NRA characterization of pretreatment operations of silicon

Nuclear Instruments North-Holland

and Methods

in Physics

NRA characterization

Research

Nuclear lnsbuments & Methods in Physics Research Sectloll 6

B64 (1992) 784-788

of pretreatment

operations of silicon *

J-J. Ganem, S. Rigo, I. Trimaille and G-N. Lu Groupe de Physique des Solides, Unirwsit& de Paris 7 et de Pierre et Murie Curie, Tour 23, 2 place Jussieu, F-75251 Paris Cedex 05, Frunce

The aim of this study is to quantify the effects of different treatments generally used in the preparation of silicon for the rapid thermal processing (RTP). The atomic quantities of oxygen were determined by nuclear reaction analysis (NRA) using the “O(d, p) 170* nuclear reaction. This work can be divided in three parts: (1) A study of the resistance to oxidation as a function of time of bare silicon deoxidized with various chemical solutions. We show that silicon deoxidized by HF 4% and rinsed with ethanol exhibits the best oxidation resistance; only 4~ 10” atoms cm-’ of oxygen were found in the sample after a three-week air exposure. (2) A rapid thermal annealing (RTA) of bare silicon at atmospheric pressure, performed in a lamp-heated furnace under Ar and 0, at 700, 775 and 900 o C for RTA times ranging from 10 to 90 s. The oxygen quantities measured by NRA show that under these conditions. the oxidation kinetic is the same for both ambients. (3) An investigation of the kinetic laws that control the rapid thermal oxidation of silicon (100). The treatment temperatures range from 1000 to 1150 0 C for durations varying from 0 to 90 s.

1. Introduction As metal-oxide-semiconductor (MOS) devices are down, ultrathin high quality dielectrics are required. Rapid thermal processing techniques [l] have shown to provide a valuable way to realize reliable, ultrathin dielectric films, and rapid thermal annealing has proved useful in many applications [l]. One of the most important problems in the integrated circuits (IQ) technology is reproducibility. Thus, a control of pretrcatments such as surface etching, preheating treatments and so on, is essential. To achieve this, we quantified pretreatments, generally used by rapid thermal processing (RTP) for the preparation of silicon. The nuclear microanalysis techniques were chosen to determine the atomic quantities of oxygen. scaled

front of the detector, a 13 km mylar foil was placed. The proton group of the fundamental state of the residual nucleus was used to evaluate the oxygen amount in the different samples. These particular cxperimental conditions were shown to provide the best estimation of the oxygen amount in the case of poor atomic quantities [12]. Absolute values of the amount of IhO were determined by comparison with standard refercnccs within a precision better than 2%.

3. Chemical cleaning of Si substrate and ageing in air

Nuclear microanalysis experiments were carried out at the 2 MeV Van dc Graaff accelerator of the Groupe de Physique des Solides at the University of Paris 7. In order to measure the amount of oxygen, we used the nuclear reaction “O(d, p)“O induced by a dcuteron beam of 810 keV with a detection angle of 90 O. In

The initial surface state of the silicon prior to oxidation is a very important parameter for the quality of the dielectric film [2-51. In this study, we used monocristalline silicon, n-type wafers with a (100) surfact plane, a 380 pm thickness, and with a resistivity ranging from 3 to 6 0 cm. In order to study the effect of initial surface state of silicon, we investigated the chemical cleaning effects on silicon (100) substrate and the ageing in air. Six types of chemical deoxidation were used: 1) HF dilute at 48% in HzO; 2) HF dilute at 48% in Hz0 followed by a rinse in

* Work supported by the Centre National de la Recherche Scientifique (GDR 86) and the Groupement de Circuits Inttgrts au Silicium (GCIS), France.

H,O; 3) HF dilute at 48% in H,O ethanol; 4) HF dilute in Hz0 (HF/H,O 5) HF dilute at 4% in ethanol;

2. Sample analysis

0168-583X/92/$05.00

0 1992 - Elsevier

Science

Publishers

B.V. All rights reserved

followed

by a rinse

ratio l/30);

in

785

J-J. Ganem et al. / NRA characterization of pretreatments of Si

0.9 nm of equivalent oxide thickness after a three-week exposure in room atmosphere).

4. Effect of preannealed

treatment

(under argon and

under oxygen)

0

0.5

1

1.S

Exposure to

2

2.5

air [Days]

16 14 12 10 8 6 4 2 0 0

5

10

15

20

25

air [Days] Fig. 1. Equivalent silicon oxide thickness as a function of air exposure times for surfaces initially etched by (0) HF 48%, (~)HF4%,(r)HF1:30,(a)HF48%andarinseinH,O, (m) HF 4% and a rinse in C,H,OH, (0) HF 4% and a rinse in C,H,OH. In (a), the exposure to air varies from 0.2 to 2 days and in (b), the exposure to air varies from 2 to 21 days. Exposure

to

6) HF dilute at 4% in ethanol followed by a rinse in ethanol. Fig. 1 shows the temporal evolution of the equivalent oxide thickness, corresponding to oxygen amounts found on silicon exposed in room atmosphere, after being dipped into the different chemical preparations. HF 48% without rinse provides the best deoxidation; the amount of oxide formed after four hours is less than one monolayer. But after 1 day at room atmosphere, the most important amount of oxygen was found (about three monolayers). This result shows that the surface silicon is the most reactive in the case of HF 48% chemical cleaning. HF dilute at 4% in ethanol and rinse in ethanol gives a good deoxidation (one monolayer silicon oxide after four hours), and an excellent temporal stability of the surface to oxidation (only

To control the parameters which have some effect on thermal oxidation, we quantified the part of low temperature preanneal treatment commonly used in the rapid thermal processes (RTP) to preheat the gas and dry the quartz chamber of the furnace. Fig. 2 shows an example of a typical two-step process generally used with this technique. Initially, the reactor is heated to 700 o C and held at this temperature for 30 s in order to stabilize the pyrometric temperature control system before the temperature is increased to the treatment temperature. At the first processing step, a 1.5 l/min gas flow (N,, Ar, O,, NH,, . . ) is introduced in the quartz chamber. All the silicon wafers were initially deoxized by a chemical cleaning; the samples were dipped in a silicon of HF/C,H,OH with the ratio 1:25 and rinsed in ethanol. Rapid thermal annealings were performed in a modified lamp-heated furnace (AET ADDAX RlOOO 4 in.) at 700, 775 and 900°C for times varying from 10 to 90 s under Ar or 0, at atmospheric pressure. The results as a whole are presented on fig. 3, for the annealing temperatures of 700, 775 and 900 “C. The experimental values show that, at 700 and 775 o C, the amounts of oxygen in the samples are equal for both annealing types. The oxidation resulting from the annealing under argon is probably due to the presence of traces of steam in the quartz chamber. This effect is not negligible and should be taken into account for applications needing a good thickness control. The partial pressure steam in the chamber is difficult to control, so it is worth to use oxygen even in the first step of the treatment.

-I---

@ypical

1

RTO e{i$ RTO

time

m-0 ____________________--__.--temp.

4 i-

J

/-

Ar 5 iimn 02 1.5 limn Fig. 2. Typical two-step process. The gases used are represented at the bottom of the diagram, with their flow rates.

XIV. THIN

FILMS

786

J-J. Ganrm

et al. / MU

of pretreatments

characterization

.-,~_~-.._II

of Si

._._

---_.-L--ILL-d 0

20

40

60

Exposure

time

80

Fig. 3. Rapid thermal oxidation kinetics of silicon under argon (0) and under oxygen and 900 o C.

An illustration of this effect is evidenced in fig. 4, where two rapid thermal oxidation cycles are compared (straight lines): one using a two-step oxidation at 700

100

(s) (01

for three treatment temperatures: 700,77$

and 1050 “C, and one using a one-step oxidation 1050 o C. The two-step RTO provides higher amount oxygen than the one-step RTO.

at of

120

LOO

80

60

One-step.

40

RTO

-

20

.

0

0

10

20

30 RTO

40

so

60

.

..I

70

time ( s )

Fig. 4. Rapid thermal oxidation kinetics of silicon under oxygen at 1050 o C for a one-step RTO (0) and for a two-step RTO (N, 0). The dashed line represents the oxidation kinetics on silicon initially etched by HF 48% and straight lines represent the oxidation kinetics on silicon initially covered by its natural native oxide.

J-J. Ganem et al. / NRA characterization ofpretreatments

5. Rapid thermal oxidation 5.1. Effect

of chemical

etching

on the oxidation

oxygen in each sample was determined by nuclear reaction microanalysis (NRA). Fig. 5 shows the equivalent dielectric thickness (assuming that lOI at. cm-* represents 0.226 nm SiO, equivalent) as a function of RTO times for the four temperatures tested. A shift at zero time is observed in the oxide. It increases with the increased oxidation temperature. This is due to the ramping up temperature in presence of oxygen in the quartz chamber of the furnace. In order to understand the mechanisms of rapid thermal oxidation, we have compared our experimental data to the predicted values of different oxidation models. For 1050, 1100 and 1150’ C treatment temperatures, the Deal and Grove model gives the best fit of the oxidation kinetics. Table 1 gives a comparison between the Deal and Grove parameters, extracted from our experimental data, and the related parameters generally found in literature [7,8] in the case of classical dry oxidation. The values of the parabolic constant, B, were found to be 1.25 to 4.2 times smaller than the parabolic constant of classical dry oxydation with a bigger activation energy. The values of the linear rate, B/A, were found to be 3.3 to 5.4 higher than the values expected in the case of dry oxidation at atmospheric pressure. Since B/A is proportional to the reaction that takes place at the interface Si/SiO,, we can surmise that an important increase of the interface reactivity occurs in the rapid thermal oxidation in the range of tempera-

kinetics

Two two-step rapid thermal oxidations were done at 1050 o C, for RTO times varying between 0 and 60 s, on bare silicon (freshly etched by pure HF) and on silicon covered with its natural native oxide (- 1 nm SiO, equivalent thickness). Absolute values of the total amount of oxygen present in the samples were determined using the nuclear reaction microanalysis (NRA). In fig. 4, dielectric thicknesses as a function of RTO time are represented for the two initial surfaces tested: open squares for bare silicon and black squares for silicon covered with its native oxide. Our results are very similar to Nulman’s [6] in the case of bare silicon. Compared to the oxidation of silicon covered with native oxide, the oxidation rate in the case of bare silicon is much more pronounced (- 50% thicker). This phenomenon shows the importance of the initial surface state and could be due to an increase of the interface Si/SiO, reactivity for a short RTO duration as well as for a longer duration (60 s). 5.2. Rapid thermal oxidation Two-step rapid ered with its natural lent) were done at RTO times ranging

787

of Si

kinetics

thermal oxidations of silicon covnative oxide (- 1 nm SiO, equiva1000, 1050, 1100 and 1150°C for from 0 to 90 s. The total amount of

-RTO

1000

-RTO

1100

-RTO

1050

-+-RTO

1150

I

20

40

60

RTO Fig. 5. Kinetics

of a two-step

rapid

thermal

oxidation

of silicon

80

100

time (s)

covered

with its natural

native

oxide for 1000, 1050, 1100 and

1150°C. XIV. THIN

FILMS

788

J-J. Ganem et al. / NRA characterization

of pretreatments of S

Table 1 Comparison of Deal and Grove parameters between two-step RTO of Si (100) and classical dry oxidation at the temperatures of 1050, 1100 ad 1150°C 1050°C

A [km] B Ikm2/hl B/A [wVhl

11Oo”c

11so”c

Exp.

Dry

Exp.

Dry

Exp.

Dry

4.6x 10-3 3.6X lo-’ 7.8x 10-l

1 x10-’ 1.5x 10-I 1.5XlO~’

3.7x lo-” 7.2~10~~ 1.9

8 x10-2 2.8X 10-I 3.5x10-’

1.4x 10-2 2.8X 10-2 2.0

5.8x IO-’ 3.5 x 10-2 3.5x 10-1

tures between 1050 and 1150 ‘C. The atomic transport mechanisms occurring during the RTO can be understood as a limited diffusion of the oxidizing species, with an increase of the interface reactivity. This effect could be due to the reaction of the photons emitted from the hallogena lamps of the rapid thermal furnace with the interface Si/SiO,. In the case of rapid thermal oxidation at 1000 o C, the Deal and Grove model fails. The best accordance with our experimental data was obtained for Cabrera and Mott model 191, taking into account the fact that the charged mobile defects on surface depend on the surface charge, therefore on the electric field E = V/x. Thus, the dependence of atomic species ./, flow versus the depth x is expressed by [lo]

J, a (l/x) exp(x,/x), since the oxidation

rate dx/dt

dx/dt

a (l/x)

where

t is the oxidation

6. Conclusions

exp(x,/x)

is proportional

or t a 1.x exp( -x,/x)

to J,, dx,

time.

and perspectives

We developed a silicon cleaning method which results in a surface slowly evoluating in air. Furthermore, it was established that during the preheating, in the range of temperatures between 700 and 900 “C, a similar oxide growth occurred under both Ar and O,, probably in relation to the presence of traces of steam and oxygen in amounts difficult to control. A direct oxygen preheating seems, therefore, to be more advisable than an argon preheating in a gas flow system. However, the best way to control this would be to experiment in a pure atmosphere by using an RTP ultrahigh vacuum system. We modelled RTO kinetics using a modified Cabrera and Mott law for an RTO temperature of 1000 o C. For higher temperatures, rapid thermal oxidation of silicon can be described by the Deal and Grove model, but with parameters that deviate from classical oxidation. This indicates an oxidation

limited by the diffusion of the oxidizing species with an increase of the interface Si/SiO, reaction. Further studies seem necessary to comprehend and fully master this phenomenon. It is therefore highly important to determine the microscopic mechanisms of this growth regime, which involves the use of isotopic labelling techniques. This study and many others will be carried out through our ultrahigh vacuum rapid thermal processor [ll], associated with nuclear techniques. Acknowledgements The authors wish to thank E. Girard and SITESA ADDAX for their help in the perfecting of our rapid thermal processing. We are very grateful to E. d’Artemare for his technical support on the Van de Graaff accelerator, and to J. Moulin, who was always ready to provide technical assistance during the measurements. References [l] R. Singh, J. Appl. Phys. 63 (1988) R.59. (21 N. Hirashita, M. Kinoshita, .I. Aikawa and T. Ajoka, Appl. Phys. Lett. 56 (1980) 29. [3] P.O. Han, M. Grundner, A. Schnegg and H. Jacob, Appl. Surf. Sci. 39 (1989) 436. [4] A. Ishizaka and Y. Shiraki, J. Electrochem. Sot. 133 (1986) 666. [5] W. Kern and D.A. Puolinen, RCA Rev. 31 (1970) 187. [6] J. Nulman, Proc. Electrochem. Sot. Meeting, J. Electrochem. Sot. (November 1986). [7] B.E. Deal and AS. Grove, J. Appl. Phys. 36 (1965) 3770. [8] L.N. Lie, R.R. Razouk and B.E. Deal, J. Electrochem. Sot. 129 (1982) 2828. [91 N. Cabrera and N.F. Mott, Rep. Prog. Phys. 12 (1949) 163. [lo] A. Atkinson, Philos. Mag. B55 (1987) 637. [l l] S. Rigo, these Proceedings (10th Int. Conf. on Ion Beam Analysis, Eindhoven, Netherlands, 1991) Nucl. Instr. and Meth. B64 (1992) 1. [12] J-J. Ganem, S. Rigo, I. Trimaille, G.N. Lu and P. Molle, these Proceedings (10th Int. Conf. on Ion Beam Analysis, Eindhoven, Netherlands, 1991) Nucl. Instr. and Meth. B64 (1992) 778.