Research Article Archive Versions 6 Vol 1 (1) : 18010102 2018
Download
Silicon Nitride Etch via Oxidation Reaction in Fluorocarbon/Oxygen Plasma: A First-Principle Study
: 2018 - 07 - 20
: 2018 - 09 - 30
4781 112 0
Abstract & Keywords
Abstract: Conducting all-in-one etch process for 3D-NAND fabrication requires close etch rate (E/R) for SiO2 and Si3N4; however, to attain comparable and high etch rate for both materials is challenging. In this work, we performed first-principle studies on the etching mechanism of Si3N4 in fluorocarbon/oxygen plasma. The feasibility of using fluorocarbon/oxygen plasma to etch Si3N4while attaining close E/R to SiO2 through the complementary nitride to oxynitiride (SiOxNy) transformation has been identified. Such transformation involves two stages: N atom elimination and Si-O bond formation. By modeling the essential chemical reactions on the Si3N4surface, we shed light upon the underlying mechanisms behind both stages. We simulated the N-elimination reactions involving the formation and desorption of NO and FNO molecules as well as the substitution with F atoms. We found that N atoms can be eliminated by forming NO molecules, especially with the assistance of F-substitution in Si-N bond breaking. The predicted O-additive energies indicates that forming SiOxNy structure after N-elimination is possible. Following that, the dependency of chemistries favoring either high E/R or active SiOxNy formation on the fluorocarbon/oxygen ratio was discussed. We hope that the work will build a foundation for future studies on pursuing all-in-one ON etch process via the surface modifications.Conducting all-in-one etch process for 3D-NAND fabrication requires close etch rate (E/R) for SiO2 and Si3N4; however, to attain comparable and high etch rate for both materials is challenging. In this work, we performed first-principle studies on the etching mechanism of Si3N4 in fluorocarbon/oxygen plasma. The feasibility of using fluorocarbon/oxygen plasma to etch Si3N4 while attaining close E/R to SiO­2 through the complementary nitride to oxynitiride (SiOxNy) transformation has been identified. Such transformation involves two stages: N atom elimination and Si-O bond formation. By modeling the essential chemical reactions on the Si3N4 surface, we shed light upon the underlying mechanisms behind both stages. We simulated the N-elimination reactions involving the formation and desorption of NO and FNO molecules as well as the substitution with F atoms. We found that N atoms can be eliminated by forming NO molecules, especially with the assistance of F-substitution in Si-N bond breaking. The predicted O-additive energies indicates that forming SiOxNy structure after N-elimination is possible. Following that, the dependency of chemistries favoring either high E/R or active SiOxNy formation on the fluorocarbon/oxygen ratio was discussed. We hope that the work will build a foundation for future studies on pursuing all-in-one ON etch process via the surface modifications.
Keywords: 3D-NAND; oxide; nitride; oxynitride; plasma etch; first-principle
1.   Introduction
As one of the most promising solid-state memory technologies, 3D-NAND has been drawing significant attention in the past years. An obstacle of engineering the structure is to achieve high quality etch profile of the well-known oxide-nitride-oxide-nitride (ONON) with the presence of dummy oxide structure on the same architecture. To precisely control the etch amount of different parts of materials, namely, to achieve the close etch rates (E/R) for SiO2 and Si3N4 is a key; however, to attain both high and similar etch rates at the same time is challenging. A potential approach to achieve this goal is surface modification. The Si3N4 structure could have a similar etch rate to that of SiO2 if the surface undergoes a transformation to silicon oxynitiride (SiOxNy). Therefore, O2 containing etching gas is worth studying. Discharge chemistries of CF4/O2 has been previously studied for the SiO2 and Si3N4 etch, while low selectivity of Si3N4 to SiO2 etch (1:1 to 3:1) was reported[1],[2] The findings, on the other hand, reveal the possibility of fluorocarbon/oxygen plasma being an etching chemistry to attain close E/R between SiO2 and Si3N4. In the later studies, a much higher selectivity of 9:1 was reported by adding N2 into the plasma[3]. Such improvement was attibuted to the reaction of NO + N(s) → N2 + O(s), where (s) indicate surface, enhancing the E/R by the ealier works[4],[5] . The accumulation of O atom on surface during etch suggests possible SiOxNy formation through creating strong Si-O bonds. Moreover, another previous work reported that the E/R of SiO2 and Si3N4 in CF4/O2 plasma becomes closer when increasing the O2 percentage[6]. This finding further implies the possibility of using fluorocarbon/oxygen to achvieve the similar E/R for the two dielectric materials. More importantly, the dependency of selectivity on both O2 and N2 flows would provide a hint on the tunable etching behavior with fluorocarbon/oxygen plasma.
To develop the idea, better understandings of etching mechanism in fluorocarbon/oxygen plasma are essential. In this work, we provided theoretical insights on the etching mechanism of Si3N4 in fluorocarbon/oxygen plasma. We focused on the N-elimination of Si3N4 etch, as the SiF4 formation for Si-elimination is expected. The purpose of this report is to point out the feasibility of using fluorocarbon/oxygen plasma to etch Si3N4 and also to obtain a similar E/R to that of SiO2 by converting the nitride to oxynitiride (SiOxNy). We hope that the study will lay a foundation for future works on pursuing fast and uniform ONON etch via surface modification. The transformation of Si3N4 structure to SiOxNy involves two stages: N atom elimination and Si-O bond formation. By modeling the essential chemical reactions on the Si3N4 surface, we were able to shed some light upon the underlying mechanisms behind both stages. To begin, the sufficiency of using atomistic model cluster on simulating the Si3N4 surface structure was verified. By employing this model cluster, we simulated the N-elimination reactions involving the formation and desorption of NO and FNO molecules as well as the substitution with F atoms. In addition, the N atom with different coordination numbers were considered to study the surface effect on the elimination reaction. The results suggest that O atoms eliminating N atoms by forming NO is feasible, especially with the assistance from F atoms. Following that, the O-additive energies were calculated. The data indicates that forming SiOxNy structure after N-elimination is possible. On top of that, the dependency of chemistries favoring either high E/R or active SiOxNy formation on the fluorocarbon/oxygen ratio was discussed.
2.   Methodologies
Atomoic structures and energies reported herein were calculated using the density functional theory (DFT). The simulations considering the H14Si4N6 cluster structure is done within the B3LYP/6-31++G(d,p) scheme[7] ,[8] ,[9],[10],[11], as implemented in the Gaussian 16[12]. (The atomic structure of cluster is detailed in the Results and Discussions). The simulation considering periodic Si3N4 slab structure is done within the generalized gradient approximation (GGA-PBE)[13] ,[14] as implemented in the Vienna Ab-inito Simulation Package (VASP)[15] ,[16]. For VASP calculation, the projected augmented wave (PAW) method[17] ,[18]with a plane-wave basis set was employed to describe the interaction between ion core and valence electrons. An energy cutoff of 350 eV was applied for the plane-wave expansion of the electronic eigenfunctions. For geometry optimization and energy calculations, all atoms were fully relaxed using the conjugate gradient method until residual forces on constituent atoms became smaller than 1×10-2 eV/Å. For Brillouin zone sampling, (3×3×1) k-point mesh in the scheme of Monkhorst-Pack[19]was used for the 1×1×3 supercell of slab structure.
The reaction energy change (ΔE) is calcualted by substracting the totoal energy of final system with the intial system, where both systems could contain a surface structure and/or gas molecules and/or atoms. The negative/positve value indicates the exothermic/endothermic reaction. The realtive values of ΔE were used to argue the predominances of related reactions.
3.   Results and Discussion
3.1.   Model Cluster Verification
Figure 1 (a) illustrates the atomic configuration of Si4N6 cluster employed in this work to model the etch front of Si3N4 surface structure. The 4-Si cluster is a pyramid geometry consisting of 4 Si, 6 N and 14 H atoms. In H14Si4N6 cluster, as the valence state of H atoms is +1, the valence state of Si and N atoms are predicted to be +4 and -3, recreating the Si3N4 structure well. Surrounding the center of cluster, four Si atoms form a tetrahedron while each Si atom binds to the other 3 Si atoms separately via a bridging N atom - 6 bridging N atoms in total. On the outer layer of cluster, 4 dangling N atoms binds to the 4 Si atoms individually, making all Si atoms 4-fold (fully) coordinated. In addition to Si atoms, the bridging and dangling N atoms are bound to another 1 and 2 H atoms, respectively, making all N atoms 3-fold (fully) coordinated. The bond distance of Si-NH2 and the rest of Si-N are 1.72 and 1.75 Å, respectively; in good agreement with the previous work[20]. It is worth noting that the hydrogenation not only saturates the surface of the cluster but also simulates the realistic silicon nitride wafer which often carries additional H atoms both on the surface and in the bulk structure from synthesis. To make sure the size effect of H14Si4N6 cluster does not compromise its sufficiency on representing the Si3N4 surface structure, we compared the Si-N bond distance and NH2 binding energy on the cluster and the periodic slab structures. Figure 1 (b) illustrates the atomic configuration of 1×1×3 supercell of hexagonal Si3N4 (001) slab structure, with lattice constant on x-y plan of a = b = 7.6015 Å and γ = 120°. The surface is terminated by H atoms added on bridging N atoms and NH2 on Si atoms; the valence state of Si and N atoms are the same as the bulk structure. The Si-N bond distance predicted for the cluster and slab structure are around 1.76 and 1.74 Å, respectively; in agreement with each other. Table 1 summarizes the predicted binding energy of the NH2 molecule on the cluster (-4.70 eV) and surface (-4.63 eV) structure. Combined with the bond distance comparison, the close values imply that the H14Si4N6 cluster structure employed in this work is sufficient as an approximation to the periodic Si3N4 slab structure. (Although the functional employed in calculation for the two systems are different, the above results should be sufficient to justify the limited size effect of H14Si4N6 on performing the surface studies in this work.)


Figure 1.   (a) The H14Si4N10 cluster used to simulate the Si3N4 etch front; Si, N, H are indicated by grey, blue, white spheres, respectively. (b) The 1×1×3 Si3N4 periodic slab structure used to compare and verify the sufficiency of H14Si3N10 cluster model of simulating the etch front.
Table 1.   Predicted binding energy between the surface Si atom and the NH2 molecule on the cluster and periodic surface using Gaussian 16 and VASP, respectively.
Eads(NH2)Si4N6 Cluster (a)Periodic Surface(b)
G16(B3LYP)-4.70N/A
VASP(PBE)N/A-4.63
3.2.   Elimination Reaction of Dangling N Atom
Using the Si4N6 cluster, we studied possible N-elimination reaction steps in fluorocarbon/oxygen plasma. We focused on the reaction incorporating O and F atoms. The total energy change (ΔE) for each reaction is calculated to verify the feasibility of proposed reaction steps. In order to comprehensively study different surface conditions, we considered both cases of eliminating dangling (N-Si) and bridging (Si-N-Si) N atom; the discussion starts with the dangling N case.
Figure 2 illustrates the chemical reactions of eliminating dangling N atoms; the related ΔE of each reaction is summarized in Table 2. Starting from the interaction between O atom from plasma and H atom of surface NH2 molecule, we calculated the ΔE of the H atom transferring to O atom to create OH molecule. The atomic configurations before and after the H-transfer are illustrated by Fig. 2 (a) and (b), respectively; the transitioning schematic of Fig 2 (a) and (b) illustrates the transfer. The ΔE of first H-transfer is predicted to be 0.42 eV, indicating the reaction is endothermic, which could be attributed to the formation of less stable NH molecule. However, considering that the ΔE is relatively small compared to the ion energy of plasma in normal conditions, this H-transfer reaction is expected to occur on the etch front. In fact, the same rationale is applied to most chemical reactions studied in this work. As being rarely comparable with the ion energy, the ΔE mostly impacts on the relative predominance of studied chemical reactions, instead of the definite feasibility. Being less stable than the first one, the ΔE of the transfer of the second H atom (i.e. the H atom of the surface NH molecule) to the O atom is predicted to be -0.24 eV. The reaction is exothermic and expected to occur more actively than the first H-transfer. Therefore, based on the calculations, the elimination of both H atoms of surface NH2 molecule in fluorocarbon/oxygen plasma by forming OH molecule is feasible. The elimination leads to the formation of dangling N atom on Si atom, as shown in Fig 2 (b). Based on the presence of dangling N atom, we studied the formation of the NO molecule. Figure 2 (c) illustrates the additive reaction of O atom on the dangling N atom to create end-on NO molecule with the N atom bound to a Si atom. The ΔE of the reaction is calculated to be -5.16 eV, implying that the additive reaction can be active. The distance between N atom of NO molecule and Si atom is 1.78 Å, close to the Si-N bond (1.76 Å) of H14Si4N6 cluster, indicating a chemisorption of NO molecule.
In order to study how NO formation could lead to the N-elimination, we studied the desorption mechanism of NO molecule from the cluster surface. Figure 2 (d) illustrates the cluster structure with the end-on NO molecule desorbed from the Si site. The binding energy of NO molecule and the 3-fold undercoordinated Si site is calculated to be -1.75 eV, implying that the desorption of the molecule could be slowed down by the strong of Si-NO bond. To consider other possible desorption mechanisms, we studied the substitution reaction of F atom from the plasma for NO molecule. The binding energy of F atom and 3-fold undercoordinated Si atom (see Fig. 2 (f)) is calculated to be -6.68 eV, making the ΔE of F-substitution reaction for NO molecule -4.77 eV. The substitution reaction could be active, effectively facilitating the N-elimination in the fluorocarbon/oxygen plasma. In addition to the substitution reaction, the F additive reaction on N atom to form FNO molecule is also considered. Figure 2 (e) illustrates the configuration of monodentate FNO on the Si site. The ΔE of the additive reaction is calculated to be -2.46 eV, suggesting the formation of FNO molecule on the etch front is possible. As a saturated molecule, FNO was expected to desorb from Si site more easily than NO; interestingly, the calculation result suggests otherwise. The adsorption energy of FNO molecule on Si site is predicted to be -1.89 eV, only 0.02 eV less negative than NO, indicating that the creation of FNO molecule may not directly affect the N-elimination; nevertheless, the indirect enhancement is possible. In that case, FNO molecule could serve as an intermediate byproduct which further decompose into a F atom and an NO molecule. As F atom binds with Si atom, the reaction can lead to the desorption of the NO molecule, thus attaining N-elimination.


Figure 2.   Atomic configurations of adsorbed specices invovled in the dangling N-elimination: (a) end-on NH2; (b) dangling N atom; (c) end-on NO molecule; (d) vacancy; (e) monodentate FNO molecule; (f) F atom.
Table 2.   Total energy change (ΔE) of chemical reactions occurring during the dangling N-elimination, involving adsorbed (-) and gas species (g). The reactions considered are illustrated by the indicated steps in Figure 2.
ReactionSteps in Figure 2ΔE (eV)
O (g) + -NH2 → OH + -NHa → b0.42
O (g) + -NH → OH + -Na → b-0.24
O (g) + -N → -NOend-onb → c-5.16
-NOend-on → NO (g)c → d1.91
F (g) + -NOend-on → -FNOmonodentatec → e-2.46
-FNOmonodentate → FNO (g)e1.89
F (g) → -Fd → f-6.68
-FNOmonodentate → -F + NO (g)e → f-2.31
3.3.   Elimination Reaction of Bridging N Atom
Figure 3 illustrates the chemical reaction steps of elimination of bridging N atom; the related ΔE of each reaction is summarized in Table 3. Same as the dangling N case, this part of study started with the elimination of H atom. We considered the reaction of O atom interacting with H atom of bidentate NH molecule, scavenging the H to form an OH molecule. The cluster structures before and after the reaction is illustrated in Fig 3 (a) to (b). The ΔE of reaction is predicted to be 0.01 eV, suggesting the reaction being almost thermal neutral. Consequently, the presence of undercoordinated bridging N atom, as shown in Fig 3 (b), in the etch process is expected. Next, we studied the O-additive reaction on the bridging N atom. Figure 3 (c1) illustrates the cluster structure having the bidentate NO molecule with the N atom bound to 2 Si atoms. The ΔE of the O-additive reaction is calculated to be -3.64 eV, indicating that the reaction would also be reasonably active, just like the formation of end-on NO molecule. It is worth mentioning that the formation of the bidentate NO is less favorable (by 1.97 eV) than the end-on NO formation. This could be due to the presence of an unpaired electron of the bidentate NO configuration, which has a doublet electronic state. We further found that the bidentate NO is actually not the optimal geometry. This bidentate adsorption configuration is less stable than the bridging NO molecule, as shown in Fig. 3 (c2). The total energy of bridging configuration is predicted to be 0.17 eV more favorable than the bidentate. Therefore, the newly formed NO molecule is expected to turn parallel to the surface, attaining a bridging adsorption configuration. The bridging NO molecule binds to 2 separated Si atoms with O and N atoms. Next, we studied the desorption of the bridging NO molecule. Figure 3 (d) illustrates the cluster structure with the bridging NO molecule desorbed. The binding energy of the bridging NO molecule to the cluster is predicted to be -3.82 eV, more favorable than the end-on NO case by 1.91 eV in magnitude. The more favorable binding is reasonable as the desorption of bridging NO molecule breaks 2 bonds to the cluster (Si-O and Si-N). Learning from the earlier case of end-on NO desorption, we considered the formation of FNO as conducive to NO removal through the introducing the F atom. This F atom subsequently displaces the NO unit via the Si-F bond formation. Figure 3 (e) illustrates the bridging FNO molecule with N and O atom bound to 2 separated Si atoms. The additive energy of F atom is predicted to be -2.76 eV, suggesting the reaction to be close or even slightly more active than the monodentate FNO case. On top of that, it is worth noting that the turning of NO molecule from bidentate to bridging could make the N atom more accessible to F atoms, increasing the probability of forming FNO molecule. Following the bridging FNO formation, we considered the substitution of Si-F for Si-O with F atom coming from FNO molecule, as shown in Fig. 3 (f). The ΔE of the substitution reaction is predicted to be -1.43 eV. Without considering the reaction barrier, the energetically favored reaction indicates that F atom could transfer from FNO molecule to Si atom, facilitating the transition of bridging to end-on NO molecule. This reaction can largely enhance the elimination of NO molecule due to the smaller binding energy of end-on NO as discussed earlier. Overall speaking, the elimination of bridging N atom is achieved via three steps: 1) bridging NO formation; 2) transiting from bridging to end-on NO; 3) desorption of the end-on NO.


Figure 3.   Atomic configurations of adsorbed specices invovled in the bridging N-elimination: (a) bidentate NH; (b) bridigng N atom; (c1) bidentate NO molecule; (c2) bridging NO molecule; (d) vacancy; (e) bridging FNO molecule; (f) F atom.
Table 3.   Total energy change (ΔE) of chemical reactions occurring during the bridging N-elimination, involving adsorbed (-) and gas species (g). The reactions considered are illustrated by the indicated steps in Figure 3.
ReactionSteps in Figure 3ΔE (eV)
O + -NHbidentate → OH + -Na → b0.01
O + -N → -NObidentateb → c1-3.64
-NObidentate → -NObridgingc1 → c2-0.17
-NObridging → NO (g)c2 → d3.82
F + -NObridging → -FNOc2 → e-2.76
-FNO → -F + -NOend-one → f-1.43
3.4.   Formation of SiOxNx
As the calculations point out that F atoms can enhance the N-elimination in fluorocarbon/oxygen plasma by substituting for the Si-N bond, without losing the generality, O atoms could provide the same kind of assistance as long as the Si-O binding energy is sufficient. Figure 4 (a) illustrates the additive reaction of O atom on the 3-fold undercoordinated Si atom. The binding energy of Si-Odangling bond is calculated to be -4.9 eV, less favorable than the Si-F bond by 1.78 eV. The ΔE of -NOend-on + O (g) → NO (g) + -O is predicted to be -2.99 eV, indicating the reaction is still exothermic although less significant than -NOend-on + F (g) → NO (g) + -F. Based on that, both F and O atoms could facilitate the N-elimination through substituting Si-F or Si-O bonds for Si-N. Once the O-substitution reaction occurs, the formation of SiOxNy could follow. Beside the binding energies, we also consider the bond order of F and O atoms. Figure 4 (b) illustrates the O-additive reaction on the bridging site between 2 Si atoms. The additive energy is predicted to be -9.4 eV. This energy almost doubles the dangling case, since 2 Si-O bonds are formed in the reaction. Considering the fact that F atom can only have coordination number equal to 1, O atoms clearly have the advantage over competing with F for the bridging site; whereby further enhance the SiOxNy formation. The results could suggest the correlation between more significant SiOxNy formation at higher O2 percentage. That could be further related to why SiO2 and Si3N4 tend to have similar E/R when O2 percentage is high in the experiments. On the other side, the stronger Si-F bond could contribute to the positive correlation between E/R and CF4 percentage. In fact, the extra enhancement of E/R from the synergy of NO and F has also been studied and reported in previous works[21].


Figure 4.   Adsorption reaction and related energy (eV) of O atom on (a) top of 3-fold coordinated Si site and (b) bridge site between two 3-fold coordinated Si atoms.
3.5.   Formation/elimination of NO2
We also considered the formation and desorption of NO2 molecule. Figure 5 illustrated the O-additive reaction on the end-on NO molecule to form a monodentate NO2 molecule on Si site. The ΔE is predicted to be -4.34 eV, implying that a fair amount of NO2 may present on the Si3N4 etch front in fluorocarbon/oxygen plasma. The desorption energy of monodentate NO2 molecule is predicted to be 2.94 eV, indicating the desorption to be endothermic. The desorption of monodentate NO2 molecule is thus less favorable than the end-on NO molecule by ~1.03 eV. The Mulliken atomic charge analysis shows the N atom of NO2 molecule carries ~ -0.7 e, suggesting that the stable adsorption configuration could be due to the charge transfer from the cluster to the N atom of NO2 molecule, forming a stable Si-NO2 bond. This finding could further justify why continuously increasing the concentration of O2 in CF4/O2 flow slows down the E/R.


Figure 5.   Adsorption reaction of O atom on the N atom of end-on adsorbed NO molecule to form monodentate NO2 molecule. The adsorption energies of O atom in eV is indicated. The desorption energy of NO2 molecule in eV is indicated in italic font.
4.   Conclusion
In this study, we employed DFT calculations to investigate the elimination of N atom from Si3N4 surface in fluorocarbon/oxygen plasma, with particular attention on the formation and desorption of the NO molecule. Based on the total energy change of reactions (ΔE), the contribution of F and O atom to the N-elimination was identified. The competition between F and O atoms over Si sites is also analyzed, followed by the discussion of the formation of oxynitride.
The etch front of Si3N4 structure is simulated using the H14Si4N6 model cluster. Before moving on to the etching mechanisms, we verified the sufficiency of the cluster on recapturing the adsorption reactions on Si3N4 surface structure. The verification was done by comparing the Si-N bond distance and adsorption energy of NH2 molecule on the cluster and periodic slab structure predicted using Gaussian16 and VASP, respectively. The negligible discrepancy (~ 0.07 eV) between the two calculations confirms the adequacy of H14Si4N6 cluster structure for the following studies. First, we confirmed that the formation of dangling N atom on Si site occurs through O atom scavenging H atom of end-on NH2 to create the OH molecule. Following that, the additive reaction of O atom on N atom to form end-on NO molecule is predicted. The desorption of NO molecule is then predicted to be endothermic. Considering the Si-F binding energy is high, we proposed that the elimination of NO occurs with the help from F-substitution for Si-NO bond. The desorption reaction involves F-substitution is predicted to be exothermic. The additive reaction of F atom on NO molecule to form monodentate FNO molecule on Si site is also confirmed to be exothermic. Considering the desorption of FNO is similar to NO molecule, we proposed that the FNO molecule could further decompose into F atom and NO molecule. The decomposition may help NO desorption as the Si-F substitutes the Si-NO bond. In the above cases of NO desorption, N-elimination is achieved.
Following the similar route, we also studied the elimination of bridging N atom (connected to 2 Si atoms). The presence of 2-fold undercoordinated (bridging) N atom is confirmed after O atom scavenges H atom from the bidentate NH molecule. Following that, we predicted the formation of bidentate NO molecule via O additive reaction to be exothermic. Moreover, the bidentate NO molecule would undergo a re-orientation to become the bridging NO molecule (Si-NO-Si). This changing of orientation could increase the exposure of NO molecule to the plasma, increasing the probability of O atom interacting with the N atom to form FNO molecule. We then calculated the additive energy of F atom on the bridging NO molecule to form bridging FNO molecule; the reaction is predicted to be exothermic. As the substitution reaction of Si-F for Si-N is also predicted to be exothermic for the bridging FNO molecule, it is possible that the F atom is subsequently transferred to the Si site. This addition-then-substitution reaction results in a transition of NO molecule from bridging to end-on adsorption, enhancing the desorption of NO molecule as well as N-elimination.
Based on the prediction of Si-O binding energies, the occurrence of O-substitution for Si-N during above N-elimination reaction steps is suggested. On top of that, the capability of O atom forming bonds with multiple Si atoms provides the further advantage over competing with F for Si sites. Combined above reasons, the formation of SiOxNy on the etch front of Si3N4 in fluorocarbon/oxygen is highly possible. This conclusion corroborates the experimental finding that increasing O2 flow in fluorocarbon/oxygen feed gas drives the E/R of Si3N4 closer to SiO2. Finally, we also considered the formation and desorption of the monodentate NO2 molecule on Si site. While the additive reaction of O atom on NO molecule is exothermic, the desorption is highly endothermic; the result qualitatively supports the experimental trend that continuously increasing O2 percentage leads to lower E/R of Si3N4.
By using DFT calculation, this work intends to shed some light on the etching mechanism of Si3N4 in fluorocarbon/oxygen plasma, especially the role played by Oatom. In addition, the feasibility of SiOxNy formation is suggested. A similar study for SiO2 etch in fluorocarbon/oxygen plasma is on-going to provide a comprehensive insights of ONON etch. We hope that these findings will provide a foundation for future study on attaining uniform and fast etch for ONON stack through the approach of modifying different surfaces into similar structure.
Acknowledgments
The authors would like to thank to Aelan Mosden, Peter Biolsi and Alexander Oscilowski from TEL Technology Center, America, LLC for their continuous support and moral encouragement
[1] Y. Horiike, and M. Shibagaki, “A New Chemical Dry Etching,” Jpn. J. Appl. Phys.,15 , 13 (1976).
[2] F. H. M. Sanders, and J. Dieleman, “Selective Isotropic Dry Etching of Si3 N 4 over SiO2” H. J. B. Peters and J. A. M. Sanders, J. Electrochem. Soc.,11 , 2559 (1982).
[3] Y. Horiike, “Applications of Plkasma Processes to VLSI Technology,” edited by T. Sugano (Wiley, New York, 1985), pp. 137-138.
[4] M. G. Blain, T. L. Meisenheimer, and J. E. Stevens, “Role of nitrogen in the downstream etching of silicon nitride,” J. Vac. Sci. Technol. A, 14(4) , 2151 (1996).
[5] B. E. E. Kastenmeier, P. J. Matsuo, J. J. Beulens, and G. S. Oehrlein, “Chemical dry etching of silicon nitride and silicon dioxide using CF4/O2/N2 gas mixtures.” J. Vac. Sci. Technol. A,14 , 2802 (1996).
[6] H. K. Lee, K. S. Chung, and J. S. Yu, “Selective Etching of Thick Si3N4, SiO2 and Si by Using CF4/O2 and C2F6 Gases with or without O2 or Ar Addition,” J. Korean Phys. Soc.,54 , 1816 (2009).
[7] A. D. Becke, “Density-functional exchange-energy approximation with correct asymptotic behavior,” Phys. Rev. A,38 , 3098 (1988).
[8] A. D. Becke, “A new mixing of Hartree–Fock and local density‐functional theories,” J. Chem. Phys.,98 , 1372 (1993).
[9] C. Lee, W. Yang, and R. G. Parr, “Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density,” Phys. Rev. B,37 , 785 (1988).
[10] T. Clark, J. Chandrasekhar, G. W. Spitznagel, and P. v. R. Schleyer, “Efficient diffuse function‐augmented basis sets for anion calculations. III.* The 3‐21+G basis set for first‐row elements, Li–F,” J. Comp. Chem., 4 , 294 (1983).
[11] M. J. Frisch, J. A. Pople, and J. S. Binkley, “Self‐consistent molecular orbital methods 25. Supplementary functions for Gaussian basis sets,” J. Chem. Phys.,80 , 3265 (1984).
[12] M. J. Frisch et al, Gaussian 16, Revision A.03, Gaussian, Inc., Wallingford CT, 2016.
[13] J.P. Perdew, K. Burke, and M. Ernzerhof, “Generalized Gradient Approximation Made Simple,” Phys. Rev. Lett.,77 , 3865 (1996).
[14] J.P. Perdew, K. Burke, and M. Ernzerhof, “Generalized Gradient Approximation Made Simple,” Phys. Rev. Lett.,78 , 1396 (1997).
[15] G. Kresse, and J. Furthmüller, “Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set,” Comput. Mat. Sci.,6 , 15 (1996).
[16] G. Kresse, and J. Furthmüller, “Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set,” Phys. Rev. B,54 , 11169 (1996).
[17] P.E. Blochl, “Projector augmented-wave method,” Phys. Rev. B,50 , 17953 (1994).
[18] G. Kresse, and D. Joubert, “From ultrasoft pseudopotentials to the projector augmented-wave method,” Phys. Rev. B,59 , 1758 (1999).
[19] H.J. Monkhorst, and J.D. Pack, “Special points for Brillouin-zone integrations,” Phys. Rev. B,13 , 5188 (1976).
[20] E. Flage–Larsen, O. M. Løvvik, C. M. Fang and G. Kresse, “β-Si3N4(0001)/Si(111) interface: Phosphorus defects, valence band offsets,and their role of passivating the interface states,” Phys. Rev. B, 88 , 165310 (2013).
[21] Y. Barsukov, V. Volynetsa, S. Lee, G. Kim, et al., “Role of NO in highly selective SiN/SiO2 and SiN/Si etching with NF3/O2 remote plasma: Experiment and simulation,” J. Vac. Sci. Technol. A,35 , 061310 (2017).
Article and author information
Yu-Hao Tsai
Yu-Hao.Tsai@us.tel.com
Du Zhang
Mingmei Wang
Publication records
Published: Sept. 30, 2018 (Versions6
References
Journal of Microelectronic Manufacturing