Negative Ion Density Measurements in Reactive Magnetron Sputtering

Full Paper

Negative Ion Density Measurements inReactive Magnetron Sputtering

Robert Dodd,* ShaoDong You, Paul M. Bryant, James W. Bradley

A combination of laser photo-detachment and conventional Langmuir probing has been usedto obtain the bulk negative ion density in both a DC and radio frequency (RF) sputtermagnetron. The argon and oxygen discharges were operated at low powers and over a rangeof pressures. The photo-detachment signal is expected to reach a limiting value; however, thesignal continues to increase with laser energy density and this can be attributed to a laserablation effect. In the RF magnetron the electron temperature (Te) in oxygen decreased withincreasing pressure, whereas the electron density (Ne) increased from 0.53 to 8.6�1014m�3. Ataround 12mTorr, a sudden increase in Ne by 3.7 is accompanied by a small drop in Te. Thenegative ion density (N�) also increases with pressure reaching a maximum of 1.7�1014m�3

between 5 and 10mTorr. Under similar conditions, the DC magnetron negative ion fraction(N�/Ne) is estimated to be �0.01, being significantly lower than in the RF magnetron whereN�/Ne� 1.

Introduction

Reactive magnetron sputtering, in which the discharge is

operated in a mixture of noble and reactive gases such as

oxygen, nitrogen or hydrocarbon have become well

established in industry and research to deposit engineering

quality thin films and coatings.[1] The operation of

magnetrons in RF is a useful tool for insulating targets[2]

and for reactive sputtering in gases like oxygen.[3] For

instance, the addition of oxygen into a magnetron

sputtering system can allow the high rate deposition of

oxide films for optical applications. The materials used

include titania (TiO2), silica (SiO2), tantalum pentoxide

(Ta2O5), indium tin oxide (ITO), etc.[4] These plasmas have

been extensively studied, through the use of Langmuir

probes, mass-energy analysis and optical emission techni-

ques,[5] to determine the concentrations and energies of

the electrons and positively charged species. However, the

presence and role of negative ions has been somewhat

neglected. Discharges containing oxygen negative ions

R. Dodd, S. D. You, P. M. Bryant, J. W. BradleyDepartment of Electrical Engineering and Electronics, Universityof Liverpool, Brownlow Hill, Liverpool, L69 3GJ, UKFax: þ44(0) 0151-794-4540; E-mail: [email protected]

Plasma Process. Polym. 2009, 6, S615–S619

� 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

(i.e. O� and O�2 ) may greatly affect the ion-assisted

deposition process of oxides at the substrate. Furthermore,

it has been found that high energy negative ions,

originating at the cathode and accelerated in the cathode

fall, can cause energetic re-sputtering of the deposited film

at the substrate.[6–8]

Despite the successes in plasma diagnostics to date the

concentrations of negative ions in the reactive sputtering

magnetron are still unknown. One effective method to

determine the negative ion density is by laser photo-

detachment.[9] This technique has been well established in

unmagnetised and non-sputtering plasmas such as the

RF plasma GEC cell.[10] Similar measurements in a pulsed

inductive RF discharge in oxygen were also carried out by

the same team[10] and by Corr et al.[11] Amemiya and

Suzuki[12] investigated the hollow cathode discharge using

this technique, as well as electrical probes, in oxygen and

neon. Other plasmas inwhich the technique has been used

successfully include Tokamak diverter simulators.[13]

In this paper, we aim to make for the first time

quantitative measurements of the concentrations of

electronegative species such as O� and O�2 in the RF

and DC magnetron using the laser photo-detachment

technique. This involves the illumination of a column of

plasma, defined by the beam diameter, by a short duration

DOI: 10.1002/ppap.200931606 S615

R. Dodd, S. D. You, P. M. Bryant, J. W. Bradley

S616

laser pulse with sufficient photon energy (hn) to exceed the

electron binding energy of the selected negative ion, A�.

This process canbewritten in the formA�þhn!Aþ e. The

photo-detached electrons temporally increase the electron

density in the laser beam path and can be detected by a

Langmuir probe.

The electron affinities of O� and O�2 are 1.46 and 0.44 eV,

respectively[12] so that the second harmonic of an Nd: YAG

laser (wavelength of 532nm, hn¼ 2.33 eV) is sufficient to

photo-detach electrons from both species. At the funda-

mental wavelength of 1 064nm (hn¼ 1.17 eV) only O�2 can

be photo-detached. The detachment fraction (DN�/N�) of

the bulk plasma negative ion density (N�) is dependant on

the incident photon flux and the photo-detachment cross

section, spd. For single species photo-detachment it can be

shown that this is given by.[9]

Figure 1. Schematic of the experimental setup. A DC or RF(13.56MHz with matching unit) power supply was used. M1and M2, turning mirrors; BS; PM; W1 and W2,windows; MFC1

Plasma

� 2009

DN�N�

¼ 1� exp � E

S

spd

hv

� �(1)

and MFC2; C1¼0.33 nF; R1¼ 10 kV; LP, Langmuir probe.
where E is the incident laser energy and S is the beam
cross-sectional area. The photo-detachment cross-sections

are 6.5� 10�18 cm2 (O�) and 1.7� 10�18 cm2 (O�2 ) at

532nm and 0.5� 10�18 cm2 (O�2 ) at 1 064nm.[12] The

opto-galvanic (OG) signal is measured by the Langmuir

probe (biased at plasma potential or higher) as a transient

electron current pulse of height, DIe (/N�). If this signal is

normalised to the pre-laser pulse value Ie (/Ne) then the

negative ion to electron density ratio a (¼N�/Ne) can be

obtained directly from DIe/Ie if DN�/N�� 1.[9]

Experimental Part

The experimental setup is shown in Figure 1. A verticallymounted

magnetron, with a 150mm diameter titanium target, was

positioned 12 cm from the grounded electrode in a 10 L chamber.

Both DC (up to 900W) and RF (up to 150W at 13.56MHz with

matching unit) power supplies were used in this paper. The

chamber can achieve a base pressure of less than 5� 10�6mbar

witha turbomolecular pump (PfeiferTMU071P) backedbya rotary

pump (Edwards E1M40). In industry, mixtures of gases are

routinely used; however, in this study the aim is to understand

the role of negative ions in the discharge and so only pure research

grade (>99.99%)gasesareused.Argonandoxygenwere fed into the

chambervia two20 sccmMKSmassflowcontrollers (MFC)working

in conjunction with a Baratron pressure gauge (MKS Type 627) to

regulate the working pressure in the range 1–20mTorr. Optical

access to the chamber is via two windows (with shutters) of

synthetic fused-silica 10mm thick positioned 10 cm from the

chamber axis. This is tominimise filmdepositionwhichmay cause

reflection or reduce the optical transmission of the incident laser

beam.

The Langmuir probe was situated 8 cm from the cathode, in the

magnetic null, and consisted of a 250mm diameter tungsten wire

with a re-entrant ceramic sleeve located outside of the laser beam.

Process. Polym. 2009, 6, S615–S619

WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

The 16mm long extension of the ‘L’ shaped probe tip (3mm short

extension) was aligned co-axially with the laser beam. For the RF

magnetron the active-compensationmethod[14]was usedwith the

ESPion acquisition system (Hiden Analytical Ltd) to obtain the

probe current–voltage characteristics. During the photo-detach-

ment the probe measurements were RF uncompensated with the

oscilloscope averaging over 512 samples to smooth out the RF pick-

up and to enhance the signal to noise ratio of the OG signal.

For the photo-detachment measurements, a pulsed Q-switched

Nd: YAG laser (Quantel Brilliant B) at wavelengths of 532 and

1064nmwas usedwith themaximumenergyper pulse of 350 and

700mJ, respectively. The laser repetition rate is 10Hzwith a 5–6ns

pulse width and a beam divergence of 0.5mrads. By using several

apertures the beam diameter can be changed from 2 to 6mm. A

power metre (PM) monitored a fraction of the average laser beam

power via the beam splitter (BS). The beam entered the chamber

throughthewindowW1afterbeingdirectedby the turningmirrors

(M1 and M2) and a periscope arrangement. The laser beam exited

the chamber through window W2 and its diameter and intensity

profileweremeasuredby a laser beamanalyser (LBA—USBeamPro

CMOSprofiler). The LBAwasalsoused to ensure that the laser beam

is co-axial with the Langmuir probe. A variable DC power supply,

capable of�120V, was used to bias the probe. When the laser was

fired, the photo-detached voltage signal was sampled from the

capacitor C1, displayed on an oscilloscope (Tektronix TDS3024B)

and captured by the computer for analysis. The measured voltage

signal was divided by the resistance R1 to obtain the OG current.

RF Magnetron Results

For these experiments the forward RF power was fixed at

40W (0W reflected) and the chamber pressure was varied

DOI: 10.1002/ppap.200931606

Negative Ion Density Measurements in Reactive Magnetron Sputtering

Figure 3. Photo-detachment fraction against laser energy densityat 532 nm and two probe biases in oxygen at 7.5mTorr with 40WRF power. Here the plasma potential is þ27.5 V.

from 1 up to 20mTorr. Prior to firing the laser the local

plasma potential was obtained from current–voltage

characteristic of the Langmuir probe in both argon and

oxygen. In this case the probe was actively RF compen-

sated.[14] To confirm the presence of negative ions, photo-

detachment measurements were obtained using a 4mm

diameter beam with average laser energy densities

increasing from 0 to 800mJ�cm�2. The probe bias was

varied from 0 up toþ80Vwith respect to ground potential.

In Figure 2 typical photo-detachment signals are shown in

oxygen at 7.4mTorr for two laser pulse energies (averaged)

at 532nm. The OG signals have short rise times of 600ns

(40mJ�cm�2) and 800ns (238mJ�cm�2), respectively fol-

lowed by long decay times of 10–15ms. By comparison

photo-detachment signals obtained inargonwere less than

2% under similar conditions used for oxygen. We detected

this smallphoto-detachedsignal inArdueto thepresenceof

sputtered or ablated oxides in the chamber form previous

runs using O2 gas. The detachment fraction at 40mJ�cm�2

corresponds to50%andlies inthe linearoperating regimeof

Equation 1whereDN�/N�� Espd/(Shn). Increasing the laser

pulse energy density to 238mJ�cm�2 increases the photo-

detachment fraction to 98% so that higher energy densities

theOGsignal is expected to reachsaturation,whenall of the

negative ions are photo-detached.

However, as shown in Figure 3 the OG signal does not

saturate and continues to increase with increasing energy

density as shown for probe biases of þ60 and þ27.5V

(plasma potential). Good agreement was obtained for both

biases below an energy density of �150mJ�cm�2. Non-

saturationof theOGsignal has also beenobservedbyKajita

et al.[15–17] and theyattributed this to a laser ablation effect.

Laser tests conducted at base pressure in our system

showed that a photo-detachment signal was observed for

an energy density greater than 200mJ�cm�2 which is in

goodagreementwithFigure3and thoseofKajita etal.[15–17]

There are several possible causes of this effect including

Figure 2. Typical photo-detachment signals at 532 nm in oxygen,40W RF power and þ30V probe bias. Here the plasma potentialwas þ27.5 V.



thermionic electron emission (due to laser and electron

current heating—if biased sufficiently above plasma

potential) and laser ablation of the probe surface. During

ablation the ejected neutrals could become ionised leading

to an increasing photo-detachment signal. The higher OG

signal for the þ60V bias could be due to the additional

electron heating of the probe enhancing the laser ablation

effect.

Taking the saturation of the photo-detachment fraction

at 159mJ�cm�2 for both probe biases the theoretical curve

forO�at532nm(Equation1)wasfitted to theexperimental

data in Figure 2. Tests performed at 1 064nm, where only

the O�2 species can be photo-detached, showed a to be less

than 2–4% of the electron density. This confirms that O� is

the dominant negative ion species and the presence of O�2

can be neglected. The fitted theoretical curve shows good

agreement to the measured curves.

Figure 4 shows the dependence of the electrondensityNe

and temperature Te with pressure as derived from the

Langmuir probe characteristics in oxygen. As expected Tedecreases (collisonal electron cooling) and Ne increases

(increasing ionisation rate) with pressure as observed in

other magnetrons and discharges. However, at around

12mTorr a rapid drop in Te by 1.7 eV and a rise in Ne (by a

factor of 3.7 from 1.7� 1014m�3) is observed. Photo-

detachmentmeasurementswerealsomadeat238mJ�cm�2

(in saturation regime) over the same pressure range. In this

case the detection probe was biased at þ3V above the

plasmapotential. These results are plotted in Figure 5, from

which the negative ion density is obtained from N�¼Ne a.

As shown, the negative ion density increaseswith pressure

until a maximum is reached somewhere between 5 and

11mTorr where N�� 1.7� 1014m�3. Between 11 and

12mTorr N� reduces rapidly to 0.7� 1014m�3 which

correlates with the rapid rise in Ne. At higher pressures

(16.7mTorr) the density reduces further to 0.3� 1014m�3

and at 20mTorr the OG signal was comparable to the noise

level.Weshouldnote that as thepressure increases from1.5

to 20mTorr, the plasma potential fell fromþ62 toþ27.5V.

To ensure we always detected the correct photo-detach-

www.plasma-polymers.org S617

R. Dodd, S. D. You, P. M. Bryant, J. W. Bradley

Figure 4. Plasma electron density (Ne) and temperature (Te) atvarious pressures in oxygen and 40WRF power, obtained from anRF actively compensated Langmuir probe. The error barsrepresent the uncertainty in the measurements.

S618

ment current the probe bias was changed at each new

pressure setting to maintain it 3V more positive than the

plasma potential.

The precise origin of the detected O� species is still

unknown. They could be produced from a number of

sources. For instance, through bulk plasma processes such

as dissociative electron attachment of ground-state or

metastableO2, as describedbyWagner andKatsch[18] for RF

plasmasor through the sputteringofnegative species at the

target as described by Mr’az and Schneider[19] The sudden

drop in the negative ion density N� observed at 12mTorr

may be a result of a lower number of negative species

released at the target as the sputter yield falls with

advancing target poisoning occurring at high O2 pressures.

This observation is different from that in[18] where the

negative ion fraction grows with O2 partial pressure in

approximately the same pressure range.

DC Magnetron Results

Tomake a comparisonwith the RF case, photo-detachment

and Langmuir probe measurements were obtained for the

magnetron operating in DC. This mode of operation is not

usual in processing due to poisoning of the target; however

Figure 5. Negative ion fraction (a¼N�/Ne) and density (N�) atvarious pressure in pure oxygen and 40W RF power. The errorbars represent the uncertainty in the measurements.



it does provide us with the opportunity to make pre-

liminarymeasurementsasa forerunning tomoreadvanced

studies in pulsed DC magnetrons. The applied power was

40Wwith the pressure set at 10mTorr. Photo-detachment

and Langmuir probe measurements were obtained in

oxygen with the probe situated at the same location as in

theRFexperiments. It is found that the onset of theablation

effect occurs at a lower laser energydensity of 100mJ�cm�2.

At the same nominal power, the deposition rate in the DC

magnetron are typically higher than for RF due to enhanced

plasma densities and higher sputter yields which correlate

to higher cathode fall potentials for DC. This leads to a

higher deposition rate of TiO2 films onto the probe surface

increasing the probe contamination. The presence of these

filmsontheprobe’s surfacemay lower the thresholdof laser

ablation.

With the laser energy density set to 238mJ�cm�2 the OG

signal was obtained at probe biases ofþ5 andþ10V above

plasma potential (�1V). It was found that a� 0.01 giving

N�¼ 0.12 and 0.28� 1014m�3 which is significantly lower

than that obtained in the RF magnetron where a� 1 and

N�� 1.7� 1014m�3 under similar conditions. The negative

ion fraction is 100 times smaller in the DCmagnetron than

in RF. In Figure 6 the Langmuir probe characteristics for

both the DC and RF are compared. It is well known that the

presence of a significant number of negative ions causes

the probe characteristic to become more symmetrical.[20]

The electron saturation current becomes comparable

to the ion saturation current as is the case for the

characteristic obtained in the RF magnetron. In the

DC magnetron the more asymmetric characteristic indi-

cates fewer negative ions in agreement with the photo-

detachment measurements. We can understand the

difference in negative ion fraction between the two

running modes as follows. In the DC magnetron, the high

cathode fall potential (325V in our case) can lead to the

generation of energetic secondary electrons in the plasma

due to ion bombardment at the target and subsequently

to the destruction of negative ions in the bulk.Whereas, for

Figure 6. Comparison of Langmuir probe characteristics in oxygenat 10mTorr and 40W in the RF and DC magnetrons.

DOI: 10.1002/ppap.200931606

Negative Ion Density Measurements in Reactive Magnetron Sputtering

the RF magnetron at low power, the self-biasing potential

on the cathode canbe typically theorder of only several 10’s

of volts (20V inour case)withamuch reducedproportionof

secondary electrons in the plasma. Consequently, this gives

rise to considerably higher negative ions fractions in the

plasma bulk.

Conclusion

Langmuir probe and photo-detachment measurements

were undertaken under similar conditions of pressure,

magnetron power and laser energy for both the DC and

RF excited magnetrons. The Langmuir probe was used to

collect the photo-detached electrons to obtain the negative

iondensity in thebulkplasma. Itwas foundthatanablation

effect becomes apparent, when the laser energy density

exceeds a threshold which increases the OG signal above

the expected saturation value at higher laser energy

densities. The threshold energy density appears to be

dependant on the probe surface condition reducing from

150mJ�cm�2 (clean probe, RF magnetron) to 100mJ�cm�2

(contaminated probe, DC magnetron) and possibly the

plasma conditions. The ablation effect can be reduced

by operating the laser energy density well below the

threshold in the linear regime. In oxygen and at 40W in

the RF magnetron the negative ion density (N�) increases

from 0.4� 1014m�3 (at 2mTorr) to a maximum of

1.7� 1014m�3 (between 5 and 10mTorr) followed by a

rapid decrease at 11mTorr to 0.7� 1014m�3. This correlates

with a rapid increase in the electron density (Ne) and a

drop in the electron temperature. Towards the higher

pressures N� decreases until at 20mTorr the negative ion

signal could no longer be observed being comparable to the

noise level.

In the DC magnetron N� found to be 0.28� 1014m�3

which is somewhat lower than that obtained in the RF

magnetron where N�� 1.7� 1014m�3 under similar con-

ditionsat10mTorrand40Wpower.However,due tohigher

plasma densities in the DC magnetron, the negative ion—

electron density ratio (N�/Ne) is 100 smaller than for the

RFcase. This is consistentwith the Langmuirprobe current–

voltage characteristicswhich show for the RF case a greater

symmetry between the magnitudes of the electron

saturation and ion saturation branches and hence the

presence of negative ions. For DC magnetron measure-

ments in O2 a more classic asymmetric single probe

characteristic is obtained. One possible explanation for

the lower negative ion density in the DC magnetron is due

to the greater flux of secondary electrons originating from

the cathode. After being accelerated through the cathode

sheath (to several 100s volts) these energetic electrons can



collidewith thenegative ionsdetaching theboundelectron.

In the low power RF magnetron, as considered here, where

the sheath potential drop is several 10s of volts, this process

would be less significant resulting in a higher negative ion

density.

Acknowledgements: The authors would like to thank the EPSRCfor funding of the project and Professor H. Amemiya for hisassistance and insightful discussions.

Received: September 12, 2008; Accepted: March 4, 2009; DOI:10.1002/ppap.200931606

Keywords: deposition; laser ablation; magnetron

[1] P. J. Kelly, R. D. Arnell, Vacuum 2000, 56, 159.[2] G. H. Yang, Y. Zhang, E. Kang, J. Phys. Chem. B 2003, 107(12),

2780.[3] H. H. Huang, C. C. Huang, P. C. Huang, C. F. Yang, C. Y. Hsu,

J. Nanosci. Nanotechnol. 2008, 8(5), 2659.[4] P. M. Martin, J. D. Affinito, C. A. Coronado, W. D. Bennett, M. E.

Gross, J. W. Johnston, D. C. Stewart, in ‘‘Surface ModificationTechnologies IX,’’ T. S. Sudarshan, W. Reitz, J. J. Stiglich, (Eds.),The Minerals, Metals and Materials Society, Warrendale, PA1966, p. 115.

[5] D. Depla, M. Stijn, Reactive Sputter Deposition, Springer Seriesin Materials Science, Springer, Berlin 2008.

[6] P. D. Rack, M. D. Potter, A. Woodard, S. Kurinec, J. Vac. Sci.Technol. 1999, 2805 (A17).

[7] L. Rieth, P. Holloway, J. Vac. Sci. Technol. 2004, 22(1), 20.[8] K. Tominga, M. Chong, Y. Shintani, J. Vac. Sci. Technol., A 1994,

12, 1435.[9] M. Bacal, Rev. Sci. Instrum. 2000, 71(11), 3981.[10] H. M. Katsch, T. Strum, E. Quandt, H. F. Dobele, Plasma Sources

Sci. Technol. 2000, 9, 323.[11] C. S. Corr, P. G. Steen, W. G. Graham, Plasma Sources Sci.

Technol. 2003, 12, 265.[12] H. Amemiya, T. Suzuki, Jpn. J. Appl. Phys. 1990, 29(9), 1772.[13] S. Kajita, S. Kado, N. Uchida, T. Shikama, S. Tanaka, J. Nucl.

Mater. 2003, 313, 748.[14] A. Dyson, P. M. Bryant, J. E. Allen, Meas. Sci. Technol. 2000,

11(5), 554.[15] S. Kajita, S. Kado, T. Shikama, B. Xiao, S. Tanaka, Contrib.

Plasma Phys. 2004, 44(7–8), 607.[16] S. Kajita, S. Kado, A. Okamoto, S. Tanaka, Jpn. Soc. Appl. Phys.

2005, 44(12), 8661.[17] S. Kajita, S. Kado, S. Tanaka, Plasma Sources Sci. Technol. 2005,

14, 566.[18] J. A. Wagner, H.-M. Katsch, Plasma Sources Sci. Technol. 2006,

15, 156.[19] S. Mr’az, J. M. Schneider, J Appl. Phys. 2006, 100.[20] H. Amemiya, J. Phys. D: Appl. Phys. 1990, 23, 999.

www.plasma-polymers.org S619

Negative Ion Density Measurements in Reactive Magnetron Sputtering

Documents

Transcript of Negative Ion Density Measurements in Reactive Magnetron Sputtering