Substrate temperature effect on migration behavior of fluorocarbon film precursors in high-aspect ratio structures

Processing of extremely high aspect ratio features is required for advanced technology systems, e.g., vertically stacked memory systems. Indeed, several publications have investigated the recent challenges facing this technology.^1,2 High aspect ratio features are typically created by dry oxide etching using fluorocarbon (FC)-based plasma processing.³ While the effect of FC contribution to etching is well understood for flat surfaces exposed directly to a plasma environment,^4–7 it is much less clear how these effects translate to high aspect ratio features. There are several key processing issues related to high aspect ratio feature creation including bowing, selectivity of stacked layers, and critical dimension variations over the depth of the structure. Etching selectivity of SiO₂ relative to other substrates is critical to these high aspect ratio processes since they involve stacks of multiple materials and are determined by the selective deposition of FC film precursors.⁸ Various CF_x radical species produce a polymerlike passivation layer on top of layers underneath SiO₂ materials which then controls how ion and other precursor species interact with the surface and thus controls the overall etch rate of the system.⁹ This FC film is based on both ion and neutral species from the plasma arriving at the surface for normal etching processes; in high aspect ratio features, the role of ions is greatly reduced as they do not penetrate very far into the gap structure.^10,11 A maximum attainable aspect ratio based solely on reactive ion etching has been described and near this maximum aspect ratio depth there are nearly no ions interacting with the surface beyond this point.⁹ Specific etching mechanisms also depend directly on aspect ratio due to effects from ion transport, neutral species transport, and surface charging which has been described in great detail in the literature.¹² The transport though high aspect ratio features leads to a strong reduction of the arrival rate of FC film precursors, e.g., CF₂, at the bottom of the features.^13–15 This reduces the etching selectivity of SiO₂ relative to other materials. This effect has been studied with respect to the investigation of bowing effects within Si and SiO₂ trench structures by Izawa et al. Specifically, they measure the sticking coefficient of fluorocarbon radicals and then calculate the reaction probability for F radicals, C rich radicals, and CF_x radicals.¹⁶ Ikegami et al. use secondary ion mass spectroscopy to measure the energy of species arriving the base and side walls of trench structures and show that the directionality is highly dependent on gas chemistry.¹⁷ Fukasawa et al. have investigated microloading effect to achieve SiO₂ selective etching in high aspect ratio features from 1.4 to 0.4 μm and measures the etch rate at the bottom of these trenches.¹⁸ Another area in particular where the etching selectivity impact has been studied is called reactive ion etching lag effect where the etch rate within a microstructure is reduced compared to surfaces which are fully exposed to reactive ion etching plasma to the point where a larger microstructure will etch faster than a smaller feature.^19,20 Modeling of these phenomena has shown that the deposition of FC film is one of the most critical parameters leading to this reactive ion etch lag.²¹ Further work has modeled how reactive ion etch lag occurs in trench structures of various size trench structures. All of these features of structure production in Si and SiO₂ materials are dependent on understanding the transport of reactive species under various processing conditions and structures.²² 

Heating of an SiO₂ structure has the potential to change the precursor surface residence time, migration behavior and therefore impact the depth-dependent or aspect ratio dependent deposition onto trench structure side walls. Previous work has discussed that increased temperature of the substrate should increase the etch rate while simultaneously decreasing the FC passivation layer thickness on the sidewall of high aspect ratio trenches as these effects are related to the Bosch process.²³ Another important consideration for this work is that the conductance of species into a trench structure is based on the surface temperature along with other factors, and therefore higher substrate temperature should be able to increase the conductance of species into a gap structure.²⁴ This work will focus on understanding the effect of substrate temperature on the behavior of the FC-based film precursors migrating into a high aspect ratio feature. This effect in high aspect ratio features has not been studied and could hold potentially beneficial effects for changing how FC behaves in these features. The study of the migration behavior of FC precursors in high aspect ratio features has been studied previously at room temperature and described by Zheng et al.²⁵ One experimental setup designed to effectively study high aspect ratio migration in a laboratory experimental system has been described previously in several publications.^25,26 This technique uses an artificial gap structure turned in a horizontal manner and is particularly well suited to investigate how neutral species contribute to FC deposition with and without ion bombardment and in the transition region between these two conditions. This experimental technique can then be translated to describing how etching will occur in a three-dimensional trench structure where the top surface nearest to the plasma is the exposed region and the region deep within the trench is the shielded area.

The experimental approach overview is schematically shown in both parts of Fig. 1. The experimental reactor used for this work is an inductively coupled high-density plasma (ICP) reactor which is schematically described in Fig. 1(a) and also has been described in detail in previous publications.^5,27,28 The reactor uses a planar coil placed on top of a quartz window and is powered through an L-type matching network at 13.56 MHz. The plasma generated is confined to a 2–3 cm region below the quartz window. The plasma conditions used were a 10%/90% C₄F₈/Ar gas mixture, 40 sccm total flow, 20 mTorr, and 400 W source power. The base pressure of the experimental chamber was pumped to 5 × 10⁻⁶ Torr or lower to ensure a low level of residual impurities in the chamber. A chiller unit cools the substrate that the samples are mounted on and a heater unit with temperature control is placed directly on top of this chilled substrate to keep experiments at a consistent temperature during plasma exposure. The base temperature for the chiller was set to ∼10 °C and the heater was set to 10, 40, and 75 °C (corresponding to substrate temperatures of 283, 313, and 348 K) based on the experiment. The lower substrate was used for measurements, which is in direct contact with the heater as the top substrate is used to help shield the trench structure from being heated by the plasma, and therefore the heater temperature should be very close to the lower substrate temperature.

Ellipsometry measurements are the primary method of characterizing the FC film thickness and refractive index. The ellipsometer utilizes an He-Ne laser with a wavelength of 632.8 nm in a polarizer-rotating compensator-sample-analyzer configuration. Evaluation of thickness and refractive index was extracted using an optical model based on the incident angle of the ellipsometer and the known optical properties of the film materials involved.

Fourier-transform infrared spectroscopy in the attenuated total reflection mode was used to check the composition of the exposed deposited FC layer as the thickness of the film inside the gap structure decreases below detectable limits very quickly. The results were consistent with previous studies on FC deposition on exposed silicon dioxide substrates with ion bombardment.^7,25

By making gap structures from SiO₂ surfaces, it is possible to study the migration behavior of CF_x species on SiO₂ sidewalls. This configuration simulates the structure of an SiO₂ trench. The experimental geometry used in the present work is displayed in Fig. 1(b). A bottom sample is 51 mm in length whereas the top sample is 45 mm in length. These samples are separated by 0.7 mm thick spacers, simulating a trench with up to approximately a 35:1 aspect ratio. Spacers 1.4 mm thick are also used to simulate a wider trench for comparison purposes. The experimental setup allows for real time monitoring of FC deposition on the bottom sample in the area directly exposed to the plasma using ellipsometry measurements. Ellipsometry measurements of the entire sample are conducted post-treatment using a single directional line scan setup at the center of the sample. All depositions were performed for SiO₂ films on Si.

The deposition of FC films was monitored using in situ ellipsometry so that there was approximately a 100 nm FC film deposited on the exposed region of the SiO₂ samples. This procedure was done at substrate temperatures of 283, 313, and 348 K and two different gap heights of 0.7 and 1.4 mm. Each experiment was run in duplicate to ensure reproducibility. Each sample was then scanned posttreatment to measure the progression of the thickness from the exposed region to the completely covered region of the sample. The full measurement of each sample resembles approximately an S-curve which has a plateaued steady-state region in the exposed area transitioning to an approximately exponential decay behavior just inside the edge of the gap structure. The inflection point of this S curve occurs approximately 0.1 mm outside of the gap structure beginning (where the top substrate starts), and for the rest of this work, the inflection point is taken as d = 0 mm. Examples of the experimental results are shown in Fig. 2. Figures 2(a) and 2(b) show the actual deposition amounts at the inflection point of the deposition curve for each gap height. As we can see from this, the deposition outside the gap structure decreases with increasing substrate temperature. The maximum FC thickness seen for the coolest temperature was approximately 110 nm thickness, and the FC total thickness for the hottest temperature was reduced to approximately 75 nm thickness. This phenomenon is well known for FC deposition behavior. However, the data are most clear when we normalize all the data to this inflection point for each data set and this is shown in Figs. 2(c) and 2(d). Specifically, Fig. 2(c) shows the deposition for a gap height of 0.7 mm and Fig. 2(d) shows the deposition for a gap height of 1.4 mm. The data shown here are the average of two repeated experiments for each experimental condition and the standard error calculated between these two experiments.

In general, we find that the behavior of the deposition trend follows approximately an exponential decay function from the beginning of the shadowed region heading further into the gap structure. There is, however, some residual deposition deep into the gap structure that shows an offset from the exponential decay behavior. In addition, the exponential curve itself varies slightly from a perfect exponential model as investigated later in this work. The results from these experiments show a clear dependence on the temperature of the substrate for the depth that the FC deposition travels into the gap structure. Increasing the gap structure height also leads to an increase in the distance that the FC deposition moves into the gap structure.

In order to understand the dependence of the deposition rate of FC with respect to substrate temperature, we can assume that based on the simplest case, the substrate temperature influences the flux of reactive species and the sticking coefficient at a given position within the gap structure. Therefore, we can write the dependence of the deposition rate dependent on position and substrate temperature as follows:

Here, the deposition rate term at a given substrate temperature and position is represented by DR_FC (d, T_sub). The term d describes the depth within the gap structure as seen in Fig. 1, T_sub is the substrate temperature, and n is the number of molecules. It should be noted here that the effective temperature of the species in the gas phase T_g within the structure is likely directly related to the substrate temperature T_sub as these species are regularly in contact with the surface as they move into the gap structure through the process of thermal accommodation. Additionally, the sticking coefficient term is given by s(T_sub), which is a function of substrate temperature, and Γ(d, T_g), which is the flux of reactive species based on position and gas temperature. There are several factors which influence whether or not a species sticks to a surface. There include reflection, desorption, and reaction which determine sticking and residence time of a species on a surface. Of these factors, desorption is the primary factor which is influenced by temperature, where species are more likely to desorb from a surface at elevated temperature. The flux at a given position within the gap structure is further determined by the number of surviving film precursor species which is given as n(d, T_sub) in this model and is a function of distance into the gap structure and substrate temperature.

Additionally, any precursors which encounter elevated temperature surfaces may be heated by thermal accommodation. Due to the clear effect of deposition based on the substrate temperature, the loss of precursor species versus distance into the gap structure may be described by a characteristic substrate temperature dependent decay distance represented by λ(T_sub). This term can be described based on the following relationship:

Therefore, based on this equation, we can express the density of species as an exponential decay function as given in the following equation:

In this equation, n₀(T_sub) is the density of precursor species at the beginning of the gap structure (d = 0). Since we assume that the deposition rate is directly proportional to the density of precursor species, we can further state that the deposition rate will also follow the same exponential decay trend starting at the entrance of the gap structure. Additionally, from this model, we expect that if the deposition follows exponential decay, then there are only precursor species with a single characteristic decay distance λ(T_sub). If we see a deviation from this exponential behavior, then it is possible that there are multiple species responsible for the deposition and these should be characterized by a different decay distance λ'(T_sub). Since the model proposed follows an exponential decay behavior, it is useful to think about the behavior of the FC deposition in terms of other well-known processes which follow a similar behavior, such as the radioactive decay of particles. Generally, a radioactive decay function is expressed in terms of time instead of distance and is shown in the following equation:

In this equation, N is the number of radioactive nuclei and the τ term is the radioactive decay time constant. From this equation, one useful quantity that is derived often is called activity and this describes the amount of radioactive decay events which occur based on a given system and is expressed as N(t)/τ. Therefore, there is more activity if the starting density is higher or if the radioactive decay constant term is shorter. In the case of our system, we may think of the deposition activity expressed shown in Eq. (5) and represent the ability to deposit FC at a certain depth within the gap structure.

We see from this equation that the activity is inversely proportional to our decay distance. This means that if we want a high activity deep into a structure such that there are still FC species available to be deposited further into the gap structure, we desire the decrease of the species density with distance to be small. For low substrate temperature, the initial loss rate of species is high (high activity) and gives rise to the cross-over seen near d = 0 in Figs. 2(a) and 2(b). The loss in FC activity with depth is reduced by having a higher decay distance which follows the modeling of the data. A graphical representation of this model is shown in Fig. 3. As the neutral species move into the gap structure, the activity, and therefore the deposition, decreases based on the temperature of the substrate.

Figure 4 clearly shows that the true dependence of the substrate temperature on this movement of the FC deposition into a gap structure is based on the aspect ratio, combining the data shown in Figs. 2(c) and 2(d). When this is normalized to the aspect ratio, which takes into account the gap distance, the amount of deposition per aspect ratio is very consistent between the two gap sizes. Increasing the temperature increases, the distance traveled into the gap structure and appears to be more dependent on aspect ratio of the gap structure as opposed to the actual gap size.

Modeling of the exponential fits was done using originlab origin graphing software. The data that are modeled are the plot of the averaged depositions versus aspect ratio. These fits are in Figs. 5(a) and 5(b) for the 0.7 mm gap and 1.4 mm gap width, respectively. The actual fit uses a slight variation on the model shown in Eq. (3) with the addition of a deposition offset term to help the model fit better. The physical representation of this offset parameter is a very thin deposition layer that extends further into the gap structure than an exponential decay fit alone can account for.

The model fits shown in Fig. 5 have very good fits initially at the beginning of the gap structure, and then there tends to be a deviation toward a faster decay than exponential once beyond approximately three aspect ratios into the gap structure. However, eventually the deposition appears to slowly reach a baseline and the data move to the slower decay side of the exponential decay function. This baseline appears to extend very far into the gap structure and represents a very small amount of deposition on the order of less than 2 nm for all cases. In order to better investigate the deviation of the deposition behavior, the model was fitted based on an increasing number of aspect ratios into the gap as a progression and is shown in Fig. 6.

Figures 6(a) and 6(b) show the exponential distance constant from the data model from zero to the plotted aspect ratio on the x-axis. For instance, at an aspect ratio of 8 on the x-axis, the model parameter has been extracted from modeling the data for only 0–8 aspect ratios. Figures 6(c) and 6(d) show the R² value of the modeling fit for the parameter extracted above. These values show that for the smaller gap structure, the data fit is generally fairly constant over the range of the fit. However, for the larger gap structure data, we see a significant drop in how well the model fits the data as the FC deposition drops off more quickly than exponential decay before leveling off toward the background value of deposition deep in the gap structure. After this drop off, the fit once again crosses the data as it moves toward the deposition that is shown deep in the gap structure. This is likely due to both a change in the dominant species which causes the deposition and the presence of a change over this aspect ratio. The initial exponential decay which follows the model closely is due to the restriction of the deposition rate on one species which only exists close to the gap entrance. At a certain depth, the species restricting the deposition rate changes. In terms of the original proposed model shown in Eq. (3), this can be represented by a dominant species over a range of aspect ratios given as λ(T) which then changes over to a different dominant species called λ′(T) which controls the deposition further into the gap.

In order to express the average exponential decay speed and deposition far into the gap structure, two modeling terms have been extracted and plotted in Fig. 7. Figure 7(a) shows the behavior of the exponential distance constant over the range of substrate temperatures explored. We see that the distance constant increases significantly with increasing substrate temperature, which indicates that the FC deposition is occurring further into the gap structure. Another way the exponential distance constant can be expressed is in terms of half-life or in this case half-distance. This is the depth in terms of aspect ratio where half of the original FC deposition occurs and it is equal to τ multiplied by ln(2). Therefore, we can calculate from the data that at the lowest substrate temperature half the FC deposition occurs at approximately 1.2 aspect ratios into the structure and for the highest substrate temperature half the FC deposition occurs at approximately 2.0 aspect ratios. The gap size has no impact on this term once it has been normalized to aspect ratio. Figure 7(b) shows the normalized offset of the exponential fit which corresponds to the residual FC deposition deep into the gap structure. This term also slightly increased with substrate temperature which suggests a greater mobility of species into the gap structure. The data also show that the larger gap structure had more FC deposition.

There are several possible reasons for the behavior of the FC deposition response to substrate temperature and gap structure size. For the exposed region, ion bombardment from direct exposure to the plasma greatly increase FC deposition which has been described in great detail in other work.^25,29 We also know from previous work that this affects the structure of the FC film and the deposition thickness.²⁵ In this work, we focus on a region with no directional contributions, and therefore, only long lived reactive species contribute to the deposition within the gap structure. Without the ion bombardment component, the deposition becomes entirely dependent on the diffusion of the neutral species within the gap structure. We relate the neutral species density within the gap structure to the substrate temperature. The substrate temperature should impact two separate phenomena related to the FC film deposition as described in Eq. (1). The surface sticking coefficient will change with substrate temperature and will impact the rate of re-emission of species from the surface once initially bonded there. Therefore, this could allow species to migrate further into the gap structure.

Traditionally, it has been shown that FC deposition in an open plasma environment decreases with increasing substrate temperature. This is believed to be primarily due to the decrease in sticking of the FC molecules to the surface with the increased thermal energy. However, there has been other work that investigated similar phenomena to what we are studying in this work where the FC deposition is not in a traditional open plasma environment. Volland and Rangelow studied the deposition rate of FC species at the bottom of a trench structure with varying aspect ratios and at a similar range of substrate temperatures studied.³⁰ This work found that the deposition at the bottom of the trench was relatively unaffected by substrate temperature. However, the critical difference in this case is that in an actual trench structure, there would still be some ions present that contribute a significant portion to the creation of new reactive sites, which facilitates the deposition process. In our work, we investigate how this deposition is affected by substrate temperature without any ions present as it represents the sidewalls of the feature more prominently than the bottom of the feature.

We find that increasing substrate temperature has a significant impact on increasing penetration depth of FC species into a high aspect ratio structure and causing deposition on an SiO₂ substrate. Our studies are based on a gap structure, which represents the sidewalls of a high aspect ratio feature. The characteristic decay distance based on aspect ratio of the FC deposition sensitively reflects substrate temperature and is consistent over multiple gap sizes. However, the modeling of the deposition data also suggests that multiple species may play a role in this deposition mechanism deep into a gap structure since the trends appear to deviate from a single exponential fit. This fitting procedure emphasizes the relative importance of distinct species that each shows different decay behavior versus depth and thus contributes differently to FC film deposition at various depths within the gap structure.

The authors thank C. Li, K. Y. Lin, and P. Luan for their helpful discussions and collaborations on this project. This work was supported by Samsung Electronics. A. Pranda and G. S. Oehrlein also acknowledge support by the National Science Foundation (No. CMMI-1449309).