Deposition of cubic boron nitride with an inductively coupled plasma

SilIIMCE

a641INGS

ELSEVIER

Surfaceand CoatingsTechnology74-75 (1995)806-812

HtllAIOL06

Deposition of cubic boron nitride with an inductively coupled plasma M. Kuhr, S. Reinke, W. Kulisch Institute of Technical Physics, University of Kassel, Heinrich-Plett-Strasse 40, D-34109 Kassel, Germa~y

Abstract

Cubic boron nitride (c-BN) films have been deposited by inductively coupled plasma chemical vapour deposition (CVD). With this technique films with a c-BN content of nearly 75% can be achieved. Investigations on the influence of the process parameters which are most important for depositing c-BN (ion flux, ion energy and substrate temperature) have been carried out. The results show good agreement with our theoretical sputter model. Furthermore, from these experiments it is possible to draw the conclusion that the physical vapour deposition and the ion-induced CVD of c-BN films rely on the same deposition mechanisms among which physical sputtering is most important. Keywords: Cubic boron nitride; Inductively coupled plasma; Langmuir probe; Sputter model

1. Introduction

Owing to the outstanding material properties of cubic boron nitride (c-BN) [-1], the deposition of thin c-BN films has attracted considerable interest during the last few years. The great majority of successful deposition methods described up to now in literature belongs to one of two groups (see [-2]): ion-assisted physical vapour deposition (PVD) (e.g. ion-beam-assisted deposition (IBAD) and ion plating) or ion-induced chemical vapour deposition (CVD) (e.g. electron cyclotron resonance (ECR) plasma and inductively coupled plasma (ICP) plus additional substrate bias). It is therefore evident that ion bombardment plays a decisive role in c-BN formation. In recent publications [3,4] we have summarized all available IBAD data, thereby showing that a well-defined c-BN domain exists which is dependent on the main parameters ion energy, ion-to-atom flux ratio and substrate temperature. Furthermore, a model was developed which assumes selective sputtering of h-BN with respect to c-BN to be the major function of the ion bombardment. Experimental evidence for this model was provided very recently [-1,5]. Nowadays, the IBAD of c-BN is well established and characterized. There are, however, several reasons for investigating (ion-induced) CVD of c-BN. First CVD experiments could extend the range of reliable data on c-BN formation down to lower energies which cannot be reached with IBAD because of the high ion currents 0257-8972/95/$09.50 © I995 ElsevierScienceS.A.All rights reserved

required according to our model. Second, the ion bombardment during c-BN formation causes some serious problems among which the nanocrystallinity and the high compressive stress of the films are most important [2]. Therefore it is worthwhile to investigate whether the ion bombardment can at least in part be replaced by chemical effects. Finally, CVD has advantages with respect to IBAD in terms of costs, throughput, and large-area and three-dimensional coating. On the contrary, it is evident from our model that a CVD technique for c-BN formation must be able to provide high plasma densities at low pressures. The ICP technique has been shown to fulfil these requirements [6,71. In this paper, we shall first describe a self-designed ICP set-up. It will be shown that high density plasmas at low pressures are possible with ICP and, most important, that the ion energy and ion flux can be varied independently and within wide ranges. We shall then discuss the influence of major process parameters such as substrate bias voltage (i.e. ion energy), plasma density (i.e. ion flux) and substrate temperature on the phase composition of boron nitride films. Finally we shall compare our results with those from PVD and other CVD techniques in terms of microscopic deposition parameters, thereby demonstrating that ion assisted PVD and ion-induced CVD of c-BN rely on the same mechanisms, and that sputtering plays a dominant role in these mechanisms.

M. Kuhr et aL/Sutface and Coatings Technology 74-75 (1995) 806-812

2. Experimental

Table 1 Summary of the major process parameters of the inductively coupled plasma process used in our investigation, and of some important properties of the resulting cubic boron nitride film

2.I. Inductively coupled plasma set-up The experiments were carried out in a stainless steel vacuum chamber schematically shown in Fig. 1. On its top plate there is a quartz tube of 10 cm height and 4 cm diameter. A 13.56 MHz r.f. signal (Advanced Energy RFX-600) is coupled inductively into this tube by means of a copper coil. In this way, a very intense plasma is created within the quartz tube which reaches almost to the substrate holder. The signal of a second 13.56 MHz generator (HiJttinger PFG 300) is coupled capacitively to the substrate holder in order to provide a substrate bias. A low working pressure (2 x 10 -2 mbar) is accomplished by means of a turbopump (Balzers TPH 330) backed by a two-stage rotary pump (Leybold Trivac D8B) which leads to a base pressure of 2 x 10 -5 mbar. With high substrate bias and low working pressure, ion energies up to several hundred electronvolts are attainable. The substrate holder can be heated to 800 °C. The gas inlet is a stainless steel ring above the substrate holder. All experiments described in the following sections have been carried out with a mixture of trimethyl borazine (TMB) (HBN(CHa)s), nitrogen and argon. Table 1 summarizes typical process parameters of a c-BN deposition.

2.2. Langnuir-probe measurements Prior to the deposition experiments, the ICP set-up shown in Fig. 1 was characterized by Langmuir-doubleprobe measurements in order to determine the plasma densities with respect to the ion fluxes achievable. The space-charge-limited current from a plasma can be calculated from the plasma density n, and the electron

,v

--

0

0

o

°

tltl I

/

1 5-" I

] pedagog J gc~s Oz~et ccLTacLc'U:4.ve co~z2aZ@zg

Fig. 1. Schematic drawing of the ICP set-up.

807

Process parameters Base pressure (mbar) Working pressure (mbar) R.f. power (inductive) (W) R.f. substrate bias (V) Ion density (cm -3) Substrate temperature (°C) Argon flow (standard cm 3 min -1) Nitrogen flow (standard

2 x 10 -6 2 x 10 -2 80 50-200 5 x 10l° 500 50 25

cm3 rain-I) TBM flow (standard) cm 3 rain -1) Film properties Growth rate (nm min-1) Stoichiometry Carbon content (%) Oxygen content (%) Maximum c-BN content (%)

,~ 1-2

< 10 0.9 ~ FB]/FN] ~<1.1 <5 <2 75

temperature Te [8]:

\mio~/ with e the electron charge, k the Boltzmann constant and mio, the ion mass. From our recently punished sputter model [4] and from the work of Weber et al. [9] it is evident that high ion fluxes are necessary for deposition of c-BN with ioninduced methods such as the ICP technique. Therefore high plasma densities at low pressures are required, while on the contrary low pressures have to be used in order to achieve sufficiently high ion energies. An example of probe measurements is given in Fig. 2. It can be seen that, even at pressures as low as 2 x 10 .2 mbar, plasma densities of about 3 x 101I cm -3 can be achieved. Besides this proof of high plasma densities, Langmuirprobe measurements yield another important result, namely for a given pressure and gas flow the plasma density is determined by the ICP. It is almost independent of the power coupled capacitively to the substrate holder. It follows that the plasma density and therefore the ion flux on the one hand, and the substrate bias voltage and therefore the ion energy on the other hand, can be chosen nearly independently. This is a great advantage of our set-up compared with some other techniques [10]. Furthermore, this is of importance in view of investigations of the ion energy and ion flux dependences of c-BN deposition. To summarize, an ICP set-up has been designed and realized that is able to provide the conditions to investigate the formation of c-BN: high plasma densities, low working pressure, independent control of ion flux and

M Kuhr et al /Smface and Coatings Technology 74-75 (1995) 806-812

808

1.0

was determined by Auger electron spectroscopy (AES). Calibration was performed by means of a standard boron nitride sample with known concentrations of oxygen and carbon, measured by energy-dispersive X-ray analysis. Fourier transform IR (FTIR) spectroscopy (Biorad FTS 60a) was used to determine the c-BN content by a procedure described in ['111. The crystal modification and the crystal size were measured with transmission electron diffraction (TED), and finally the morphology was investigated with scanning electron microscopy.

o.s !

3. Results and discussion

30

I

~ro~e

~-

I

I

I

;l,[eC~S2ZTeTTCe72~S

I /

/

2.5 -- pressure: 2 • lO-2mbctr

I

@/75 sccm Ar

/

pccra~ze ~er,' gas

©

o

O?

2.0 --~

!.5

50 seem AR + 25 . . . . N2/)O?~,..,

•

O

O

CO

0 0

30

60

SO

120

150

i 180

3.1. General fihn properties 210

rf power [W] Fig. 2. Dependence of the plasma density on the inductively coupIed r,f. power for different process gas mixtures.

ion energy and a iarge range of deposition temperatures. Moreover, it is relatively cheap compared with other set-ups with high plasma densities (e.g. ECR).

2.3. Characterization methods The thicknesses and the refractive indices of the films was measured by ellipsometry (Plasmos SD 2100). The composition (stoichiometry and contamination content)

With the ICP set-up shown in Fig. 1, and the parameters given in Table 1, it is possible to deposit boron nitride films with a considerable content of the cubic phase. TED measurements (Fig. 3) confirm the existence of c-BN. They further reveal a crystal size of about 5 nm and a weak texture of the c-BN films. A maximum c-BN content of 75% was obtained as measured with FTIR (Fig. 4) which seems to be typical for CVD techniques [9,10,12], in contrast with IBAD deposition where c-BN contents greater than 90% are possible ['13]. The reason for the limitation to 75% c-BN could be the incorporation of hydrogen during the deposition which is inevitably present in CVD. The deposition of c-BN critically depends on the

Fig. 3. TED picture of a boron nitride film with a c-BN content of nearly 50%,

809

M. Kuhr et al./Smface and Coatings Teclmology 74-75 (1995) 806-812 i.2

I

1,0

c-BN

t

1

t

I

I

cot~#e~ .~ 73%

~0

.~0.6

0 ©

Z

~0o4 <

I

©

0°2 0o@ I750

1500

z-I I I c-BN/t~-BN +BNI

A-BN

~o.8

2000

I

J

c-BN

!250

1000

Wavenumber

[c:-ri- i ]

750

Tes~ztte~n9

+/: I

i/

500 Vc

Fig. 4. IR spectrum of a boron nitride film with a c-BN content of 73% (a.u., arbitrary units).

process parameters. With wrong sets of process parameters, hexagonal boron nitride (h-BN) layers are obtained. As already mentioned, the ion energy and ion flux play decisive roles in the process of c-BN deposition. Their influence will be discussed below. The stoichiometry of the films is also important. With AES measurements we observed a range 0.9 < ['B]/[-N] < 1.1 for the formation of c-BN. The stoichiometry in turn depends on the ion bombardment (sputtering effects) [11] and the flow of the process gases. Furthermore the contamination content must be kept below a certain level, which can be estimated to 10% C and 5% O. Typically, our c-BN films possess an oxygen content less than 2% and a carbon content less than 5%. The main problem with the ICP deposition of c-BN is the stress of the films as is the case with all ioninduced techniques of c-BN deposition. Films with c-BN contents of 30-75% show microcracks or even peel off. No ellipsometric and stress measurements could be performed with such films. From the literature it is known that such microcracks could be avoided with the use of boron-rich or silicon boron nitride interlayers [-14]; we shall investigate the deposition process with such interlayers in the future.

Vm

Vx Vng

b i a s volLage Fig. 5. Schematic diagram of the bias voltage sequence.

content decreases again. At very high Vb (Vb > Vx), films are obtained which show no IR peak due to c-BN but are quite different from the low ion bombardment h-BN. In order to indicate that this material is not the standard h-BN we refer to it as x-BN. It should be mentioned that an sp2BN phase close to the no-growth region was also found in other investigations (PVD as well as CVD) [12,15,16]. Fig. 6 shows the surface of an x-BN sample, h-BN films deposited at low biases are smooth and without any features. In contrast, Fig. 6 shows the x-BN surface to have a distinct morphology. Furthermore, whereas h-BN films are transparent in the visible region, the x-BN films are opaque, which renders ellipsometric determination of thickness and refractive index impossible. Finally, there are distinct differences between the FTIR spectra of h-BN and x-BN. This is evident from Fig. 7 where a spectrum of an x-BN sample is compared with those of commercial h-BN powder

3.2. The bias voltage sequence In all our experiments, the following sequence has been found inevitably as a function of the bias voltage (Fig. 5): h-BN-+c-BN/h-BN~x-BN -+no growth

(2)

This means that for any given set of parameters, at low voltages, h-BN is obtained. If Vu is increased over a certain threshold V~, c-BN-h-BN mixtures are deposited. For Vb > V~, the c-BN content of the films first increases. However, under any conditions a maximum c-BN content is obtained at Vm, while for Vb > Vm the c-BN

Fig. 6. Scanning eIectron micrograph of an x-BN film.

M. Kuhr et aI.ISurface and Coatings Technology 74-75 (1995) 806-812

810

I

I

I

60

t

i

I

I

l

plasma density

5O

g

• P , , ~ 180W ® P,.~= 80W

!i

+o

©

I

0 ~=

GOlf'

o P,~=

40If

ao

© z

Ill IIII IllIIII

2o ,i

lO

"h-bn p owd~~l 2000

1750

1500

1250

wavenumber

tO00

750

..... o,, i

J~ I 'k,,-=tadl ~;

@. 0

500

/~

5O

100

Io 150

fir . . . .

o I o+~'""l 200

250

,\

I 300

"" 350

bias volLage [V]

[ore- 1]

Fig. 7. IR spectra of an x-BN and a h-BN fiIm, and of commercial h-BN powder (a.u., arbitrary units).

Fig. 8. Dependence of the bias voltage sequence on the inductively coupled r.f. power.

(Alfa products; 325 mesh) and an h-BN film deposited in the low voltage region. All three spectra show the two peaks common for h-BN, and there is no hint of c-BN in the spectrum of the x-BN film. However, it can be seen in Fig. 7 that the B-N stretching peak at 1380 cm -~ is extremly broad in the case of x-BN. Such a high full width at hag-maximum of this peak is only found for x-BN films. At the present stage, neither the structure of this x-BN nor the reason for its existence is clear. However, AES measurements of these films yield an interesting result. All x-BN films are boron rich ([B]/EN] ~ 1.1-1.2). Stoichiometry on the contrary has been shown to be an important requirement for c-BN formation (see above). A closer examination of the AES data reveals that preferential sputtering of nitrogen with respect to boron may be the reason for this overstoichiometry [11]. These boron-rich films are stable against ion bombardment at high ion energies and ion fluxes; therefore the existence of x-BN is not in contrast with the sputter model [5]. As has already been pointed out, the bias voltage sequence is observed irrespective of the other process parameters. However, these parameters, and among them namely the substrate temperature T~ and the plasma density % strongly influence the position of the important points (i.e. Vu, Vt, Vm and V~) on the voltage scale, and also the maximum c-BN content at Vm. The influence of T~ and ne on the bias voltage sequence - which is a direct consequence of our sputter model - will be described in the following sections.

This result shows that low ion energies can be compensated by high ion fluxes and vice versa and is in a good agreement with our sputter model. Fig. 9 shows a schematic F-E (F is the number of ions per boron atom and E is the ion energy) diagram according to [2] with a domain where films with c-BN content can be obtained. Below the boundary Fh, o n l y h-BN is deposited whereas, above Fx, no film growth takes places. As discussed above, there is an x-BN domain close to the no-growth boundary. With each bias voltage sequence a cut is performed through the different regions. In order to compare our data with other CVD and PVD data we introduced the parameter F* (F* is the number of ions per incorporated boron atom) E2] which can be calculated according to Eq. (1) and from the known composition of the deposited film. The results are shown in Fig. 10. Since with our ICP set-up the ion flux and ion energy can be varied independently and in wide ranges, we are able to cover a large part of the F*-E diagram. Fig. 10 shows a well-defined boundary over a large energy range which separates the h-BN region from the

3.3. Influenceof the plasmadensityn~ The influence of the plasma density no on the bias voltage sequence is shown in Fig. 8. As predicted by our model, the sequence shifts to lower Vu if the inductively coupled r.f. power is increased since this results in an increase in n~ and therefore in the ion flux (see Fig. 2).

15

-/--2--%

E 0 -~io

C o

,',

"

O _Q rO

5 h-BN

0

F

q

100

200

300

ion energy (eV) Fig. 9. Schematic drawing of the c-BN domain in the F-E diagram,

M, Kuhr et al./Swface and Coatings Tecknolo~, 74-75 (1995) 806-812 tOO0

70 S Bouchier Wada I PVD[ Ulrich

O • O 0

2

This Work I

"m

O

A0

I

60

t~'i~ethyt borazine -argon + nitrogen

5O 4J

© '-'U ol O

i

00 @

-open symbols: h-BN "~] holf symbols: h - B N / c - B N I full symbols: c-BN

lit

•. O0

,$

0 ~I~! ~, • ~ 0 ~0" ~"'--._ 0 0 , I 0 I I t 1 II

100

i0

Ion

i000

Energy (eV)

I

/

:>

'7L'YL72

i

@ •

o

I

ICP

~kl [ Yokoyama J CVI).~

i00

I

811

/

0 o 30

\

/l<

Z

\\N

I o 20

\\

Fig. 10. Dependence of the boundary between h-BN and c-BN on ion energy Eion and the number of bombarding ions per incorporated B atoms for CVD as well as PVD deposition methods. The CVD data were taken from this work, from Yokoyama et al. [17] and from Dworschak et al. [10]. The PVD data were taken from Ulrich et aI. [18], Bouchier and coworkers [19,20], Tanabe et al. [211 and Sueda et al. [15]. The full line represents the dependence of the inverse sputter yield y - t of Si under argon bombardment 1-22], fitted to the data at 500 eV.

10

12 i0 8

c-BN region. An inverse sputter yield can be fitted as discussed in ['2] which indicates that the transition boundary indeed represents a sputter limit as expected from our model. From Fig. 10, two major conclusions can be drawn. First the energy range of c-BN growth reaches down to energies well below 100eV. Lower energies can be compensated by high ion fluxes and vice versa. Second PVD and ion-induced CVD of c-BN rely on the same mechanisms. Therefore chemical mechanisms can be excluded in the ion-induced CVD of c-BN.

3 6 Ts

Ar

I

N2

•

800

32

] 32

•

70o

32

I 52

0

550

52

32

•

500

50

25

coc] i[ oo @oo ] o 2

0 lO0

120

i~0

160

180

200

bias voltage iV] Fig. 11. Dependences of bias voltage sequences and growth rates on the substrate temperature and the Ar-N2 gas mixtures.

3.4. Influence of the substrate temperature T~ Fig. 11 shows the influence of the substrate temperature T~ on the bias voltage sequence. The following observations can be made: with increasing temperature, (1) the sequence shifts to lower energies, (2) the maximum c-BN content increases and (3) the growth rates decrease. Observations (1) and (3) are easily explained by our model ['4]. At higher temperatures, desorption processes become more and more important which leads to a decrease in the deposition rates. As a consequence, the number of incorporated boron atoms per unit time decreases whereas the number of incoming ions remains constant; as a result, F* increases and the sequence shifts to lower energies according to Fig. 8. The explanation for the increase in the maximum, however, needs further investigations. An increase in the surface mobility, or a decrease in hydrogen incorporation are two possible mechanisms. Fig. 11 also shows one set of data for an increased Ar-to-N2 ration in the gas phase. Compared with the data for an Ar-to-N2 ratio of 1 at the same temperature,

the rates are lower, and the maximum of the c-BN content is shifted to lower energies. These effects can be explained by the Langmuir measurements in Fig. 2. The plasma density increases for higher Ar-to-N2 ratios. Therefore the number F* of ions impinging on the surface and alongside increase, leading to enhanced sputtering. We discussed in [2] the fact that the sputter yields of nitrogen and argon are similar and therefore the ion beam composition cannot be the reason for this result.

4. Summary An ICP set-up has been presented which allows the deposition of BN films with a maximum c-BN content up to 75%. Our experiments show a distinct bias voltage sequence. Higher ion fluxes and higher substrate temperatures lead to a shift in the important points of the bias voltage sequence to lower ion energies which can be well explained by the sputter model [2]. Quantification

812

M. Kuhr et al./Surface and Coatings Technology 74-75 (1995) 806-812

of our results in terms of ions per incorporated boron atoms and comparison with other deposition techniques shows that the deposition mechanisms of PVD and ioninduced plasma-enhanced CVD are the same. Sputtering seems to be the most important mechanism, as proposed by our model [21.

Acknowledgements We would like to thank Dr. Pongratz and Dr. Wurtzinger, University of Vienna, for the TED measurements. The authors gratefully acknowledge financial support of the present work by the Deutsche Forschungsgemeinschaft carried out under the auspices of the trinational " D - A - C H " German, Austrian and Swiss cooperation on the "Synthesis of Superhard Materials". This work was also supported financially by the European Community ( B R I T E - E U R A M Project BE 4484-89) which is gratefully acknowledged.

References [11 G. Demazeau,Diamond Relat. Mater., 2 (1993) 197. [2"] R. Reinke, M. Kuhr, W. Kulisch and R. Kassing, Diamond Relat. Mater., 4 (1995) 272. [3] S. Reinke, M. Kuhr, R. Beckmann, W. Kulisch and R. Kassing, Proc. 3rd Int. Syrup. on Diamond Materials, Honolulu, HL May

1993 ElectrochemicalSociety,Pennington, NJ, 1993.

[4] S. Reinke, M. Kuhr and W. Kulisch, Diamond Relat. Mater., 3 (1994) 341. [-5] S. Reinke, M. Kuhr and W. Kulisch, Proc. Int. Co@ on Plasma Swfaee Engineering, Garmisch Partenkirehe~l, 1994, Swf. Coat. Technol., (1995). [63 J. Ocompo, Ew'. Semieond., (September 1991) i5. [-7] J. Amorim, H.S. Maeiel and LP. Sudano, J. Vae. Sei. Technol. B, 9 (1991) 362. [8] B.N. Chapman, Glow Discharge Processes, Wiley, New York,

1982. I9] A. Weber, U. Bringmann, R. Nikulski and C.P. Klages, Proe. 11th Int. Syrup. on Plasma Chemistr}; Loaghborough, 1993,

IUPAC, 1993. [,10] W. Dworschak, K. Jung and H. Erhardt, DiamoM Relat. Mater., 3 (1994) 337. [111 M. Kuhr, S. Reinke and W. Kulisch, Diamond Relat. Mater., 4 (1995) 375. [12] T. Ichiki,T. Momoseand T. Yoshida,J. Appl. Phys., 75(3)(1994). [,'13"1 D.J. Kester and R. Messier, 3. Appl. Phys., 72 (1992) 504. [14] M. Okamoto, H. Yokoyamaand Y. Osaka, Jpn. J. Appl. Phys., 29 (1990) 930. [15] M. Sueda, T. Kobayashi, H. Tsukamoto, T. Rokkaku, S. Morimoto, Y. Fukaya, N. Yamashita and T. Wada, Thin Solid Films, 228 (1993) 97. [,,16] T. Ikeda, Y. Kawate, and Y. Hirai. J. Vac. Sei. Technol. A, 8 (1990) 3168. [-17] H. Yokoyama, M. Okamoto and Y. Osaka, Jpn. J. Appl. Phys., 30 (1991) 344. ['18"1 S. Ulrich, I. Barzen,J. Schererand H. Ehrhardt, paper presented at Diamond Films '94, I1. Ciocco,25-30 September, 1994. [-19] D. Bouchierand W. MNler, Swf. Coat. Tectmol. 5I (1992) 190. [201 D. Bouchier, M.A. Sen6, M.A. Djouadi and P. M61ier, Nucl. Instrum. Method, B, 89 (1994) 369. [21"1 N. Tanabe, T. Hayashi and M. Iwaki, Diamond Relat. Mater., 1 (1992) 883. [223 N. Matsunami, Y. Yamamura, Y. Itikawa, N. Itoh, Y. Kazumata, S. Miyagawa, K. Morita, R. Shimizu and H. Tawara. At. Data Nuel. Data Tables, 31 (1984) i.