Selective isotropic etching of SiO2 over Si3N4 using NF3/H2 remote plasma and methanol vapor

In this study, an isotropic etching process of SiO2 selective to Si3N4 using NF3/H2/methanol chemistry was investigated. HF was formed using a NF3/H2 remote plasma, and in order to remove the F radicals, which induces spontaneous etching of Si-base material, methanol was injected outside the plasma discharge region. Through this process, etch products were formed on the surface of SiO2, and then the (NH4)2SiF6 was removed by following heating process. When the H and F radicals were abundant, the highest SiO2 etch per cycle (EPC) was obtained. And, the increase of H2 and methanol percentage in the gas chemistry increased the etch selectivity by decreasing the F radicals. The etch products such as (NH4)2SiF6 were formed on the surfaces of SiO2 and Si3N4 during the reaction step and no noticeable spontaneous etching by formation of SiF4 was observed. By optimized conditions, the etch selectivity of SiO2 over Si3N4 and poly Si higher than 50 and 20, respectively, was obtained while having SiO2 EPC of ~ 13 nm/cycle. It is believed that the cyclic process using NF3/H2 remote plasma and methanol followed by heating can be applied to the selective isotropic SiO2 etching of next generation 3D device fabrication.

SiO 2 films have been widely used in Si-based semiconductor devices together with Si 3 N 4 films due to their excellent interfacial properties with Si, excellent dielectric properties, and easy processing 1 . These materials will be continuously used in next generation 3D semiconductor devices such as gate all around-field effect transistor (GAA-FET), and 3D memory devices, and, for these devices, highly selective isotropic etching processes in addition to highly selective anisotropic etching processes are required 2,3 . Currently, for selective isotropic etching of SiO 2 and Si 3 N 4 , wet etching processes using hydrofluoric acid (for etching of SiO 2 ) and phosphoric acid (for etching of Si 3 N 4 ) are used 4,5 . However, wet etching processes may not be applicable for fabrication of next-generation devices of 10 nm or less due to the issues such as pattern leaning and collapsing caused by unbalanced capillary forces during drying of rinse liquid. In addition, since the etch rate is fast, it is difficult to control the process, and the problems might occur such as surface chemical damage by the etchant during etch process [6][7][8] . Therefore, it is necessary to develop selective isotropic dry etching processes that may solve these problems.
For a long time, gas-based SiO 2 etching processes have been applied to remove native SiO 2 layer formed on silicon wafer and thin SiO 2 layer at the bottom of contact hole and, to thin or remove SiO 2 hardmask layer. For isotropic dry etching of SiO 2 selective to Si 3 N 4 and/or Si, a vapor process using HF vapor as HF source [9][10][11][12][13][14] and plasma processes forming HF by discharging F-based gases such as NF 3 , SF 6 , CF 4 , OF 2 , etc. and H-based gases such as NH 3 , H 2 , H 2 O, etc. have been studied [15][16][17][18][19][20][21] . In these processes, since SiO 2 is hardly etched by anhydrous HF alone, a solvent that ionizes HF to react with SiO 2 is required, and the solvents such as H 2 O, alcohol, NH 3 , etc. have been used. By using H 2 O or alcohol as a solvent, SiO 2 reacts with ionized HF (HF 2 -, HF, H + , and F -) in the solvent and forms volatile SiF 4 4,5,9,13 . For NH 3 as a solvent, SiO 2 is transformed into (NH 4 ) 2 SiF 6 salt by the reaction of HF and NH 3 , and the formed (NH 4 ) 2 SiF 6 is decomposed into NH 3 and SiF 4 by heating at a temperature of 100 °C or higher [14][15][16][17][18][19] . Among these, the etching by HF ionization is known to be difficult to control the etch process because the etch rate is fast similar to the wet process, etch rate varies depending on the quality of the SiO 2 film, and additional rinse process may be necessary due to residual etchant 10,11,14 . So, HF/NH 3 vapor process and the NF 3 /NH 3 plasma process are more widely used. However, ammonium salts such as NH 4 F, NH 4 HF 2 formed in the process using NH 3 15-19 can be a source of contamination in the chamber because they exist in the form www.nature.com/scientificreports/ of solid powder at room temperature 22,23 . Therefore, to avoid direct reaction between NH 3 and HF in gaseous or plasma state, alternate SiO 2 dry etching processes have been investigated through reaction with NF 3 gas after inducing NH 3 formation through N 2 /H 2 plasma 24 or through the formation of (NH 4 ) 2 SiF 6 on SiO 2 by sequential supply of NH 3 gas after adsorption of HF from HF vapor or from plasmas generated with SF 6 /H 2 or NF 3 /H 2 25,26 . The above processes still use NH 3 gas or NH 3 formed through plasma which might generate particles, therefore, in this study, selective isotropic SiO 2 dry etching process not generating noticeable NH 3 , by using H 2 /NF 3 plasma as HF source and methanol (CH 3 OH) vapor as a solvent, has been investigated. In order to minimize the generation of F radical during dissociation of NF 3 in the plasma which induces spontaneous etching of Si-base material 27 , H 2 /NF 3 gases were supplied to the remote plasma area and methanol were injected to the processing area located outside the plasma discharge region. And, dual grids separating the discharge area and the processing area were located to block the ions from the discharge area. Through a cyclic process, saturated etch products such as (NH 4 ) 2 SiF 6 were formed on the surface of SiO 2 , and then the etch products were removed by heating for precise control of etched SiO 2 thickness. The SiO 2 etch characteristics were measured according to process conditions such as gas ratio, pressure, temperature, and process time, and the etch selectivities over Si 3 N 4 and poly Si were compared. The etch mechanism was investigated by analyzing plasmas, surface composition, etc.

Methods
200 nm thick SiO 2 deposited by low pressure chemical vapor deposition (LPCVD) on silicon wafers was used to measure the etch depth during the dry etching. 50 nm thick Si 3 N 4 deposited on silicon wafers and 50 nm undoped poly silicon deposited on silicon oxide by LPCVD were used to measure the etch selectivity over SiO 2 . In addition, trench patterned silicon wafers (80 nm top critical dimension (CD), 150 nm pitch, and 440 nm depth) sequentially deposited with 5 nm thick LPCVD-Si 3 N 4 and 10 nm thick LPCVD-SiO 2 sequentially were used to investigate the etching effect on patterned samples. The quality of the Si 3 N 4 used in this experiment was examined by wet etching in a HF solution (HF:deionized water = 1:100, RT) and the wet etch rates of Si 3 N 4 and SiO 2 used in this experiment are shown in Supplementary Fig. S1. As shown in Fig. S1, the etch selectivity of SiO 2 over Si 3 N 4 was ~ 50.
The isotropic dry etching system used in this study is shown in Fig. 1a,b. NF 3 /H 2 gas mixture was supplied to the plasma area, and the plasma was generated by an inductively coupled plasma (ICP) source operated with 13.56 MHz radio frequency (RF) power. Two anodized aluminum grids with crossed holes were located below the plasma source to block ions and flow radicals only to the processing area, and methanol gas was supplied to the substrate through a shower ring located between the substrate and grids. The substrate temperature was controlled using a bath circulator (JEIO TECH, HTRC-10) and the substrate temperature was measured using a thermocouple in contact with the wafer. The base pressure of chamber was maintained at ~ 1 × 10 -4 Torr using a dry pump (ALCATEL ADIXEN, ADS 602P).
The etching process was consisted of a reaction step using remote plasma and a removal step using heating as shown in Fig. 1c. During the reaction step, reacted etch products were formed on the surface of SiO 2 through radicals supplied from NF 3 /H 2 plasmas through the grids and methanol vapor injected through the shower ring and, during the removal step, the formed etch products were removed by heating on a hot plate. For the reaction step, NF 3 /H 2 plasma was formed with 500 W of RF power to the ICP source while maintaining the total operating pressure at 200 mTorr. Pressure ratios between NF 3 and H 2 gases in the plasma area and those between NF 3 /H 2 gas mixture and methanol vapor were varied while keeping 200 mTorr. The substrate temperature was varied from 0 to 60 °C. For the removal step, formed etch products were removed by heating for 10 min. immediately after the plasma process on a hot plate (MTOPS, HSD180) preheated to 150 °C in a fume hood located next to the etcher. Also, the temperature and moisture of the etch room were controlled by using an air conditioning system.
Etch thicknesses of SiO 2 , Si 3 N 4 , and poly Si were measured using an ellipsometer (Nano-View SE MG-1000). To understand the etching mechanism, an optical fiber was connected to a quartz window located on the top of the grid as shown in Fig. 1a, and gas species generated in the plasma were analyzed using optical emission spectroscopy (OES; Avantes, AvaSpec-3648). Radicals and recombined gas species outside the discharge region were analyzed through a residual gas analyzer (RGA; SRS, RGA200) located at the processing area below the grids. Gas phase Fourier transform-infrared spectroscope (FT-IR; MIDAC corporation, I2001) was connected to the exhaust end of the dry pump and the gases generated during the reaction steps were analyzed at the exhaust area while a 5 × 5 cm 2 sample located on the substrate holder was being processed. In addition, etch products formed on the blank samples and patterned samples were observed through field emission-scanning electron microscope (FE-SEM; HITACHI S-4700). Surface compositions and elemental bonding states of the SiO 2 and Si 3 N 4 after each step were analyzed through X-ray photoelectron spectroscopy (XPS; Thermo VG, MultiLab 2000, Al Kα source) after peak shift based on C 1 s (~ 285 eV) and through Fourier transform-infrared spectroscopy (FT-IR; Thermo Electron Nicolet 5700), respectively. Also, the surface roughness for each step was analyzed using an atomic force microscope (AFM; Park System XE-100).

Results and discussion
For the cyclic etching, SiO 2 , Si 3 N 4 , and poly Si were exposed to NF 3 /H 2 remote plasma and methanol vapor for the formation of etch products (reaction step), and the etch products formed on the materials surface after the reaction were removed by heating at 150 °C for 10 min (removal step). The etch depth/cycle (EPC) of SiO 2 , Si 3 N 4 , and poly Si and their etch selectivities were measured as functions of various reaction step parameters and the results are shown in Fig. 2. Figure 2a shows the EPCs and etch selectivities measured a function of NF 3 /H 2 gas ratio. The partial pressure ratios between NF 3 /H 2 and methanol were kept at 1:1 for total operating pressure of 200 mTorr. The RF power, substrate temperature, and process time were maintained at 500 W, 20 °C, and 20 min, respectively. As shown in Fig. 2a www.nature.com/scientificreports/ etch selectivity was increasing with increase of H 2 in the NF 3 :H 2 . While keeping the ratio between NF 3 :H 2 ~ 1:3, the partial pressure ratio between NF 3 /H 2 and methanol was also varied and the results are shown in Fig. 2b. Other process conditions were the same as those in Fig. 2a. (The EPCs as a function of the different partial pressure ratios between NF 3 /H 2 and methanol for different ratios of NF 3 :H 2 are also shown in Supplementary  Fig. S2). The increase of partial pressure ratios of methanol from 1:0 (NF 3 /H 2 only) to 1:5 decreased etch rates of all materials but increased the etch selectivity of SiO 2 over Si 3 N 4 higher than 50 from the gas ratio of 1:5. With the ratio of NF 3 :H 2 ~ 1:3 and the ratio of (NF 3 /H 2 ): methanol ~ 1:3, the substrate temperature was varied from 0 to 60 °C while maintaining the other process conditions the same as Fig. 2a, and the etch results are shown in Fig. 2c. As shown in Fig. 2c, the decrease of substrate temperature increased EPC of SiO 2 possibly due to the increased adsorption on the materials surfaces while keeping the etch selectivity of SiO 2 /Si 3 N 4 higher than 50. When the plasma process time was varied for the optimized condition of the ratio of NF 3 :H 2 ~ 1:3, the ratio of (NF 3 /H 2 ):methanol ~ 1:3, and 0 °C, the SiO 2 EPC was saturated after ~ 10 min while keeping the etch selectivity over Si 3 N 4 higher than 50. In the case of etch selectivity of SiO 2 over poly Si, the highest etch selectivity higher than 20 was also observed at the substrate temperature of 0 °C and the process time of 10 min for the ratio of NF 3 :H 2 ~ 1:3 and the ratio of (NF 3 /H 2 ): methanol ~ 1:3. SiO 2 surface was exposed to the optimized reaction conditions of NF 3 /H 2 (1:3):methanol ~ 1:3, the substrate temperature of 0 °C, and process time of 10 min (reaction step), and the etch products formed on the surface was observed by SEM before and after the reaction, and after the removal by heating at 150 °C for 10 min using a hot plate (removal step), and the results are shown in Fig. 3a-c. As shown in Fig. 3b, after the reaction step, etch products were observed on the SiO 2 surface and, after the removal step, the etch products formed on the SiO 2 surface were appeared to be removed. Tilted SEM images of the experimental results for Fig. 3b,c are shown in Fig. S3. While repeating the reaction and removal, the EPC and total etch depths were measured as a function of etch cycle and the result is shown in Fig. 3d. As shown in Fig. 3d, the EPC of SiO 2 was similarly remaining at ~ 13 nm/cycle, therefore, the total etch depth was linearly increased with cycle number not only for SiO 2 but also for all materials while keeping the SiO 2 etch selectivity over Si 3 N 4 over 50.
To understand the EPC behavior observed in Fig. 2 as functions of various reaction parameters, the plasma characteristics in the plasma area and the species characteristics in reaction area under the grid were investigated  Figure 4a,c show the Ar normalized optical emission intensity ratios of species observed as a function of (a) NF 3 :H 2 gas ratios and (c) NF 3 /H 2 :methanol ratios. As shown in Fig. 4a, with increasing H 2 ratio, the increase of H intensity was observed while, with increasing NF 3 , the increase of F and N-related (NH or N 2 ) peaks were observed due to the increase of related molecules in the gas mixture. Also, as shown in Fig. 4c, the increase of NF 3 /H 2 in NF 3 /H 2 :methanol increased F, H, and N-related peaks while the increase of methanol in NF 3 /H 2 : methanol increased the CO also due to the increase of related molecules in the gas mixture. In addition, even though no hydrogen exists for NF 3 :H 2 = 1:0 in Fig. 4a, H in addition to CO was detected in the plasma area and, as shown in Fig. 4c, even with decrease of H 2 with increasing methanol, H did not decrease linearly possibly due to the formation of H through decomposition of methanol penetrated into the plasma area. Especially, as shown in Fig. 4a, c, the highest HF (or CN, CH but from Fig. 4b, it appears to be more related to HF) peak intensity was observed with the adequate mixture of NF 3 and H 2 possibly due to the highest formation HF from sufficient H and F in the plasma. For RGA measured at the reaction area, masses such as H 2 (2 amu), H 2 O (18 amu), HF (29 amu), CO or N 2 (28 amu), CH 2 OH/CH 3 OH (31/32 amu), CF 3 (69 amu), NF 3 (71 amu), etc. were observed. However, highly reactive H (1 amu) and F (19 amu) radicals observed by OES in the plasma area were hardly measured in the www.nature.com/scientificreports/ reaction area by RGA due to recombination during the gas transportation from the plasma area to the reaction area. Figure 4b,d show the mass peak intensities observed as a function of (b) NF 3 :H 2 gas ratios and (d) NF 3 / H 2 :methanol ratios for the conditions in Fig. 4a,c, respectively. As shown in Fig. 4b,d, with the increase of H 2 in NF 3 :H 2 and with the increase of NF 3 /H 2 in NF 3 /H 2 :methanol, the increase of H 2 was observed in the reaction area by RGA, while the increase of NF 3 in NF 3 :H 2 and the increase of methanol in NF 3 /H 2 :methanol increased CO/N 2 possibly due to the reaction of CH 3 OH with F from NF 3 and dissociation of CH 3 OH. HF molecules were observed by RGA in the reaction area as shown in Fig. 4b,d, and the HF molecules showed the highest peak with the adequate mixture of NF 3 :H 2 at the reaction area and at the ratio of NF 3 /H 2 :methanol = 1:0 similar to the SiO 2 etch results shown in Fig. 2a,b, respectively. Therefore, it is believed that, HF molecules formed by reaction of F with H (from H 2 and decomposed CH 3 OH) in the plasma area and diffused to the reaction area are related to the SiO 2 EPC observed in Fig. 2a,b. While etching SiO 2 with the conditions in Fig. 2a,b, the recombined species and etch products were observed using gas phase FT-IR located at the exhaust area. The species observed by gas phase FT-IR are shown in Fig. 5a for NF 3 :H 2 gas ratios and (b) for NF 3 /H 2 : methanol ratios. Species such as CH 3 OH, CO, CH 4 , NF 3 , NH 3 , and HF were observed at the exhaust area by FT-IR. However, species such as H 2 , N 2 , O 2 etc. could not be observed or measured by using FT-IR because these species do not absorb IR due to no changing dipole moments during the vibration that allow for the absorption of photons. As shown in Fig. 5, the variation of species observed by gas phase FT-IR with NF 3 :H 2 gas ratios and NF 3 /H 2 :methanol ratios was similar to RGA data possibly due to the transport of recombined species without significant reaction from reaction area to the exhaust area. The FT-IR results obtained during the etching of Si 3 N 4 and poly Si (not shown) under the same conditions as those obtained during the SiO 2 etching were the same as Fig. 5a,b, and no SiF 4 or SiH 4 was also observed under all conditions. Therefore, it indicates that, during the reaction step, no noticeable spontaneous etching of SiO 2 , Si 3 N 4 , and poly Si with the reaction gases such as F and H in the reaction area was occurred.
The reaction compounds on the surfaces of SiO 2 and Si 3 N 4 after reaction step and removal step for the different NF 3 :H 2 ratios were observed by SEM and the results are shown in Figs. 6 and 7 for SiO 2 and Si 3 N 4 , respectively. The process conditions are the same as those in Fig. 2a. As shown in Fig. 6, after the reaction step, the reaction products formed on the surface were observed and, for high etch rate condition of NF 3 :H 2 ~ 1:1, the thickest reaction product was observed. And, after the removal step, no noticeable reaction products on the SiO 2 surface could be observed for all etch conditions. In case of Si 3 N 4 , as shown in Fig. 7, the thickest reaction product was also observed at the highest Si 3 N 4 etch condition of NF 3 :H 2 ~ 1:1, and after the removal step, no reaction products were observed on the Si 3 N 4 surface for all etch conditions. Because no etch products such as SiF 4 were observed during the reaction step as observed by gas phase FT-IR while forming the thickest reaction product at the condition of highest etch rate conditions, the reaction products were formed on the surfaces of Si 3 N 4 and SiO 2 during the reaction step without spontaneous etching of Si 3 N 4 and SiO 2 by F radicals and the reaction products were removed by heating during the removal step during the removal step. In fact, even with high F concentration for the condition of NF 3 :H 2 = 1:0 as shown in Fig. 4a, no spontaneous etching of Si 3 N 4 and SiO 2 was observed as shown in Fig. 2a. From the data from OES, RGA, and gas phase FT-IR in Figs. 4 and 5, it is believed that F radicals formed from dissociation in the plasma area react with methanol (CH 3 OH) in the reaction area to form HF according to the formula below 28,29 ; (1) F + CH 3 OH → CH 2 OH + HF (−166.8 kJ/mol)    Fig. 2a, www.nature.com/scientificreports/ Therefore, no spontaneous etching of SiO 2 and Si 3 N 4 is appeared to be observed even with high concentration of F radicals in the plasma area. Some methanol also can participate in the discharge and is decomposed into H and CO, so it could act as H donor for HF formation for F radicals in the plasma area 30,31 . The increase of etch selectivity of SiO 2 over Si 3 N 4 obtained with the increase ratios of methanol in NF 3 /H 2 :methanol is also believed to be related to the formation of HF through the reaction of F with methanol.
To find out the materials characteristics of the reaction products formed on the surfaces of SiO 2 and Si 3 N 4 , the surface characteristics were measured by XPS, FT-IR, and AFM. The process conditions are the same as those in Fig. 3. Figure 8 shows (a, d) atomic composition and narrow scan data of (b, e) N 1 s and (c, f) F 1 s for the surfaces of (a-c) SiO 2 and (d-f) Si 3 N 4 measured by XPS before (reference) and after reaction step, and after removal step with the optimized etch condition (XPS wide scan data of the surfaces of SiO 2 and Si 3 N 4 during the process steps are shown in Supplementary Fig. S7). As shown in Fig. 8a, for SiO 2 , after the reaction, the atomic percentages of Si and O were decreased from 36.1 to 12.7% and from 59.1 to 2.6%, respectively, while showing 54.7% of F after the reaction. The atomic percentage ratio between N and F in the etch product after the reaction was similar to (NH 4 ) 2 SiF 6. In the case of Si 3 N 4 , as shown in Fig. 8d, after the reaction, the atomic percentages of Si and N were slightly decreased from 42.6 to 35% and from 12.8 to 6.9% respectively, while showing 12.8% of F after the reaction due to the etch product formation similar to SiO 2 even though the etch product thickness on Si 3 N 4 was much thinner than that on SiO 2 . However, after the removal step, the surface compositions of both SiO 2 and Si 3 N 4 were almost recovered similar to those of references due to the removal of reaction products formed on the surfaces. Also, as shown in Fig. 8b,c,e,f, after the reaction step, the formation of etch products composed of N-H-Si-F could be identified for both SiO 2 and Si 3 N 4 by forming N-H bonding at 402 eV and Si-F bonding at 685 eV after the reaction step and, after the removal step, the removal of the etch products by removing those peaks. In addition, on the samples surfaces, ~ 10.9, 7.8% of carbon for SiO 2 and Si 3 N 4 and ~ 7.4% of oxygen for Si 3 N 4 were continuously observed possibly due to the contamination and oxidation, respectively, due to the air exposure during the transport to XPS.
The binding states of etch products formed on the surfaces of SiO 2 and Si 3 N 4 after the reaction/removal steps using the optimized conditions of Fig. 3 were also observed using FT-IR and the results are shown in Fig. 9a,b, respectively. As shown in Fig. 9a, for SiO 2 , the absorption peaks such as N-H stretching (3330 cm −1 ), N-H bending (1454 cm -1 ) related to NH 4 + bonding, and SiF 6 2− bonding (717 cm −1 , 478 cm −1 ) were observed after the  www.nature.com/scientificreports/ reaction and, by comparing the XPS results in Fig. 8, it could be concluded that the etch products formed on the SiO 2 surface after the reaction were related to (NH 4 ) 2 SiF 6 . However, in case of Si 3 N 4 , as shown in Fig. 9b, possibly due to very thin etch product formation, no noticeable absorption peaks such as Si-F bonding related to etch products could be observed. In addition, after the removal steps, as shown in Fig. 9a,b, no absorption peaks could be also observed due to the removal of the etch product formed on the surfaces of both SiO 2 and Si 3 N 4 . The XPS and FT-IR were also measured for poly Si and the results are shown in Supplementary Fig. S8. From the above observations, it is believed that the alkaline salt etch products such as (NH 4 ) 2 SiF 6 are formed by following reactions; In the above reactions, the formation of H 2 O for SiO 2 etching and the formation of NH 3 for Si 3 N 4 etching can be important and the differences in etch rates, that is, differences in the etch product thickness could be related to the differences in the free energy of formation of H 2 O (− 265 kJ/mol) and NH 3 (− 33.2 kJ/mol).
For the optimized conditions in Fig. 3, the change of surface roughness values during the each step of cyclic processing was observed using AFM and the RMS surface roughness values observed for SiO 2 , Si 3 N 4 , and poly Si are shown in Fig. 10a up to 5 cycles of etching and the 2D AFM surface roughness images of references, after reaction step, and after removal step for SiO 2 and Si 3 N 4 during 1st cycle of etching are shown in Fig. 10b-g. As shown in Fig. 10a,b-d, in the case of SiO 2 , after the reaction, the RMS surface roughness was significantly www.nature.com/scientificreports/ increased from 0.77 to 2.86 nm due to the formation of etch product such as (NH 4 ) 2 SiF 6 . However, after the removal step, the RMS surface roughness was decreased to ~ 1.0 nm, and, even after 5 cycles of etching, the RMS surface roughness remained similar to that of reference. In the case of Si 3 N 4 , as shown in Fig. 10a,e-g, due to the very thin etch product formation during the reaction step, only slight increase of surface roughness from 0.22 to 0.42 was observed during the reaction step and, after the removal step, the RMS surface roughness was also decreased similar to that of the reference and no significant change was observed up to 5 cycles of etching.
In the case of poly Si, possibly due to no formation of etch product similar to (NH 4 ) 2 SiF 6 during the reaction step, no significant change in RMS surface roughness was observed during the each step of etching. Therefore, on the surfaces of SiO 2 , Si 3 N 4 , and poly Si, no noticeable physical damage was observed by the cyclic etching. The optimized SiO 2 cyclic etching selective to Si 3 N 4 was applied to a patterned sample and the results are shown in Fig. 11. Figure 11a shows the schematic drawing of a patterned Si sample having the aspect ratio of ~ 10 and sequentially deposited with 5 nm thick Si 3 N 4 and 10 nm thick SiO 2 . Actual SEM images of Si pattern after deposition of 5 nm thick Si 3 N 4 and after following deposition of 10 nm thick SiO 2 are shown in Fig. 11c,d, respectively. After the deposition of 5 nm thick Si 3 N 4 , the top/bottom CDs were ~ 60 nm/105 nm, and after the following deposition of 10 nm thick SiO 2 (reference), the top/bottom CDs were changed to ~ 80 nm/115 nm. The Si trench pattern sample deposited with Si 3 N 4 /SiO 2 was etched using the optimized condition in Fig. 3 and Fig. 11e,f show the SEM images after the reaction step and after the following removal step, respectively. As shown in Fig. 11e, after the reaction, due to the formation of reaction product, the top/bottom CDs were increased to ~ 100/125 nm, and the reaction product thickness was thicker on the top area (the byproduct thickness at the top of the trench was ~ 27.85 nm and that at the bottom of the trench was ~ 12.67 nm as shown in Fig. 11e). As shown in Fig. 11f, after the removal step, the etch product appeared to be removed and, the top/bottom CDs were decreased to ~ 60/111 nm. Due to the thinner etch product formation at the trench bottom area, three cycles were required to remove all SiO 2 deposited on the Si pattern sample as shown in Fig. 11g,h. The measured top/bottom CDs are shown in Fig. 11b and, after the 3 cycles, the top/bottom CDs were the same as those of Si pattern deposited with 5 nm thick Si 3 N 4 , therefore, isotropic SiO 2 etching selective to Si 3 N 4 could be verified in a high aspect ratio patterned sample.

Conclusions
In this study, isotropic dry etching of SiO 2 selective to Si 3 N 4 and poly Si was carried out using a two-step cyclic process composed of the etch products formation on SiO 2 surface by using NF 3 /H 2 remote plasma and methanol vapor, and the thermal desorption of the etch products by heat treatment. By using the cyclic process, ~ 13 nm/ cycle of SiO 2 etch per cycle (EPC) with the etch selectivity over Si 3 N 4 higher than 50 for Si 3 N 4 and that over poly Si higher than 20 could be obtained. The highest EPCs of SiO 2 and Si 3 N 4 were obtained when both H and F are abundant and the higher etch selectivity of SiO 2 over Si 3 N 4 could be obtained when H is richer than F in the system. Especially, methanol acted as H donor for HF formation for F radicals in the plasma area, therefore, the increase of methanol in the gas mixture increased the etch selectivity of SiO 2 over Si 3 N 4 even though the EPC of SiO 2 decreased with increasing the methanol percentage. The etching of both SiO 2 and Si 3 N 4 was processed by formation and removal of etch products such as (NH 4 ) 2 SiF 6 not by forming volatile SiF 4 . When a silicon nanoscale trench patten sequentially deposited with Si 3 N 4 and SiO 2 was etched by the isotropic etching method, only SiO 2 was etched on the nanoscale trench pattern by exposing silicon pattern surface deposited with Si 3 N 4 . Therefore, it is believed that this process can be applied for the selective isotropic etching of SiO 2 for the next generation 3D devices, etc. Especially, by the formation of HF in the plasma without using the NH 3 for the isotropic etching of SiO 2 , the formation of particles caused by NH 3 is expected to be minimized.