High-Selectivity Silicon Nitride Etch Process

		At a Glance
	To meet the high-selectivity requirements for LOCOS and STI applications, very high-selectivity silicon nitride etch processes have been developed on an isotropic etcher using a downstream, inductively coupled plasma (ICP) system. Etch selectivity of LPCVD nitride to thermal oxide greater than 40:1 was achieved at a nitride etch rate of 500 Å/min. The fluorine-based etching chemistry was optimized for nitride-to-oxide selectivity of greater than 12:1 with a nitride etch rate of ~1200 Å/min.

Silicon nitride is widely used as a mask materialin isolation technologies such as local oxidation of silicon (LOCOS) and shallow trenchisolation (STI) and in self-aligned-contact (SAC) technology. Removal of this maskinglayer with a dry etch process offers many advantages over current wet etch technology. Theremoval of an oxynitride layer formed on top of the nitride mask during field oxidation,for example, would require two separate wet etch baths: a buffered-oxide etch bath foroxynitride removal, prior to a hot phosphoric acid bath for the nitride layer removal. Wetetch processes typically require frequent chemical changes to maintain process stabilityand have the added expense of associated disposal or recycling of chemicals, thusincreasing the system's cost-of-ownership. 

In contrast, single-wafer, dry-etch processing can provide better process, particulateand environmental control than a batch-loaded, wet-etch bath. These advantages candirectly improve the yield of small geometry IC devices. However, to achieve the highselectivity required by nitride dry etch processes for advanced devices fabrication withvery thin pad oxide films is very challenging. While the addition of chlorine has beenshown to produce high selectivity of films,^{1, 2} a non-chlorine dry etch processthat achieves the same or better process result is more desirable because of corrosion andcontamination issues associated with chlorine. A non-chlorine, high-selectivity nitrideetch process has therefore been developed.

Experiment

Experiments were carried out on Mattson Technology's Aspen LiteEtch system, whichemploys an inductively coupled plasma (ICP) source operating in a downstream mode.Operation without energetic ion bombardment minimizes the oxide etch rate, making it morefavorable for use in nitride etch.

LPCVD nitride and thermal oxide films on 150 mm, blanket silicon wafers were used forestablishing process baseline and trends. The pre-etch and post-etch film thicknesses weremeasured on a Tencor UV1050. Goodness-of-fit values were monitored, both before and afteretch to ensure valid measurements. Customers provided 150 mm wafers from split lots forcomparison against their standard wet bench etching performance. Electrical and yield datawere collected from these runs.

Optimizing the etch process

NF₃/O₂ and CF₄/O₂/N₂ chemistrieswere investigated as etchants for the bulk nitride. The fluorine species of NF₃provided the efficiency and simplicity in chemistry to understand the etching mechanism.³Comparisions were made to CF₄, a less expensive and preferred gas forsemiconductor manufacturing. As expected, similar process trends were observed between thetwo chemistries because of their similar etching mechanisms.

The first major etch process optimization was achieved by generating nitrogen oxide(NO) species in fluorine-based etching chemistries. Ni-tride-to-oxide selectivity as afunction of the O₂/NF₃ gas flow ratio was studied.³ Byincreasing the O₂/NF₃ ratio at a fixed NF₃ flow rate (50sccm), selectivity was increased by a factor of seven.

Nitride-to-oxide selectivity and etch rates of nitride and oxide films as a function ofN₂ flow is shown at three O₂/CF₄ gas flow ratios inFigure 1. Both selectivity and nitride etch rates increase with increasing O₂/CF₄ratio; however, the oxide etch rate saturates at an O₂/CF₄ ratio of~20%. Nitride etch rates maximize below 50 sccm N₂ flow, whereas oxide etchrates decrease with increasing N₂ flow. This results in an increase in thehigh-selectivity trend at a high O₂/CF₄ flow ratio and a decrease inthe maximum value of selectivity at a low O₂/CF₄ flow ratio.

From these trends, we conclude that high N₂ and O₂ flowconditions that provide many NO species offer high nitride-to-oxide selectivity.

During the etch process, concentration of reactive species, gas-surface reactioncharacteristics and byproduct volatility determine the etch rate.⁴ Excessiveamounts of O₂ in NF₃ and O₂/N₂ in CF₄dilute the concentration of reactive fluorine species and therefore lower the oxide etchrate. However, these gases increase the nitride etch rate because of the presence of NOmolecules in the plasma that react with the N atoms in the nitride film after Si atomshave been removed by reactive F species.^3,5 High ratios of O₂/NF₃and (O₂/N₂)/CF₄ with O₂/N₂ = 1 arethus preferred for a high nitride etch rate and high selectivity.

Optimization: temperature

Fig. 2. Increasing methanol flow increases selectivity while decreasing both nitride and oxide etch rates.

The second major improvement to nitride-to-oxide selectivity and etch rates is achievedthrough their strong dependence on wafer temperature. Test results indicate thatselectivity increases with decreasing temperature and that the oxide etch rate increasesmuch faster with increasing temperature than the nitride etch rate. By fitting the logscale of the etch rates against the reciprocal of absolute temperature (T) (Arrhenius'sLaw), we will obtain activation energies for nitride and oxide etching. These values arecompared with those obtained in an atomic fluorine etch environment⁶ in thesame temperature range as shown in Table 1. 

With the presence of O₂ in the NF₃ plasma, the nitride activationenergy is reduced, and the oxide activation energy is increased. This larger difference inthe activation energies results in higher nitride-to-oxide selectivity when processing atlow temperatures.

Optimization: pressure

Table 1. Comparison of Activation Energy Values
	NF3/O2	F-only7
Nitride Eact	0.32 .02 kcal/mol	3.55 .28 kcal/mol
Oxide Eact	5.4 .5 kcal/mol	3.36 .40 kcal/mol

The third improvement on selectivity and etch rates is achieved by changing processpressure.³ The selectivity and etch rates of nitride and oxide films were takenat different pressures using the NF₃/O₂ process. Both selectivityand etch rates increase with increasing pressure. The nitride etch rate increases muchfaster than the oxide etch rate, resulting in the observed higher selectivity. The etchrate increase is attributed to reactive species residing longer at the wafer surface(assisting gas-surface reaction) and to a greater abundance of reactive species. Reactivespecies for nitride etch include fluorine and NO species.

Flow rates of process gases examined showed improvements in etch rates but had noeffect on nitride-to-oxide selectivity.³ Selectivity is independent of the NF₃flow rate at each O₂/NF₃ ratio, whereas etch rates increase withincreasing NF₃ flow rates. The etch rate increase is attributed to morereactive species being delivered to the wafer surface for the etching reaction, consistentwith the discussion by Chapman.⁷ When the flow of an etchant gas is introducedat a 'low flow rate region' where the etch rate is 'mass transportlimited,' the etch rate increases with increased gas flow, and selectivity isdetermined primarily by the ratio of the generation rate of active species.⁷

High-selectivity nitride etch (HSNE)

Using the methods described thus far, by providing reactive NO species influorine-etching chemistries, lowering wafer temperature and raising process pressure, wecan only achieve a 12:1 nitride-to-oxide selectivity and an ~1200 Å/min nitride etchrate. To achieve >40:1 nitride-to-oxide selectivity, a 'deposition' gas wasadded into the etching processes. A deposition gas can include a gas used for thin filmdeposition (for example, NH₃), a gas used for polymer formation in anisotropicoxide etching (high C/F) ratio gases (CH₃F, CH₂F₂) or ahydrocarbon gas (such as methanol CH₃OH).

Processes with the addition of deposition gases, such as NF₃/O₂/NH₃,NF₃/O₂/CH₃F and CF₄/O₂/N₂/NH₃,were tested at 10°C and resulted in essentially the same selectivity and etch rates.These experiments were done on a standard Aspen LiteEtch tool. CF₄/O₂/N₂with the addition of CH₃OH, CH₃F and CH₂F₂,respectively, tested at ambient temperature with slight hardware modifications, achievedsimilar selectivity and etch rates.

NF₃/O₂/NH₃ etch process trends as a function of NH₃flow rate at 10°C were studied.³ With increasing NH₃ flow,selectivity increases, and the etch rates of nitride and oxide films decrease almostlinearly. At 50 sccm NH₃ flow, the oxide etch rate crosses the zero axis,resulting in infinite selectivity. The nitride etch rate is ~500 Å/min.

Two-sided etch

Fig. 3. SEM cross sections of pre-etch (top) and post-etch (bottom) patterned wafers indicate no pinholes, indicating high selectivity, no etch residues/particles and maximum pad oxide remaining.

A CF₄/O₂/N₂/CH₃OH etch process wasdeveloped at ambient temperature. LPCVD nitride films used as a masking layer for LOCOSand STI applications are deposited on both sides of a wafer. These films are preferablyremoved simultaneously in a dry etcher for high throughput. The setup for the simultaneousdry etch is referred to as two-sided etch (TSE) and is accomplished with the heater blockremoved. The wafer is supported on pins so that the wafer is etched at ambient temperature(~25°C). With a temperature increase from 10°C to 25°C, the oxide etch rate increasesfaster than the nitride etch rate, resulting in decreased nitride-to-oxide selectivityfrom >50:1 to ~25:1. The nitride etch rate remains at ~500 Å/min. 

To effectively minimize the oxide etch rate, a baffle is placed in the chamber. Thisincreases nitride-to-oxide selectivity to greater than 40:1. With this hardwareconfiguration, the se-lectivity and nitride etch rate for the bulk etch (no CH₃OH)are ~20:1 and 1200 Å/min, respectively. 

Figure 2 shows nitride-to-oxide selectivity and etch rates as a function of CH₃OHflow with the modified hardware. The selectivity increases with increasing CH₃OHflow rate. When CH₃OH flow rate reaches 50 sccm, the selectivity becomesgreater than 40:1. Nitride etch rate reaches a local maximum around 10 sccm CH₃OH,whereas oxide etch rate decreases with increasing CH₃OH flow.

Results from device wafers

The processes with >12:1 nitride-to-oxide selectivity and a nitride etch rate of~1200 Å/min can be used for bulk nitride etch (referred to as bulk etch), where 80%-90%of the top, thick nitride layer is removed first. Because only the thick field oxide isexposed to the plasma during the bulk etch, greater than 12:1 selectivity is satisfactory.However, when the etching process reaches the underlying thin pad oxide (typically in the100 Å range for 0.25 µí technology), selectivity greater than 40:1 is required.Processes described under the 40:1 selectivity and 500 Å/min discussion can be used forremoving the 10-20% remaining nitride and for overetching. A two-step, complete etchrecipe was developed, consisting of bulk etch and HSNE in consecutive order. 

The two-step etch process was performed on device wafers at two different customersites, comparing performance against standard wet etch processing. Selectivity and nitrideetch rate results are summarized in Table 2. The first set of wafers were taken from splitlots and used CF₄/CH₃OH chemistry with the modified hardware. Figure3 shows SEM cross sections of pre-etch and post-etch wafers. No pinholes or surfaceresidues were observed on post-etch wafers. Nitride film on both sides of the wafer werecompletely removed. To monitor etch endpoint for the backside film, optical emissionsignals ~540 nm (bandwidth 0 nm) were also collected (Fig. 4).

Fig. 4. Endpoint traces of device wafers using two-sided etch hardware indicate complete removal of the backside nitride film.

Repeatability tests within-cassette and cassette-to-cassette have shown nitride etchrates are within <,+/- 2% for bulk etch and HSNE. The oxide etch rates are within ~Å/min for the HSNE process. These data indicate process stability. Electrical and yielddata have shown comparable results with the customer's standard wet etch process.

A two-step etching using NF₃/NH₃ chemistry at 10°C was performedon the second customer's device wafers from split lots for electrical data comparison. Thetested device parameters include BV_DSS, diode leakage, L_eff, W_effand Q_BD. The electrical data have shown equivalent results with a tighterdistribution compared to the customer's standard wet etch process. Cross-section SEMs haveshown comparable post-etch profiles for LOCOS structures for both dry and wet processeswith no indication of pin holes across the entire wafer.

Marathon tests have shown nitride etch rate uniformity to be within %, oxide etchrates to within Å/min and particle generation of <0.08 particle/cm².Remaining particles can be removed by subsequent wet pad oxide etching or wet cleaning.This two-step, HSNE process, for bulk etch and HSNE, is shown to be stable, repeatable andessentially particle-free.

CF₄/O₂/N₂/CH₂F₂ chemistryprovides a higher nitride etch rate than CF₄/O₂/N₂/CH₃OHchemistry. Latest results on 200 mm blanket wafers using CF₄/O₂/N₂/CH₂F₂have been evaluated for the HSNE process. With the modified hardware, the nitride etchrate is ~600 Å/min, and nitride-to-oxide selectivity is >40:1.

Acknowledgments

The authors appreciate helpful discussions from Allan Wiesnoski, Bob Guerra and SteveSchultheis of Mattson Technology, Didier Florin of Verity Instruments and Dr. Matt Holdenof Nanometrics Inc. The authors also greatly appreciate help from Lindsey Mitobe and TedPettit of Mattson Technology for completion of this work.

References

1. N. Hayasaka, H. Okano, Y. Horiike, Solid State Technology,31(4), 127(1988).

2. D.E. Ibbotson , J.A. Mucha, D.L. Flamm, J.M.Cook, Appl. Phys. Lett.46, 794 (1985).

3. Ying Wang, Leroy Luo, to be published, J. Vac. Sci. Technology,May/Jun A (1998).

4. J.W. Coburn, 'Plasma Etching and Reactive Ion Etching,'American Institute of Physics, (1982).

5. M.G. Blain, T. L. Meisenheimer, J.E. Stevens, J. Vac. Sci.Technol., 14 2151 (1996).

6. Lee M. Loewenstein, J. Appl. Phys., 65 386 (1989).

7. Brian Chapman, 'Glow Discharge Processes: sputtering and plasmaetching,' John Wiley & Sons, 1980.