Performing selective etch of Si₃N₄ and SiO₂ using a single-wafer wet-etch technology

D. Martin Knotter, Philips Semiconductors; and Nigel Stewart, Ian Sharp, and Dana Scranton, Semitool

Etching silicon nitride (Si₃N₄) films on semiconductor wafers is and will continue to be an essential part of microelectronics manufacturing for the foreseeable future. The properties of Si₃N₄ are desirable for a variety of well- known reasons, including its ease of deposition, etch-masking properties, and stability. The stability of Si₃N₄ films allows the material to be used for such elements as gate spacers, etch and CMP stops, antireflective coatings, barriers, and hard masks. In next-generation devices, Si₃N₄ will continue to be a critical material for device fabrication.

As devices approach the 65- and 45-nm design rules, maintaining good etch selectivity between Si₃N₄ and other films, such as Si and SiO₂, will become increasingly critical to the integrity of the finished product. Additionally, the capability of adjusting selectivity to achieve process goals will introduce potential enabling processes for next-generation devices. This article discusses a single-wafer wet-etch technology for selective etching
of Si₃N₄.

Etching Si₃N₄ Films

There are two conventional ways to etch Si₃N₄ films. Wet etching is accomplished by immersing wafers in a mixture of H₃PO₄ and H₂O heated to 150°–180°C. Water in solution with H₃PO₄ is important for maintaining etch stability and performance. The nominal etch rate depends on the properties of the deposited Si₃N₄ film, bath loading with silica or other etch by-products, and the temperature and percentage of water in solution.
Typical etch rates are on the order of 4 to 11 nm/min, with selectivity on the order of >10:1 for SiO₂ and 50:1 for silicon. Figure 1 shows typical etch rates for hot H₃PO₄ solutions as a function of water concentration at 1 atmosphere of pressure.

Figure 1: Typical etch rates as a function of water concentration in hot phosphoric acid solutions.

The second method for etching Si₃N₄ uses plasma. Plasma etching using various gas species, such as CH_xF_y, is a relatively fast process, with etch rates on the order of 20 nm/min. With Si₃N₄:SiO₂ selectivity on the order of 2:5, the selectivity of plasma etching is generally inferior to that of wet etching. However, because plasma etching has the advantage of isotropy, it produces nearly vertical etch profiles. It generally creates a passivation polymer to reduce lateral etching, which generally requires a wet clean to remove the polymer. The high energies of plasma etching can increase local temperatures, consuming thermal budget, and cause device damage from the impact of high-energy particles or charging.

Si₃N₄ wet-etch processes have generally been performed in batch immersion platforms, while plasma etching has been performed in single-wafer platforms. This article describes a new approach in which Si₃N₄ is etched at elevated pressure and temperature in a single-wafer wet-process tool. This approach produces etch rates and selectivity suitable for advanced processes. The productivity that can be achieved using a high-volume single-wafer tool makes the process feasible, especially for 300-mm manufacturing.

Etch Selectivity and Rate

Etch rate and selectivity are two critical parameters in the Si₃N₄ etching process. Solutions composed of hydrofluoric acid (HF) etch Si₃N₄, but they also etch SiO₂ at rates that are higher than those for Si₃N₄, resulting in poor selectivity. However, because of the characteristics of dilute HF mixtures and the kinetics of the reactions that take place between HF, water, Si₃N₄, and SiO₂ at elevated temperature and pressure, the etch selectivity between Si₃N₄ and SiO₂ can be adjusted to deliver selectivity better than 10:1.

The reactive species used to etch Si₃N₄ are monofluoride species (F^–, and HF), while the reactive species used to etch SiO₂ are difluoride species (HF₂^– and H₂F₂). The reactivity and selectivity of HF solutions can be altered by introducing high concentrations of protons, resulting in a shift of the HF equilibrium from ionic fluoride species (HF₂^– and F^–) to neutral fluoride species (H₂F₂ and HF). This shift diminishes the reactivity of Si₃N₄ by 150 times and SiO₂ by approximately 2500 times. The addition of acidic organic solvents can also alter the reactivity.¹ In such cases, the solution becomes less polar, and the HF species prefer to be in a neutral form. While the decrease in reactivity is most pronounced, the nonlinear nature of the shift in reactivity between Si₃N₄ and SiO₂ is the key to attaining desirable etch selectivity.

As suggested by the reaction kinetics model in Figure 2, diluting HF with water causes the solution to contain more monofluoride species. This gives rise to a trend toward higher etch selectivity ratios, but at the expense of the etch rate. The etch rate can be increased by elevating the temperature of the solution. Heating an HF:H₂O solution also results in the creation of higher concentrations of monofluoride species as a consequence of entropy effects. Hence, by heating the HF solution, the Si₃N₄ etch rate increases at a faster rate than the SiO₂ etch rate.² The objective then becomes to select an appropriate HF dilution and temperature for performing the process to achieve a suitable etch rate and selectivity.

Figure 2: Reaction kinetics of Si3N4 and SiO2 with HF species.

To assess the effect of HF concentration on etch selectivity and rate, a batch spin-spray system was used to perform a series of tests at different HF dilutions. Spray tools, because they tend to have a high mass-transport rate by virtue of their relatively turbulent and rapid boundary-layer flow, are suitable for examining the limiting case of high mass transport.

In preparation for the experiments, a 160-nm-thick layer of Si₃N₄ was deposited on 200-mm wafers using SiH₂Cl₂ and NH₃ in a low-pressure chemical vapor deposition (LPCVD) horizontal furnace from ASM (Bilthoven, The Netherlands). Process temperature was 785°C and process pressure was 300 mTorr. The SiO₂ film was thermally grown to a thickness of 600 nm on silicon at a temperature of 950°C in oxygen. Nine-point measurements using a UV1250 spectroscopic ellipsometer from KLA-Tencor (San Jose) determined layer thicknesses.

The spin-spray etch experiments were performed at 85°C under conditions of constant flow and time using a solution without HF and solutions containing 0.01, 0.02, 0.03, and 0.04% HF, respectively. The test results in Figure 3 show the Si₃N₄ etch rate and selectivity to SiO₂. SiO₂ was also etched, facilitating the presentation of selectivity.

Figure 3: Selectivity of Si3N4:SiO2 and Si3N4 etch rate in four HF dilutions at 85°C.

The results of this test validate the kinetics theory advanced above. Increasing the HF dilution improves selectivity while lowering the etch rate. For example, the selectivity ratio for Si₃N₄:SiO₂ is approximately 6:1 at an HF concentration of 0.04% and approximately 85:1 at an HF concentration of 0.0075%. Meanwhile, an etch rate of approximately 2.0 nm/min is attained at an HF concentration of 0.04%, while an etch rate of 0.6 nm/min is attained by reducing the HF concentration to 0.0075%. The etch rate decreases by a factor of 3.3 as HF concentration decreases by a factor of 5.3. Hence, between the higher and the lower HF concentration, selectivity increases by a factor of 14 while the etch rate decreases by a factor of 3.3. While these etch rates are too slow to be viable for most production processes, the quantification of the theory and the demonstration of the substantially nonlinear selectivity trend is valuable for understanding the mechanics of the etch process.

Achieving a Production-Worthy Process

To achieve a production-worthy process for advanced devices or 300-mm wafer production, an etch rate of at least 10 nm/min and a selectivity for Si₃N₄:SiO₂ of at least 10:1 is desirable. Extrapolating from the test data summarized in Figure 3 and considering the effects of temperature, it was determined that an HF concentration of approximately 0.08% in combination with an elevated temperature would achieve a desired etch rate and selectivity.

Figure 4: Elevated pressure and temperature process chamber used for HF processing.

An analysis of the target process conditions suggested that an 0.08% HF solution would have to be run at approximately 125°C to achieve an etch rate at or above 10 nm/min. However, dilute HF at 125°C results in a boiling solution. To etch at 125°C, it is necessary to maintain pressure above ambient. To that end, a process chamber based on the Capsule design from Semitool (Kalispell, MT) was developed. Pictured in Figure 4, this chamber can operate at approximately 2 bar and 125°C without boiling the solution.

As shown in Figure 5, three HF dilutions were tested in the elevated pressure chamber to derive a correlation for etch rate as a function of HF concentration. From the data, it is apparent that increasing the temperature to increase the etch rate was a viable option and that it was more beneficial than only increasing the HF concentration.

Figure 5: Si3N4 etch rates at 125°C and 2 bar pressure as a function of HF concentration.

Nevertheless, at 125°C, the etch rate fell short of the desired target of 10 nm/min. Thus, a process was tested in which the temperature was increased to 130°C. At 130°C, the Si₃N₄ etch rate was found to be 13 nm/min. Moreover, SiO₂ etched under the same conditions resulted in an etch rate of 1.4 nm/min, resulting in selectivity of approximately 10:1. The etch-rate data in Figure 6 for both Si₃N₄ and SiO₂ processed at 130°C indicate that the rates are essentially constant over time.

Since etch processes generally tend to be diffusion limited, the mass transport of reaction by-products and the replacement of etchant species is critical to maintaining a sustained and uniform process. In diffusion-limited processes, the hydrodynamics of flow can play a significant role in etch uniformity across the wafer and in microscopic features.

Figure 6: Si3N4 and SiO2 etch rates in 0.08% HF solution at 2 bar pressure under three rotational flow conditions.

The data in Figure 6 demonstrate the influence of mass transport on the etch rate. Usually, the wafer is rotated at an angular velocity sufficient to achieve good mass transport while simultaneously delivering dilute HF to the wafer at a flow rate of approximately 300 ml/min. Two cases of rotational flow were examined, one with fluid delivery directly at the center of the wafer and the other with fluid delivery 5 cm off-center. By way of contrast, a third case involved wafers that were subjected to neither rotation nor flow.

When rotational flow was applied, the etch rate remained nearly constant, although across-wafer etch uniformity was 3.5% in the case of the centered nozzle and 2% in the case of the off-centered nozzle. The etch rate for the wafers that were not subjected to rotation or flow decreased substantially from 13 to 8.3 nm/min. These results clearly demonstrate that the etch process is limited by diffusion or mass transport. In fact, the etch process's dependence on mass transport means that uniform fluid flow, or flow that maximizes mass transport across the entire wafer surface, is essential for achieving good etch uniformity and selectivity for Si₃N₄ and SiO₂.

The curve fit of the data in Figure 5 is relatively linear, highlighting the important role played by water in the etching of Si₃N₄. Water alone has a measurable impact on the etch rate of Si₃N₄ in the presence of low pH, as extrapolated from the etch data gathered in this experiment and the literature.³ At an HF concentration of 0, the approximate etch rate is predicted to be 2.4 nm/min. The mechanism for etching Si₃N₄ in ultradilute HF is essentially the same as etching it in H₃PO₄ solutions diluted with water. Figure 7 depicts the etch mechanism that takes place in the presence of water in a low-pH solution.

Figure 7: Mechanism for determining the reaction that etches Si3N4 in H2O or HF. The reaction is driven by the protonation of reactive sites to form [SiN*] and leads to the subsequent replacement of NH3 by F– or H2O.

Activated surface sites, [SiN*], are formed by the reaction of the reactive surface site with a proton, transforming NH₂ groups into a strong leaving group, NH₃, which are then replaced by F^– or H₂O groups. The relative surface concentration of [SiN*] is a function of pH. The acidic solutions discussed in this article have a pH of <3. The concentration of [SiN*] will be independent of pH if the ionization constant (pKa) value is on the order of 5 for surface NH₂ groups.⁴ Therefore, the etch rate of Si₃N₄ in hot H₃PO₄ (with water concentrations between 8 and 15%) and acidified water using, for example, HF can be represented by the equation:

R = k[H₂O%][SiN*] = k^´[H₂O%] = k₀e^(–Ea/RT)[H₂O%]

where k is the reaction rate constant, k₀ is the preexponential factor, Ea is the activation energy of the rate-determining reaction step (J/mol), R is the molar gas constant (8.314 J/mol K), and T is temperature in kelvin.

This equation can be validated by curve-fitting published data on Si₃N₄ etch rates in boiling and nonboiling H₃PO₄ solutions to an Arrhenius relationship.³ The resulting curve, shown in Figure 8, produces a strong correlation of etch rate for solutions, normalized by water content, to etch temperature. The correlation coefficient, R², is greater than 0.99 for this relationship.

Figure 8: Published etch-rate data for Si3N4 etched using LPCVD at 880°C in hot H3PO4 (normalized for water concentration).3

Extrapolating the curve to the case of 100% water at 125°C yields an etch rate of 4 nm/min, compared with the 2.4 nm/min rate extrapolated from the experiments in the Capsule chamber. Furthermore, extrapolating the curve to the case of 100% water and 85°C yields an etch rate of 0.2 nm/min, compared with 0.3 nm/min. These findings support the etch model described in this article and further explain the mechanisms of etch rate and selectivity that favor a dilute acidic solution, such as ultradilute HF, for etching Si₃N₄ films.

The results of tests using ultradilute HF at an elevated temperature and pressure demonstrate that Si₃N₄ etching can be performed with suitable selectivity to SiO₂ at relatively high etch rates. These etch rates, coupled with modern high-volume single-wafer platform technology, can provide suitable process capability and productivity for advanced device manufacturing. Therefore, it appears feasible to replace H₃PO₄ with dilute HF as an etchant for Si₃N₄.

Figure 9: Etch selectivity of Si3N4:SiO2 as a function of the molar concentration of HF and pH at 25°C.

This discussion has focused on developing and demonstrating a process that can provide a suitable etch rate and selectivity between Si₃N₄ and SiO₂. The kinetic model described here also suggests that selectivity can be adjusted based on HF concentration and temperature. Figure 9 shows etch selectivity as a function of molar concentration and pH for HF-based solutions. This graph demonstrates that selectivity can be chosen based on molar concentration, providing the means to achieve a wide range of process objectives.

Conclusion

The scanning electron microscope (SEM) images in Figures 10a and 10b demonstrate how a dilute HF process at an elevated temperature can be used. The figures show a poly layer and an Si₃N₄/SiO₂ hard mask with photoresist. In Figure 10a, the hard mask has been structured in the dilute HF process using a solution that provides 1:1 selectivity. In Figure 10b, the structure has undergone plasma etch, in which the structuring of the Si₃N₄/SiO₂ hard mask provides the desired poly etch profile. These images illustrate the versatility and feasibility of using the dilute HF process to etch Si₃N₄.

Figure 10: SEM images showing (a) poly structure with Si3N4/SiO2 hard mask under resist, and (b) structured poly layer after resist removal and plasma etch (process not optimized).

The application of dilute HF solutions at elevated temperature and pressure provides a means for adjusting the selectivity of Si₃N₄ to SiO₂ to achieve various process objectives. In the era of thinner films and challenging device architectures, such a process will provide precise process performance at a low cost of ownership.

Acknowledgments

The authors wish to acknowledge Nicole Wils from Philips Research (Eindhoven, The Netherlands) for supplying the poly structure/hard-mask SEM images, Rene Vroom from Philips Semiconductors' MOS34 for providing wafers with Si₃N₄ layers, and James Heffernan from Semitool UK (Cambridge) for providing logistics.

References

1. CA Deckert, "Pattern Etching of CVD Si₃N₄/SiO₂ Composites in HF/Glycerol Mixtures," Journal of the Electrochemical Society 127, no. 11 (1980): 2433–2438.

2. V Harrap, "Equal Etch Rates of Si₃N₄ and SiO₂ Utilizing HF Dilution and Temperature Dependence," in Semiconductor Silicon 1973, ed. HR Huff and RR Burgess (Pennington, NJ: Electrochemical Society, 1973), 354.

3. W van Gelder and VE Hauser, "The Etching of Silicon Nitride in Phosphoric Acid with Silicon Dioxide as a Mask," Journal of the Electrochemical Society 114, no. 8 (1967): 869–872.

4. DM Knotter and TJJ Denteneer, "Etching Mechanism of Silicon Nitride in HF-Based Solutions," Journal of the Electrochemical Society 148, no. 3 (2001): F43–F46.

---

D. Martin Knotter, PhD, is principal scientist for cleaning/wet processes at Philips Semiconductors in Nijmegen, The Netherlands. When he joined the company in 1990, he belonged to the surface chemistry group. In 1995 he joined the process module group, working on advanced cleaning processes. He received an MS in chemistry from the University of Amsterdam and a PhD in organometallic chemistry from the University of Utrecht, both in The Netherlands. (Knotter can be reached at +31 24 3532225 or martin.knotter@philips.com.)

Nigel Stewart is a mechanical design engineer at Semitool Europe (Cambridge, UK), which he joined in 2001. From 1985 to 2001, he was employed at Westcode Semiconductor. He received a higher national diploma in mechanical engineering and a higher national certificate in electronics engineering. (Stewart can be reached at +44 1223 518879 or nstewart@semitool.com.)

Ian Sharp is Semitool's European sales manager for the greater European region. At Semitool, which he joined in 1989, he has held positions as a process engineer, product manager, and general manager. In addition, he has worked as a process engineer in the III-V sector. Sharp holds multiple U.S. patents. He received a BS in chemistry from the University of Glasgow in Scotland. (Sharp can be reached at +44 1223 518463 or isharp@semitool.com.)

Dana Scranton, PhD, is vice president of surface preparation technology at Semitool (Kalispell, MT.) He has been associated with the semiconductor industry for 19 years, 15 of which he has spent with Semitool. Scranton has held various positions in engineering, marketing, sales, and general management. He received BS and MS degrees in mechanical engineering from the University of Wyoming in Laramie and a PhD in mechanical engineering from LaSalle University in Philadelphia. (Scranton can be reached at 406/751-6360 or dscranton@semitool.com.)