Samsung Electronics Co., Ltd. v. Demaray LLC
`Samsung Electronic's Exhibit 1039
`Exhibit 1039, Page 1
`
`

`

`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`49
`
`1. Introduction
`
`great care has to be taken with the time constants
`associated with the changes of gas pressure in a vac-
`uum chamber.
`Reactive sputtering, where a metal target is sput-
`tered by an inert gas (e.g. argon) in the presence of a
`A further solution lies in separating the deposition
`reactive gas (e.g. oxygen), to produceafilm of electri-
`and reactive process in a cyclic process where a thin
`cally insulating material has proved to be difficult to
`metal film is deposited and then converted. In our
`versionsof this process, the substrate is moved (succes-
`introduce as an industrial process. The simple concept
`sive plasma anodisation, or SPA) or the magnetron is
`is that metal is sputtered from the target and is inci-
`dent on the substrate/growing-film-surface,
`together
`made to change function from a provider of sputtered
`with reactive gas from the residual atmosphere, to form
`metal
`to one of oxygen and energetic argon ions
`(successive pulsed plasma anodisation, or SPPA). In
`a compound. The reaction occurs when these species
`meet on the surface. Energy is provided from the
`both cases, the unbalanced magnetron, the provider of
`substrate temperature,
`the high energy of sputtered
`argon-ion energy,
`is required to operate in the pres-
`material or by ions used to bombard the surface. Mag-
`ence of a partial pressure of oxygen.
`netron sputtering has allowed the sputtering to be
`When the reaction product is insulating and the
`undertaken at sufficiently low pressures such that the
`power is DC, other problems appear. A region of the
`mean free path for sputtered material is greater than
`target-cathode remains covered with reaction product,
`the target-to-substrate distance and the energy of this
`the racetrack region only can be balanced to be sput-
`material and the efficiency of its transfer from the
`tering metal faster than it is reacting with the propor-
`target can be maintained. The magnetron can also be
`tion of reactive gas in the system. This insulating region
`arranged,
`through unbalancing the magnetic field,
`is subject to bombardmentof ions and becomeshighly
`and/or operating it in a system to create a closed field,
`charged, due to charge accumulation or through sec-
`to direct a dense plasma to the surface of the growing
`ondary-electron emission. This leads to rapid localised
`film. This plasma can cause an insulating or isolated
`discharges, arcs, which disrupt the sputtering discharge
`surface to acquire a floating bias, which leads to it
`and causes arc-evaporation from regions of the target,
`being bombarded with low energy argon ions, to pro-
`with particle ‘spitting’ and consequent contamination
`vide energy for excitation energies for chemical and
`of the film. If the process is not properly designed, and
`structural reactions. The growth of the film can be
`carefully balanced, an insulating film may be formed on
`controlled in this manner and it does not require the
`the anodeof the discharge and on the chamberwalls to
`film or substrate to be electrically conducting. If addi-
`give a discharge which changes character with time.
`tional bias is required, RF can be used. The problems,
`These problems can be overcome by using RF in reac-
`which have to be solved, occur at the cathode surface.
`tive sputtering and target biasing. RF introduces its
`The presence of reactive gas in a region of high
`own problems and is not suitable for the high-rate,
`energy and freshly exposed metal surfaces, which is the
`large-area processes that are required.
`sputtering race-track of the magnetron, leads to rapid
`More recently, a solution to these problems has
`reaction. If the surface reacts faster than it sputters, a
`appeared. The insulating surface has to be discharged
`surface compound film is formed, which in general
`before the charge accumulates to an extent where an
`Sputters at a much lower rate than the metal, a process
`arc is formed. This has been done by providing a
`which has becometo be called ‘poisoning’. The transi-
`reverse potential pulse to the surface within the charge
`tion between metal sputtering and ‘poisoning’ is depen-
`accumulation time. This can be a short pulse applied to
`dent upon the power used to sputter and the partial
`a DC supply or AC power applied between two cath-
`pressure of the reactive gas and occurs at a different
`odes. The frequency required turns out to be approxi-
`level when approached from the metal or ‘poisoned’
`mately 40 kHz, which can be provided, without the
`condition. This process leads to a hysteresis being seen
`tuning systems necessary with RF, and at low cost. The
`between the sputtering parameters and the flow of
`latter process has the further advantage of creating a
`reactive gas into the chamber andthere is a region of
`continuously cleaned anode, because for the other half
`this flow where the process is unstable. It becomes one
`of the cycle it is the sputtering cathode. This emerging
`of the rapid deposition of a partially reacted metal or
`technique has been utilised recently by a few workers
`[1-10]. The results were very promising. For example,
`the slow deposition from a poisoned target of a fully
`Szczyrbowski and Braatz [7] have reactively deposited
`reacted film. This problem can besolved by controlling
`the state of the sputtering surface, and hence the
`films of SiO, at high rates using 40 kHz-AC power
`partial pressure of the reactive gas, through the light
`applied between two Si magnetrons. In addition to the
`emitted by the sputtering metal or other gases in the
`excellent optical and mechanical properties of the de-
`discharge, or in some cases, from the voltage of the
`posited films, no arcing was observed during the entire
`Sputtering cathode. These techniques provide a fast
`lifetime of the target, which was more than a week.
`Schiller et al.
`[11] have reactively deposited films of
`feedback to the control of the flow of the gas. However,
`
`Ex. 1039, Page 2
`Ex. 1039, Page 2
`
`

`

`50
`
`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`Al,O, from two Al magnetrons using an AC powerat
`different frequencies ranged from 50 Hz to 164 kHz.
`They reported a significant decrease in the defect den-
`sity of the deposited films with increasing frequency.
`The curve, for non-absorbing films, saturated at fre-
`quencies greater than 50 kHz, which was an indication
`of an arc-free process beyond this frequency. They also
`reported a deposition rate of approximately 60% of
`that of metallic Al. The rate was almost independentof
`the frequency in the range they investigated. Scherer et
`al. [9] have also adopted the two cathodes technique to
`reactively deposit films of Al,O;, SiO, and Si,N, using
`40 kHz-AC power. They reported deposition rates com-
`parable to those obtained with the DC power.
`A further problem is encounteredif it is desired to
`reactively sputter two metals at the same time to pro-
`duce mixed metal oxides of controlled composition.
`The metals operate in different ranges of partial pres-
`sure of the reactive gas for their optimum deposition,
`which also changes with the powerthat is used, andit
`is difficult to operate a system with both sources oper-
`ating in the desired metal sputtering mode.
`This paper reports on the deposition of mixed metal
`oxides using a simultaneous oxidisation stage, con-
`trolled by spectral radiation from the discharge (plasma
`emission monitoring, or PEM), or the cathode voltage
`in constant-power sputtering. This is used in conjunc-
`tion with movement of the sample between that source
`and one sputtering the other metal. The process was
`optimised with witness pieces admitted through an air
`lock into a continuously operating system. The second
`metal was oxidised in the unbalanced magnetron dis-
`charge ofthefirst, ic. a SPA process, whose sputtered
`metal was simultaneously converted. Power (40 kHz)
`was applied between the two cathodes and a DC bias
`used to adjust the ratio of metals in the mixed oxide.
`The stoichiometry was determined by the partial pres-
`sure, i.e. the PEM signal, in that chamber, the power
`applied and the rate of rotation.
`
`2. The techniques
`
`2.1. Mid-frequency AC powered magnetronsin floating
`mode with a DCbias
`
`For the coating composition to be easily controlled,
`the amount of power(or current) received by one/both
`of the floating magnetrons has to be independently
`varied. This was achieved by DC-biasing one of these
`magnetrons. In most cases, the two floating magnetrons
`were operated at a constant AC power, using the
`40-kHz supply, whereas the DC power, applied to one
`of the magnetrons, was varied. Such a method of com-
`bined AC and DC power application enjoys the fol-
`lowing major advantages:
`
`1.
`
`2.
`
`3.
`
`It paves the way to produce coatings of virtually
`any desired composition by simply varying the ratio
`of the applied DC and AC powers.
`It retains the advantages of applying mid-frequency
`AC power, between two floating magnetrons.
`It also retains the advantages of applying DC power,
`such as high deposition rates onto large area sub-
`strates.
`
`The AC powersupply wasisolated from the DC with
`a capacitor and the DC from the AC with a LCLfilter.
`Fig. 1 shows a schematic diagram of this filter. It should
`be emphasised, that the plasma was earthed through
`the chamber wall to provide the DC return path for the
`current. Initial experiments were carried out with two
`Al targets, sputtered in an Ar atmosphere. An oscillo-
`scope was used to measure the DCpotential developed
`on the magnetrons, relative to earth, in the following
`cases:
`
`1. On one of the magnetrons when the applied power
`
`Plasma
`Nene
`
`
`
` a
`
`
` MDX 2500-W
`
`vy
`68nF <2 Built-in
`1500 V ~7~ capacitor
`
`PE 2500 Generator
`(40 kHz AC Power Supply)
`
`Magnetron Drive
`(DC Power Supply)
`
`Fig. 1. The filter used to protect the DC powersupply from AC currents.
`
`Ex. 1039, Page 3
`Ex. 1039, Page 3
`
`

`

`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`51
`
`(a) On one of the magnetrons when the applied power was 100 W AC
`floating between the two magnetrons
`
`Voltage
`04)
`
`Voltage
`(Vv)
`
`Voltage
`
`part of the circuit, is to be the same in both directions
`when a DC bias is applied. Secondly, without a DC
`bias, the capacitor is charged/discharged following the
`voltage waveform of the AC power supply. Alterna-
`tively, when a DCbias is applied to one of the mag-
`netrons, the capacitor is charged/discharged (via the
`biased magnetron) following the voltage difference V,
`—V,, where V,
`is the fixed DC voltage and V,
`is the
`alternating AC one. This leads to a slight change in the
`AC current in the AC part of the circuit from that
`when only the AC power was applied. As a result, a
`slight change in the voltage of the non-biased mag-
`netron should also occur to maintain AC current mag-
`nitude in both directions. In other words, for periods
`when the AC current flow is in the same direction to
`the DC current, the capacitor will charge due to the
`latter. This accumulated charge is discharged onto the
`biased magnetron. In order to maintain AC current
`magnitude in both directions,
`the non-biased mag-
`netron must develop an appropriate bias, as can be
`noticed by comparing the positive voltage regions of
`Fig. 2a,c. Consequently, Fig. 2a can be regarded as a
`representative voltage waveform for such a circuit.
`Comparing Fig. 2a,c, it is suggested that the AC cur-
`rent (in the AC part of the circuit) is indeed approxi-
`mately equal in both cases. The magnitudes in both
`directions are approximately equal, within the biased
`magnetron arrangement.
`On the other hand, whenever the DC potential of
`the magnetron is negative with respect to the plasma
`potential (earth potential in this case), sputtering will
`occur. It is easy to see, then, from Fig. 2b, that the
`biased magnetron is sputtering continuously, albeit to
`various degrees. Fig. 2c shows that
`the non-biased
`cathode is not sputtering continuously; although it at-
`tains a potential of approximately — 550 V, these peri-
`Fig. 2. A reproduction of the photographs taken from the oscillos-
`ods are short compared to the ‘off times’, when its
`cope for the DC potentials, relative to earth, developed onafloating,
`potential goes positive with respect to plasma potential.
`biased and non-biased Al magnetron.
`In summary, we have:
`
`
`
`(Vv) 5
`
`5
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`Time(Us)
`(b) On the biased magnetron when the 100 W ACfloating
`was combined with 300 W DC
`
`5
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`Time (Hs)
`(c) On the non-biased magnetron when the 100 W AC floating was
`combined with 300 W DC
`
`
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`Time (Hs)
`
`was 100 W AC floating between the two mag-
`netrons (Fig. 2a).
`2. On both the biased (Fig. 2b) and the non-biased
`(Fig. 2c) magnetrons, when the 100-W ACfloating
`was combined with a 300-W DC powerto bias one
`of the magnetrons (a commoncase).
`
`A similarity can be noticed between Fig. 2a,c. To
`understand the reason behind that, the readeris re-
`ferred to Fig. 1. The capacitor shown in the AC part of
`the circuitry is an internal component of the output
`stage of the AC power supply. Its purpose is to block
`any DC component of current should there be any,
`emanating from the AC powersupply, and has twofold
`ramifications. The first is that the current, in the AC
`
`1. The biased magnetron is tied to DC potential. The
`current to it is the sum of the DC current driven
`from the DC powersupply and the current that is
`driven from the AC powersupply. The existence of
`a DC bias meansthat the plasma hasto be earthed,
`by contact with the chamberwalls, to allow the DC
`current to flow.
`2. The non-biased magnetron will adjust its internal
`DC voltage so as the AC current flowing around
`the AC part of the arrangement is equal in magni-
`tude for both current directions (i.e. to both mag-
`netrons). This results in a constant AC power ap-
`plied to each cathode, provided the target materials
`are identical.
`
`Ex. 1039, Page 4
`Ex. 1039, Page 4
`
`

`

`52
`
`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`2.2. Plasma emission monitoring (PEM)
`
`Vacuum chamber
`
`Process
`
`
`
`
`
`
`
`Magnetron
`
`power supply
`
`
`Piezoelectric
`.
`valve
`Reactive ~->
`atte
`
`gas
`
`
`Magnetron
`controller
`
`
`
`BNC
`
`BNC
`
`Voltage
`controller
`
`--> Gas pipe
`—>— Electrical signal
`
`This method of control has been described in detail
`in a previous paper[6]. In the course of this work, the
`plasma contained mainly the emission lines corre-
`sponding to argon, oxygen and the target material.
`Therefore, to control the emission line/s of one of
`these elements, the optical filter had to be chosen so as
`either the wavelength/s of light
`it
`transmitted was
`unique to this element, or the emission intensity at the
`selected wavelength/s was sufficiently higher than the
`corresponding ones of the other two elements.
`For example, for the reactive sputter-deposition of In
`oxide, a band passfilter had been used for controlling
`on In emission line at 451.1 nm. Sufficient signals up to
`approximately 150 mV were obtained at applied powers
`of the order of 300 W. At this wavelength, there is
`neither an argon nor an oxygen line, sufficiently in-
`tense, to interfere with the In one. A more universal
`technique was required for the large number of materi-
`als investigated in this work. A high-passfilter, with a
`cut-off wavelength at approximately 620 nm, was suc-
`cessfully used to control the reactive sputter-deposition
`of W, V, Moand Ti oxide. At wavelengths greater than
`approximately 620 nm, the intensities of the emission
`lines of these metals are very weak. Thus, the transmit-
`ted signal by this filter is either due to argon or
`oxygen-lines. The signal fell as oxygen was admitted
`and it is concluded that the strong signal was due to
`argon, and the control was carried out on the argon
`lines, the fall in intensity being attributed to changes in
`the discharge current and coupling to the components
`of the atmosphere as the oxygen was added. Signals up
`to approximately 1.5-2 V were obtained at applied
`powers of the order of 300 W.
`
`2.3. Voltage control
`
`Voltage control has been described in detail in a
`previous paper [6]. This method of control has been
`used in this work in the cases of Al, Zn, Cu and Pb
`when they were sputtered in an Ar/O, atmosphere.
`Fig. 3 shows a schematic of the voltage control loop.
`A user connector,
`located on the rear panel of the
`Advanced Energy™ power supplies which was used in
`this work, provides a 0-5-V DC analogue signal repre-
`senting the cathode voltage (i.e. the output voltage of
`the power supply). The DC signal was 0 V when the
`output voltage of the supply was also 0 V and it was 5
`V at the full-scale output voltage of the supply. This
`0-5-V DC signal was used as an input to a voltage
`controller. The signal was taken through two controls;
`one of which backs it off against another potential to
`provide a zero reference. The difference from this zero
`signal was then amplified by a variable gain amplifier to
`give an output ranged from 0 to 1 V. The outputsignal
`
`Fig. 3. Voltage control (VC) system used in controlling the reactive
`magnetron sputtering processes.
`
`from the voltage controller was then applied to a
`standard pressure controller (process controller) which
`was connected to a piezoelectric control valve of a very
`fast response. The ‘zero reference’ was the signal cor-
`responding to the voltage seen when the target was
`fully poisoned, the ‘1’ was that for metal sputtering.
`Any intermediate degree of target poisoning (i.e. an
`intermediate value of cathode voltage) can be repre-
`sented, in this technique, by a value of input voltage to
`the controller in the range 0-5 V, and a value of
`output voltage in the range 0-1 V. The input to the
`voltage controller was taken from the DC power sup-
`ply, when it was used, as it represented the dominating
`power applied to the main magnetron relative to the
`floating power applied by the AC power supply. This
`arrangement provided better control.
`
`2.4, Substrate condition probe
`
`The information that was required was the ion cur-
`rent density to the substrate and its floating potential.
`In other words, the number and energy of ions that
`bombardedthe growingfilms relative to the number of
`atoms deposited. To obtain this information, we used
`what we termed a ‘substrate condition probe’. Fig. 4
`shows a cross-sectional and a bottom view (ie.
`the
`surface with a direct contact with the plasma) of this
`probe. It essentially consisted of a central cylindrical
`head, whose diameter was 6 mm, surrounded by a
`25 X37 mm guard. The guard, which was entirely
`isolated from the head, was utilised to minimise the
`plasma edge-effect from the probe head. The probe
`was placed in the plane of the substrate following the
`same procedureof placing a substrate; it was mounted
`in the jig, which wasin turn inserted into the platen.
`The I-V characteristics were then obtained by bias-
`ing the probe head. The current
`to the guard was
`excluded. Probe measurements were performed using
`
`Ex. 1039, Page 5
`Ex. 1039, Page 5
`
`

`

`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`53
`
`(a) Cross-sectional view.
`
`[err
`
`(probe head)
`
`Copper nut Copper cylinder
`
`Screw
`Washer
`
`Ceramic insulator
`
`Aluminium body
`(probe guard)
`
`Voltage (V)
`-60
`-90
`-8¢
`-70
`-100
`-110
`-120
`-130
`-140
`-150
`-160
`
`e100, Centre
`4 2000, Centre
`2-300, Centre
`
`|-o- 3001, Erosion (qwa/yu)AyIsuapyuatin
`
`
`
`
`
`
`(b) Bottom views: Surface with
`contact with plasma
`0.5mm—j—
`1$mm
`
`;—
`
`i.t-
`
`Aluminium guard
`
`Copperprobe head
`
` im—
`
`Fig. 4. A cross-sectional and a bottom view of the probe.
`
`In and Sn targets attached to the main and secondary
`magnetrons, respectively. Keeping the working pres-
`sure at 2107? torr,
`two sets of experiments were
`carried out. The first was when the probe was held
`opposite to the centre of the In magnetron using dif-
`ferent DC powers (i.e. at 100, 200 and 300 W). In
`addition, the characteristics of the probe when it was
`facing the erosion zone of the In magnetron, when the
`applied power was 300 W,wasalso plotted for compar-
`ison. The results are shown in Figs. 5 and 6. In the
`second set of experiments, the probe was held opposite
`to the centre of the In magnetron throughout. The
`applied powers were 50 W and 100 W AC floating
`between the two magnetrons, and 300 W DC combined
`with 100 W AC.The results are shown in Fig. 7. The
`following remarks can be deduced from these figures:
`
`Fig. 6. The negative part of the I-V characteristics of the probe, at
`different DC powers to an In magnetron, when it was held opposite
`the centre of the magnetron. The curve when the probe was held
`opposite the erosion zoneis also plotted.
`
`1. The ion-current to the probe increased with the
`applied DC power to the magnetron. On the other
`hand, at a fixed DC power(e.g. 300 W), such a
`current was higher when the probe was held oppo-
`site the centre of the magnetron than when it was
`opposite the erosion zone. Similarly, the ion-cur-
`rent to the probe also increased with the applied
`floating AC power.
`2. At a fixed magnitude of power (e.g. 100 W), the
`ion-current to the probe was lower in the DC case
`than in the floating AC one. Furthermore,
`the
`ion-current to the probe in the case of the 100-W
`AC combined with 300-W DCwasthe highest (Fig.
`7).
`3. The floating potential of the probe was almost
`independent of the applied AC powerandslightly
`dependent on the DC power. However, the order
`of magnitudeof these floating potentials wasslightly
`lower when DC powers were applied.
`
`The above conclusions are in very good agreement
`with the results obtained by Window and Savwvides[12]
`and the results of Glocker[8]. In addition, the ion-cur-
`rent and floating potential, when the probe was oppo-
`site to the erosion zone, are less than the correspond-
`ing values when the probe was opposite to the centre of
`
`*
`65 |
`
`[-o- Floating atSOWACaFloatingat100WAC—a—Floatingat100ACandbiasedwith 300WDC
`Voltage (V)
`55
`»
`-160
`-150 140-130-120 -110
`-100
`90 80-70-60 50-40-30 20, 100s
`
`++ a 0
`0i BO
`
`
`
`
`
`Currentdensity(mA/cm2)
`
` aaneed
`-160
`-140
`-120
`-100
`80
`60
`-40
`20
`0
`20
`40
`60
`Voltage (V)
`
`Fig. 5. The I-V characteristics of the probe when an In magnetron
`was held at 300 W DC and the probe was held opposite to the centre
`of the magnetron.
`
`e
`
`*.
`
`45
`35
`
`
`
`25: of
`
`
`
`
`
`
`
` (qup/y)Ssuapyuasan5
`
`Fig. 7. The negative part of the I-V characteristics of the probe at
`different AC powers to a non-biased and biased floating In mag-
`netron when the probe was held opposite the centre of the mag-
`netron.
`
`Ex. 1039, Page 6
`Ex. 1039, Page 6
`
`

`

`54
`
`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`the magnetron under the same conditions(i.e. applied
`power). This result
`is also in good agreement with
`those of Howsonetal. [13] and Spenceretal. [14].
`It is informative to compute the arrival ratio of Ar
`ions to metal atoms at the substrate. It was found that
`this was 0.7 and the energy delivered to the substrate
`by ions per In atom was approximately 21 eV [1-6].
`On the other hand, in the case when the applied
`power to the magnetron was 300 W DC combinedwith
`100 W AC, J§=7.1 mA/cm’, V,;= —17 V and V, =0
`V (Fig. 7), where J?, V; and V, are the ion current
`density to the substrate, the floating potential of the
`substrate and the plasma potential, respectively. Con-
`sidering the case of indium oxide, which had a thick-
`ness of 152 nm, it was found that N?/N,, = 5.7, where
`N§ and N§ are the numberof ions bombarding 1 cm?
`of the substrate per second and the number of de-
`posited metal atoms on 1 cm? of the substrate per
`second, respectively. Consequently, the energy deliv-
`ered to the substrate by ions per In atom was approxi-
`mately 100 eV.
`Although the two magnetrons were unbalanced in
`the system used in this work, the measured floating
`potentials of the substrate were relatively low, whereas,
`the measured ion current densities were moderate.
`This could be due to the fact that target-to-substrate
`distance is less than the null-point of the magnetrons,
`which means that, at such low distance,
`ions cannot
`acquire high kinetic energies when they impinge on the
`substrate, with the lower floating potential that the
`substrates have. The small target-to-substrate distance
`also affects, but less severely, the ion current density,
`as the substrate can not collect all ions available be-
`cause it is not in the way of the focused beam leaking
`from the cathode, rather it is in the base of that beam.
`The average ion densities in the AC plasma are approx-
`imately four times that of the DC plasma. This last
`difference, between the AC and the DC plasmasis
`significant. According to optical emission measure-
`ments, the plasma extinguishes on each half-cycle and
`has to be reignited. The increase in ion densities in the
`ACplasmawasattributed to target voltage spikes dur-
`ing the reignition of the plasma on each half-cycle, asit
`is evident on the negative-going part of each cycle.
`Such spikes cause rapid electron acceleration in the
`pre-sheath region leading to significantly more efficient
`ionisation of gas and hence much higher plasma bom-
`bardment.
`
`3. Experimentaldetails
`
`3.1. The sputtering system
`
`The chamber comprised a 42-cm diameter stainless
`steel chamber, 12-cm deep internally, giving a short
`
`pump down time with the turbomolecular pump, backed
`by a two-stage rotary pump, compared with conventio-
`nal bell-jar systems [3]. The chamber base accommo-
`dated two identical magnetrons. The magnetron, which
`had both the oxygen inlet to the chamber and the
`optical fibre input tip of PEM control loop attached to
`its pod will be, henceforth, called the ‘main magnetron’.
`The other magnetron was connected to the argon inlet
`to the chamber.
`An axially mounted aluminium platen was located
`above the magnetron cathode surfaces, and it was onto
`this platen, which was electrically isolated, the sub-
`strates were loaded from the airlock allowing a target-
`to-substrate distance of approximately 40 mm. The
`centrally oriented metal shaft was attached to the platen
`so that it could be rotated around this axis with a DC
`motor at a rotation speed of up to 60 rev. /min.
`The partial pressure of the sputtering gas, argon, was
`produced through a mass flow controller balanced by
`the vacuum pumping and measured with the system
`pirani.
`For the admission of reactive gas a solenoid valve
`was replaced by a piezoelectric valve, having a faster
`response in order to cope with much faster changes in
`the desired supply of reactive gas required to maintain
`a certain cathode status, compared with that of inert
`gas. This was controlled to produce a pre-determined
`optical emission signal or cathode potential in much
`the same wayas is used to control pressure. In addi-
`tion, the total distance between the reactive gas pipe
`exit
`in the chamber and the piezoelectric valve was
`minimised to help reducing the time constant of the
`pipe. These modifications, allied with the pipe outlet
`being very close to the target in the very confined
`volume provided by the gettering enabled very efficient
`control of the reactive deposition processes to be ob-
`tained.
`
`3.2. The airlock system
`
`the magnetrons
`The system was designed so that
`could be operated continuously during the changing of
`substrates. We have found this to be of prime impor-
`tance for iterative reactive processing,
`in which the
`partial pressure of the reactive gas is varied gradually
`until the desired film properties are attained. In order
`that this could be done the system was airlocked, that
`is, the main deposition chamber always remained in
`operation whilst
`the samples could be loaded/un-
`loaded via a separately pumped airlock. The airlock
`was 10 cm in diameter and 4.6 cm deep, had a 0.361 |
`volume, and could typically be evacuated from atmo-
`sphere to approximately 40 mtorr in approximately 2
`min, via two-stage rotary pump.
`Samples were mountedsingly in a jig, which was then
`attached to the end of a loading arm, which was moved
`
`Ex. 1039, Page 7
`Ex. 1039, Page 7
`
`

`

`linearly through double Wilson-type vacuum seals,
`mounted off axis (2.2 cm from the centre) in the
`perspex window plate, which allow visual location of
`the substrate.
`
`55
`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`
`
`--+~ Bottom surface (treated) —e— Matt blackpaper —m Back surface |+ Ordinary glass substrate ~o~ Top surface (untreated)
`
`_,
`12+
`ne
`10 —
`9=
`
`<ertootoesasssetsonscetesetttesggnestttttbonsss**terseeroeeestessesetecee®®
`
`E +
`(%)KomDPwRDw Ht
`
`Reflectance
`
`weSsS
`
`wPy3
`
`400
`
`450
`
`500
`
`+
`550
`Wavelength (nm)
`
`+ 4
`
`k
`
`600
`
`650
`
`700
`
`750
`
`800
`
`Fig. 8. Reference spectra of uncoated glass substrates in different
`cases. The reflectance spectrum of matt black paper is included for
`comparison.
`
`coating. In spectrophotometric measurements carried
`out in this work, both transmittance and reflectance
`spectra were measured in the spectral range 350-750
`nm with a scanning speed of 400 nm/min.
`In order to obtain values for the absolute reflection
`coefficients that were required,
`it was necessary to
`make corrections of the measured reflectance spectra
`of the deposited films. The corrections involved:
`
`4, Optical measurements
`
`4.1. Calculations of refractive indices from reflectance
`spectra
`
`In this section, the case of transparent non-absorbing
`films deposited on transparent substrates will be con-
`sidered. It should be mentioned first that the wave-
`length » of the incident light is chosen so that it is
`comparable to the film thickness d,,,, to allow interfer-
`ence effects to occur [15].
`The maximum and minimum reflectance of a thin
`film on an infinitely thick substrate are given by:
`
`Rmax _
`
`2
`
`2
`Ney —n
`subs
`amb‘*
`£ im
`Nem + Aamb”™ subs
`
`n
`
`(1)
`
`2
`
`where Roa» Rminr “alms Msubs 200 Nyy, are a re-
`flectance maximum, reflectance minimum, the refrac-
`tive index of the thin film, the refractive index of the
`substrate and the refractive index of the ambient
`medium, respectively. By solving Eq. (1) for ng,,, we
`get:
`
`n —n
`1. Measuring the reflectance spectrum of the bottom
`(2)
`subs + amb
`Rain =
`AeupsT“amb
`surface of an uncoated glass substrate in order to
`subtract it from the measured valuesof reflectance.
`This was achieved by treating the bottom surface of
`an uncoated glass substrate with emery paper so
`virtually eliminating reflection from it (Fig. 8). The
`reflectance spectra of the untreated (or top) sur-
`face and that of an ordinary glass substrate were
`then measured. The reflectance spectrum of the
`back surface was obtained by subtracting the re-
`flectance values of the top surface from those of
`1+ YRinax
`the ordinary substrate, and was averaged to be 5%.
`Agim =|Mamb™subs
`1-— yR max
`2. Measuring the reflectance spectrum of a single
`crystal silicon wafer and comparing it with a calcu-
`lated one [16] to derive a correction curve of the
`measured reflectance in order to obtain the abso-
`lute reflectance of the coatings (Fig. 9). The correc-
`tion ratio was, on average, 0.76.
`
`1/2
`
`The relative precision of n,, is given by
`
`
`
`Angi =_
`Nem
`
`
`
`ARwax
`VRmax
`2(1 — Rinax) Roax
`
`(3)
`
`(4)
`
`For example, if R,,,, = 36%, it is then sufficient to
`measure R,,,, to an accuracy of approximately 2% (i.e.
`ARjax = 9-7%) so that the relative error for 75), is not
`larger than 1%.
`In this work, films were deposited on glass substrates
`with 1,,,,= 1.525 and the spectrophotometric mea-
`surements were carried out in air (i.e. n,,.4 = D.
`Transmittance and reflectance spectra of transparent
`films, produced in this work, were measured using a
`Hitachi U-2000 double-beam spectrophotometer with a
`simple reflection attachment, which allowed compar-
`ison of the sample with freshly prepared aluminium
`
`Thus, the relation between the measuredreflectance,
`Rms" and the actual one, R, is
`
`R= 0.76(R™*"" — 5%)
`
`or
`
`R=0.76R™*" — 4%
`
`(S)
`
`(6)
`
`if Ruse’ is the measured value of a
`As a result,
`reflectance maximum, the corresponding actual value,
`max?
`R
`is given by
`
`Ex. 1039, Page 8
`Ex. 1039, Page 8
`
`

`

`I. Safi / Surface and Coatings Technology 135 (2000) 48-59
`
`—o- Experimental —e— Calculated —a— Correction ratio 0.770
`
`=aryhi
`
`
`
`oneswopaais0;)
`
`0.760
`
`
`+
`+
`+
`+
`400
`500
`550
`600
`300
`350
`450
`650
`700
`750
`800
`Wavelength (nm)
`
`0.755
`
`56
`
`
`
`Reflectance(%)
`
`30
`
`Fig. 9. Calculated and experimental reflectance spectra of a single
`crystal silicon wafer. Also shown is the correction ratio as a function
`of wavelength. Data of the calculated curve are from [16].
`
`Roman * 0.76R™2"* — 4%
`
`Thus, by substituting the value of R,,,,:
`
`Nem =
`
`ait 0.76RB=* — 0.04
`1 — 0.76R™2" — 0.04
`
`1/2
`
`(7)
`
`(8)
`
`Instead of ellipsometry, Eq. (8) was used to calculate
`refractive indices of films produced in the course ofthis
`work, for the following reasons:
`
`1.
`
`It is independent of optical thickness of coa