Electrode-selective deposition/etching processes using an SiF4/H2/Ar plasma chemistry excited by sawtooth tailored voltage waveforms

We report on the electrode-selective deposition and etching of hydrogenated silicon thin films using a plasma enhanced chemical vapour deposition process excited by sawtooth-shaped tailored voltage waveforms (TVWs). The slope asymmetry of such waveforms leads to a different rate of sheath expansion and contraction at each electrode, and therefore different electron power absorption near each electrode. This effect was employed with an SiF4/H2/Ar plasma chemistry, as the surface processes that result from this gas mixture depend strongly on the local balance between multiple precursors. For a specific gas flow ratio, a deposition rate of 0.82 Å s−1 on one electrode and an etching rate of 1.2 Å s−1 on the other were achieved. Moreover, this deposition/etching balance is controlled by the H2 flow rate, which limits the deposition rate at low flows. When the H2 injection is sufficiently high, the processes are then limited by the dissociation of SiF4, and the relative rate of the surface processes on the two electrodes are reversed, i.e. a higher net deposition rate is observed on the electrode where the fast sheath contraction occurs due to the electronegative character of the plasma.

More recently, a particular group of temporally asymmetric TVWs, resembling a sawtooth waveform, have also attracted considerable research interest. Such waveforms present no amplitude asymmetry in the applied voltage, but are distinguished by having differing rise and fall rates during each fundamental RF period, leading to different sheath motions in front of the powered and grounded electrodes. The effect of this slope asymmetry has been thoroughly studied [11,12] and depending on the gas properties, different electron heating modes can lead to drastically different discharge dynamics. For an electropositive discharge operating in the α-heating mode, the electron heating can be directly linked to the velocity of sheath expansion [13]. In this case, when employing sawtooth-type waveforms, intense local power absorption would occur near the electrode where the sheath expands rapidly, but weak power absorption would be observed near the other electrode where the sheath expansion is slow. For electronegative discharges, electron heating can operate in the drift-ambipolar mode due to the presence of a strong electric field in the plasma bulk. Significant electron acceleration from the plasma bulk towards one electrode occurs during the fast sheath contraction [14], and therefore a reversed discharge asymmetry can be observed for sawtooth waveforms compared to electropositive discharges. Although the slope asymmetry effect caused by sawtooth TVWs has been shown to be useful to understand fundamental aspects of single-gas discharges, the practical applications of such discharges have not been clear. In this letter, we report such an application, and combine the use of sawtooth TVWs and an SiF 4 /H 2 /Ar plasma chemistry to show the possibility to differentially deposit and etch silicon thin films on opposing electrodes of a single reactor.
The experiments were carried out in a laboratory scale PECVD reactor, which consists of a grounded cylindrical confinement chamber surrounding two electrodes of 100 mm diameter, for which the inter-electrode distance d i is adjustable. Both electrodes are heated by resistive coils controlled by feedback loops, with temperatures monitored by embedded thermocouples. The TVW generation system (composed of a feedbackcontrolled arbitrary function-generator and amplifier) has been described in detail previously in [10]. The voltage applied to the powered electrode can be expressed as where A is the voltage amplitude prefactor, set to obtain the desired peak-to-peak voltage (V PP ) on the powered electrode, n is the number of harmonics (here set to four), and ω is the angular frequency, corresponding to the fundamental frequency of 13.56 MHz. No multi-frequency matching network was used in this study; however, practical TVW systems employing multiple sources and matching units have been demonstrated by Franek et al [15]. As presented in figure 1, sawtooth waveforms can be generated with ϕ equaling to either 0.5π or 1.5π; the former gives a sawtooth-up waveform comprising a slow rise and a fast fall, while the latter gives a sawtooth-down waveform with a reversed rate of rise and fall. It is worth noting that in a geometrically symmetric reactor, switching the waveform from sawtooth-up to its counterpart simply results in reversing the role of each electrode.
Process conditions were chosen similar to those in the work of Dornstetter et al [16] which thoroughly investigated the deposition of silicon thin films from an SiF 4 /H 2 /Ar gas mixture using a standard 13.56 MHz radio frequency (RF) excitation source. In this study samples have been deposited on the grounded electrode with an SiF 4 /Ar flux ratio of 3.6/88 sccm and an H 2 flow rate varied between 0 and 6 sccm using sawtooth TVWs, A working pressure of 3 Torr was used in all the experiments. The V PP and d i were kept constant at 285 V and 30 mm. The temperatures of the powered and grounded electrode were set to 80 °C and 150 °C, respectively. Ex situ spectroscopic ellipsometry (UVISEL-Horiba Jobin Yvon) measurements in the range from 1.5-4.5 eV were performed to estimate the thicknesses of the deposited films. To determine the consumption of feed gases during processing, a Residual Gas Analyzer (RGA, Microvision 2-MKS ins.) was deployed in the downstream of the discharge.
As shown in figure 2(a), for both types of sawtooth waveforms, the deposition rate r d increases with H 2 flow rate up to a value of 3 sccm, and then slightly decreases for greater H 2 flows. One can note that the sawtooth-down waveform leads to a higher r d than its counterpart at low H 2 flow rates, while the reverse situation is observed for H 2 flow rates above 3 sccm. Most interestingly, there is almost no deposition process occurring for H 2 flow rates below 1.5 sccm in the case of sawtooth-up waveform. Since the deposition of this material results from a competition between deposition and etching processes, a second series of two-step depositions was performed to allow us to see if etching is occurring. For these experiments, before the plasma processing by the SiF 4 /H 2 /Ar gas mixture, an underlying hydrogenated amorphous silicon (a-Si:H) film of 100 nm thickness was deposited using a standard 13.56 MHz RF source, an SiH 4 / H 2 mixture with a flow ratio of 10/20 sccm and a pressure of 400 mTorr. This sublayer allows us to observe that indeed, an etching of the underlying a-Si:H layer is occurring at low H 2 flow rates, indicated by the shaded region in figure 2(b). For the sake of simplicity, we classify all these results by a net deposition rate r d,net . Similar to the results of the one-step depositions of figure 2(a), r d,net increases with the H 2 flow rate initially, followed by a slight decrease at higher H 2 injection rates. One can see that when there is no H 2 injection, etching processes are obtained for both types of waveform. However, once a small amount of H 2 is added to the mixture, completely different surface processes for the sawtooth-up and sawtooth-down waveforms are observed. In the case of only 1 sccm H 2 , the sawtooth-up waveform leads to a strong etching effect, while almost no thickness variation of the underlying a-Si:H layer is observed for the sawtooth-down waveform. When the H 2 flow is increased to 1.5 sccm, an etching process (r d,net = −1.2 Å s −1 ) is still observed for the sawtooth-up waveform, whereas a deposition process (r d,net = 0.82 Å s −1 ) is observed for its counterpart. By further increasing the H 2 flow rate up to 2 sccm, the etching effect is essentially suppressed for the sawtooth-up waveform, while for the sawtooth-down waveform, a further increase in r d,net up to 1.2 Å s −1 is seen.
In brief, by applying the two types of sawtooth TVWs, conspicuously different surface processes are observed during the SiF 4 /H 2 /Ar plasma processing, despite the fact that all other processing conditions are kept constant. It should be noted that a difference in the maximum energy of ions impinging on the growing surface is expected (up to ~40 eV) when switching the driving voltage waveform. However, previous studies on the PECVD of silicon thin films [17][18][19], have shown that while such variations in ion energy can modify film quality, they have very little effect on the deposition/etching rate, as was observed in this study.
Recalling that in a geometrically symmetric reactor, switching between these two waveforms is equivalent to reversing the roles of the two electrodes, figure 2 therefore indicates that differing deposition or etching processes could be achieved on each electrode of a CCP system, as determined by the relative H 2 flow rate. In short, one can realize either a deposition or etching process on one electrode (or substrate) without influencing the other, or even a deposition process on one electrode but an etching process on the other.   One additional note should be made concerning the selective process observed above. As the reactor in this study is geometrically asymmetric, the maximum ion energy observed on the substrate for both types of waveform is lower than would be observed a symmetric reactor. However, as ion energy alone is not a determining factor for deposition rate [17][18][19], this would at most result in a shifted process window for selective processing when transferring this process from a small, asymmetric reactor to a large, symmetric one.
To further understand the changes in the plasma that enable this selective processing, we have measured the effluent gas composition by RGA and the DC self-bias voltage during processing. The use of RGA allows us to quantify the consumption of SiF 4 and H 2 molecules in the plasma, and the results as a function of H 2 flow rate are presented in figure 3. When no H 2 is injected, SiF 4 is barely consumed at all, relative to its actual flow (3.6 sccm). For low values of H 2 flow, H 2 is essentially fully consumed, and SiF 4 is consumed at about half the rate of H 2 . For H 2 flow values above about 3 sccm, the consumption of H 2 remains constant, as does that of SiF 4 . A second complementary measurement is that of the η during processing. Figure 4 shows the experimentally obtained η plotted as a function of H 2 flow rate. Due to the geometrical asymmetry of the reactor, all the measured values of η are negative. As previously shown for the case of single-frequency discharges [16], for both types of sawtooth waveform, η decreases in absolute value with increasing H 2 flow up to 3 sccm (presumably due to an increase in the electronegativity of the plasma), and stays roughly constant at higher flow rate. However, the absolute values of η for each waveform remain quite different, and in fact, reverse in relative magnitude at around 1 sccm.
These results allow us to make some assertions about what is occurring in the plasma with increasing H 2 flow. From figure 3, we see that at low H 2 flow rates, the consumption of SiF 4 increases with H 2 flow. This is consistent with the phenomenological model described in [16] to understand the otherwise complex SiF 4 /H 2 /Ar discharge. In that model, the role of H 2 is to remove the fluorine produced by the SiF 4 dissociation through the formation of HF molecules. Without the scavenging of fluorine through this process, fluorine atoms etch any silicon surface. Therefore, the presence of H 2 is the limiting factor in the consumption of SiF 4 .
The additional complexity in this experiment is the spatial variation in the fluorine removal process due to local variations in the electron power absorption. These variations (caused by the slope asymmetry of the sawtooth waveform) can be partially observed through the measurements of η. In figure 4, for no H 2 flow, the η value for the sawtooth-down waveform is more negative than that for the sawtooth-up. This indicates that for these conditions, the discharge acts as an electropositive one (like argon) [12]; an electron power absorption maximum occurs near the electrode experiencing a fast sheath expansion, as described by the hard-wall model proposed by Godyak [20]. This leads to a much higher charge density within that sheath. Considering the expression given by Heil et al [7] and the situation of sawtooth TVWs [12], the η can be estimated by: where ε is a symmetry parameter that can be determined by the surface area ratio between the powered and grounded electrodes, as well as the respective mean charge densities in the sheath regions near said electrodes. Therefore, a more negative η is generated by applying the sawtooth-down waveform, consistent with the results in [12].
With increasing H 2 flow rate, a significant increase of η is found for the sawtooth-down waveform, while less so for the sawtooth-up waveform. This trend is actually the superposition of two effects, both due to the plasma taking on a more electronegative character. Firstly, the absolute value of η is decreasing for both waveforms, as previously observed for the same plasma chemistry excited with a single frequency (13.56 MHz) waveform [16]. Secondly, the relative values of η for the sawtooth-up and sawtooth-down waveforms switch places, and the sawtooth-up waveform takes on the more negative value (as observed for CF 4 [21,22]). This is due to the change in where the most intense electron power absorption occurs as the plasma acts more electronegative. Opposite to the case of electropositive Ar, the most intense power absorption for an electronegative gas is near the electrode where a fast sheath contraction occurs. This is due to the electron attachment process leading to a depletion of local electron density and conductivity. Hence, a strong reversed electric field is generated near the edge of the rapidly collapsing sheath, which accelerates the electrons from the plasma bulk towards the grounded electrode. These energetic electrons will be trapped by the potential wall formed in front of the electrode [21,22], which further increases the probability of the electron attachment process, leading to a higher charge density in that sheath.
Together, these plasma results are consistent with the processing results they produce. For the case of a sawtooth-up waveform, a qualitative comparison of the hydrogen and fluorinated species concentration close to each of the two electrodes with increasing H 2 flow rate is presented in figure 5. Under the condition of no H 2 injection, more intense power absorption occurs close to the powered electrode, experiencing the fast sheath expansion. Net etching processes are observed on both electrodes as there are no H 2 molecules to scavenge the active fluorinated species. Once H 2 is injected, the plasma switches to behaving like a more electronegative one ( figure 4), a phenomenon observed but for which the explanation is not yet clear. The intense power absorption occurring close to the grounded electrode (experiencing fast sheath contraction) would result in a higher active fluorinated species concentration in front of the substrate. However, the injected H 2 is still insufficient to scavenge these species, leading to either no deposition (figure 2(a)) or intense chemical etching [23] ( figure 2(b)). The situation is simply reversed for a sawtooth-down waveform; due to a weaker electron power absorption, a lower fluorinated species concentration is present near the grounded electrode, easily scavenged by the available H 2 . The etching effect is much weaker, leading to either deposition (figure 2(a)) or at least the absence of net etching process (figure 2(b)) at low H 2 flows.
With increasing H 2 injection, the etching effect is gradually suppressed, and when the H 2 flow rate is higher than 2 sccm, deposition processes can be observed on both electrodes ( figure  2(b)). The resulting surface process is no longer limited by the presence of H 2 , but by the dissociation of SiF 4 . In this case, the more intense dissociation of SiF 4 close to the grounded electrode will now lead to a higher deposition rate for this electrode (which previously observed a lower net deposition rate, figure 3).
In conclusion, we have experimentally shown that the use of sawtooth TVWs and an SiF 4 /H 2 /Ar gas mixture can result in controllably differing deposition or etching processes on each electrode in a CCP chamber. This is due to two effects: the multi-precursor nature of the deposition/etching process and the localized electron power absorption caused by the slope asymmetry effect of sawtooth TVWs. At low H 2 flow rates, when processing identical a-Si:H layers but with opposite sawtooth TVWs, one can achieve a deposition process on the electrode where the weak electron power absorption occurs during the fast sheath expansion, and an etching process on the other electrode that experiencing the fast sheath contraction. Moreover, this deposition/etching balance on each electrode can be directly controlled by the H 2 flow rate. At higher H 2 flow rates, a deposition process is obtained on both electrodes. However, the discharge now operates in the drift-ambipolar heating mode (at the process conditions used in this work). As limited by the dissociation of SiF 4 , the relative net deposition rate on each electrode is now reversed. In short, this work encourages the prospect that one could choose a set of process conditions to achieve a wide variety of desired depositions on one electrode, while leaving the other electrode pristine.