The contamination of SF6 decomposition products produces particle on wafer and suffers yield loss. At the first, it is necessary to know wafer structure and plasma etching condition.
Figure 3.2 shows wafer structure, the isotropic etching to define poly silicon depth on this step. The plasma chemistry contacts poly silicon film and nitride film at the same time, but only removes poly silicon film and keeps depth to target. The SF6 is popular gas to etch poly silicon with isotropic etch in DPS poly etcher. The recipe has two steps:
First step – Break-through CF4/source power/bias power for remove native oxide on top of poly film
Second step – Main etch SF6/Low source power for depth control.
Figure 3.3 in this step, appeared semi-translucent hexagonal defects on the wafer by KLA inspected. If put the high level defect wafer in cassette box for 2, 24, 48 hours; the defect level is still high on the wafer. But take the high level defect wafer out cassette box for 2 hours, the defect level is down to half quickly. Therefore, the part of semi-translucent hexagonal defect is volatile in air sometime and remains in box. We have tried to remove they with popular clean skills, can reduce defect counts but can’t remove completely.
(a) SPM and APM clean: “SPM” means sulfuric acid (H2SO4) and (H2O2) on ratio 7:1.
“APM” means ammonia (NH4OH) and (H2O2) on ratio 1:4. SPM used to dehydrate the organic particle (like photo-resist) and APM used to remove particle on silicon or oxide surface. Figure 3.4 shows the efficiency of SPM and APM. The left lot – defect counts reduce from 1253 to 25 at the 1st time clean. The right lot – defect counts reduce from 21762 to 300 at the 1st time clean and to 21 at the 2nd time clean. Figure 3.5 shows another experiment.
(b) Correlation between defect and chamber condition: One, chamber idle time test – chamber run intermittent for different idle time, the performance didn’t improve. Two, checking defect level before and after preventive maintenance (wet clean) – the defect level didn’t change at all. Three, implementing wafer season between production – using bare silicon wafer season between every wafer production, defect level was still high. Four, defect v.s. cbamber parts – checking the usage of chamber parts; liked ceramic dome, focus ring, gas distribution plate (GDP); still wasn’t any correlation.
(c) Couldn’t see hexagonal defect on daily particle monitor wafers. In order to monitor particle level of chamber, it is known to test oxide wafer by production recipe in IC’s manufacturing. From daily defect monitor data, couldn’t see hexagonal defect on daily particle monitor wafers.
(d) Comparison poly etching by different recipes on DPS chambers, the serious semi-translucent hexagonal defect most appeared on low source power recipe. The high source power or Ar gas injected recipe only appeared little hexagonal defect, the most of time couldn’t see any hexagonal defect on wafers; even film structures were
same.
(e) Add N2 plasma treatment in the end of poly etching: Figure 3.6 shows the result of adding N2 plasma treatment in the end of poly etching. The semi-translucent hexagonal defects still remain on wafer.
(f) Add O2 plasma treatment in the end of poly etching:
5mT/50 O2/150W/5 sec
Figure 3.7 shows the result of adding O2 plasma treatment in the end of poly etching.
The removed efficiency is good in treatment of 5, 10, 15 seconds. Therefore, the O2 plasma may be a best solution; using O2 plasma treatment can remove over ninety-nine percent defect. Another question must be considered, did it influence poly etching when implement the O2 plasma treatment? Figure 3.8 shows the O2 plasma treatment didn’t influence poly etching performance, the depth control stable in different treatment time.
The etch rate of poly etching is same in different treatment time.
3.3 Analysis
From experiments of above, characteristics of semi-translucent hexagonal defects were summarized:
(a) Random small and volatile defect was observed after recess process. Semi-translucent hexagonal defects can’t volatilize in environment of box, but volatilize in open area.
Defect level is decreasing during idling. Possible volatile compounds need to be high boil point.
(b) Volatile defect can’t be test by EDX, vaporizing during analysis.
(c) Volatile defect can’t come from chamber parts, such as chamber wall, gas distribution plate (GDP), focus ring and byproduct to be deposited on parts. Because semi-translucent defect appeared didn’t correlate with chamber idle time, PM cycle, changing time of parts and chamber condition during production. Chamber condition was changed after seasoning during production.
(d) Compound of semi-translucent hexagonal defect needs nitrogen source to react with SF6, which may explain why we never saw volatile particle on oxide particle monitor wafer.
(e) From experiment “d”, semi-translucent hexagonal defect may form by radicals or species of low decomposition rate of SF6. The high source power or Ar gas injected can induce SF6 to decompose smaller radicals or species.
Summarizing all analyses of above, preferring to say the “F11NS2” (or (F5S)2NF) is the most possible compound in the volatile residue. The reason for the assumption were:
(a) “F11NS2” is a stable compound, which may form from dissociated SF6 gas in the plasma. From equation 2.1 to 2.9 can summary as below:
SF6 + SF5* + N = (F5S)2NF (3.1) That may explain why serious semi-translucent hexagonal defect most appeared on low source power recipe and only little hexagonal defect on high source power or Ar gas injected recipe.
(b) “F11NS2” needs nitrogen source to react with SF6 and SF5, that may explain why we never saw volatile particle on oxide particle monitor wafer.
(c) From table 3.1 shows boil point of “F11NS2” is 51.6 ° C, that may explain Semi-translucent hexagonal defects can’t volatilize in environment of box, but volatilize in open area. Defect level is decreasing during idling.
From table 3.1, two candidates can meet the requirement, “F11NS2” and “F2N2S3”
(d) The etch byproduct of SiN is the nitrogen source in our candidate compounds, because the slow SiN etch rate we should not expect the compound of high nitrogen ratio may form during process. The purpose of this process is control poly depth and not removed nitride, the selectivity of poly to nitride is over 100. So, “F2N2S3” and
“F4N4S4” are not good candidates.
3.4 Conclusion
The “F11NS2” (or (F5S)2NF) is the most possible compound of the volatile residue.
Semi-translucent hexagonal defect was formed on nitride wafer in DPS poly etcher with SF6 plasma. Adding O2 plasma treatment in the end of poly etching is the best solution to remove particle. The “F11NS2” (or (F5S)2NF) is oxidized by oxygen radicals. Under SF6 plasma with
experience of another recipe in DPS poly etcher, adding argon in plasma also can reduce semi-translucent hexagonal defect. Because excited state of argon can impact SF6 molecule and produce small fractions of decomposition. Concentration of SF5 may lower than haven’t argon in plasma.
Chapter 4
Conclusion and Suggestion for Future Work
In this thesis, we investigated the effects of SF6 decomposition products in DPS poly etcher. The evaluation of the effects of SF6 decomposition induced corrosion metal in DPS poly etcher was discussed in chapter 2. When transport boxes are contaminated with SF6 decomposition from DPS poly etcher will induce cross contamination and corrosion of metal.
The evaluation of semi-translucent hexagonal defect from SF6 decomposition in DPS poly etcher was discussed in chapter 3. Some important conclusion are summarized below:
(a) In this thesis, SF6 decomposition products come from same source, same recipes and same tools of DPS poly etcher. The condition of SF6 plasma is low power, low pressure, high density, only pure SF6 gas. From equation 2.1 to 2.4, the decomposition includes more excited state of SF6, SF5 or SF4. In the other word, the fractions of SF6 decomposition are large and dissociation rate is low because the electron energy is in low level. If injecting argon into plasma, this phenomenon will decrease gradually because excited state of argon has very large energy to impact molecule of SF6.
(b) SF6 decomposition products contaminate the transport boxes and induce metal corrosion. From equation 2.5 to 2.10, SF5 or SF4 absorb on the surface of box and react with moisture. HF is existence on the boxes and attack metal film cause corrosion metal. That may explain why the corrosion metal appears irregularly, sometime and somewhere. In some Fabrication, corrosion metal has correlation with different season because different concentration of H2O in air.
(c) SF6 decomposition products react with nitride film in DPS poly etcher, cause semi-translucent hexagonal defect on wafer surface. The “F11NS2” (or (F5S)2NF) is the most possible compound of the volatile residue. From equation (2.9) and (3.1), that may prove SF6 decomposition is on the wafer and carried out into the boxes.
(d) Adding O2 plasma treatment can eliminate defect very successfully. That can reduce concentration of SF5 or SF4 very efficiently.
In the end of this thesis, we try to discuss about the future of IC’s manufacturing. From this thesis, both of by-products and decomposition products in etching easy induce defect or cross contamination in unfriendly environment. How to improve it? There are some suggestions can be used:
(a) Both of by-products and decomposition products in etching can’t be carried out chamber. One, using some ramp down procedure or treatment by N2/O2 plasma can remove most of by-products on surface of wafer. Two, keeping pressure of transfer chamber higher than process chamber can avoid by-products diffusing out chamber.
(b) Replace whole cassette loadlock with single wafer loadlock, can avoid cross contamination when whole cassette is in loadlock. Because transferring out wafer needs waiting all wafers process completed, it is easy to induce cross contamination in the waiting time. But in single wafer loadlock, the previous etching wafer put on cassette in open area outside the loalock.
(c) Don’t use boxes to carry wafers. Two advanced transfer system have be implemented to IC’s manufacturing. One, automatic transport by unmanned vehicle.
Wafer transfer between stocker and machine uses automatic vehicle (AGV), doesn’t use man to hand-carry. Two, SMIF (standard mechanical interface) is popular transfer system for IC’s manufacturing.
(d) It is necessary to control environment of fabrication very accurately. The concentration of H2O is very important parameter. Many of chemistry of by-product have relation with moisture, can form particle on wafer and kill yield. The concentration of ammonia is another important parameter, may cause quality of photo mask changed and induce particle on photo mask.
Reference
[01] P.L. Pai, C.H. Ting, W.M. Lee, R. Kuroda, “Metal Corrosion in Wet Resist-Stripping Process” Interface ’89, KTI microelectronics Seminar.
[02] R.R. Rogers, S.R. Wilson, “Localized Corrosion of Aluminum – 1.5% Copper Thin Films Exposed to Photoresist Developing Solution”, J. Vac. Sci. Technol. A9(3) May/Jun 1992, P1616 –1621
[03] K.H. Baek, C.I. Kim, K.H. Kwon, T.H. Kim, “Passivation Role of Fluorine on the Anticorrosion of AlCu Films After Plasma Etching” J. Vac. Sci. Technol. A 16(3), May/Jun 1998
[04] S. Bothra, H. Sur, V.Liang, R. Annapragada and J. Patel “Corrosion of Tungsten due to plasma charging in a metal plasma etcher”, VLSI Technology, Inc. 1109, Mckay Dr, San Jose, CA 95035
[05] S. Kirk, T. Maw, “Corrosion Prevention in The Bond Pad Cleaning Process” EKC Technology, Inc.
[06] S. Thomas, H.M. Berg, “Micro-Corrosion of Al-Cu Bonding Pads”, Motorola, Inc., P.153 [07] S. Graham, “Corrosion Issues in Solvent Processing”, Semitool Application Note,
February 1996
[08] C. Beryer, H. Jenett, D. Klockow, “Influence of Reactive SFx Gases on Electrode Surfaces After Electrical Discharges under SF6 Atomsphere” IEEE Transactions on Dielectrics and Electrical Insulation, Vol.7 No. 2, April 2000
[09] M. Piemontesi, L. Niemeyer, “Surface Reactions of SF6 Decomposition Products”, IEEE annual report – conference on electrical and dielectric phenomena, San Francisco October 20-23, 1996
[10] L. Niemeyer, F.Y. Chu, “SF6 and the Atmosphere” IEEE Transactions on Electrical Insulation, Vol.27 No.1, February 1992
[11] R.J. Van Brunt, J.T. Herron, “Fundamental Processes of SF6 Decomposition and Oxidation in Grow and Corona Discharges” IEEE Transactions on Electrical Insulation, Vol.25 No.1, February 1990
Compand Mp Bp Compand Mp Bp
Table 3.1 Possible SF6 decomposition compounds during DPS poly etcher plasma unit: °C
Fig 2.1 Wafer map of metal corrosion inspected by KLA TENCOR tools, KLA map of slot 1,11,25 in whole cassette of 25 wafers
Slot 1
Slot 25 Slot 11
Fig 2.2 the pictures of corrosion metal, the top picture inspected by optical microscope (OM), the others inspected by second electron microscope (SEM)
LOT ID Corrosion Operation Date metal etch Alarm Post clean
lot 1 serious M2 E 0306 22:11 A-DPS NO dSP
lot 2 serious M2 E 0307 00:34 B-ECR NO ACT-935
lot 3 serious M3 E 0307 02:46 A-DPS NO ACT-935
lot 4 serious M2 E 0307 15:13 C-ECR NO ACT-935
lot 5 serious M3 E 0308 14:23 C-ECR NO ACT-935
lot 6 serious M3 E 0308 10:00 B-ECR NO dSP
lot 7 serious M2 E 0310 16:45 B-ECR NO ACT-935
LOT ID Corrosion Metal sputter Time PR coating Exposure Time
lot 1 serious C 0301 00:20 A C 0302 19:24
lot 2 serious B 0228 16:00 A B 0302 00:16
lot 3 serious A 0301 10:39 A A 0303 12:08
lot 4 serious C 0302 19:11 B C 0303 16:54
lot 5 serious C 0301 23:43 B C 0304 23:30
lot 6 serious A 0302 03:19 A B 0305 06:43
lot 7 serious B 0303 12:30 A B 0310 10:25
Fig 2.3 Corrosion metal appeared all kinds of metal etchers, post clean process, metal sputtering deposition, photoresist coater, photoresist exposure in all time.
Fig 2.4 Corrosion metal was composed of aluminum, oxygen and fluorine by EDX
Fig 2.5 Corrosion metal wasn’t eliminated after metal etching back again
Al,TiN
Si,O F,Al,O
Metal etching back again
Fig 2.6 Wafer A – just metal etched and didn’t asher, the KLA map appeared metal corrosion on All wafer. This type corrosion metal was composed of chlorine, oxygen and aluminum; also inspected by SEM (second electron microscope)
#B #C
Fig 2.7 The KLA maps of “wafer B – wafer held with cassette in open area” and
“wafer C – wafer held with cassette in clean box”. Didn’t find corrosion metal on all wafers
Fig 2.8 Wafer D – wafer held with cassette in contaminated fluoride box, the KLA map appeared metal corrosion on top/bottom of wafer. This type corrosion metal was composed of fluorine, oxygen and aluminum; also inspected by SEM (second electron microscope)
Figure 2.9 Wafer E – box be contaminated with SF6 high density plasma of poly etcher, there were serious metal corrosion on wafer. Inspected by SEM and composed of aluminum, oxygen, fluorine by EDX.
Figure 2.10 Wafer F – box be contaminated with NF3 high density plasma of poly etcher, there were metal corrosion on wafer. Inspected by SEM and composed of aluminum, oxygen, fluorine by EDX.
Wafer I
Wafer G Wafer H
Figure 2.11 Wafer maps of wafer G – box be contaminated with NF3 median density plasma of poly etcher, wafer H – box be contaminated with C4F8 high density plasma of oxide etcher, wafer I -- box be contaminated with C4F8 median density plasma of oxide etcher; we couldn’t find any corrosion on three types of boxes.
4/9 使用期
Metal 專用 Box 一般 Box
生一 專用
M1.M2.M3 Etch
白底紅
Fig 2.12 In order to avoid contamination of metal, dedicated special boxes for metal wafer transport is a must.
Fig 3.2 wafer structure, isotropic dry etching to define poly 2 depth at this step.
P2
P1
Particle
nitride
Poly si
oxide Fig 3.1 DPS poly etcher chamber
Fig 3.3 Semi-translucent hexagonal defect on the wafer by KLA inspected
Fig 3.4 The efficiency of particle reduction with SPM + APM. The left lot – defect counts reduce from 1253 to 25 at 1st time clean. The right lot – defect counts reduce from 21762 to 300 at the 1st time and to 21 at the 2nd time clean.
After RRV
After RRV
Fig 3.5 The efficiency of particle reduction with SPM +APM once and twice.
Fig 3.7 The result of adding O2 plasma treatment in the end of poly etching. The removed efficiency is good for treatment 5, 10, 15 seconds.
O2 plasma treatment 15 sec O2 plasma treatment 10 sec
O2 plasma treatment 5 sec No O2 plasma treatment
Fig 3.6 The result of adding N2 plasma treatment in the end of poly etching.
RC3 Depth
200 250 300 350
POR 5SEC 10SEC 15SEC
Fig 3.8 The O2 plasma treatment don’t influence poly etching performance. The depth is stable in different treatment time.