`
`Enhancement of aluminum oxide physical vapor deposition with a
`secondary plasma夞
`
`Department of Nuclear, Plasma and Radiological Engineering, University of Illinois, Urbana-Champaign, Urbana, IL 61801, USA
`
`Ning Li*, J.P. Allain, D.N. Ruzic
`
`Received 15 March 2001; accepted in revised form 26 July 2001
`
`Abstract
`
`Reactive sputtering of aluminum oxide in a planar magnetron system is conducted with a mixture of O and Ar reacting with
`2
`and bombarding an aluminum target. The aluminum target is powered by a pulsed directed current (DC) bias which functions to
`discharge the accumulated ions on the insulating AlO film surface during the positive duty cycle and suppresses arc formation.
`x
`A seven-turn helical antenna sits below the magnetron sputtering system in the vacuum system and delivers radio-frequency (RF)
`power to generate a secondary plasma in the chamber. This plasma can efficiently ionize the sputtered flux, achieving ionized
`physical vapor deposition (IPVD). A gridded energy analyzer (GEA) and a quartz crystal microbalance (QCM) are located in
`the substrate plane to allow the ion and neutral deposition rates to be determined. Electron temperature and electron density are
`measured by a RF compensated Langmuir probe. A RF power of 500 W significantly increases the deposition rate of AlO up to
`x
`half of the Al deposition rate in metallic mode at the total pressure of 1.33 Pa (10 mtorr). At 3.33 Pa (25 mtorr), the ionization
`fraction of Al atoms reaches 90%. In addition the RF power extends the range of O partial pressure in which the sputtering
`2
`occurs in the metallic mode. SEM photos show that the secondary RF plasma makes the films smoother and denser due to a
`moderate level of ion bombardment. The deposition rates and ionization fractions fluctuate as a function of O partial pressure.
`2
`These variations can be explained by the combined variation of sputtering at the target, electron temperature and electron density.
`䊚 2002 Elsevier Science B.V. All rights reserved.
`
`Keywords: Physical vapor deposition; Ion bombardment; Radio-frequency; Reactive sputtering; Aluminium oxide
`
`1. Introduction
`
`Reactive sputtering is currently preferred to depositing
`dielectrics such as oxides and nitrides, as well as
`carbides and silicides, among which aluminum oxide is
`of particular interest because of its wide applications in
`material coating technologies for the protection of metal-
`lic components operating in hostile corrosive or oxida-
`tive environments w1x. In addition, aluminum oxide films
`
`夞Paper presented at 28th International Conference on Metallurgical
`Coatings and Thin Films, 30 April–4 May 2001, San Diego, CA,
`USA.
`* Corresponding author. 214 NEL, 103 S. Goodwin Ave., Urbana,
`IL 61801, USA. Tel.: q1-217-333-6291; fax: q1-217-333-2906.
`E-mail addresses: ningli@uiuc.edu (N. Li), druzic@uiuc.edu
`(D.N. Ruzic).
`
`show promise as possible substitutes of SiO films in
`2
`microelectronic devices w2x. Such applications are pos-
`sible due to aluminum oxide’s low refractive index,
`high resistivity, high mechanical hardness, high wear
`resistance, and high dielectric constant w1,3–7x.
`The implementation of reactive sputtering in DC
`magnetron systems has been frequently reported for
`deposition of highly insulating materials w4,5,8,9x. Sev-
`eral problems exist
`in using reactive DC magnetron
`sputtering for deposition of insulating films. One, in the
`process of reactive sputtering, the reactive gas forms a
`thin insulating film on the metal target surface, leading
`to charge build-up from highly-energetic incident ions.
`Breakdown of this charged layer occurs in the form of
`arcs leading to the ejection of material from the target
`which compromises the growing film quality. A second
`
`0257-8972/01/$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved.
`PII: S 0 2 5 7 - 8 9 7 2Ž 0 1. 0 1 4 4 6 - 3
`
`TSMC et al. v. Zond, Inc.
`GILLETTE-1010
`Page 1 of 10
`
`
`
`162
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`Fig. 1. Schematic of the magnetron system. A seven-turn coil sitting inside a commercial size magnetron delivers a secondary ICP. Sputtered Al
`atoms are partially ionized by the ICP and deposit on the substrate. The sensor simulates the position of a wafer to measure the deposition rate
`and ionization fraction.
`
`problem is an enhanced reduction in the metal deposition
`rate due to a variety of reasons, including preferential
`sputtering of the oxide species formed on the metal
`target surface, negative ions resputtering to the substrate
`film, and cluster sputtering w10x. A third problem is the
`higher secondary electron coefficient of the insulating
`films formed on the target w10x. The large secondary
`electron emission contributes to the increase of the
`discharge current, which in turn decreases the target
`voltage in order to maintain a fixed input power. The
`lower voltage decreases the sputter yield due to the
`lower incident energy.
`Alternate processes have been investigated because
`the use of reactive DC magnetron sputtering has not
`been optimized for the deposition of such films. Pulsed
`DC power reactive magnetron sputtering seems to be
`the most promising processes w1,5,9,11,12x.
`one of
`Frequencies in the range of 10–200 kHz, applied to the
`target, can successfully eliminate arcing while sustaining
`
`relative high-metal deposition rates w1x. Also the control
`of partial reactive operating pressure, or flow rates w13x
`and arc suppression techniques w1x are beneficial
`in
`pulsed DC sputtering.
`Besides the requirement of high deposition rates, the
`sputtered metal atoms are required to successfully fill
`high aspect ratio (AR) features (AR)1) with insulating
`films in ultra-large scale integration (ULSI) technolo-
`gies. The angular distribution of the sputtered atoms is
`roughly a cosine distribution, and is further broadened
`by gas phase scattering, yielding insufficient bottom
`coverage and voids during filling of high aspect ratio
`features. This problem is solved by ionizing the metal
`flux and applying a bias on the substrate, accelerating
`the metal
`ions through the plasma sheath near the
`substrate surface w14,15x. Since the sheath thickness is
`thinner than the mean free path and the electric field is
`normal to the substrate, a narrow angular distribution is
`achieved. Ionized physical vapor deposition (IPVD) is
`
`Fig. 2. Pulsed power wave form can effectively eliminate the arcs. Variation of the output affects the deposition process and film quality. Typical
`working points in this work: frequencys100 kHz, pulse widths20%, powers2 kW, V ;250 V.
`
`pp
`
`GILLETTE-1010 / Page 2 of 10
`
`
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`163
`
`Fig. 3. Coverage of Al and Al O at the substrate and on the target as a function of the partial pressure of O gas in the chamber. As the O
`2
`3
`2
`2
`component is extremely low, both the target and the substrate are mainly covered by pure Al, i.e. no Al O is deposited on the substrate. Later,
`2
`3
`within a certain range of O partial pressure, only the substrate is covered by Al O while the target is still clean, This is called the metallic
`2
`2
`3
`mode. And the deposition rate begins to drop. As the O percentage exceeds some critical value, both the target and the substrate are covered by
`2
`Al O which is called poison mode. An abrupt drop of deposition rate occurs at this transition.
`2
`3,
`
`accomplished by adding an inductively coupled plasma
`(ICP) coil between the substrate and target w16–18x.
`Previous work has shown IPVD is advantageous over
`PVD both in high deposition rates as well as feature
`target sputtering w17,18x. It
`fillings in metal
`is also
`practical to use IPVD to make insulating films such as
`Al O in combination with a pulsed power w19,20x.
`2
`3
`This paper investigates the benefit brought by IPVD
`in pulsed DC magnetron reactive sputtering of aluminum
`oxide films. Deposition rates are enhanced with the RF
`power at relative high total pressure. Electron density
`and electron temperature changes caused by the reactive
`gas partial pressure are measured by a Langmuir probe.
`SEM images of deposited films show differences in film
`morphology as a function of RF power. X-Ray photoe-
`lectron spectroscopy (XPS) characterizes the composi-
`tion of the films.
`
`2. Experiment
`
`A commercial DC planar magnetron-sputtering tool
`is used along with a secondary-plasma to study IPVD
`processes and applications w17,18x. As in conventional
`PVD, a 33-cm diameter pure aluminum target (cathode)
`is mounted with a rotating magnet being able to process
`200-mm wafers. The vacuum can be established to a
`y5
`y7
`base pressure of 6=10
`Pa (4.5=10
`torr) by a two-
`stage roughing pump and two CTI Cryogenic cry-
`opumps. Ar and O are introduced to the chamber as
`2
`the sputtering gas and reactive gas, respectively. Both
`flow rates are controlled by MKS mass flow meters.
`A seven-turn water-cooled helical resonator coil deliv-
`ers a 13.56-MHz RF power to the plasma and generates
`
`a secondary inductively coupled plasma. This plasma
`can efficiently ionize the sputtered flux, achieving IPVD.
`The total length of the unwound aluminum, helical coil
`is equal to one-half of the wavelength of the RF. The
`mid-point of the coil can be grounded, but was not for
`this experiment. The coil is placed between the target
`and substrate with a diameter of 310"20 mm. The
`matching network consists of three capacitors connected
`in parallel. The first is held constant at 600 pF, while
`the second and third capacitors can be varied between
`0–600 pF and 0–150 pF, respectively. Adjusted to proper
`values, the capacitors work effectively to minimize the
`reflected RF power supply. The net RF power absorbed
`by the plasma is the subtraction of reflected power from
`the incident power (500y24) W means 500 W forward
`and 24 W reflected. A cross-section of the chamber
`geometry, the seven-turn ICP coil and the diagnostic are
`shown in Fig. 1.
`The aluminum target is powered by an ENI RPG-100
`pulsed directed current DC power unit, which functions
`to generate the first sputtering plasma in the vicinity of
`the aluminum target. This unit discharges the accumu-
`lated ions on the insulating AlO film surface during
`x
`the positive duty cycle and suppresses arc formation.
`The unit utilizes a maximum arc count function to shut
`down power output when this count is reached, typically
`approximately 1000 counts. Several parameters of the
`output waveform can be adjusted to minimize arcing,
`thus protecting the substrate and deposited film from
`small metal droplets. The tradeoff is that low pulsed-
`power frequency tends to increase deposition rate (more
`like DC), while a higher frequency improves plasma
`chemistry, enhancing film uniformity, resistivity and
`
`GILLETTE-1010 / Page 3 of 10
`
`
`
`164
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`Fig. 4. Typical I–V traces of IPVD in the reactive sputtering at the total pressure of 3.33 Pa. Curves (a–d) demonstrate different conductivity of
`the plasma under different reactive sputtering modes. The two individual data, (e) and (f) were taken with Ar only.
`
`AlO composition. The typical values chosen for pulsed
`x
`power frequency and pulse width are 100 kHz and 2100
`respectively. The reverse voltage (positive) is
`ns,
`approximately 10% of the negative voltage. In this
`experiment, the power is kept at 2 kW, the negative
`voltage is in the range of 200–250 V. Fig. 2 depicts the
`pulsed width configuration.
`A water-cooled quartz crystal microbalance (QCM)
`and a three-layer-gridded energy analyzer (GEA) located
`in the plane of horizontal substrate surface are utilized
`to measure deposition rate and ionization fraction of the
`
`metal species. The top grid is grounded and the middle
`and bottom one are biased with the same voltage. The
`plasma has a positive potential with respect
`to the
`grounded top grid due to the formation of the plasma
`sheath, roughly in the order of 20 V. Since the mean
`free path in this experiment is far less than the distance
`between the substrate and target, oxygen negative ions
`formed and accelerated near the target cannot reach the
`substrate with enough energy to transit
`the plasma
`sheath. Therefore, negative ions cannot reach the QCM.
`When the bottom two grids are biased y30 V, both
`
`Fig. 5. Deposition rate variation with O partial pressure change under total pressure of 1.33 Pa. The process is divided into three regions according
`2
`to the variation. The error bars for the deposition measurements are approximately "10%.
`
`GILLETTE-1010 / Page 4 of 10
`
`
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`165
`
`Fig. 6. Deposition rate vs. O partial pressure change under total pressure of 3.33 Pa. The error bar for the deposition measurement is approximately
`2
`"10%.
`
`aluminum positive ions and neutral atoms diffuse to the
`crystal sensor and deposit on it. The deposition rate of
`the two species, D qD
`, can be determined by
`ions
`neutrals
`an XTCy2 deposition controller. While the bottom two
`grids are biased q30 V, positive ions are repelled and
`only neutral aluminum species deposit, thus obtaining
`. Therefore, the ionization fraction is calculated
`D
`neutrals
`by the two sets of data:
`Ionization fractions
`
`(1)
`
`the film is believed to be Al or Al O since their density
`2
`3
`and Z-ratios differ. If the film was conductive, the Al
`values were used. If the film was an insulator,
`the
`Al O values were used. The density and Z-ratio of
`2
`3
`AlO with various x values are unknown. We believe
`x
`the error introduced by this procedure falls within the
`reported error bars.
`A Hitachi S4700 is used for SEM images to assess
`film morphology. XPS was used to determine the com-
`position of AlO at the 40% O (RF on) point, where x
`2
`x
`was found to be 1.75, a bit more O rich than the ideal
`ratio of 1.5 for Al O , without presputtering.
`2
`3
`
`3. Results
`
`There are two modes of operation for reactive sput-
`tering that should be considered in depositing a com-
`pound film. Fig. 3 is a qualitative graph which explains
`how the film produced and the target material change
`as a function of increasing O partial pressure. Two
`2
`deposition conditions for Al O are apparent: ‘metallic’
`2
`3
`mode, where the target is still pure metal, and ‘poison’
`mode where the target has become covered with Al O .
`2
`3
`Note that
`in general, as depicted in Fig. 3, higher
`deposition rates are achieved in the ‘metallic mode’ than
`in ‘poison mode’. For fixed total pressure, as the reactive
`gas flux is varied,
`there is a transition between the
`poison and metallic modes which can exhibit hysterisis.
`The addition of an inductively-coupled (ICP) coil to
`generate a secondary plasma produces more Ar ions,
`more Al ions and more O radicals, which changes the
`transition point between the modes and the variation of
`
`Dion
`D qD
`ion
`neutral
`Deposition rates are taken at the throw distance of 15
`cm and are calibrated with film thickness profilometer
`measurements on a silicon substrate. Calibration of the
`measured data accounts for the transparency of the grids,
`chamber geometry, and scattering. Calibration was done
`for a pure Al deposition and for an AlO deposition at
`x
`40% oxygen partial pressure. This procedure
`is
`explained in detail in earlier papers w17x. Film thickness
`is measured based on the following formula w21x:
`p F yF
`Ž
`.
`N dat q
`q
`F
`pd F Z
`f
`c
`q
`where Z represents Z-ratio and is a function of film
`density and shear modulus and T , film thickness; N ,
`f
`at
`frequency constant of AT cut quartzs166 100 Hz cm;
`d , density of single crystal quartzs2.649 gycm ; d ,
`3
`q
`f
`film density; F , quartz crystal with uncoating frequen-
`q
`cy; and F , quartz crystal with coating frequency.
`c
`For a quartz crystal frequency change of 2% F , the
`q
`Al film thickness change is 1.456 times that of an
`Al O film. Therefore, different calibrations are used if
`2
`3
`
`B
`T s
`C
`f D
`
`B
`B
`E
`C
`C
`F
`arctan Z tan
`D
`D
`G
`
`c
`
`EE
`FF
`GG
`
`(2)
`
`GILLETTE-1010 / Page 5 of 10
`
`
`
`166
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`Fig. 7. Electron temperature and electron density change with the O partial pressure with total pressure of 1.33 Pa and 3.33 Pa. These are the
`2
`two main factors affecting the ionization fraction of the Al atom.
`
`deposition rate with the O partial pressure. With the
`2
`RF power on, the metallic mode occurs at a wider range
`of reactive gas partial pressure than without the RF
`power. Since the number of Al ions also increases with
`the addition of RF, the deposition rate goes up as the
`RF power is applied.
`Depending on the working condition, the impedance
`of the target–plasma–ground circuit differed as shown
`in the current–voltage (I–V) traces (Fig. 4). A higher
`magnetron cathode current was induced at a given
`voltage when the RF power to the ICP coil
`is on,
`regardless of whether one is in metallic or poison mode,
`due to increased ionization of both aluminum atoms and
`gas atoms caused by the secondary plasma. Since a
`higher ion-induced secondary electron emission rate
`exists for aluminum oxide compared to pure aluminum
`w10x, magnetron cathode current
`increases with any
`coverage of aluminum oxide (Q)0) on the target.
`Therefore,
`the behavior in metallic mode vs. poison
`mode is quite distinct.
`
`In addition to the four curves shown in Fig. 4, two
`data points (shown in Fig. 4) were taken with Ar only.
`Note that a higher current was still obtained with the
`RF on, though not as high as when O was present. The
`2
`even higher current with O present is due to RF power
`2
`producing O radicals which can oxidize the target away
`from the sputtering track and therefore raises the sec-
`ondary electron component of the current. Once the
`target was heavily covered with aluminum oxide (far
`into poison mode, 100% O ), the secondary electrons
`2
`dominated the current, and the addition of RF power
`did not change the current significantly.
`Deposition rates were measured under different O2
`partial pressures and total (ArqO ) pressures. These
`2
`rates are presented in Figs. 5 and 6. Fig. 5a,b are the
`deposition rates vs. oxygen partial pressure at 1.33 Pa
`(10 mtorr) total pressure, with the RF power on and
`off, respectively. Each figure is divided into three regions
`as explained in the figure. The placement of the dashed
`lines is approximate and could vary "5%. Several
`
`GILLETTE-1010 / Page 6 of 10
`
`
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`167
`
`the first point in Fig. 5a to the first point in Fig. 5b,
`when the RF is on, More sputtering occurs due to more
`q
`Ar
`. In addition, aluminum atoms are ionized by the
`secondary plasma and deposited directionally to the
`substrate. Therefore for a pure aluminum target,
`the
`deposition rate with a secondary plasma is greater than
`without a secondary plasma.
`The second region contains a higher percentage of
`O partial pressure, where aluminum oxide began to
`2
`deposit on the substrate while the target was mostly
`clean (metallic mode). In this region the electron tem-
`perature (Fig. 7a) increased as more O was present
`2
`dramatically increasing the ionization fraction (Fig. 8a).
`q
`In addition, the O also contributes to the sputtering
`process. According to transport of ions in matter (TRIM)
`w23x simulation of sputtering to an Al O layer with the
`2
`3
`˚
`thickness of 1000 A, the sputter yield of aluminum by
`q
`q
`O bombardment is much larger than by Ar bombard-
`ment. For example, for a 300-eV normal incidence, the
`q
`aluminum sputtering yield due to O bombardment of
`Al O is 0.968 compared to 0.422 due to argon bom-
`2
`3
`bardment. Consequently more energy is transferred to
`target atoms from oxygen bombardment than from argon
`bombardment, and sputtering of aluminum as well as
`aluminum from Al O , is more efficient. As the Al O
`2
`3
`2
`3
`q
`coverage grows the effect of ionization and O sputter-
`ing on deposition rate reaches a maximum and prefer-
`ential sputtering effects take over. It is more likely to
`sputter O than Al from fully covered Al O targets. At
`2
`3
`that point the Al deposition curve begins to decrease.
`Fig. 5a shows the maximum value happened at O2
`partial pressure of approximately 40% and reached
`nearly half of the aluminum deposition rate in pure Ar.
`In the poison mode region, the third region of Fig.
`5a, both the target and substrate were covered by AlOx
`and few Al atoms were sputtered. Formation of oxides
`on the target leads to a substantial decrease in aluminum
`sputtering due to preferential sputtering of the oxygen
`as well as an effective increase of the surface binding
`energy of the target atoms w10x. These factors explain
`the substantial drop of deposition rate, as low as one-
`tenth of aluminum deposition rate in pure Ar.
`Fig. 5b is the deposition rate variation with partial
`pressure of O at a total pressure of 1.33 Pa (10 mtorr)
`2
`while the RF power was off. Compared to Fig. 5a the
`variation was simple and was only affected by the
`AlO coverage of the target. The deposition rates were
`x
`lower than the Fig. 5a case when AlO was deposited.
`x
`Also, the RF power extended the region of O partial
`2
`pressure in the desirable metallic mode deposition of
`AlO as shown in Fig. 5a. This is because (1) the RF
`x
`q
`power generated more Ar
`ions, which are accelerated
`by the cathode voltage to sputter the target surface and
`slow down the formation of AlO on the target surface,
`x
`
`Fig. 8. Ionization fraction changes with the O partial pressure at the
`2
`incident RF power of 500 W.
`
`combined factors determine the variation of deposition
`rates with the reactive partial pressure: the number of
`aluminum atoms sputtered from the target, the aluminum
`oxide coverage on the target, the electron temperature,
`and electron density of the ICP plasma.
`In the first region of Fig. 5a, the O partial pressure
`2
`low (less than 20%), both the target and
`was still
`substrate were covered mainly by pure aluminum, and
`only aluminum was deposited on the substrate (as
`described in Fig. 3). Total pressure was fixed, so as the
`argon density is decreased,
`the O partial pressure
`2
`increases, resulting in a lower sputtering rate of the
`aluminum target and a dramatic drop in the deposition
`rate. The sputtering rate drops because even some
`oxygen coverage of the surface effectively increases the
`surface binding energy of aluminum atoms. This effect
`is exacerbated due to preferential sputtering of Al over
`Al O and decreases the yield w22x further. Comparing
`
`2
`
`3
`
`GILLETTE-1010 / Page 7 of 10
`
`
`
`168
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`Fig. 9. Comparison of SEM pictures between RF power on and RF power off, at different O partial pressures and different total pressure (Arq
`O ).2
`
`2
`
`even though more O radicals are now present; and (2)
`the substrate does not undergo any sputtering due to the
`low sheath voltage and this permits more oxidation of
`the deposited film. In turn, with lower AlO target
`x
`coverage, oxygen sputtering dominates and more alu-
`
`minum is sputtered as explained earlier. The fact that
`RF power dissociates more O molecules into O atoms,
`2
`which are consequently responsible for the formation of
`aluminum oxide is most apparent at the substrate where
`no sputtering occurs. That
`is why aluminum oxide
`
`GILLETTE-1010 / Page 8 of 10
`
`
`
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`
`169
`
`deposition is obtained at lower O partial pressure, with
`2
`a secondary plasma present.
`Fig. 6a,b are the deposition rates vs. oxygen partial
`pressure at 3.33 Pa (25 mtorr) total pressure with the
`RF power on, while Fig. 6b is the same condition with
`the RF power off. Fig. 6a shows deposition rate variation
`with oxygen partial pressure at a total pressure of 3.33
`Pa (25 mtorr) with the RF power on. The variation is
`similar to that of Fig. 5 — except that the metallic
`mode region is wider. With the RF on, the mechanisms
`involved are the same as in the 1.33-Pa (10-mtorr) case.
`With the absence of a secondary plasma (RF off), Oq
`sputtering dominates in the first region since at a higher
`q
`pressure, a greater number of O will exist. Thus, an
`increase in deposition rate is seen in Fig. 6b. However,
`since more of these oxygen ions will react with the
`target, the deposition rate will reach a maximum and
`based on preferential sputtering, the deposition rate will
`begin to decrease.
`Ionization fraction of the Al atoms at 3.33 Pa (25
`mtorr) total pressure was higher than that of 1.33 Pa
`(10 mtorr), as Fig. 8 shows. The ionization fractions go
`to a maximum value of 90% (3.33 Pa) ionization
`fraction in metallic mode due to the increase of electron
`temperature (see Fig. 7). The electron temperature did
`not vary significantly beyond that point, but the ioniza-
`tion fractions dropped because the electron density
`reached a minimum approximately 80% O partial pres-
`2
`y
`sure. The presence of O ions likely decreased the
`density of electrons at the highest O partial pressures.
`2
`At 100% O partial pressure,
`there was a rise in
`2
`ionization fraction at both 1.33 and 3.33 Pa due possibly
`to the dynamics of the electronegative gas.
`Fig. 9 shows SEM images of film formed on the
`silicon substrates. Fig. 9a is for the cases of 20% O2
`partial, and Fig. 9b is for 100% O partial pressure.
`2
`Both cases included data with and without RF power at
`a total pressure of 0.667 Pa (5 mtorr). At 0.667 Pa (5
`mtorr) total pressure, with 20% O , (Fig. 9a) the
`2
`deposited film was AlO when RF was present, while
`x
`pure aluminum film was only obtained when the RF
`was absent. As discussed in the deposition rate figure,
`RF power shifts the onset of metallic mode AlOx
`deposition. When RF is on, the secondary ionization
`increased positive ion bombardment to the film, thus
`densifying it. Without RF power, the film shows porous
`columnar structure. In the 100% O case SEM images
`2
`shown in Fig. 9b, the film made without the RF looks
`like molten droplets; but the RF power and subsequent
`ion bombardment keeps the column structure, and film
`is smoother. Fig. 9c shows the morphology of films
`made at 3.33 Pa (25 mtorr) total pressure with 40%
`O , both with a heated substrate of 2208C. The ion
`2
`bombardment produced by the RF power makes the film
`
`denser and smoother, and the grain structure is finer and
`more complete.
`
`4. Conclusions
`
`IPVD is advantageous over conventional PVD in
`reactive sputtering for three reasons. First,
`the ICP
`extends the metallic mode operating range of reactive
`sputtering of AlO . Ionization fraction and deposition
`x
`rate can be maximized by controlling the reactive gas
`partial pressure. This variation of deposition rate and
`ionization fraction with the reactive gas partial pressure
`was due to the combined reasons of target coverage,
`number of Al atoms sputtered off, electron temperature
`and electron density of the inductively coupled plasma.
`Second, the deposition rate of AlO was higher with the
`x
`RF power on than without. Lastly, the increased ioni-
`zation fraction contributes to ion bombardment of the
`film on the substrate and creates a denser and smoother
`film.
`
`Acknowledgements
`
`NSFyDOE Basic Plasma Science Initiative, DE-
`FG02-97ER54440 supplies part of the funding and MRC
`(Now TEL, Arizona) donated the commercial magne-
`tron. CVC Corp. (now Veeco & CVC) partially funded
`the work and donated the RPG-100 pulsed power supply.
`SEM work was carried out in the Center for Microanal-
`ysis of Materials, University of Illinois, which is par-
`tially supported by the US Department of Energy under
`grant DEFG02-96-ER45439. Vania Petrova is acknowl-
`edged for assistance in SEM work.
`
`References
`w1x P.J. Kelly, J. Vac. Sci. Technol. A 17 (3) (1999) 945.
`w2x J. Kolodzey, Appl. Phys. Lett. 71 (26) (1997) 3802.
`w3x C. Deshpandey, L. Holland, Thin Solid Films 96 (1982) 265.
`w4x M. Scherer, J. Schmitt, R. Latz, M. Schanz, J. Vac. Sci. Technol.
`A 10 (1992) 1772.
`w5x P. Frach, U. Heisig, Chr. Gottfried, H. Walde, Surf. Coat.
`Technol. 59 (1993) 177.
`w6x K.K. Shih, D.B. Dove, J. Vac. Sci. Technol. A 12 (1994) 321.
`w7x B. Lux, C. Colombier, H. Altena, K. Stjemberg, Thin Solid
`Films 138 (1986) 49.
`w8x D.A. Glocker, J. Vac. Sci. Technol. A 11 (1993) 2829.
`w9x W.D. Sproul, M.E. Graham, M.S. Wong, S. Lopez, D. Li, R.A.
`Scholl, J. Vac. Sci. Technol. A 13 (1995) 1188.
`w10x W.D. Westwood, in: S.M. Rossnagel, J.J. Cuomo, W.D. West-
`(Eds.), Handbook
`wood
`of Plasma Processing, Noyes
`Publications, Park Ridge, New Jersey, USA, 1990, pp. 233–257.
`w11x A. Belkind, A. Freilich, J. Vac. Sci. Technol. A 17 (4) (1999)
`1934.
`w12x J.M. Schneider, W.D. Sproul, R.W.J. Chia, M.-S. Wong, A.
`Matthews, Surf. Coat Technol. 96 (1997) 262–266.
`w13x M. Kharrazi Olson, J. Vac. Sci. Technol. A 16 (2) (1998) 639.
`w14x P.F. Cheng, S.M. Rossnagel, D.N. Ruzic, J. Vac. Sci. Techol. B
`13 (2) (1995) 203–208.
`
`GILLETTE-1010 / Page 9 of 10
`
`
`
`170
`N. Li et al. / Surface and Coatings Technology 149 (2002) 161–170
`w15x S.M. Rossnagel, J. Hopwood, Appl. Phys. Lett. 63 (1993) 24.
`w20x J.M. Schneider, W.D. Sproul, A. Matthews, Surf. Coat. Technol.
`w16x S.M. Rossnagel, Thin Solid Films 263 (1995) 1.
`98 (1998) 1473–1476.
`w17x M.M.C. Allain, D.B. Hayden, D.R. Juliano, D.N. Ruzic, J. Vac.
`w21x Leybold XTCyC XTCys User Guide, 5–11, part number 074-
`Sci. Technol. A 18 (3) (2000) 797.
`183y041795.
`w18x D.B. Hayden, D.R. Juliano, M.N. Neumann, M.C. Allain, D.N.
`w22x R. Kelly, N.Q. Lam, Rad. Effects 19 (1) (1973) 39.
`Ruzic, Surf. Coat. Technol. 120_121 (1999) 401–404.
`w23x J.P. Biersack, W. Eckstein, Appl. Phys. A 34 (1984) 73.
`w19x J.M. Schneider, W.D. Sproul, A.A. Voevodin, A. Matthews, J.
`Vac. Sci. Technol. A 15 (1997) 1084–1088.
`
`GILLETTE-1010 / Page 10 of 10
`
`