Acid Etching Chemistry Characterizes Silicon Wafer Surface Metals
Fred Meyer and
John B. White,
Silicon Engineering
Technology Center Laboratory,
MEMC Southwest Inc.,
Sherman, Texas
-- Semiconductor International, 7/1/1999
Acids and acid mixtures such as HF, HF: HNO3 and HF: H2O2 have been used in the extraction process for contamination concentration techniques. The extraction solution's chemical composition in this process is critical in ensuring the efficient extraction of metal contaminants from the wafer surface. By using thermodynamic calculations, the effect of this chemical composition can be predicted for each element. Results of HF: H2O2 intentional contamination studies agree with the thermodynamic predictions, showing excellent recoveries for the metals of interest, except for copper.
Introduction
Due to the impact on device performance, surface metal contamination on silicon wafers is a critical concern. The need to monitor and control surface metal contamination has led to the development of sensitive acid etching techniques that extract and concentrate surface metals into small drops for analysis. Extraction techniques, such as automatic acid drop wafer scanning, can be coupled with inductively coupled plasma mass spectrometry (ICP-MS) and total reflective X-ray fluorescence spectroscopy (TRXRF) to determine contaminant concentration1.
Acids such as HF, HF: HNO3 and HF: H2O2 have been used in the extraction process for contamination concentration techniques. However, proper selection of the extraction chemistry is necessary to ensure reliable recovery of these contaminants.
'Surface' metals can be in four distinct areas representing potentially different chemistry requirements for removal. The metal can be on the native oxide's surface, trapped within it, at the silicon-oxide interface or in the silicon bulk. Also, metals can exist in several forms including elemental metal, oxide, hydroxide, fluoride, metal complex, silicate, silicide, etc. They can exist in these forms reacted with the silicon or silicon oxide, absorbed onto the surface or even as a multi-component particle.
The large number of possible chemical states for contaminants appears to make extraction a difficult procedure. Potentially good choices for the extraction solution include HF, HF: HNO3 and HF: H2O2. With the exception of copper, HF alone provides good recoveries for all the metals of interest. The oxidizing power of HNO3 and H2O2 make them good choices for use with HF and can even provide good recoveries for copper. H2O2 is often chosen over HNO3 to minimize bulk silicon etching.
HF Chemistry
A wafer exposed to HF reacts with the surface silicon oxide, SiO2, per Equations 1 and 22. Initially, the contamination affected is on the native oxide surface, trapped in the silicon oxide or at the SiO2-Si interface.
SiO2 (s) + 6 HF (aq) * H2SiF6 (aq) + 2 H2O (l) (Equation 1)
H2SiF6 (aq) * SiF4 (g) + 2 HF (aq)
(Equation 2)
The conversion of water soluble hexafluorosilicic acid3 (H2SiF6), to volatile silicon tetrafluoride raises the question as to surface contaminants' volatility and potential loss from the wafer surface. The fluorides of contaminants of interest, aluminum, calcium, chromium, cobalt, copper, iron, magnesium, nickel, potassium, sodium and zinc and the fluorosilicates of aluminum, calcium, cobalt, copper, iron, magnesium, nickel, potassium, sodium and zinc, vary in solubility in acid solutions.
Elemental Metal
As stated, HF alone is insufficient for collection of contaminants for analysis, since Cu is known to be reduced to the metallic state in an acidic solution. Table 1 lists the results of thermodynamic calculations for the reaction sequence in Equations 3-5, where M is the metal of interest. A negative Gibbs function for the reaction DGrxn shows that the deposition of the metal from the acid, e.g., HF, is thermodynamically favored. The Gibbs function of reaction, DGrxn, is the sum of the products' Gibbs functions of formation, DGf, minus the sum of the reactants' Gibbs function of formation, DGrxn = S (mDGf products) - S (nDGf reactants), where m and n are moles of products and reactants, respectively.
|
Table 1 Thermodynamic Evaluation of Metals in Acidic
Solutiion Mn+ + (n/2)H2r nH+ + M | |||||||
|
|
Reactants |
Products | |||||
| Mn+ | n | n/2 | DGf(Mn+) | DGf(H2) | DGf(H+) | DGf(M) | DGrxn |
| KJ/mol | KJ/mol | KJ/mol | KJ/mol | KJ/mol | |||
| Ca2+ | 2 | 1 | -553.58 | 17.6 | 0 | 0 | 536 |
| Al3+ | 3 | 1.5 | -485 | 17.6 | 0 | 0 | 459 |
| Mg2+ | 2 | 1 | -454.8 | 17.6 | 0 | 0 | 437 |
| Mn2+ | 2 | 1 | -228.1 | 17.6 | 0 | 0 | 211 |
| Zn2+ | 2 | 1 | -147.06 | 17.6 | 0 | 0 | 129 |
| Fe2+ | 2 | 1 | -78.9 | 17.6 | 0 | 0 | 61 |
| Co2+ | 2 | 1 | -54.4 | 17.6 | 0 | 0 | 37 |
| Ni2+ | 2 | 1 | -45.6 | 17.6 | 0 | 0 | 28 |
| Fe3+ | 3 | 1.5 | -4.7 | 17.6 | 0 | 0 | -22 |
| Cu+ | 1 | 0.5 | 49.98 | 17.6 | 0 | 0 | -59 |
| Cu2+ | 2 | 1 | 65.49 | 17.6 | 0 | 0 | -83 |
| Co3+ | 3 | 1.5 | 134 | 17.6 | 0 | 0 | -160 |
(n/2) H2 * n H+ + n e- (Equation 3)
Mn+ + n e- * M (Equation 4)
With the overall reaction,
Mn+ + (n/2) H2 * n H+ + M (Equation 54)
From Table 1, it is clear that Fe3+, Cu+, Cu2+ and Co3+ are predicted to reduce to elemental metal based on the negative Gibbs function. However, data also suggest that Fe3+ and Co3+ are not the favored ions based on their less favorable DGf compared to Fe2+ and Co2+. Thus, only copper is expected to reduce to elemental metal and in this form can deposit on the silicon surface. Table 1 does not consider the possibility of metal complexes with anions in solution that could also prevent deposition as metal.
Metal Oxides
Metal oxides can be present on the surface as a particle embedded in the native silicon oxide or embedded under the native silicon oxide. For example, Fe oxide is deposited onto the silicon surface in an SC1 (NH4OH: H2O2: H2O) solution5. Metal oxides can be removed by the acid per the reactions below, where M is the metal of interest; MO is the metal oxide, etc. The equilibrium constant, Kc, can be used to determine if the metal oxide will precipitate out of solution at given acid and metal concentrations.
MO + 2 H+ * M2+ + H2O
Kc = [M2+]/[H+]2
(Equation 6)
M2O + 2 H+ * 2 M+ + H2O
Kc = [M+]2/[H+]2
(Equation 7)
MO2 + 4 H+ * M4+ + 2 H2O
Kc = [M4+]/[H+]4
(Equation 8)
M2O3 + 6 H+ * 2M3+ + 3 H2O
Kc = [M3+]2/[H+]6
(Equation 9)
Table 2 provides these reactions' thermodynamic analysis for several metals. The value for the equilibrium constant, Kc, was calculated and compared to Qc, the reaction quotient calculated for a 5% HF, 500 ppt metal solution. Qc is much less than Kc in every case, suggesting the metal oxide will remain in solution. As already stated about elemental metals, metal ions Cu+ and Cu2+ are thermodynamically favored to deposit or plate back onto the silicon surface. Also, the reactions of Cu2O, MnO2 and Fe2O3 with the acid are not favored thermodynamically.
|
Table 2 Thermodynamic Evaluation of Metal Oxides in
HF | ||||||||
|
Metal Oxide (MO) |
Metal Ion (MI) |
DGf(MO) |
DGf(H+) |
DGf(MI) |
DGf(H2O) |
DGrxn |
Kc |
Qc(5 % HF,500 ppt metal) |
|
KJ/mol |
KJ/mol |
KJ/mol |
KJ/mol |
KJ/mol |
|
| ||
|
CaO |
Ca2+ |
-604 |
0 |
-553.6 |
-237.1 |
-187 |
5.3E+32 |
1.4E-05 |
|
MgO |
Mg2+ |
-569.6 |
0 |
-454.8 |
-237.1 |
-122 |
2.8E+21 |
2.3E-05 |
|
MnO |
Mn2+ |
-363 |
0 |
-228.1 |
-237.1 |
-102 |
7.7E+17 |
1.0E-05 |
|
CoO |
Co2+ |
-213.4 |
0 |
-54.4 |
-237.1 |
-78 |
5.0E+13 |
9.6E-06 |
|
FeO |
Fe2+ |
-244.3 |
0 |
-78.9 |
-237.1 |
-72 |
3.7E+12 |
1.0E-05 |
|
NiO |
Ni2+ |
-216.1 |
0 |
-45.6 |
-237.1 |
-67 |
4.7E+11 |
9.7E-06 |
|
ZnO |
Zn2+ |
-318.3 |
0 |
-147.1 |
-237.1 |
-66 |
3.5E+11 |
8.7E-06 |
|
CuO |
Cu2+ |
-130 |
0 |
65.49 |
-237.1 |
-42 |
2.0E+07 |
8.9E-06 |
|
Na2O |
Na+ |
-376.5 |
0 |
-261.9 |
-237.1 |
-384 |
2.5E+67 |
8.9E-06 |
|
Cu2O |
Cu+ |
-146 |
0 |
49.98 |
-237.1 |
9 |
2.8E-02 |
8.9E-06 |
|
MnO2 |
Mn4+ |
-466 |
0 |
134 |
-237.1 |
23 |
1.0E-04 |
9.6E-06 |
|
Al2O3 |
Al3+ |
-1576 |
0 |
-485 |
-237.1 |
-105 |
2.5E+18 |
2.1E-05 |
|
Fe2O3 |
Fe3+ |
-741 |
0 |
-4.7 |
-237.1 |
20 |
2.9E-04 |
1.0E-05 |
Oxides can react with HF to form metal fluorides, but this varies with different metals. For example, the formation of both CuF2 and ZnF2 is favored thermodynamically:
CuO (s) + 2 HF (aq) + H2O (l) * CuF2 * 2 H2O (s) DGrxn = -64 KJ/mol
(Equation 10)
CuO (s) + 2 HF (aq) * Cu2+ (aq) + 2 F- (aq) + 2 H2O (l) DGrxn = -279
(Equation 11)
ZnO (s) + 2 HF (aq) * ZnF2 (aq) + H2O (l) DGrxn = -61
(Equation 12)
ZnO (s) + 2 HF (aq) * Zn2+ (aq) + 2 F- (aq) + 2 H2O (l) DGrxn = -303
(Equation 13)
Metal Silicates
The reaction of metal silicates with the acid is illustrated with calcium in the equations below. If the calcium silicate's reaction with HF is favored thermodynamically and kinetically, the CaO can be expected to react and remain in solution per the previous calculations for oxides.
CaSiO3 (s) + 6 HF (aq) * CaO (s) + H2SiF6 (aq) + 2 H2O (l) (Equation 14)
CaO (s) + 2 H+ (aq) * Ca2+ (aq) + H2O (l) (Equation 15)
H2SiF6 (aq) * SiF4 (g) + 2 HF (g)
(Per Equation 2)
With the overall reaction,
CaSiO3 (s) + 4 HF (aq) + 2H+ (aq) * Ca2+ (aq) + SiF4 (g) + 3 H2O (l)
(Equation 16)
and DGrxn= -232 KJ/mol.
HF: H2O2 Chemistry
Extraction chemistry for HF: H2O2 includes HF chemistry considerations outlined previously. It also includes extracting contaminants from the surface by complexing the plated metals, keeping metals in solution through oxidation and removing bulk contaminants such as silicides through the etching of the silicon substrate. Typical acid extraction solutions include an oxidizing agent such as H2O2 or HNO3 and a complexing agent such as HF or HCl. Again, H2O2 is often chosen over HNO3 to minimize bulk silicon etching. The extraction solution consisted of 5% HF: 6% H2O2 (w/w) in 18 M+ DI water.
H2O2 Reactions
H2O2 can be both oxidized and reduced in acid solution as shown in the following reaction equations:
Oxidation
H2O2 (aq) * O2 (g) + 2 H+ (aq) + 2 e-
(Equation 17)
Reduction
H2O2 (aq) + 2 H+ (aq) + 2 e- * 2 H2O (l)
(Equation 18)
Complexing Action
Low pH, acidic hydrogen peroxide solutions (e.g., hydrogen chloride or acetic acid) are effective in desorbing Au and Cu from silicon and dissolving Ag, Ni, Co, Pb, Mg, Ni, Te and W6. This is done by keeping metal ions in a soluble, oxidized state as indicated in the following example for oxidation of Cu and the reduction of H2O2:
H2O2 (aq) + 2H+ (aq) + 2e- * 2 H2O (l) DGrxn = -340 KJ/mol
(per Equation 18)
Cu (s) * Cu2+ (aq) + 2e- DGrxn = 65.5
(Equation 19)
And the overall reaction is then,
Cu (s) + H2O2 (aq) + 2H+ (aq) * Cu2+ (aq) + 2 H2O (l) DGrxn = -275
(Equation 20)
If Fe is present on the surface as a metallic element, the following reaction is also important:
Fe (s) + H2O2 (aq) + 2H+ (aq) * Fe2+ (aq) + 2 H2O (l) DGrxn = -419
(Equation 21)
Oxidative Component
Nitric acid and hydrogen peroxide were tested as the extraction chemistry's oxidative component (Table 3). Cu, Fe and Ca data are presented for comparison, since they are common contaminants and vary in difficulty to recover from the wafer surface. For each chemistry, four successive extractions were made on a relatively unclean silicon wafer, and the resultant recoveries observed. The initial form and concentrations of the contaminants are not known. However, if the wafers had received an HF process as a final step prior to analysis, it is expected that some Cu metal could be present in elemental form. In addition, Cu and Fe may be present in the bulk silicon (e.g., iron silicide).
| Table 3 Comparison of HF: H2O2 and HF: HNO3 Extraction Solutions (Concentration 1E10 atoms/CM2) | |||||
|
Solution(with 5% HF) |
1st Extraction |
2nd Extraction |
3rd Extraction |
4th Extraction | |
|
Cu |
6 % H2O2 0.5% HNO3 |
3.30
2.50 |
0.26
1.50 |
0.19
1.50 |
0.101.30 |
|
Fe |
6 % H2O2 0.5% HNO3 |
0.59
1.60 |
0.28
0.19 |
0.36
0.42 |
0.160.20 |
|
Ca |
6 % H2O2 0.5% HNO3 |
3.80
91.00 |
0.063
0.66 |
< 0.049
0.84 |
0.120.94 |
The recovery data in Table 3 suggest that Ca was associated with the silicon surface, since it was effectively removed in the first extraction. Fe and Cu recovery data show the metal was not contained in the first extraction and show significant differences between the two oxidizing agents. This could be due to different initial metal concentrations, oxidation strengths, complexing abilities, silicon etch rates, etc. The Cu concentrations in Table 3 for the second, third and fourth extractions appear constant but are nearly a factor of 10 different for the two oxidizing agents. The Fe concentrations in Table 3 for the second, third and fourth extractions, however, are essentially the same for both agents and appear constant.
Silicon Etching
Oxidizing and complexing agents H2O2 and HF etch the silicon substrate and remove contaminants associated with bulk silicon, such as silicides. The etching mechanism has been explained by an oxidation-reduction reaction sequence for HNO3 and HF7. A similar mechanism can be shown for H2O2 and HF.
Anodic Reaction
Si * Si4+ + 4 e- (Equation 22)
Si4+ + 2 H2O * SiO2 + 4 H+
(Equation 23)
Cathodic Reaction
H2O2 + 2 H+ + 2e- * 2 H2O
(from Equation 18)
Overall Reaction
Si + 2 H2O2 * SiO2 + 2 H2O
(Equation 24)
The HF then dissolves the SiO2 per the reactions in Equations 1 and 2.
Intentional Contamination Chemistry
The acid mixture's ability to efficiently extract elements from the surface has been studied with intentional wafer contamination8. A spiked mixture of ammonium hydroxide, hydrogen peroxide and DI water (SC1 or standard clean 1) was used to deposit metals on the surface. Acid extraction was done using HF: H2O2, and trace metals analysis was performed using ICP-MS. Multiple extractions were performed on each wafer to determine extraction efficiency for the metals of interest. If all the metal is removed in the first extraction, then the acid mixture is 100% efficient.
The SC1 bath dissolves any existing native SiO2 layer on the wafer surface. It then simultaneously etches and oxidizes the surface. Small molecular contaminants such as metal silicides, dissolved metallic contaminants in the surface and small particles are released during this etching process. Particles are electrostatically lifted from the wafer surface during this process. The newly formed oxide will contain contaminants from the bath. The oxide layer can contain contaminants in the form of ions, elemental metal, metallic oxides and metal silicates.
Concentrations of these contaminants in the oxide depend on their chemistries in an SC1 solution. In a cleaning process it is hoped that these concentrations are lower than those originally on the wafer surface. However, in most cases it is necessary to follow an SC1 clean with an acidic clean to remove metallic contaminants such as Fe and Zn.
Experimentally, higher deposition rates of Al, Fe and Zn have been observed from SC1 baths. These can result from the formation of insoluble oxides or insoluble hydroxides in solution. Modeling has been done9, 10 of the deposition of Fe from SC1 solutions. This shows that both Fe2O3 precipitates at nucleation sites on the oxidizing silicon surface and precipitation of Fe(OH)2+ occur when Fe is present in SC1 solutions. Also shown is that ZnO deposits from SC1 when Zn is present. Insoluble hydroxides of Al, Mg and Zn are known to exist, but the thermodynamic details of their chemistry in SC1 have not been elaborated. It is likely these metals will exhibit high deposition rates on the wafer surface in a spiked SC1 bath.
|
Table 4 SC1 Intentional Contamination of 150mm Wafer
(IE10 atoms/cm2) | |||||
|
Element |
Mass |
Extraction 1 |
Extraction 2 |
% Extraction Efficiency | |
|
Na |
23 |
0.90 |
0.19 |
83 | |
|
Al |
27 |
73.00 |
0.85 |
99 | |
|
Ca |
40 |
1.40 |
0.57 |
71 | |
|
Cr |
52 |
0.33 |
0.034 |
91 | |
|
Fe |
56 |
27.00 |
0.41 |
98 | |
|
Ni |
58 |
0.43 |
0.012 |
97 | |
|
Co |
59 |
0.27 |
0.019 |
94 | |
|
Cu |
63 |
0.27 |
0.13 |
67 | |
|
Zn |
66 |
16.00 |
0.59 |
97 | |
|
Ag |
107 |
0.096 |
0.0047 |
95 | |
After removal of the initial oxide layer, NH4OH acts as a metal scavenger through complex reactions forming soluble products with metals such as Cd, Co, Cr, Cu, Ag, Au and Pt. Elements that preferentially form complexes with the NH4OH exhibit low deposition rates into the oxide formed on the wafer surface. Two caveats are needed here. First, NH4OH concentration must be great enough to dissolve all the metal. This is especially true of Cu, which forms an insoluble oxide when this concentration is low. Secondly, the thermodynamics of the formation of complexes in the presence of H2O2 must be evaluated for each metal.
Intentional Contamination Results
Results are summarized in Table 4 with concentrations in atoms/cm2. As predicted, Al, Fe and Zn show high deposition rates, while Cr and Co show low rates. Of particular note are excellent extraction efficiencies using HF: H2O2 for all elements except Ca and Cu. Since Cu can diffuse into Si at room temperature, it is possible the less efficient extraction for Cu is due to Cu in the bulk Si11.
Conclusion
The chemical composition of the extraction solution in this process is
critical in ensuring efficient extraction of metal contaminants from the wafer
surface. The effect of this chemical composition can be predicted for each
element using thermodynamic calculations. Results of the HF:
H2O2 intentional contamination agree with thermodynamic
predictions showing excellent recoveries for metals of interest, with the
exception of copper.
References
1. F. Meyer, J. White, M. Radle, Silicon Wafer Surface Metals Characterization Using Automatic Wafer Scanning and Inductively Coupled Plasma Mass Spectrometry, Submitted to Semiconductor International, September 1998.
2. D. R. Turner (Bell Telephone Laboratories), On the Mechanism of Chemically Etching Germanium and Silicon, JECS, 1960.
3. B. Tuck, Review: The Chemical Polishing of Semiconductors, Journal of Material Science, 10, 1975.
4. S. Dhanda, P. Gupta, B. B. Triplett, C. R. Helms, Optimization of Pre-Gate Cleaning Sequences & Fluid Purity, SRC Contract 350 Final Report, Part 2, January 1997.
5. S. Dhanda, et al.
6. W. Kern, D. A. Puotinen, Cleaning Solutions Based on Hydrogen Peroxide for use in Silicon Semiconductor Technology, RCA Review, June 1970.
7. B. Tuck
8. F. Meyer, J. White, M. Radle
9. C.R. Helms and Heungsoo Park, Proceedings of the 3rd International Symposium on Cleaning Technology, The Electrochemical Society, 1993.
10. S. Dhanda, et al.
11. F. Meyer, J. White, M. Radl
|
Fred Meyer received his Ph.D. in molecular physics in 1978 from the University of North Texas. He has been characterizing contaminants and defects in or on Si for Texas Instruments and for MEMC Southwest since 1984. Prior to that, he has five years experience in Si crystal growth. |
|
John White received his B.S. in chemistry in 1980 from the University of Texas at Dallas. He has 18 years experience in the semiconductor industry including materials characterization and process/quality engineering for Texas Instruments and MEMC Southwest. |