Study of the chemical composition of silicon nitride films obtained by chemical vapour deposition and plasma-enhanced chemical vapour deposition

Study of the chemical composition of silicon nitride films obtained by chemical vapour deposition and plasma-enhanced chemical vapour deposition

Surface and Coatings Technology, 45 (1991) 137 146 137 Study of the chemical composition of silicon nitride films obtained by chemical vapour deposi...

401KB Sizes 0 Downloads 43 Views

Recommend Documents

No documents
Surface and Coatings Technology, 45 (1991) 137 146

137

Study of the chemical composition of silicon nitride films obtained by chemical vapour deposition and plasma-enhanced chemical vapour deposition C. Oliveri and F. Baroetto Co.Ri.M.Me., Consorzio per la Ricerca Microelettronica nel Mezzogiorno, Stradale Primosole 50, 05121 Catania (Italy)

C. Magro SGS-Thomson s.r.l., Stradale Primosole 50, 95121 Catania (Italy)

Abstract Silicon nitride films were deposited by pyrolytic chemical vapour deposition (CVD) and plasmaenhanced CVD techniques. The effects on the stoichiometry of variations in the reactant gas flow were evaluated. Electron spectroscopy for chemical analysis, Fourier transform infrared (FTIR) and secondary ion mass spectrometry (SIMS) were used in order to clarify the differences in the composition of the films. The analytical information obtained with these techniques included the silicon-to-nitrogen ratio, N - - H and Si--H group content, surface oxide and bulk contamination. It was found that the classic FTIR approach was unable to detect the low Si--H content in CVD films. This low S i - - H level could, however, be detected by SIMS without interference at 31 a.m.u, as a 30Sill + species.

1. Introduction Silicon nitride films are extensively used in microelectronics as a passivation layer on top of finished devices, as interlayer dielectrics in multilevel m~tal structures, in trench filling, as spacers and as an oxidation mask for obtaining mean local oxidation of silicon (LOCOS) structures. These applications require high performance materials, well characterized, and good control of the repeatability of the deposition process. Analytical techniques such as electron spectroscopy for chemical analysis (ESCA), Fourier transform infrared (FTIR) and secondary ion mass spectroscopy (SIMS) are becoming routine tools for process control. In this work these techniques were used to determine the stoichiometry of films obtained under different experimental conditions. Chemical analysis showed that these materials, produced by chemical vapour deposition (CVD) and plasma-enhanced chemical vapour deposition (PECVD), were substantially different from a chemical composition point of view and also t h a t their composition was heavily influenced by the deposition parameters, even though sometimes they are referred to by the same general name. This point is reinforced by the fact t h a t among the numerous parameters of fundamental importance for deposition, only the partial pressures o f Elsevier Sequoia/Printed in The Netherlands

138 the r eactan t s were varied during the experiments. Since the polarizability, dielectric constant and refractive index are related to the stoichiometry, they are dependent on the composition. From the applications point of view, aspects other t han the composition of the nitride play an important role in some key processes. For example, etching, either dry (in CF4/SF ~ plasma) or wet (in boiling 85% H3PO4) , may be hindered because of the surface oxide layer (50-150 • thick). In other cases this oxide itself may be used as an etch mask because of the selectivity of the etchant to the underlying bulk film. The differences in composition and in deposition process result in different uses of the two types of films. Their purity and good stoichiometry make CVD films useful as inert barriers for oxidizing species during processes involving local oxidation of silicon (LOCOS). PECVD gives bad results in this type of application, so it is used in applications where high thickness and low temperature processing are needed, for example in the final passivation of aluminium metallized devices. A not he r example of the effects of the composition of the nitride is in the stability of the electrical performance of high voltage ( > 1000V) p-channel m e t a l - o x i d e - s e m i c o n d u c t o r (PMOS) transistors. The hydrogen, present in the film as S i - - H and N - - H in varying proportions depending on the deposition conditions, has an effect related to the reduction of the breakdown voltage (BV) of the devices, of the order of 100 200 V, which occurs after the passivating deposition. In common with other authors [1], we observed a lesser BV reduction in silicon-rich films with a higher S i - - H content. Despite the general opinion of many [2, 3] authors as to the negative effects of hydrogen content in nitride films, we conclude that this is largely true, but not necessarily in every case.

2. E x p e r i m e n t a l details The set of experiments consisted of both plasma-assisted and pyrolytic depositions using pure electronic grade reagents with a range of flow ratios as follows: (i) PECVD: ammonia-to-silane ratios of (a) 4:1; (b) 6.4:1 and (c) 10:1; (ii) CVD: ammonia-to-dichlorosilane ratios of (d) 2:1; (e) 3:1 and (f) 10:1. PECVD samples were prepared in a JANUS series react or made by ASM. A 60-wafer graphite boat was used for the deposition run. The boat was powered by an ALSATHERM r f. generator with an output frequency of 40 kHz. The power density was 0.030 W cm -2. The deposition t em perat ure was 380 °C at a pressure of 1400 mT. Typical flow rates used for the experiments were in the order of 100-1000 sccm. CVD samples were prepared in a SEMY Eng. furnace with wafers carried on a quartz boat introduced by a soft-lander. The operating t em pe r at ur e was 790 °C and the pressure was 200 mT. The flow was varied in the range 10-100sccm. A BIORAD model FTS 40 FTIR spectrometer was used for infrared analysis. The films were deposited on high resistivity silicon test wafers; the background s p e c t r a were subtracted from the data,

139

TABLE 1 Atomicratiosofsilicon, nitrogen and o x y g e n i n t h e filmsstudied PECVD NH3:SiH4

CVD NH3:SiH2C12

4:1

6.4:1

10:1

2:1

3:1

10:1

Si3N2.1O0.5

Si3N2.500.~ Si3N2.201.6 (no HF dip)

Si3N2.~Oo.~

Si3N3.2O0.4

Si3N300.3 Si3N2.7Ol.s (no HF dip)

Si3 N3.200.5

ESCA analysis was performed by using an ESCA ES 300 KRATOS spectrometer, operating in FRR mode. A1 K radiation was used (1486.6 eV). The SIMS spectra and depth profiles were recorded in a SIMSLAB from VG IONEX. The primary ion beam source was an Ar ÷ gun operating at 5 keV. The primary ion current was kept at a low level (typically 10 nA). For the spectra the t a r get was set so as to obtain the maximum count in defocused quadruple mode (zero resolution).

3. R e s u l t s

We determined the stoichiometric fractions of silicon, nitrogen and oxygen by ESCA (Table 1), starting from a quantitative evaluation of expanded Si2p, N 2s and O 2s peaks. The composition o f the as-deposited sample was measured in cases (b) and (e): their surfaces were strongly oxidized in both cases (see the general ESCA scans in Fig. 1). After a i rain dip in diluted HF the analysis showed an oxygen amount one third less t han in u n t r e a t e d samples. The samples must have been reoxidized during drying and handling before the analysis, but it seems clear t hat a certain amount of oxygen was contained in the bulk of the nitride. A sample sputtered for 2 rain by an Ar ÷ 8 keV beam at 10 -7 Torr and analysed without exposure to the atmosphere registered a comparable oxygen c o n t e n t . Furthermore, a SIMS depth profile of a 44 a.m.u, species associated with SiO ÷ (not reported here) confirmed the presence of the oxygen. This was despite the fact t hat oxidizing species were absent during deposition and t hat the venting and stand-by of the tubes was all under nitrogen atmosphere. In all the samples the silicon-to-nitrogen ratio was always lower t han the nominal SigN4. The results for the CVD samples were more nearly stoichiometric th an those for the PECVD ones and they had a more nearly constant composition within the range of variation in reactants and partial pressures. The FTIR data confirmed t hat CVD produced a purer film with a ~low hydrogen content which was detectable only as N - - H (Fig. 2(a)). In films made by PECVD the S i - - H peaks increased with Sill 4 partial pressure. For sample (a) in particular, it was also related to the greater amount of silicon

140

PECVD

N

O

l J i

CVD

Sl

i

2,~ ~ 2p N SI

4:1

= i

L l

F

~=\!~ .

no~ldlp

,

10:1 O

II v

6.4:1

I

i

'

!

3:1

,o,

tl

io K.E.(eV)

s~o

2:1

L 1~oo

K.E.(eV)

1~00

Fig. 1. ESCA spectra: general scans of PECVD and CVD samples before and after dip in HF. c o m p a r e d with n i t r o g e n (Fig. 3). The S i - - H p e a k r e m a i n e d c o n s t a n t on going from sample (b) to sample (c). The N - - H peaks i n c r e a s e d with N H 3 p a r t i a l pressure on going from (a) to (c) and also on going from (d) to (f) [4, 5] (Figs. 2(b) and 3(b)). SIMS for samples made by CVD and P E C V D gave a similar p a t t e r n for the ions observed (Fig. 4). These d a t a t a k e into a c c o u n t the

141

4O00.0

3000.0

2000.0 WAVENUMBERS

1000.0

400.0

(a)

2:1

3600.0

34()0.0

3200.0

3000.0

WAVENUMBERS

(b) Fig. 2. (a) FTIR spectra from CVD samples (1000/~ t h i c k n e s s on high r e s i s t i v i t y bare silicon); (b) e x p a n d e d N - - H peaks for CVD samples. '~ ~ , ~:

142 • 6.4:1

4000.0

3000.0

/

2000.0

1000.0

400.0

WAVENUMBERS

(a)

N-

° i ,

3600.0

,

3ooo.o

2s~o.o

2ooo.o

%

18oo.o

WAVENUMBERS

(b) Fig. 3. (a) FTIR spectra from PECVD samples (1000/~ thickness on high resistivity bare silicon): (b) expanded N - - H peaks for PECVD samples (normalized with respect to Si--H peaks).

143

PECVD (b) SI~

I

|

I-H ~r

| |

H~

SI2N "~

1 / lil30Si--H + *,

SI2+ k

1 JLAI

|

| | O,|

I.I

aO,O

30.0

40.0

W.O

W,O

70.0

IO.O

W.O

~DO.O

Atomic Haso L~te

(a)

il

CVD

: Ill tll.I o.o

=.o

,,,.o

i.o

28S1=92 % 29SI-4.6 % 30SI-3,4 %

Ce)

Ill dJ HItu Ill,llJ ~,J,lJ ~.o

mo

=.o

~.o

=.=

=.o

-,.o

Atomic Mass Units

(b) Fig. 4. SIMS s p e c t r a for s a m p l e s (b) a n d (e): p r i m a r y i o n s were of a 5 keV, 10 n A A r + beam.

isotopic a b u n d a n c e of silicon, so the g r o u p s of p e a k s f r o m 28 to 31 a.m.u, m u s t be e s s e n t i a l l y r e l a t e d to Si + a n d S i l l +. In fact, the p e a k at 31 a.m.u, m a y be r e l a t e d to the 30Sill ÷ species a n d it is useful as a m a r k e r for S i - - H c o n t e n t . T h e d e p t h profile of the 28 a.m.u, species s h o w e d a n i n c r e a s i n g c o u n t w h e n

144

passing th r ough the interface between the deposited sample and the bare silicon substrate. For the species at 29 a.m.u. (28Sill ÷ and 29Si ÷) there was a decrease in counts after the interface; this means t h a t the peak was related to the hydrogenated species whose concent rat i on was higher in the nitride th an in bare silicon. This depth profile was collected by monitoring of neutral species (SNMS), which have a higher population t han ions, among the sputtered particles; this leads to a better quantitative evaluation since it is based on more counts. The 31 a.m.u, species was absent from the depth profile because of its weak intensity. In fact, given the hypothesis t hat its originated only from the 29Si ÷ species, the trend should be the same as with the 28Si ÷, with proportions of 92% and 4.6% owing to their isotropic abundance. Now we can conclude t ha t the 31 a.m.u, peak is doubtless due to the 30Sill ÷ species (Fig. 5). The quantitative evaluation of this species contained in our PECVD samples is in agreement with the FTIR S i - - H peak intensity. A more accurate quantitative determination is in progress. For the evaluation of N - - H content

536

I

SI3N

SI

-

sub

1700

3.84

~o X L.Z 0

1-92 ~SI,

28SI--H

I I

I 28 63 TIME

5 727

L~ID1

s

Fig. 5. S N M S d e p t h profile for s a m p l e (b) (1700/~ t h i c k on b a r e silicon): p r i m a r y i o n w a s of a n 8 keV, 90 n A Ga + b e a m .

145 we have not found a marker species; the 15 a.m.u, peak has a low intensity and is not unfailingly related to the species NH ÷.

4. D i s c u s s i o n The stoichiometry (silicon-to-nitrogen ratio) of all the films, especially in sample (a), showed considerable nitrogen depletion. With r e a c t a n t partial pressure above the 6.4:1 ratio there was no meaningful effect on the stoichimetry of the films. This means t hat in this range of values the limiting factor was not the partial pressure of ammonia. With CVD, probably because of the slower deposition rate and the lower deposition pressure, there were no visible effects attributable to variations in the partial pressures of the reactants. This low sensitivity was probably due to the thermodynamic control of the formation of the activated reagent species, which was driving the pyrolytic reaction, in contrast with the kinetic control in PECVD. The presence of oxygen in the bulk of both CVD and PECVD layers was probably associated with its high reactivity. Consequently, the residual partial pressure, caused only by outgassing or leakage in the deposition system, was enough to give a sufficient and constant source of oxygen. The surface oxide was formed during exposure to the atmosphere, especially during extraction of the hot boat from the furnace. The higher hydrogen content in the films made by PECVD was probably due to the low temperature of deposition at which there would have been incomplete decomposition of the starting materials. The activated radical species formed in the plasma arise by a partial breaking of the N - - H and S i - - H bonds previously contained in ammonia and silane. FTIR spectroscopy showed the presence of N - - H in all the samples, and of S i - - H in the PECVD ones. With the CVD films the peak at a wave number of 2300 cm 1 could not be observed but similar patterns could be observed in the SIMS spectra of both types of films. A semiquantitative comparison made on the basis of 31 a.m.u, peak counts fits well with the FTIR data, as shown in Fig. 6.

5. C o n c l u s i o n The true composition of silicon nitride obtained by CVD and PECVD is quite different from the nominal Si3N 4. The presence of hydrogen and oxygen affects the composition so t ha t it can never reach the fully stoichiometric ratio, as happened both in our samples and in those described by other authors. CVD layers were only weakly influenced by large variations in the r e a c t a n t partial pressure. In every case pyrolytic CVD films were more nearly stoichiometric and purer t ha n PECVD ones. In PECVD, the silane-to-ammonia ratio may be used to modify the total quantity of hydrogen and its distribution as N - - H or Si--H. Over this range we observed no remarkable effects on composition. By SIMS analysis we were able to detect S i - - H even

146 ~

'

I

'

I

. . . .

I

. . . .

I

. . . .

I

. . . .

I

. . . .

I

. . . .

I

. . . .

I

. . . .

I

. . . .

I(m/e=31)

[ (m/el28)

1

4:1

PECUD 6.':1 : 1

18:1

CUD(3

,

,



I

,

,

,

,

. . . .

I



,

.

.

.

.

.

.

I

,

,

,

,

,

,

,



I

:1)

. . . .

Fig. 6. Relative intensity of 31 a.m.u, peaks in SIMS of PECVD and CVD samples (f).

w h e n its c o n t e n t was so low t h a t it was n o t d e t e c t a b l e by F T I R spectroscopy.

References V. S. Dharmadhikari, J. Vac. Sci. Technol., A6(3) (1988) 1922-1923. N.-S. Zhou, S. Fujida and A. Sasaki, J. Electron. Mater. 14 (1985) 55 72. S. V. Nguyen and K. Albaugh, J. Electrochem. Soc., 136(10) (1989) 2835-2840. V. G. Mokerov, B. K. Medvedev, V. V. Saraikin, N. M. Manzha and A. S. Ignat'ev, Soy. Tech. Phys. Lett., 7(10) (1982) 534-535. 5 P. P a n and W. Berry, J. Electrochem. Soc., 132(12) (1985) 3001-3005. 1 2 3 4