Mechanism of Ni removal from Si materials using hydrogen cyanide aqueous solutions

Article abstract of DOI:10.1149/1.2372692

The mechanism of Ni removal from Si O2 -covered Si specimens by HCN aqueous solutions has been investigated by means of total reflection X-ray fluorescence spectroscopy and X-ray photoelectron spectroscopy measurements. Ni contaminants on the Si O2 surface are present in the form of SiO-NiOH. The removal mechanism consists of two steps, i.e., an initial fast process followed by a slow process. The rate-determining steps of both the processes are of the first order with respect to the concentration of cyanide ions (C N-). The fast and slow processes are tentatively attributed to the removal of SiO-NiOH on terraces and in sub-nanometer pores, respectively. The cleaning ability of the HCN aqueous solutions is much better than ammonia aqueous solutions, because of high reactivity to form nickel-cyanide complex ions and avoidance of readsorption of Ni (CN) 4 2- complex ions in the solutions.

Full text of DOI:10.1149/1.2372692

H16
Journal of The Electrochemical Society, 154 ͑1͒ H16-H19 ͑2007͒
0013-4651/2006/154͑1͒/H16/4/$20.00 © The Electrochemical Society
Mechanism of Ni Removal from Si Materials Using Hydrogen
Cyanide Aqueous Solutions
z
*
Yueh-Ling Liu, Masao Takahashi, and Hikaru Kobayashi
Institute of Scientific and Industrial Research, Osaka University, and CREST, Japan Science and
Technology Organization, 8-1 Mihogaoka, Ibaraki, Osaka 567-0047, Japan
The mechanism of Ni removal from SiO2-covered Si specimens by HCN aqueous solutions has been investigated by means of
total reflection X-ray fluorescence spectroscopy and X-ray photoelectron spectroscopy measurements. Ni contaminants on the
SiO2 surface are present in the form of SiO-NiOH. The removal mechanism consists of two steps, i.e., an initial fast process
followed by a slow process. The rate-determining steps of both the processes are of the first order with respect to the concentration
of cyanide ions ͑CN−͒. The fast and slow processes are tentatively attributed to the removal of SiO-NiOH on terraces and in
sub-nanometer pores, respectively. The cleaning ability of the HCN aqueous solutions is much better than ammonia aqueous
2−
4
solutions, because of high reactivity to form nickel-cyanide complex ions and avoidance of readsorption of Ni͑CN͒ complex
ions in the solutions.
© 2006 The Electrochemical Society. ͓DOI: 10.1149/1.2372692͔ All rights reserved.
Manuscript submitted May 15, 2006; revised manuscript received August 2, 2006. Available electronically November 17, 2006.
Complete removal of metal contaminants is required for the fab-
liminary cleaned by the RCA cleaning method.2 A rinse in ultrapure
water ͑UPW͒ was carried out after each cleaning step.
rication of Si devices because heavy metals such as Cu, Fe, Ni, and
Co form trap states in the Si bandgap.1 In the fabrication process of
Si devices, surface contaminants are usually removed by the RCA
method, i.e., alkaline and acid solutions containing hydrogen
peroxide.2 The conventional RCA method requires high concentra-
tion ͑e.g., ϳ5%͒ cleaning solutions and heating at 50 ϳ 80°C.
Moreover, the RCA method causes surface etching, which makes the
device fabrication processes complicated and possibly forms rough
surfaces and defect states. To avoid etching, RCA cleaning solutions
using dilute varieties are recently employed in LSI fabrication pro-
cesses.
For intentional Ni contamination, SiO2/Si specimens were im-
mersed in 20 ppm nickel͑II͒ nitrate ͓Ni͑NO3͒2, 99.95%, Kanto
Chemical Co., Inc.͔ aqueous solutions at room temperature for
10 min. The pH of the Ni͑NO3͒2 aqueous solutions was set at 9 by
addition of ammonia aqueous solutions, and then the specimens
were rinsed with UPW.
HCN gas was generated by addition of concentrated sulfuric acid
͑H2SO4, EL grade, Kanto Chemical Co., Inc.͒ to solid sodium cya-
nide ͑NaCN, 97%, Wako Co., Inc.͒. The generated HCN gas was
introduced to a reservoir cooled below −20°C by liquefied nitrogen,
to solidify HCN with the freezing point of −13.3°C.15 Then, the
solidified HCN was dissolved in UPW.
The concentration of the HCN aqueous solutions was determined
by the Liebig titration method using silver nitrate ͑AgNO3, 99.8%,
Wako Co. Inc.͒ aqueous solutions.15 Before titration, a 6 M NH4OH
solution was added to the HCN aqueous solution to adjust pH, to
prevent vaporization of HCN from the solution. Then, a 10% potas-
sium iodide ͑KI, Wako Co. Inc.͒ aqueous solution was added as an
indicator. When all CN− ions in the solutions were consumed for the
formation of silver-cyanide complex ions, Ag͑CN͒−2, milky-white
precipitates of silver iodide ͑AgI͒ were formed, indicating comple-
tion of the titration.
Solutions containing chelating agents such as EDTA are found to
have an effect to prevent readsorption of metal species in cleaning
solutions.3,4
We have recently developed semiconductor cleaning solutions
which contain cyanide ions ͑CN−͒.5-8 Due to the strong reactivity of
CN− ions with metal species, semiconductor cleaning can be per-
formed at room temperature using low concentration ͑e.g., 0.2%͒
solutions. Due to the large complex stability constant of CN− ions
compared with other chelating agents,9 metal-cyanide complex ions
in the cleaning solutions are not readsorbed on the surface, which
makes it possible to remove metal contaminants completely and
to use the cleaning solutions repeatedly. We have already found
that CN− ions selectively react with semiconductor defect states
such as Si dangling bonds, resulting in the defect passivation.10-14
The defect passivation by CN− ions greatly improves various-
type Si solar cells10-13 and decreases the density of a leakage cur-
rent flowing through an ultrathin SiO2 layer of metal-oxide-
semiconductor ͑MOS͒ diodes.14 Therefore, cyanide solutions can be
applied to not only semiconductor cleaning but also defect passiva-
tion.
Results and Discussion
Figure 1 shows the XPS spectrum in the Ni 2p3/2 region for
Ni-contaminated SiO2/Si specimens. Two intense peaks were ob-
served at 857.0 and 862.5 eV, attributable to Ni͑OH͒2.16 This result
shows that Ni contaminants on the SiO2 surface are present in the
chemical state similar to Ni͑OH͒2.
Figure 2 compares the cleaning ability of HCN aqueous solutions
with ammonia aqueous solutions. pH of the ammonia aqueous solu-
tions was set at 10.7, i.e., the same as that for the HCN aqueous
solutions, for direct comparison. The Si cleaning was performed at
room temperature for 3 min for both the solutions. It is clearly seen
that HCN aqueous solutions are more effective for Ni removal than
ammonia aqueous solutions. Although Ni2+ ions form complex ions
with both CN− and NH3, and Ni͑CN͒ and Ni͑OH͒ are dissolved
In the present study, HCN aqueous solutions were found to pos-
sess a great effect of removing Ni contaminants and the removal
mechanism is investigated by means of total reflection X-ray fluo-
rescence spectroscopy ͑TXRF͒ and X-ray photoelectron spectros-
copy ͑XPS͒. It was found that the removal proceeds with an initial
fast process followed by a slow process, and their activation ener-
gies and preexponential factors are obtained.
2
2
in both HCN solutions and ammonia aqueous solutions,17,18 the sta-
bility constants of nickel-cyanide complex ions are much larger than
those of other complex ions,9 as displayed in Table I. The great
cleaning ability of HCN solutions results from high reactivity of
CN− ions to form nickel-cyanide complex ions and avoidance of
readsorption.
Experimental
Boron-doped p-type CZ Si͑100͒ wafers with 10–15 ⍀ cm resis-
tivity were thermally oxidized to form ϳ5 nm thick silicon dioxide
͑SiO2͒ layers. The wafers were cut into 2 ϫ 3 cm pieces and pre-
It is likely that Ni contaminants are adsorbed on SiO2 surfaces by
reacting with surface Si-OH groups.19 This mechanism is based on
the behavior of silica gel where Si-OH groups act as weakly acidic
*
Electrochemical Society Active Member.
z E-mail: h.kobayashi@sanken.osaka-u.ac.jp
Downloaded on 2015-03-08 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑1͒ H16-H19 ͑2007͒
H17
Figure 1. XPS spectrum in the Ni 2p3/2 region for the Ni-contaminated
SiO2/Si specimens.
Figure 2. Surface Ni concentrations measured before and after cleaning in
ammonia aqueous solutions and HCN aqueous solutions with pH set at 10.7
for 3 min.
ion exchangers.20-22 Hoshino et al.23 have reported that metal hy-
droxide reacts with Si-OH, resulting in the formation of Si-O-metal
species
Ni͑OH͒ + Si–OH → SiO–NiOH + H2O
͓1͔
2
͓Ni2+͔
ln
= − k t
͓7͔
Ј
Therefore, the removal of Ni contaminants by the HCN aqueous
solutions is likely to proceed with the following reaction formula
͓Ni2+͔
0
The surface Ni concentrations after cleaning the specimens with
an initial Ni concentration of ϳ1014 atoms/cm2 were measured as a
function of the cleaning time, t. The semi-logarithmic plots of
−͑m−2͒
SiO–NiOH + mCN− + H2O ꢀ Si–OH + Ni͑CN͒
m
+ 2 OH− ͑m = 2, 3, or 4͒
͓2͔
͓Ni2+͔/͓Ni2+
͔ vs t at 19°C for various HCN concentrations are
0
We have already shown that in the case of Si/SiO2 cleaning with
HCN aqueous solutions, readsorption of metal species in HCN so-
lutions does not occur.6,8 This result shows that the reverse reaction
in Eq. 2 can be ignored. Thus, the reaction rate is written as
shown in Fig. 3. The experimental results can be well fitted by two
straight lines. From the slops of the straight lines, k can be esti-
Ј
mated.
From Eq. 5, we have
d͓Ni2+͔
Ϸ − k͓Ni2+͔͓CN−͔m
͓3͔
ln k = m ln͓CN−͔ + ln k
͓8͔
Ј
dt
Therefore, from the slope of the plot of ln k vs ln͓CN−͔ shown in
where k is the reaction rate constant and Ni2+ denotes SiO-NiOH.
Since ͓CN−͔ is much higher than ͓Ni2+͔, ͓CN−͔ is almost constant
with time. Therefore, Eq. 3 is simplified to
Ј
Fig. 4, m can be determined to be unity for both the fast and slow
reactions. This result indicates that the rate-determining steps for
both the reactions are of the first order with respect to ͓CN−͔.
The reaction formula for the removal of Ni contaminants by
HCN aqueous solutions is written as
d͓Ni2+͔
Х − k ͓Ni2+͔
͓4͔
Ј
dt
where
Step 1 SiO–NiOH + CN− → SiO–NiCN + OH−
͓9͔
k = k͓CN−͔m
͓5͔
͓6͔
Ј
Step 2 SiO–NiCN + CN− + H2O → Si–OH + Ni͑CN͒ + OH−
2
Integration of Eq. 4 leads to
͓Ni2+͔ = ͓Ni2+͔0e−k t
where ͓Ni2+͔ is the initial surface Ni concentration. Therefore, we
have
͓10͔
Ј
Ni͑CN͒ is soluble, and thus it desorbs from the surface.17,18 In
2
0
the solutions, the following reaction proceeds
−
Step 3 Ni͑CN͒ + CN− → Ni͑CN͒
͓11͔
2
3
−
The fourth CN− ion reacts with Ni͑CN͒ in HCN solutions to
3
form Ni͑CN͒2−, which is the most stable Ni-cyanide complex ion
4
Table I. Stability constants of Ni2+ complex ions with various
ligands.9 ␤n denotes the stability constants of complex ions with n
ligands.
9
with a high stability constant of 2.0 ϫ 1030
−
2−
4
Step 4 Ni͑CN͒ + CN− → Ni͑CN͒
͓12͔
3
Ligand
Stability constants ͑logarithmic values͒
Readsorption of Ni contaminants to the surface does not occur
because of the high stability of Ni͑CN͒2− in HCN aqueous solutions.
CN−
NH3
Ethylenediamine
͑en͒
30.3 ͑␤4͒
8.89͑␤4͒
4
Similar experiments were carried out at various temperatures to
obtain the temperature dependence of the rate constant, ks. ks is
generally given by the Arrhenius equation
2.21͑␤1͒
7.35͑␤1͒
4.50͑␤2͒
13.54͑␤2͒
6.86͑␤3͒
17.71͑␤3͒
Downloaded on 2015-03-08 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H18
Journal of The Electrochemical Society, 154 ͑1͒ H16-H19 ͑2007͒
Figure 3. Surface Ni concentration di-
vided by the initial surface Ni concentra-
tion on the SiO2 surfaces vs the time of
immersion in the HCN aqueous solutions
at 19°C with pH of 10.7. The concentra-
tions of the HCN solutions were ͑a͒ 10.7,
͑b͒ 6.6, ͑c͒ 3.0, and ͑d͒ 1.4 mM.
For the fast reaction, the value of A is very large, indicating that
CN− ions easily react with Ni contaminants. If SiO-NiOH is present
on terraces, CN− ions easily attack the species, leading to the large A
value.
Ea
k = A exp −
͓13͔
͓14͔
ͩ ͪ
RT
or
The slow reaction becomes dominant when the surface Ni con-
Ea
R
1
centration is less than ϳ1012 atoms/cm2. For the slow reaction, on
ln k = −
+ ln A
ͩ ͪ
T
where A is the preexponential factor and Ea is the activation energy.
From the slope and the intercept of the semilog plot of the rate
constant vs the reciprocal temperature ͑Fig. 5͒, Ea and A are deter-
mined to be 43 kJ/mol and 1.4 ϫ 1010 for the fast reaction, and
24 kJ/mol and 2.3 ϫ 104 for the slow reaction, respectively.
Figure 4. Log-log plots of k ͑cf. Eq. 8͒ vs the concentration of CN− ions in
the HCN aqueous solutions ͑a͒ for the fast reaction and ͑b͒ for the slow
reaction.
Figure 5. Semilog plots of the rate constant vs the reciprocal of temperature,
where the open squares and the solid line are for the fast reaction, and the
full squares and the broken line are for the slow reaction.
Ј
Downloaded on 2015-03-08 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑1͒ H16-H19 ͑2007͒
H19
the fast and slow reactions are of the first order with respect to the
concentration of CN− ions. The activation energy and the preexpo-
nential factor of the rate constant for the fast reaction ͑or slow re-
action͒ are determined to be 43 kJ/mol and 1.4 ϫ 1010, respectively
͑or 24 kJ/mol and 2.3 ϫ 104, respectively͒. Ni contaminants for the
fast and slow reactions are attributed to SiO-NiOH on terraces and
sub-nanometer pores, respectively.
the other hand, A has a low value, indicating that the reaction prob-
ability of CN− ions is low. If the slow reaction species is due to Ni
species in sub-nanometer size pores, CN− ions have difficulty in
approaching, resulting in the small A value. The fast and slow re-
moval species are tentatively attributed to SiO-NiOH on terraces
and in sub-nanometer pores, respectively, although no direct evi-
dence has been obtained in the present study. Ni-O bonds in the
slow reaction species are likely to be weakened by the adsorption in
the pores, resulting in the relatively low activation energy of
24 kJ/mol.
Osaka University assisted in meeting the publication costs of this article.
References
In the present study, Ni contamination with the 1014 atoms/cm2
order concentration has been investigated to elucidate the removal
mechanism, and dilute HCN aqueous solutions between 10.7
and 1.4 mM concentrations have been employed in order to
decrease the removal rate ͑cf. Fig. 3͒. Even in the case of the
10.7 mM ͑0.029 wt %͒ HCN solutions, the Ni concentration of
1011 atoms/cm2 is decreased by one order of magnitude by immer-
sion for ϳ3 min. When 30 mM HCN aqueous solutions are used for
the cleaning, the Ni contaminants can be removed to less than
ϳ3 ϫ 109 atoms/cm2 ͑i.e., below the lower detection limit of
TXRF͒. Therefore, we think that HCN solutions can be used for
semiconductor cleaning in LSI fabrication processes. For practical
use, the pH of the HCN aqueous solutions should be adjusted to
above 10 to avoid vaporization of poisonous HCN gas. In fact, the
saturated vapor pressure of HCN aqueous solutions with pH above
10 is less than 10 ppm, and in addition, the cleaning ability increases
with pH due to an increase in the electrolytic dissociation probabil-
ity of HCN.8 Etching of SiO2 does not occur at all even in such a
high pH HCN aqueous solution.7 Moreover, it is found that HCN
aqueous solutions have the capability of removing various metals
such as Cu, Fe, Cr, Co, Mn, and Ca from SiO2 surfaces. It is also
evident that the HCN aqueous solutions can dissolve metal Ni from
the following experimental result. After metal Ni was immersed in
50 mM HCN aqueous solutions, the absorption spectra of the HCN
aqueous solutions showed a peak at ϳ270 nm, attributable to Ni-
cyanide complex ions.24 However, the detailed behavior of HCN
aqueous solutions for Ni removal from HF-etched bare Si surfaces
and changes in Si surface roughness caused by HCN solutions are
not known.
1. S. M. Sze, Physics of Semiconductor Devices, p. 21, Wiley, New York ͑1981͒.
2. W. Kern and D. A. Puotinen, RCA Rev., 31, 187 ͑1970͒.
3. H. Akiya. S. Kuwano, T. Matsumoto, H. Muraoka, M. Itsumi, and N. Yabumoto, J.
Electrochem. Soc., 141, L139 ͑1994͒.
4. H. Morinaga, M. Aoki, T. Maeda, M. Fujisue, H. Tanaka, and M. Toyoda, Mater.
Res. Soc. Symp. Proc., 477, 35 ͑1997͒.
5. N. Fujiwara, Y.-L. Liu, T. Nakamura, O. Maida, M. Takahashi, and H. Kobayashi,
Appl. Surf. Sci., 235, 372 ͑2004͒.
6. M. Takahashi, Y.-L. Liu, N. Fujiwara, H. Iwasa, and H. Kobayashi, Solid State
Commun., 137, 263 ͑2006͒.
7. Y.-L. Liu, N. Fujiwara, H. Iwasa, M. Takahashi, S. Imai, and H. Kobayashi, Surf.
Sci., 600, 1165 ͑2006͒.
8. N. Fujiwara, Y.-L. Liu, M. Takahashi, and H. Kobayashi, J. Electrochem. Soc.,
153, G394 ͑2006͒.
9. E. Hogfeldt, Stability Constants of Metal-ion Complexes Part A: Inorganic
Ligands, Pergamon Press, New York ͑1982͒.
10. H. Kobayashi, S. Tachibana, K. Yamanaka, Y. Nakato, and K. Yoneda, J. Appl.
Phys., 81, 7630 ͑1997͒.
11. H. Kobayashi, A. Asano, S. Asada, Y. Yamashita, K. Yoneda, and Y. Todokoro, J.
Appl. Phys., 83, 2098 ͑1998͒.
12. E. Kanazaki, K. Yoneda, Y. Todokoro, M. Nishitani, and H. Kobayashi, Solid State
Commun., 113, 195 ͑2000͒.
13. H. Kobayashi, Y. Kasama, T. Fujinaga, M. Takahashi, and H. Koinuma, Solid State
Commun., 123, 151 ͑2002͒.
14. A. Asano, Asuha, O. Maida, Y. Todokoro, and H. Kobayashi, Appl. Phys. Lett., 80,
4552 ͑2002͒.
15. H. W. MacFarren, Text Book of Cyanide Practice, McGraw-Hill, New York ͑1912͒.
16. N. S. Mclntyre, T. E. Rummery, M. G. Cook, and D. Owen, J. Electrochem. Soc.,
123, 1165 ͑1976͒.
17. G. G. Hawley, The Condensed Chemical Dictionary, Van Nostrand Reinhold, New
York ͑1981͒.
18. Encyclopaedia Chimica, Kyouritsu, Tokyo ͑1987͒.
19. L. M. Loewenstein, F. Charpin, and P. W. Mertens, J. Electrochem. Soc., 146, 719
͑1999͒.
20. L. M. Loewenstein and P. W. Mertens, J. Electrochem. Soc., 145, 2841 ͑1998͒.
21. D. L. Dugger, J. H. Stanton, B. N. Irby, B. N. McConnel, W. C. Cummings, and R.
W. Maatman, J. Phys. Chem., 68, 757 ͑1964͒.
Conclusion
22. P. W. Schindler, B. Fuerst, R. Disk, and P. U. Wolf, J. Colloid Interface Sci., 55,
469 ͑1976͒.
23. T. Hoshino, M. Hata, S. Neya, and H. Morinaga, J. Electrochem. Soc., 151, G590
͑2004͒.
Ni contaminants on Si/SiO2 are present in the form of SiO-NiOH
and this species can be removed by immersion in the HCN aqueous
solutions. The Ni removal proceeds with the initial fast reaction,
followed by the slow reaction. The rate-determining steps for both
24. S. B. Black and R. S. Schulz, Int. J. Environ. Anal. Chem., 72, 129 ͑1998͒.
Downloaded on 2015-03-08 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).