exposure on the chemistry and intensity of carbon and silicon.
No chemical changes are observed.
Figure 4. Energy spectra for atomic and molecular oxygen. Ninety percent
of the ions have energies smaller than 28 and 33 eV for O+ and O2 + ,
respectively.
Figure 7. XPS spectra of the surface of forsterite (Mg2 SiO4 ), before and
after 30-s exposure to O2 plasma.
Figure 5. Effect of O2 plasma from constricted-anode source on surface
hydrocarbons as a function of exposure time. The treatment reduces the
surface concentration of carbon from 44 to 10% (atomic percent). The
intermediate values are 29, 15, and 11%.
relative to forsterite’s other constituents – only the removal of
surface hydrocarbons.
Discussion and Conclusions
Surf. Interface Anal. 2010, 42, 40–44
c 2009 John Wiley & Sons, Ltd.
Copyright www.interscience.wiley.com/journal/sia
43
While plasma cleaning is fairly novel within the field of surface
physics, it has become a standard practice for sample preparation
in electron microscopy.[12] In this application it is observed that,
in addition to cleaning the samples, argon/oxygen preprocessing
also appears to reduce hydrocarbon buildup at the electron beam
spot.[13 – 15] These plasma cleaners typically utilize a radio frequency
plasma in a cavity where the sample is inserted.[13,15] The main
drawback of the radio frequency (RF) plasma for surface analytical
application is the high energy of the ions that can preferentially
sputter some components of the sample thus changing its surface
composition.
Another method uses ozone or a combination of ozone and
ultraviolet (UV) radiation to remove hydrocarbons from samples
in surface analytical instruments.[16,17] These sources appear to
work well, the ozone/UV source being most efficient, but are
somewhat more complicated in nature (ozone generation) and
require significantly longer exposure times (hours) compared to
the seconds needed with our plasma source.
The constricted-anode plasma source can also operate with
other gases, besides oxygen. Activated hydrogen, nitrogen, and
other molecular compounds may also be useful as chemical
C. A. Dukes and R. A. Baragiola
cleaning or etching agents for various applications. We have
tested this plasma source with air, which forms a white–blue
plasma under similar conditions to the oxygen, and found that it
removes hydrocarbons from surfaces but required longer exposure
times. We plan future studies using the source with hydrogen, for
the removal of both hydrocarbons and air-formed metal oxides.
In summary, we have expanded and tested the design of
the constricted-anode plasma source and adapted it for in situ
use on a surface analytical system. We find that oxygen plasma
removes effectively and efficiently residual hydrocarbons and
adventitious carbon from the surface of different types of
materials. Measurements confirm that the source chemically
etches surfaces with low-energy ions, radicals, and activated
molecules, without altering the intrinsic surface stoichiometry
of nonoxidizing materials.
Acknowledgements
This work was supported by the NASA LASER (Grant: NNX08AX11G)
and Cosmochemistry (Grant: NNX08AG72G) programs.
[3] S. Kohiki, K. Oki, J. Electron Spectrosc. Relat. Phenom. 1984, 33, 375.
[4] P. Swift, Surf. Interface Anal. 1982, 4, 47.
[5] J. Geller, in Surface Analysis by Auger and X-ray Photoelectron
Spectroscopy (Eds: D. Briggs, J. T. Grant), The Cromwell Press:
Trowbridge, 2003, pp 89.
[6] M. A. Lieberman, A. J. Lichtenberg, Principles of Plasma Discharges
and Materials Processing, John Wiley and Sons, Inc., New York, 1994.
[7] A. Anders, S. Anders, Plasma Sources Sci. Technol. 1995, 4, 571.
[8] A. Anders, M. Kühn, Rev. Sci. Instrum. 1998, 69, 1340.
[9] A. Anders, G. Y. Yushkov, Rev. Sci. Instrum. 2007, 78, 043304.
[10] V. I. Miljević, J. Appl. Phys. 1986, 60, 4109.
[11] V. I. Miljević, Rev. Sci. Instrum. 1990, 61, 312.
[12] C. G. Morgan, M. M. Gleason, R. Vane, Microsc. Microanal. 2007,
13(SO2), 1736.
[13] J. T. Grant, S. D. Walck, F. J. Scheltens, A. A. Voevodin, Mater.Res.Soc.
Symp. Proc. 1997, 480, 49.
[14] P. E. Fischione, J. Ringnalda, Y. Feng, T. Krekels, M. Hayles,
H. O. Colijn, M. J. Mills, J. M. Wiez̄orek, Mater. Res. Soc. Symp. Proc.
1997, 480, 225.
[15] T. C. Isabell, P. E. Fischione, Mater. Res. Soc. Symp. Proc. 1998, 523,
31.
[16] D. W. Moon, A. Kurokawa, S. Ichimura, H. W. Lee J. Vac. Sci. Technol.,
A. 1999, 17, 150.
[17] T. Momose, E. Hayasaka, K. Saitou, K. Nagayama and S. Abe J. Vac.
Sci. Technol., A. 1998, 16, 961.
References
[1] T. L. Barr, S. Seal, J. Vac. Sci. Technol., A. 1995, 13, 1239.
[2] S. Evans, Surf. Interface Anal. 1997, 25, 924.
44
www.interscience.wiley.com/journal/sia
c 2009 John Wiley & Sons, Ltd.
Copyright Surf. Interface Anal. 2010, 42, 40–44