Plasma Chemistry Study in an Inductively Coupled Dielectric Etcher

Transcription

1 Plasma Chemistry Study in an Inductively Coupled Dielectric Etcher Chunshi Cui, John Trow, Ken Collins, Betty Tang, Luke Zhang, Steve Shannon, and Yan Ye Applied Materials, Inc. October 26, /28/2008

2 Introduction Content Experimental Set-up Inductively-coupled dielectric etcher Plasma diagnostics Langmuir probe measurements Ion Mass Spectrometry (IMS) Optical Emission Spectrometry (OES) Effects of Different Parameters Source Power Pressure Surface temperatures Carrier Gas flow and type Location, Wafer type Plasma sources Summary

3 Introduction (1) An independent plasma source in conjunction with capacitively coupled rf bias has allowed Independent control of density and ion energy In-situ dry clean of process chamber These have been the major focus until recently for plasma source development Control of chemical species continues to be a major challenge This is the key to high selectivity of dielectrics to mask and underlayer with high etch rate and wide process window This requires control of both gas phase and surface reactions

4 Introduction (2) A selective dielectric etch involves careful balance of polymer deposition, chemical etching and physical sputtering. The etch relies on proper combination of fluorocarbon radicals and ions (CxFy). Progress has been made in characterization of CxFy species using LIF, laser absorption and etc. Etch rate and selectivity need to be correlated to the fluxes of these radicals and ions to the wafer. This study is a survey of CxFy ions by Hiden IMS and CxFy radicals by OES in an ICP dielectric etcher in conjunction with plasma measurements by Langmuir probe.

5 Dissociation Paths in C4F8 Plasmas H. Kazum et al., Plasma Sources Sci. Technol. 5 (1996),

6 CxFy Reaction Paths on Si Surface Si + CxFy Si Surface Reaction C2 + SiF + other by-products Adsorption of Reactants CxFy CxFy Desorption of Reactants Desorption of By-Product C2 + SiF Reaction Path: 1. Adsorption of reactants 2. Desorption or boiling off of reactants 3. Surface reaction to form C2, SiF and other by products 4. Desorption of by-products Reaction Rate is determined by Rate-Limiting (slowest Step).

7 Inductively Coupled Dielectric Etcher Temperature Control of all Plasma Surfaces Plasma Uniformity Control with Dual-Antenna Source Si Roof acting as RF Window & Electrode Movable Ion Mass Spec. or Langmuir probe S/N N/S Si Hot Ring Gas Feed Plasma Region SiC Collar N/S Temp Controlled S/N OES Window Electrostatic Ceramic Chuck Bias DC Hot Si Ring

8 Plasma Measurements Diagnostics Langmuir probe : Hiden Ion Mass Spec. (IMS): Optical Emission Spec. (OES): Te and ne Relative spectra of ion species Relative spectra of excited neutrals Experimental Parameters Total source power = W Pressure = 7-80mT C4F8 flow = sccm Carrier gas = He, Ar, Xe Si roof temperature = C Baseline conditions are shown by the blue numbers above.

9 Effect of Source Power: OES At low source power (<1000 W at 30mT), the spectra are dominated by Ar emission lines with very low C2 or SiF. This is an characteristics of E-mode plasma (capacitive coupling from the bias power). Change pf E- to H-mode (ICP) happens between 1000 to 1500W. The threshold power for mode change is strongly dependent on pressure, gas chemistry and chamber geometry. Intensity (arb.) RIE 500Wout 1000Wout 1500Wout C3 SiF C2 C2 C RIE 500Wout 1000Wout 1500Wout F Ar lines Wavelength (nm) Wavelength (nm)

10 Effect of Pressure: Langmuir Probe Plasma density (Ne) decreases with increasing pressure. Electron temperature (Te) decreases with increasing pressure. 5 Te vs. Pressure 15C4F8/100Ar/1400Wout 2.E+11 Electron Density vs. Pressure 15C4F8/100Ar/1400Wout ElectElectron Temperature (ev) Te (ev) Electron Density (cm-3) 2.E+11 1.E+11 5.E+10 Ne (cm-3) 0 0.E Pressure (mt) Pressure (mt) (Data were collected using Smart Probe by Scientific Systems.) / Alex Paterson

11 CxFy Ion Count 1.E+07 1.E+06 1.E+05 1.E+04 C+ and F+ decrease while CF+ increases with pressure. CF2+ and CF3+ slightly increase but much less than CF+. Higher pressure can reduce dissociation due to lower Ne and Te (Langmuir probe data) but not effective enough in the highly dissociation plasma. Effect of Pressure: Ion Mass Spec. C F CF CF2 CF3 C2F Pressure (mt) CF2 13% CF3C2F4 4% 1% 5mT CF 82% CF2CF3C2F4 3% 3% 0% 50mT CF 94%

12 Effect of Radial Distance: Ion Mass Spec. O+/O2+ ratio decreases with radial distance, possibly due to charge exchange between O+ and O2 and wall recombination of O and O+. O+/O2+ ratio decreases with pressure indicating reduced O2 dissociation. O+ / O O+/O2+(3.5cm from wall) O+/O2+ (at wall) Pressure (mt)

13 Effect of Si Surface Temperature: OES Below 200C, C2*/Ar* increases with Si temperature meaning the rate-limiting-step is desorption of polymer from Si surface. Above 200C, the trend reverses due to the limitation of CxFy absorption. Si surface acts like a good polymer sink or reflector but not a F scavenger. Normalized Intensity Si Roof Si Ring 1.00 Si Ring@300 o C Si Roof@120 o C Silicon Temperature (C) SiF/Ar*10 C2/Ar*10 F/Ar*10

14 Effect of C4F8 Flow: Ion Mass Spec. C4F8 flow of 15 and 30sccm results in similar ion mass spectra meaning not an effective knob for ion species control in an already highly dissociated plasma. CF2 12% Baseline 15C4F8/100Ar/in probe 15C4F8/100Ar C4F8/Ar CF3 C2F4 5% 1% Higher Flow 30C4F8/100Ar/in probe 304F8/100Ar C4F8/Xe CF2 8% CF3 C2F4 4% 1% CF 82% CF 87%

15 Effect of Carrier Gas: He, Ar and Xe The different ionization threshold and mass of He, Ar and Xe affect fundamentally plasma generation and loss resulting in different plasma density and electron energy distribution function (EEDF). Langmuir probe measurement shows that the EEDF is non-maxwellian in C4F8/Xe mixture with depleting high energy tails. The average electron energy is lowest in Xe and highest in He while the electron density has opposite trend. The low ionization energy threshold of Xe (~11eV) allows the high ionization rates with less hot electrons and results in a high density but low electron energy plasma.

16 EEDF in C4F8/Ar and C4F8/Xe Plasmas C4F8/Xe C4F8/Ar

17 Effect of Carrier Gas: Electron Density and Average Energy 8 C4F8 Mixed With Noble Gases He Ne Ar Xe ne (10^11cm3) Av Elect Energy (ev) Plasma Potential / 10 (Volts) Mass of Noble Gas

18 Effect of Carrier Gas: CxFy Ion Mass Spec. Percentage of F is 50 times less in C4F8/Xe than in C4F8/He. Percentage of heavy CxFy ions is far more in Xe than in Ar & He. These are due to low dissociation in C4F8/Xe which has the lowest average electron energy. Nomalized by Total Ion Count C4F8/He C4F8/Ar C4F8/Xe C F CF CF2 CF3 C2F4

19 Effect of Carrier Gas: OES Normalized By F* C2*/F* ratio is 8 times higher in C4F8/Xe than in C4F8/Ar. This demonstrates, again, less dissociation in C4F8/Xe Xe Ar Si(288.16) SiF(440.05) CO(313.44) C2(516.52) C2(563.55) H(656.29) F(703.75) O(777.54)

20 Effect of Carrier Gas: Process C4F8/Ar or C4F8/Xe process was applied to the pre-patterned via holes. Sharp facet of via tops with C4F8/Xe implies that physical sputtering is dominant over chemical etching of the holes. Rounded facet with C4F8/Ar suggests that isotropic etching maybe by excessive F atoms shown clearly by high C2*/F* ratio in OES spectra. 15C4F8/150Xe 15C4F8/150Ar

21 Summary Chemical species were experimentally investigated in an C4F8- based ICP as a function of many parameters. Clear E- to H-mode transition from low to high ICP power was observed accompanied by drastic change in chemical species spectra. In typical ICP operation, CxFy molecules are highly dissociated and the plasma is characterized by C2* rich OES and CF+ dominated IMS spectra. High pressure can slightly reduce dissociation but not effective enough to change the nature of high dissociation. Hot Si surface inside chamber plays more a role of polymer sink or reflector but a less role of F atom scavenger. The most effective knob among all is the type of carrier gas. Using Xe as a carrier gas can dramatically reduce the dissociation of fluorine-carbon molecules in an ICP oxide etcher. Dissociation control using different plasma generation mechanisms are currently in progress.

22 Future Perspective Gas phase and surface reaction mechanisms of CxFy plasmas need to be thoroughly studied. Semiconductor industry faces serious challenges due to feature shrinking and low-k/cu Dual-Damascene in interconnect circuitry. This plasma community can really help the industry with new innovations in terms of chemical species control Different plasma generation mechanisms Introduction of new chemistry Introduction of new surfaces Many more

23 Ion Mass Spectra vs. Wafer Type CF+ ions are the dominant reactive ion species. Minor loading effect of CxFy+ observed from Si to patterned SiO2 wafer. More release of by-product CO+ seen with patterned SiO2 wafer. Normalized by Total Ion Count C4F8/Ar Si Wafer C4F8/Ar Patterned Oxide Wafer C F CF CF2 CF3 C2F4 CO SiF Reactive Ions Byproduct ions

24 Challenges due to Device Evolution Feature Shrink beyond 0.1μm VLSI device features shrinks rapidly, leading to thinner mask thickness higher aspect ratio (HAR) of dielectric patterns. This requires highly selective etching to dielectric materials to mask. Evolution of Materials: SiO2 & Al low-k & Cu The VLSI industry has clearly taken the direction of Low-k & Cu Dual Damascene (DD) for lower response (RC) time in interconnect circuitry. This has introduced more challenges for dielectric etch due to more complex structures and constantly evolving low-k materials.

25 Intensity (arb.) Optical Emission in ICP and CCP In ICP, CxFy plasmas are characterized by C2 rich OES which is a symbol of high dissociation. Typically, C2*/Ar* ratio is many times higher in ICP than CCP plasmas C2*/Ar* ratio is used as a empirical method to guide process development for selective etch. OES from an ICP Chamber C2 (516) OES Emission from an MERIE Chamber Ar (750) Ar (750) F (703) Intensity (arb.) C2 (516) F (703) Wavelength (nm) Wavelength (nm)

26 Probe Data Interpretation The total probe current was used. No correction was made for ion current. The Electron Energy Distribution Function (EEDF) was calculated from the Druyvestyn formula. f 0 ( ε) = 8m. e q e d Area d V 2 I The electron density and the average electron energy were calculated from integrating the EEDF. The effective electron temperature for a non - Maxwellian EEDF is T eff = 2. 3 E av