Temperature and oxygen concentration effects on anisotropy in chromium hard mask etching for nanoscale fabrication

Patterned chromium is a popular material in micro- and nanofabrication due to its beneficial properties and wide range of applications. For example, the high optical density, down to the deep ultraviolet regime, makes chromium a good material for photomask and reticle fabrication.¹ It is electrically conductive, wear-resistant, and easy to deposit by physical vapor deposition methods.^2,3 Furthermore, it forms a dense and thin oxide layer that self-passivates the bulk from further corrosion.⁴ Cr–Cr₂O₃–Cr junctions have been shown to be active for single-electron transistor fabrication.^5,6 Chromium is also a popular adhesion interlayer, for example, between gold and other materials. Moreover, chromium is a good hard masking material for dry etching, as it is relatively sputter resistant⁷ and has high selectivity in fluorine and bromine chemistries. This allows chromium to be an etching mask for many materials, for instance, silicon,⁸ SiO₂,^9,10 ZnO,¹¹ Si₃N₄,¹² TiO₂,¹³ MoSi,¹⁴ and GaAs.¹⁵ At the same time, chromium is patternable in oxygen and chlorine plasma chemistry.^16,17

In this work, we study chromium etching for use as a hard mask in creating UV-transparent nanoimprint templates, typically quartz or fused silica, for sub-10 nm applications. Deep nanoscale resolution has been demonstrated using nanoimprint lithography (NIL).^18,19 This allows for applications in bit patterned media^20,21 and semiconductor fabrication.²² Large area template lithography at such features sizes is being enabled by evolving technologies like directed self-assembly,²³ extreme UV,²⁴) or mix and match scanning probe lithography.²⁵ Because UV NIL is a one-to-one molding process, the template topography must be etched with single-digit nanometer (<10 nm) control in all dimensions. To achieve this control in the template etching step, a robust anisotropically dry-etched hard mask is required. Chromium has shown promise for such applications.^26–28 However, to achieve repeatability and manufacturability, a new level of mechanistic understanding of the plasma-etching process of chromium is required.

Over the last four decades, chromium dry etching studies have focused on patterning UV photomask absorber layers using chlorine/oxygen-based plasma chemistry.^1,16 Because there is a 4:1 size reduction from mask to projected feature, the minimum mask feature size can be relatively large—112 nm mask features are targeted in 2021.²⁹ Chromium etching process windows are significantly wider in this size regime. A common approach to achieve controlled and straight sidewalls is by etching the film past the time needed to clear the bulk of the film.³⁰ During this over etch, the feature sidewalls tend to straighten but at the expense of expanding the trench size and shrinking the feature. This feature shrink during etching is offset by printing lithographic features larger than the desired feature size, i.e., biasing the mask to compensate for the critical dimension (CD) target loss. This approach, however, is problematic for nanoimprint template fabrication where the features are in the sub-10 nm regime.³¹ 

Our work focuses on understanding the chromium etching process to control pattern transfer at sub-10 nm dimensions. Previously, we investigated chromium etching on blanket films³² over a wider range of temperature and oxygen concentrations than previously reported in the literature. To look for benefits that can arise at cryogenic etching temperatures—for instance, cryogenic etching of silicon using SF₆/O₂ chemistry promotes the formation of an etch inhibitor and etching anisotropy^33,34—studies were conducted at temperatures as low as −100 °C. Our blanket film etching tests showed us that by changing temperature and oxygen levels, we could move into different etching mechanism regimes and better control etching rates. Here, we build on this knowledge and focus on patterned films. In particular, we are interested in how etching profiles change with the oxygen concentration and how lowering the temperature drives the profile toward anisotropy. To achieve this, we study the patterning of sub-20 nm chromium films with trench sizes down to 15 nm as a function of oxygen concentrations and temperature. The chromium is masked with a highly selective resist, hydrogen silsesquioxane (HSQ), which allows us to characterize the chromium pattern anisotropy independent of mask erosion. High-resolution images of SEM cross sections, using the conductive polymer-metal coating technique for high contrast cross sections,³⁵ are used to quantify and qualify the etch results.

Overall we see a trend toward high anisotropy with low temperature. In addition, observation over a wide range of feature sizes, gas concentrations, and temperature gives us insights on mechanisms of profile control for different chromium etching Conditions. Our analysis suggests that anisotropy is supported by sticking of oxygen and chlorine on the sidewall and ion-induced redeposition of nonvolatile etch compounds. Profiles change with local oxygen and chlorine concentration as a function of temperature, etch depth, and position in the trench. These changes can lead to sidewall passivation and opportunities for anisotropic etching of chromium.

The chromium etch process in oxygen chlorine plasma was first described by Abe et al.,¹⁶ and a reaction process was proposed by Nakata et al.¹⁷ Chromium is etched by oxygen and chlorine, forming the volatile chromyl chloride CrO₂Cl₂ according to the reactions

This ternary etch system is rare in plasma etching where most volatile reactants are binary halides. In ternary systems, the two reactants can form intermediate products, bringing additional complexity to the etching process.

In our previous work,^26,32 we studied the etching of chromium films, without patterning, over ranges of chlorine/oxygen gas concentrations (0–100%) and substrate temperature (−80 to +40 °C) previously unpublished. We found that etching mechanisms change significantly with chlorine to oxygen ratios. In particular, etching rates first increased with the introduction of more oxygen into the gas mixture and then decreased. Maximum etching rates were found at oxygen concentrations between 20% and 30% for temperatures of −20 to +40 °C. In addition, we saw that etching rates generally slowed with a decrease in temperature, which is helpful when trying to control etching in sub-10 nm films or features. However, at the lowest oxygen concentrations, we saw a deviation from this trend—below oxygen concentrations of 12%, etching rates were higher with lower temperature. We surmised this was due to sticking coefficient (SC) increases at low temperature, which can strongly affect etching rates when the oxygen/chlorine ratio is low. For patterned films, our paper showed the growth of a film or condensation product at −100 °C.

The aim of this work is to extend the investigation into patterned films so that we can use gas ratios and temperature as a means to better control chromium dry etching anisotropy at the nanoscale. In addition, we initiate an understanding of the mechanisms which affect profile control over the studied parameter ranges.

All samples were etched using a load-lock equipped inductively coupled plasma (ICP) reactor (Oxford Plasmalab 150) at a coupling frequency of 13.56 MHz. The bottom electrode is temperature-controlled (±1 °C). All samples were etched using helium backside cooling. For improved thermal chip-to-carry-wafer contact, a perfluoropolyether vacuum oil (Fomblin 25/6) was used. Fomblin was carefully spread over the backside of the sample and not exposed to the plasma. Before etching, the chamber was conditioned with the etch gas environment. All etches were done using a chlorine/oxygen gas mixture with the following parameters:

Oxygen concentration: Oxygen concentrations (C_O2) in this work are given as a percentage (%) of the oxygen partial pressure (p_O2) over the total pressure of oxygen and chlorine (p_Cl2+ p_O2) in the etch chamber and are calculated by

Three oxygen concentrations were used and fixed at 1.5%, 41%, and 81%. These three oxygen concentrations represent the “low,” “mid,” and “high” referenced in our previous work.³² 

Temperature: we studied +20 °C and −50 °C to investigate the effect of changing temperature on anisotropy.

RF forward power: RF power that is coupled capacitively through the substrate creates a negative bias (DC bias) on the electrode with respect to the plasma (literature convention, however, is positive values, which are used here). High powers create narrower ion angular distributions that increase directionality.^36,37 Ion bombardment, however, can lead to mask erosion due to sputtering. To minimize this effect on the mask and carrier wafer, a low power, 5 W, was chosen for etching (see the supplementary material⁶¹ for the dependence of DC bias and RF forward power).

ICP power: An electromagnetic induction coil is supplied with 13.56 MHz RF power using an RF tuning network with automatic matching close to the coil. The ICP power was set to 500 W. This value corresponds to the value used in our previous work^26,32 and is comparable to work from other authors.³⁸ 

Etch steps: The entire etching process consisted of three steps: (1) cooling for at least 2 min to give sufficient time for temperature stabilization; (2) gas flow stabilization for 10 s at the required gas concentration with a 40 sccm total gas flow and a pressure of 5 mTorr to equalize conditions prior to the next step; (3) 5 s plasma ignition step at 500 W ICP power and 30 W RF power to strike the plasma; and (4) etching at 500 W ICP power and 5 W RF power.

To be able to study profile control over this large range of oxygen concentrations, we minimized mask contributions to the profile by avoiding erodable carbon-based polymer resists. Instead, we used HSQ, patterned by electron beam lithography, to create the chromium etching mask. HSQ contains only silicon, oxygen, and hydrogen bonds and, after lithographic exposure, is highly resistant in the chromium etching process studied. Film thickness was approximately 10 nm forming to give mask aspect ratios ≤1. For a majority of the studies, the chromium film thickness was 20 nm. These films were masked with HSQ patterned trenches with widths between 30 and 100 nm, a size range that can be reliably made by e-beam lithography and measured accurately by SEM cross-sectional analysis. To demonstrate feasibility on smaller features, 10 nm chromium films with 15 nm trenches were used.

These two types of samples were prepared as follows:

Sample type 1—HSQ patterns on 20 nm chromium/SiO₂/silicon stack: Samples were prepared using 100 mm. single-side polished prime grade silicon wafers with 250 nm (±5%) dry thermal oxide as the substrate material. Chromium was deposited by electron beam evaporation (Semicore SC600) with a deposition rate of 2 Å/s (chromium actually sublimates but is usually referred to as evaporated) to obtain a film thickness of 20 nm (±1.5 nm). Thickness was monitored in situ by a quartz crystal. For patterning, HSQ resist was spin coated and exposed using electron beam lithography (Vistec VB300). The resulting line patterns of 600 μm in length had different dimensions ranging from 200 to 60 nm full pitch. Post exposure, the resist was developed in a NaOH-based developer. Subsequently, chips 15 × 15 mm² were prepared. The postdeveloped HSQ mask was measured to have an average thickness of 23 nm.

Sample type 2—HSQ on a 10 nm chromium/SiO₂/silicon stack (provided courtesy of Seagate Technology): Samples were prepared on a silicon substrate with 300 nm thermal silicon dioxide and a top layer of 10 nm evaporated chromium. The HSQ mask was 10 nm thick with 30 nm pitch features.

To increase SEM imaging resolution, cross-sectional samples were prepared using conductive polymer-metal coating (CPM).³⁵ This method uses a layer of 50 nm Aquasave (Mitsubishi Rayon) and a 7 nm Au/Pd sputter coating. Samples were then cleaved for cross-sectional evaluation and imaged using a Zeiss ULTRA 55 field emission scanning electron microscope with an acceleration voltage of 7 keV. Images were taken at 200 k magnification (M) with a pixel size of 558.2 pm, for sample type 1 and M = 600k (pixel size of 186.1 pm) for sample type 2.

Measurements of profile characteristics were made using the measurement scheme illustrated in Fig. 1. This approach allows us to measure the vertical etching depth, D = e − d, where e is the chromium layer thickness and d is the remaining chromium thickness measured in the center of the trench. The lateral etch depth is calculated by U = (a − c)/2, where a is the width of the HSQ line and c is the minimum width of the remaining chromium feature. Note that measurements do not take into account the exact profile shape. The lateral etch depth, U, was always considered as the deepest part of the lateral etch no matter the mechanistic origin. A positive profile is defined as being sloped without lateral undercut. In all cases, HSQ mask erosion was undetectable and considered negligible.

We define and discuss two etching depth regimes. The etching regime is the regime from the start of the etch to the point where an approximately 6 nm film is left at the bottom. The over etch regime starts once chromium is completely removed at the bottom. As these regimes change, the mechanism changes. We start by showing results in the etching regime.

Figures 2 and 4 show results from three feature sizes etched at three oxygen concentrations, 1.5% (low O₂), 41% (mid O₂), and 81% (high O₂), and two temperatures, +20 °C and −50 °C. The profile shape is both oxygen and temperature dependent. This can be seen clearly in Fig. 2 which shows images of 30 nm features and a plot of the measured anisotropy as a function of oxygen and temperature. Figure 2(a) shows 30 nm wide trenches etched ∼14 nm deep into a 20 nm film. Figure 2(b) plots anisotropy, A = 1–U/D (see Fig. 1 for definitions). Each point in the plot represents an average of 3 to 8 samples. Error bars represent the standard deviation from the mean.

As reflected from Fig. 2, profile shape/anisotropy changes as a function of oxygen concentration. In addition, profile shape changes significantly between substrate temperatures +20 °C and −50 °C. The highest anisotropy is found at low-temperature and mid-oxygen concentration. On the other hand, undercut is especially prominent at temperatures of +20 °C and at high- and mid-oxygen concentrations. Overall, etching at −50 °C shows better anisotropy than etching at +20 °C.

This trend extends to smaller features. Figure 3 shows 15 nm trenches etched at −50 °C for the same three oxygen concentrations—features etched at +20 °C are not shown because the undercut was so severe that the HSQ collapsed. Like for the 30 nm features, 15 nm features etched at −50 °C have the highest anisotropy at the mid-oxygen concentration of 41%. This SEM image was consistent with other features imaged in this size regime.

For larger trenches, the trend toward anisotropy is not as obvious. Figure 4 shows 50 nm features over the same three ranges of oxygen concentrations and two temperatures. Mask undercut is present for all etching conditions. At +20 °C the profiles are similar to the 30 nm features in Fig. 2(a). However, the 50 nm feature profiles at −50 °C are not as anisotropic as those in Fig. 2(a) [for details see Supplement A (Ref. 61)]. In contrast to +20 °C, at −50 °C, the 50 nm features have a concave profile—they are less undercut immediately below the mask as compared to deeper into the film. Nevertheless, there is definitely an increase in anisotropy observed for 50 nm features at low-temperature and high levels of oxygen. For instance, note the limited undercut at −50 °C and high O₂ levels. The mechanisms related to such profiles will be discussed in Sec. V.

Comparing Figs. 2 and 4, clearly, anisotropy changes with changing feature size. To further illustrate the changes with feature size, Fig. 5 shows 30, 50, and 100 nm trenches in the etching regime at low-temperature and low-oxygen regimes. At 100 nm, there is pronounced footing at the edge of the features. This foot extends to about 15 nm into the trench. This suggests that feature edge effects may dominate the profile for small features.

All the previous results considered the chromium etching profile evolution prior to hitting the SiO₂ interface. Results for the over-etching regime (etching after reaching the SiO₂ interface), can be seen in Fig. 6 for 100 nm features at −50 °C and +20 °C and at different oxygen concentrations. The over etch times were defined in terms of blanket films with features being etched 20% to 40% more time than that needed to etch a blanket film under the same conditions. Note that at our etch parameters, the underlying SiO₂ appears to have infinite selectivity to chromium. Figure 6 shows that etching at +20 °C generates undercut with a positive profile for all oxygen regimes. At −50 °C, with increasing oxygen, the profile changes from straight to curved to retrograde. We surmise that reactant sticking coefficients and surface diffusion rates, both temperature dependent, play an important role in promoting transport of species to the sidewalls. Our hypothesis is that retrograde undercut in the high-oxygen regime is indicative of chlorine surface diffusion at the SiO₂ interface. We will discuss this further in the subsequent section.

Studies in the over-etching regime (Fig. 6) reinforce that lower temperature is preferred for better control of undercut rates. Using the correct overetch time for each regime, overall anisotropy is better at −50 °C and mid-oxygen levels. Full anisotropy, however, was only reached in the etch regime and the thinner (10 nm) films. We now address the mechanisms which contribute to profile evolution under the studied etching conditions.

Understanding what contributes to the exact profile shape is complex.

Here, we aim to explore the literature to develop some first-order hypotheses on the main mechanisms contributing to the shape in this chromium etching study. We will provide additional experimental studies in subsequent work.

Past studies on chromium dry etching typically have focused on near room temperature etching (e.g., Refs. 39 and 40). The process is primarily isotropic with small changes induced by ion bombardment. For patterned features, the directional nature of the ion bombardment can affect undercut and thus CD bias. For instance, a past study showed that RF power increases vertical etch rates but usually with the formation of a foot at the profile edge.⁴¹ The foot is removed by over-etching at the cost of CD (due to mask erosion).

While the literature on chromium etching anisotropy is sparse, we can draw analogy to extensive mechanistic studies of halide-based etching (fluorine, chlorine, or bromine) of silicon and SiO₂.^42–45 There are generally two dominant mechanisms—if the etching reaction is spontaneous, as is the case with silicon and fluorine chemistry, a surface inhibitor mechanism is necessary to prevent lateral etching.⁴³ Vertical etching can proceed when the directional ions remove the inhibitor from the bottom of features. The sidewall passivant sees limited ion bombardment giving rise to anisotropic etching. If the reactions are not spontaneous, such as that of silicon with HBr, etching is enhanced at the feature bottom because the ions promote lattice damage⁴⁶ and/or reaction products.⁴³ 

While these are the two primary mechanisms, in reality, anisotropic profile evolution is much more complex. For instance, both mechanisms may be active to different degrees in the same etching chemistry. In addition, other mechanisms must be considered to understand profile shapes.^45,47,48 These include charging,⁴⁹ surface diffusion,^50,51 redeposition,⁵² re-emission,⁵⁰ ion and neutral shadowing,^53,54 as well as sidewall passivation.⁴⁷ In our case, another important item to consider is what happens during intermediate product formation in this ternary etch system.

We first studied chromium etching as a function of oxygen concentration and temperature in blanket films.³⁵ This work showed Arrhenius-type behavior for oxygen concentrations between 23% and 73% of oxygen at temperatures measured between −80 and +40 °C. Therefore, one might expect isotropic etching even at lower temperatures. However, we clearly see that the etching is not isotropic over this entire temperature range. To understand the profile evolution which includes undercut of the mask, we must consider all mechanisms which can supply etch and passivant species to the sidewall. This includes primary reaction from the gas phase and secondary reactions after interactions with another surface.

Figure 7 illustrates several of the processes we will consider. Illustrated in the middle of the figure is product formation. CrCl₂O₂ is the volatile reaction product.¹⁷ Being a ternary etching system with a five-atom product, each reactive chromium site requires two chlorine and two oxygen atoms to become gas phase stable. Things to consider during product formation include flux of ions and neutrals, sticking coefficients of chlorine and oxygen, rates of surface diffusion, and rates of reaction of volatile productions [labeled as 1–6 in Fig. 7(c)].

Several processes can be sources as reactants at the sidewall and contribute to mask undercut including surface diffusion at the bottom of the feature [Fig. 7(c)] and surface diffusion from the mask [Fig. 7(e)] as well as re-emission of reactive species [Fig. 7(d)]. These processes can be balanced with ion-induced desorption of low-volatility compounds and inhibitors at the sidewalls such as redeposition [Fig. 7(a)]. Temperature can affect transport and reaction rates causing one mechanism to dominate over another.

Because we see increased anisotropy with a decrease in temperature, we surmise a sidewall etch inhibitor is forming. The primary reactants must be the source for the passivant since no polymerizing-type passivant is part of the gas feed. The next step is to consider if the passivant is a condensation product from the gas precursors or a secondary reactant formed after interacting with the etching surface. The reaction product CrO₂Cl₂ is volatile above −78 °C at 5 mTorr.³² As such, we conclude that passivation of condensed chromyl chloride is not possible at −50 °C. We thus consider all the effects of surface interactions and how this might lead to a passivant and/or undercut of the surface. Our conclusions are summarized in Table I and explained in detail in Secs. V B–V D.

It is also important to note that there are several mechanisms that we do not consider. In particular, this includes redeposition of the masking material. The HSQ mask selectivity to chromium was between 15:1 (at 1.5% oxygen) and 50:1 (at 41% oxygen), with no visible lateral erosion. Therefore, any effects due to mask erosion are neglected. In addition, physical sputtering of chromium is excluded because chromium etches with very low activation energy between 0.04 and 0.12 eV.³² This is much smaller than the energies required to dissociate bonds (1.6 eV).⁵⁵ This leads to etching rates which are much faster than the sputtering rate at 5 W RF power (DC bias between 40 V, at 81% oxygen, and 48 V, at 1.5% oxygen). Note that ion energy has not been measured but can be assumed to be similar to Ref. 38 and on the order of about 30 eV maximum.

What we will need to consider is ion enhanced removal of precursors and reactive products like nonvolatile intermediates. Kwon et al.⁴⁰ have shown that in a chlorine/oxygen plasma environment, the chlorine and oxygen separately bind to chromium but not to each other. Therefore, the formation of chromium oxides and chromium chlorides can be assumed. The final volatile product CrO₂Cl₂ must, therefore, be formed by a precursor. In addition, nonvolatile chromium oxides, chlorides, and oxychlorides which are not or insignificant precursors to chromyl chloride may also be present in the surface layer. These may be susceptible to removal and redeposition via ion bombardment.

Considering all these mechanisms, we now discuss which mechanisms can contribute to the low-temperature anisotropic profile formation in the features imaged in Figs. 2 and 5.

In this section, we evaluate whether reductions in surface diffusion and reflection/re-emission can contribute to the increased anisotropy we see as the etching temperature was lowered from 20 to −50 °C. Both of these mechanisms are important because they can promote reactive flux directed toward the underside of the mask, like that seen in Figs. 2(a) and 4. Reductions in temperature can affect both mechanisms. In surface diffusion, transport of adsorbed species is attributed to a weakly bound (physisorbed) state, since chemisorbed species are expected to be less mobile.⁵⁶ The transport of neutral species by re-emission from the surface takes place when an etching precursor desorbs before the reaction toward the final volatile product.⁵¹ 

Simulations are typically used to understand how these two mechanisms contribute to profile evolution, however, that is outside the scope of our current paper. Instead, we analyze our results in light of simulations conducted in an F/Si etching system, where Singh et al.^50,51 have simulated re-emission and surface diffusion effects of reactive species and how they bring flux to the sidewalls. We believe that their approach is relevant here since, like that of silicon etching by fluorine, we observe the presence of an isotropic reaction with similar activation energies (0.11 eV for the F/Si system⁵⁷ and 0.58 eV for the Cl/Si system⁵⁷).

To estimate the effect of surface diffusion in O₂/Cl₂ chromium etching, we have to consider the travel distance of reactive species on the surface at both temperatures studied. If anisotropy is caused by surface diffusion-limited reactive flux to the sidewall, the ratio of the characteristic length of the trench to the diffusion length on the surface, the Damköhler number “Da” has to be regarded.⁵⁰ For Da >1, the surface diffusion of a radical does not cover the entire trench width, and the etch would be surface diffusion-limited. For Da <0, the radical density would be constant. Our analysis for 30–100 nm trenches is detailed in Supplement C (Ref. 61) and uses estimated values from the literature. Overall, we estimate a maximum value of Da >0.18, indicating that species can easily diffuse to the sidewall. Therefore, even at −50 °C, we are not in a surface diffusion-limited regime; diffusion lengths much larger than the feature sizes can still supply reactive flux to the sidewall. (We made some simplifying assumptions that the diffusion is confined to the surface. It could be, for example, a multistep subsurface diffusion process, like known form chlorine–oxygen copper reactions.⁵⁸ In addition, flux can be provided directly from the gas phase itself.) Therefore, reduced surface diffusion does not contribute to the increased anisotropy we observed at low temperature.

We now investigate whether changes in re-emission (Fig. 7), which supplies reactants to the sidewall promoting undercut, might promote low-temperature anisotropy. We again draw an analogy to the work of Singh et al.⁵¹ in silicon etching. Singh et al. considered re-emission flux as a function of the sticking coefficient at the trench bottom. From Singh's model, a low sticking coefficient results in the highest undercut directly underneath the mask and a flatter profile bottom, as schematically shown in Fig. 8(a). This shape arises from the characteristic shape of the neutrals re-emitted angles from the trench bottom. As temperatures are lowered, sticking coefficients increase to give a concave-shaped undercut [Fig. 8(b)] with the shallowest undercut directly beneath the mask and a bowl-shaped bottom. The high sticking coefficient favors etching at the trench bottom as opposed to re-emission of etching precursors to the sidewall.

While we observe similar profile shapes as observed by Singh et al., it is with the opposite temperature trend. The concave-shaped undercut with a bowl-shaped bottom is observed in chromium etching at low-temperature [Fig. 8(b)], whereas it is observed at high temperature for silicon etching of Singh et al.

From this discussion, we conclude that the observed low-temperature anisotropy is due to neither reduced surface diffusion—diffusion lengths are large compared to feature lengths—nor reduced re-emission—we see opposite profile trends to what would be expected from reduced re-emission due to higher sticking coefficients. We now evaluate whether an inhibitor flux to the sidewall might be responsible for improved low-temperature anisotropy.

Typically, in low-activation energy dry etching processes, a sidewall inhibitor is required to promote high anisotropy. As mentioned previously, the only source of such an inhibitor is from redeposition of an etch product removed from the trench bottom by thermal or physical activation [Fig. 7(a)]. For instance, Lii investigated the effect of redeposition flux in silicon etching.⁵² They found that nonvolatile precursor emitted from the trench bottom can redeposit on the sidewall and inhibit radical fluorine attack and thus sidewall etching.

There are multiple potential products that can be redeposited in the chromium etching process. The nonvolatile intermediates possible are chromium oxides, Cr_xO_y, chromium chlorides, Cr_xCl_y, or some chromyl chlorides, Cr_xO_yCl_z (e.g., CrOCl is of low volatility⁵⁹). The source of these products is most likely the trench bottom. Note that redeposition of chromium-containing products by dissociation of CrO₂Cl₂ in the plasma is considered negligible given the small chip size and a residence time of about 0.4 s at 5 mTorr in the chamber at the etch conditions. XPS analysis by Kang et al.³⁹ revealed the formation of chromium chloride and chromium oxide on the surface during oxygen and chlorine plasma etching. They also found that the ratio of such surface compounds changes from oxides to chlorides with decreasing oxygen concentration. Such nonvolatile intermediates can be emitted from the trench bottom, activated by ion bombardment to form a deposit on the sidewall. This seems like a possible mechanism of anisotropy in this work.

In addition, such a redeposition can explain the increased anisotropy we see at low temperature. With the low-activation rate Arrhenius-type behavior we previously measured,³² primary etching will decrease with decreases in temperature. In addition, sticking coefficients are higher for both of the primary reactants and the passivant. We believe that this will increase sidewall passivant rates in a low-temperature re-emission mechanism. Considering the changes we see in the profile at −50 °C as oxygen concentration is increased (Figs. 2 and 4), that is, higher anisotropy at the mid- and high-oxygen concentration regimes, we surmise the redeposition precursors are the Cr_xO_y compound from the trench bottom. The nature of the redeposition product and the effect on formation with changes in power will be further explored in a subsequent paper.

Finally, we want to comment at unusual profile footing observed in Fig. 5 for the large features and how it might affect profile shape for the small features. This shape was only apparent in the low-oxygen and low-temperature regimes. There are three mechanisms to consider: (1) shadowing of ions, (2) shadowing of neutrals, and (3) growth of an etch inhibitor. In considering these, we think the most likely cause of the footing at low-oxygen concentration is excess chromium chloride formation when oxygen neutrals are shadowed at the trench edges. We come to this conclusion by first ruling out the ion shadowing as discussed in more detail below.

First, we consider if ion shadowing contributes to the footing. From Fig. 5, we see that the angle between the footing and mask edges in the 100 nm trench is about 30°. In addition, the mask aspect ratio is low, ranging from 0.8 at 30 nm to 0.25 at 100 nm. Ion angular distributions are, however, usually very narrow—the vast majority of ions striking below 10°, even at low energies with a low DC bias of about 50 eV.⁶⁰ Yet, we see a foot up to 30° angles with the mask edge. Such a large angle precludes ion shadowing as the cause of the footing.

Neutrals, however, with an isotropic angular distribution, could explain this phenomenon. At the low oxygen regime, an oxygen shadowing effect could reduce the formation of the CrO₂Cl₂ volatile product close to the trench edges. Due to their angular spread, the flux of neutral species is about half at the edge compared to the trench center.⁵⁴ Depletion of oxygen is also evidenced by the very low etch rates observed in the low-oxygen regime—it is close to the theoretically lowest possible etch rates for CrO₂Cl₂ (details in the supplementary material⁶¹). Furthermore, in our blanket film etching studies, we found that chromium chloride films start growing on the chromium surface in the absence of oxygen.³² Oxygen depletion due to neutral shadowing could promote a chlorine-rich surface near the sidewalls, where chloride formation could inhibit etching at trench edges. Therefore, because we are working in a low-oxygen regime, have neutral shadowing, and know that chromium chloride films can grow when oxygen levels are depleted, we conclude that the foot is not due to ion shadowing, but some combinations of these other two effects.

In this work, we investigated the mechanisms of patterning sub-20 nm chromium thin films in O₂/Cl₂ ICP dry etching at low pressure and low RF forward power. HSQ masked chromium patterns with trench widths down to 15 nm were evaluated from high-resolution SEM cross sections. Effects of variations in oxygen concentrations and substrate temperature were studied. We found that the patterning of chromium at −50 °C produces higher anisotropy than etching at +20 °C, offering an opportunity to obtain straight sidewalls and maintain CD. Etching anisotropy was high at smaller feature sizes, and sub-10 nm etching indicated that high anisotropy was possible at 50% oxygen concentration. The complex nature of the ternary chromium etch system shows a variety of mechanisms responsible for the development of certain profile shapes. A variety of mechanisms were examined for their role in promoting the high anisotropy seen at higher oxygen levels and smaller features or other profile features. These included surface diffusion, re-emission, and formation and redeposition of nonvolatile products. Profile analysis strongly suggests that surface re-emission of reactive species is the main process for undercutting of chromium films at +20 °C. Anisotropy at low temperature is consistent with surface redeposition of intermediate nonvolatile chromium oxides and chromium chlorides. Overall, chromium etching at low temperature is found to be promising for deep nanoscale and single-digit nanofabrication. Future work will focus on controlling these processes to promote anisotropic etching over a wider range of film thicknesses and etching conditions.

This work was completed at the Molecular Foundry and supported by the Office of Science, the Office of Basic Energy Sciences, and the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. D. Staaks, D. L. Olynick, S. D. Dhuey, and S. Sassolini were supported by this contract. The authors are especially grateful to Seagate Technology LLC for support of D. Staaks. The work of D. Staaks was partially supported by the European Union (EU, Contract No. 318804)—Single Nanometer Manufacturing for beyond CMOS devices (SNM).