Discharge Instability at Patterned Conductive Layers on Insulating Substrates during Pulsed-Plasma Chemical Vapor Deposition under Near Atmospheric Pressures

Discharge instability at patterned conductive layers on insulating substrates during pulsed-plasma chemical vapor deposition under near atmospheric pressures was studied by observing the distribution of plasma damage spots in the grown film. Instability of the discharge is found to be enhanced at conductive layers, and the density and the size of the plasma damage spots vary with the shape and the size of the conductive pattern. Based on these observations, we propose a simple model, uniformly diluted charge (UDC) model, to explain the damage distribution within the conductive pattern. [DOI: 10.1380/ejssnt.2013.47]


I. INTRODUCTION
Plasma-enhanced chemical vapor deposition (PECVD) under near atmospheric pressures (NAP) is expected as a novel film deposition technique for large scale electronic devices, such as thin film transistors (TFTs) for flat panel displays.Several advantages over low-pressure PECVD have been reported for the use of NAP, which include the high deposition rate and the low process temperatures [1].While most of the NAP-based plasmas utilize mixing of helium gas to maintain the glow discharge [2], the technique developed by Yuasa et al. utilizes a pulsed biasing [3], which thereby has essentially no limitations in the choice of gases.For TFT fabrication, we have been trying to apply this pulsed NAP plasma based CVD process of thin films such as Si [4] as channel layers and silicon nitride (SiN x ) [5] as gate insulating layers.In the former application, high growth rates and high crystallinity with no incubation layer, both of which have been obtained by use of NAP-PECVD, seems quite suitable for bottom-gated TFTs fabrication.Another advantage in this technique is no need to use toxic ammonia for nitride formation.Moreover, the atmospheric process is suitable for plasma enclosure between a pair of parallel electrodes, and the discharge operation by bipolar pulses allows deposition of a pair of equivalent films on both electrodes [6].This suggests possibility of a cleaning-free roll-to-roll process using the pulsed PECVD under NAP.
One of the challenges for this pulsed PECVD under NAP condition, however, is the instability of the discharge on conductive substrates.Figure 1 shows a comparison between the plasmas on insulative substrate and on conductive substrate.On conductive substrate, the localized strong light emissions are observed.If we want to grow a gate-insulating layer of a bottom-gated TFT using NAP-PECVD, the conductive bottom gate electrode would be a cause to induce plasma instability on it and to affect the device performance.Little is known, however, about the details of this kind of instability.In this study, we have conducted a series of NAP-PECVD experiments of SiN x films on glass substrates, which are patterned with Al electrodes of various sizes and shapes.The plasma instability was monitored through density, size, and distribution of the plasma damage spots formed on the grown SiN x film.These properties were found to be strongly affected by the size and the shape of the conductive patterns.A simple model (uniformly diluted charge (UDC) model) will be presented to account for the features, based on the concept of counter electric field developed in the dielectric barrier discharge (DBD) as well as a possible dilution of the surface charges transferred from the plasma.

II. EXPERIMENTAL
The schematic image of the plasma unit for the pulsedplasma CVD is shown in Fig. 1(c).More details of experimental setup have been described elsewhere [4,5].After cleaning the glass substrate (Corning #1737) by organic solvents, an aluminum (Al) layer with thickness of 70 nm was evaporated onto it.The Al layer was then patterned into rectangles with several sizes of x width (20, 40, 60, 80 and 100 µm) and y length (420, 620, 820 and 1020 µm).Finally, SiN x thin films were formed by using the pulsed-plasma CVD with the condition shown in Table I.Obtained films were evaluated by Nomarski microscope with the dark-field mode and by Atomic Force Microscope (AFM).

III. RESULTS AND DISCUSSIONS
A dark field images of Nomarski microscopy of the SiN x films deposited on the Al pattern is shown in Fig. 2(a).In this paper, we define the direction parallel to the short end to be x width and that parallel to the long edge to be y length, both axes originating at the very center of the rectangles as depicted in Fig. 2(b).It is confirmed that many bright spots of various sizes exist on the patterns.The dark field images of Nomarski microscopy display stepped part bright.We thus understand that these bright spots originate from plasma damages.We tried to identify the origin of these plasma damages by performing the AFM observations, as shown in Figs.2(c) and (d).Although the size and the density are different, it is common that in both regions, namely at the center (2(c))  In general, plasma damages are prone to be concentrated at edges of the patterns.For example, the concentration of small spots at the central portion of the patterns varies with the pattern width; the wider the width, the lower becomes the central density.
Figure 3(b) shows the distribution of the small spots on the x axis area (y ∼ 0).For wider patterns with x width exceeding 60 µm, the spot density at around x = 0 is lower than the edge regions.The decay length from both of the edges is around 40 µm, and the density be- comes constant for smaller x absolute away from these edge regions.In the sample (II) with x width of 40 µm, the spot density near x = 0 is increased from the rest.We understand this as to be caused by additive impacts from both of the edges.For larger spots in wider samples (x width larger than 40 µm), on the other hand, their appearance is limited to the vicinity of y end sections.We understand that this happens because the regions near the short y ends are close to three edges, i.e., one short end and two long edges.A more serious damage occurred because of the superposition of the edge effects from these three directions.
In Fig. 4(a), patterns with a constant x width (60 µm) and different y lengths (420, 620, 820 and 1020 µm) are compared.On shorter patterns such as 420 µm in y length, the small spot density near the y center is lower and the area with large spots is smaller, being limited to the y end sections.As the length of the pattern increases, however, the small spot density near the y center (out of the image in these longer samples) increases.Figure 4(b) shows the distribution of the small spots on the x axis for the samples identical to those shown in Fig. 4(a).In Fig. 4(b), we confirm here again that the spot density increases as the y length of the pattern becomes longer.
The aforementioned relation between the Al patterns and the plasma damage can be summarized as follows; (i) the plasma damage occurs only at Al patterns, (ii) the damages tend to be concentrated in the end sections of the patterns, and (iii) both the density and the size of the damage spots increase with the y length of the patterns.
To understand these features, knowledge on dielectric barrier discharge (DBD) [7] can be instructive.A DBD is one of the original atmospheric plasma techniques, and is known to consist of aggregate of micro streamer dis- charges.A DBD is believed to be stable because once a micro streamer occurs at a certain position, charges (electrons, generally) coming from the discharge stick to the surface of the dielectric barrier, create a counter field, and relocate or turn off the micro streamers.Therefore, micro streamer never stays at the same position on the dielectrics, which avoids serious plasma damages in DBD.In the present experiment, the glass substrate is considered to be a good dielectric barrier, and blocks the plasma instability.One possible mechanism for (ii) is, therefore, the trapping of streamer discharge at the edges that migrated from the glass portion of the substrate.This mechanism, however, cannot account for the damage distribution within the Al pattern.
Another and more convincing mechanism is the dilution of surface charges within the Al pattern.On conductive layers like Al, the charges transferred from the plasma immediately diffuse to other portions within the pattern.Then, the charge density on the pattern decreases, and the counter electric field also decreases.As a result, the streamer discharge continues for a longer duration at a same position, leading to occurrence of severer damages on the film surface.
Based on this consideration, we here propose a simple model to account for the obtained distribution of the plasma damages within a patterned electrode.We start with a case that a single streamer discharge occurs at a certain point within an electrode.We further assume, for simplicity, that no surface charges were present before the discharge within and outside the patterned electrode.This is validated considering that surface charges, if present in sufficient amount, would establish a counter electric field and thereby impede onset of the discharge http://www.sssj.org/ejssnt(J-Stage: http://www.jstage.jst.go.jp/browse/ejssnt/) under consideration.
The absence of the surface charge outside the pattern has a strong impact on the distribution of the counter electric field within the pattern.At the pattern center, the counter electric field is strong because of the integration of the surrounding charges.At the pattern edges, on the other hand, the counter electric filed is weakened because of the absence of surface charges outside the pattern.Be-cause of this weakened counter electric field, discharges in the vicinity of pattern edges can last for longer durations, which eventually results in plasma damages.To model this, we assume that surface charges with density σ exist uniformly within a pattern with size a × b.The normal component E z of the field at a given position within the conductive pattern can then be calculated using the following equation: ( The potential φ at the same position, a measure of the stopping power against discharge, is obtained by integrating E z from the counter plasma electrode to the sample surface in the normal direction as: Figures 5(b) and 5(c) show calculated absolute values of the potential distribution for a 60×620 µm 2 pattern (sample (VII)).We notice that a weak potential region occurs around the periphery of the pattern, which shows a good correspondence with the observed distribution of the small spots shown in Fig. 5(a).This uniformly diluted charge (UDC) model not only explains the enhanced plasma damage at edges, but also the pattern size effect of the damage.The damage density is up to 0.7 µm −2 , and the maximum pattern size is 100×1020 µm 2 .Therefore, the number of damage spots in one pattern is up to 7×10 4 .On the other hand, because frequency of applied pulsed-voltage is 30 kHz and processing time is 5 min, total number of voltage pulses is 9×10 6 .So that, a streamer discharge occurs one time on one pattern during about one hundred voltage pulses, which indicates that multi streamers hardly occur in one pattern at the same time.In this model, the charge dilution is stronger as the area of the conductive pattern is larger.As a result, the inverse electric field becomes weaker for larger patterns.In other words, larger patterns require more charges to stop the discharge.The discharge therefore last for longer duration, and the plasma damage becomes serious.The feature (iii) of the plasma damage, i.e., onset of severer damage on longer patterns, can be explained in terms of this pattern area effect.Figure 6 shows the absolute values of the potential distributions calculated for several patterns, assuming that the surface charge densityσ is inversely proportional to the pattern http://www.sssj.org/ejssnt(J-Stage: http://www.jstage.jst.go.jp/browse/ejssnt/) area.When we focus in the impacts of the pattern length (III, VI, VII, VIII), the potential is weaker as the pattern is elongated, which is consistent with the experimental results showing enhanced damages on longer patterns.When we focus in the impacts of the pattern width (I, II, III, IV, V), however, the consistency is lost.The calculated potential becomes stronger with the narrowing of the pattern, which is contrary to the experimental results showing enhanced damages on narrower patterns.This is a limitation of the present UDC model, and we have to consider another effect to explain the pattern width effect.One possibility is role of heat diffusion.If the Joule heating by the discharge current is the dominant cause for the plasma damage, escape rate of the Joule heat may also have a significant impacts on the occurrence of the damage; shorter the pattern width, the severer the plasma damage because of the concentrated heat.
One implication of the UDC model is that shrinkage of the pattern area may suppress formation of damages.In order to test this hypothesis, a sample with smaller sized patterns (20×80, 20×20, 16×16, and 12×12 µm 2 ) was prepared and damage distributions in them were observed.Figure 7 shows the microscopy image.Large spots can no longer be observed.It should especially be noted that there is no visible damages in the patterns less than 20×20 µm 2 in size.It is thus demonstrated that shrinkage of the conductive region is beneficial in suppressing the plasma damages.This result indicates that the charge dilution effect becomes more operative than other effects as the pattern area shrinks.

IV. CONCLUSIONS
The plasma instability during deposition of silicon nitride films by pulsed-plasma chemical vapor deposition technique under near atmospheric pressures was researched.The plasma instability was evaluated by observing the density and the size of the damage spots caused by the instable plasma.The plasma damages appeared only on Al patterns, especially at their pattern edges.The spot density showed a positive correlation with the pattern length.Based on this result, we proposed a uniformly diluted charge (UDC) model for the distribution of the plasma damage.Based on the model, it is suggested that shrinkage of the conductive region should be effective in suppressing the plasma damages, which is actually confirmed experimentally.We believe that the knowledge obtained in this work will contribute to betterment of the PECVD using near atmospheric pressure plasma technique to fabrications of electron devices.

FIG. 1 :
FIG. 1: The images of plasmas on (a) insulative substrate and (b) conductive substrate, in the latter of which is seen two arc-like beams.(c) is the cross sectional image of the plasma unit, in which (a) and (b) were observed from side.

FIG. 2 :
FIG. 2: (a) A dark field image of the surfaces of the SiNx film deposited on Al pattern.(b) A schematic of microscope image area and the axes definition of x and y.(c) An AFM image of the SiNx surface at the pattern center.(d) An AFM image of the SiNx surface near the shorter edge of the pattern.
Figure 3(a) shows the microscopy images of the SiN x films deposited on the Al patterns of several widths.The length of the patterns is fixed at 1020 µm.
FIG. 3: (a) The dark field images of the surfaces of the SiNx films deposited on several x width Al patterns.The y lengths of the patterns are fixed at 1020 µm, and each x width is (I): 20 µm, (II): 40 µm, (III): 60 µm, (IV): 80 µm and (V): 100 µm.(b) The number densities of the small spots at near the length (y) center of the Al patterns which are same as (a).

FIG. 4 :
FIG. 4: (a) The dark field images of the surfaces of the SiNx films deposited on several y length Al patterns.The widths of the patterns are fixed 60 µm, and each length is (VI): 420 µm, (VII): 620 µm, (VIII): 820 µm and (III): 1020 µm.(b) The number densities of the small spots at near length (y) center of the Al patterns which are identical to the samples shown in (a).

FIG. 5 :
FIG. 5: Comparison between the experimental results and simulation on 60×620 cm 2 pattern (sample (VII); (a) microscopy image, (b) calculated absolute value of potential distribution with 2-D display and (c) calculated absolute value of potential distribution with 3-D display.Here, the charge density σ is set to 1×10 −4 C/m 2 .
FIG.6: Calculated absolute values of potential distributions on several size patterns.Here, the total charge within one pattern is set to 1×10 −11 C and respective charge density σ was calculated by dividing with the pattern area.

TABLE I :
The conditions of SiNx films deposition.