Ž .Diamond and Related Materials 10 2001 1369�1374

Formation and characterisation of c-BN thin films depositedby microwave PECVD

Ali Soltani�, Philippe Thevenin, Armand Bath´CLOES-MOPS, Uni�ersite de Metz & Supelec, 2 rue Edouard Belin, 57078 Metz, Cedex 03, France´ ´

Abstract

Boron nitride thin films were deposited on silicon substrates at low temperature by plasma-enhanced chemical vapourŽ .deposition PECVD , with a microwave surfatron launcher. An organometallic compound, borane-dimethylamine, was used as

the boron precursor. To promote the growth of the cubic phase, a negative self-bias was applied to the sample holder by means ofa 13.56-MHz RF signal. A very high fraction of c-BN was achieved as revealed by infrared spectrometry. The assignation of thecharacteristic TO absorption band was ascertained by the observation of its LO associated mode by performing the IRtransmission spectra at non-normal incidence. The deposited films are well adherent and do not delaminate even after 6 monthsin the atmosphere, and they exhibit a very low roughness, as observed by atomic force microscopy. � 2001 Elsevier Science B.V.All rights reserved.

Keywords: Cubic boron nitride; BN phases; Infrared; Chemical vapour deposition

1. Introduction

Boron nitride can be synthesised in two major crys-Ž .talline polytypes, the hexagonal h-BN and the cubic

Ž . 2 3one c-BN , related respectively to the sp and the sphybridisation, of the chemical bonding of both atomicspecies. The cubic form is very attractive, due to itsextreme properties, similar to those of diamond interms of hardness, thermal conductivity, chemical in-ertness and optical transparency. It is even more stableagainst oxidation up to higher temperature, and can be

� �doped whether p or n type 1 , making it a candidatefor application in power electronics.

The synthesis of c-BN can be performed in variousways, classically at high temperature and high pressure,but to obtain thin films, PVD and CVD are the most

� Corresponding author.Ž .E-mail address: asoltani@ese-metz.fr A. Soltani

widely used methods. In this case, the growth process� �requires bombardment with energetic particles 2 , and

in general the structure is left in compressive stressafter deposition and delamination or cracking of thefilm can occur.

In this paper we present our first results on thesynthesis of cubic BN thin films, and their characterisa-tion. A plasma enhanced CVD method was used forthe sample preparation, and a radio frequency self-biasing voltage was applied to the substrate holder inorder to obtain c-BN. The characterisation involvesinfrared transmittance spectroscopy at normal andoblique incidence, this latter configuration permittingus to observe the longitudinal optical mode coupledwith the transversal one. We have also measured thesurface roughness with atomic force microscopy, andperformed optical bending measurements to obtaininformation about the internal stress in the depositedlayer.

0925-9635�01�$ - see front matter � 2001 Elsevier Science B.V. All rights reserved.Ž .PII: S 0 9 2 5 - 9 6 3 5 0 0 0 0 4 1 6 - 7

( )A. Soltani et al. � Diamond and Related Materials 10 2001 1369�13741370

Fig. 1. Experimental deposition set-up.

2. Experimental details

The depositions were performed with a plasma en-Ž .hanced chemical vapour deposition PECVD set-up,

and a schematic diagram is given in Fig. 1. The activepart of the reactor is composed of a conical quartz wall,and the plasma is sustained by a surfatron launchersituated in its upper cylindrical part. This is connectedto a magnetron working at a frequency of 2.45 GHz,through a wave-guide. The microwave generator canprovide up to 1200 W. The power is transferred to theplasma by tuning the stubs and the short circuit posi-tion, in order to minimise the reflected power down tozero. The reactor was maintained at a pressure of 0.2Torr during the deposition, with a mixture of argon andnitrogen. The gases were introduced at a constant fluxof 10 sccm each, by mass flow controllers.

The samples were heated from the front face by fourhalogen lamps, and a K thermocouple was located inthe substrate holder. The temperature was maintainedat 220�C during the deposition, with the real tempera-ture at the surface estimated at 250�300�C. The boronwas introduced in the reactor by bubbling nitrogen, at aconstant flow rate of 1 sccm in borane-dimethylamine

Ž . Ž .CH �NH-BH BDMA , maintained at 37�C, 1�C3 2 3above its melting point. The pressure inside the vesselwas measured with a capacitive type gauge, and wasgenerally regulated at 3 Torr by acting on a variablevalve at its output in order to obtain a controlled andreproducible BDMA flux.

The substrate holder was made of molybdenum andwas self-biased with a radio frequency signal at 13.56MHz, applied through a matching box. During thedeposition, the samples were located 12 cm below thewave-guide, slightly outside the plasma; the variousparameters are given in Table 1.

The films were deposited on monocrystalline mirror-polished p type silicon substrates, with a resistivity of

Table 1Deposition conditions

Total gas pressure 0.06�0.2 Torr3Ar or N gas flow 20�100 cm �min23N bubbler 0.5�2 cm �min2

Self-bias substrate �30 to �200 VŽ .RF power 13.56 MHz 7�25 W

Ž .Microwave power 2.45 GHz 200�500 WSubstrate temperature �300�C

( )A. Soltani et al. � Diamond and Related Materials 10 2001 1369�1374 1371

Fig. 2. IR spectra of a h-BN film prior and after the CF , O etching4 2process.

Ž .10 ��cm, and 100 oriented. The infrared transmit-tance spectra were measured on the films with aMattson 3000 FTIR spectrometer, covering a regionranging from 400 to 4000 cm�1. A wide band pass

Ž .optical grid polariser SPECAC IGP225 could beadded, to select the S or P component at non-normalincidence. Surface roughness measurements were ob-

Ž .tained by atomic force microscopy AFM .Prior to the deposition, the reactor was actively

cleaned with a plasma composed of a mixture of CF ,4O and Ar gas for 1 min. This cleaning procedure is2very efficient in removing deposits on the wall and theholder. An example is given in Fig. 2, with two IRspectra realised on a test sample, approximately 192nm thick, of a h-BN layer, before and after the etchingprocess. The characteristic peaks at 1373 and 786 cm�1

are drastically reduced, and the resulting surface ap-pears identical to the virgin one.

Fig. 3. AFM image of a c-BN layer surface.

3. Results and discussion

Fig. 3 represents a typical surface topography of thec-BN film obtained by AFM over an area of 1�1 �m2.

No particular morphology can be observed in thepicture, even at a larger scale. Its aspect appears to bevery smooth; the maximal height from the bottom tothe top is 1.78 nm. The average roughness deducedfrom these data is 0. 18 nm.

In Fig. 4, two IR transmittance spectra are reportedwith a resolution of 2 cm�1 on a sample obtained witha self-biasing of �160 V. The plot is restricted between500 and 2000 cm�1, because no additional absorptionpeaks were observed outside this region. The lowercurve was recorded 30 min after the deposition, andexhibits two peaks, one at approximately 1360, and asharper one at 1079 cm�1. They are attributed to twotransversal optical modes, to a turbostratic h-BN in-plane vibration, and to c-BN reststrahlen band, respec-tively.

The upper curve was measured 24 h later, on the

Ž . Ž .Fig. 4. IR spectra at normal incidence of a c-BN film, measured 30 min lower curve and 24 h upper curve after the deposition.

( )A. Soltani et al. � Diamond and Related Materials 10 2001 1369�13741372

Fig. 5. IR spectra recorded at 0 and 60� incidence to the normal,with natural light.

same sample kept at room temperature and in theatmosphere. The measurements were further per-formed in the following weeks, giving the same results,without any additional modification. We can see thatthe hexagonal component vanishes and the cubic BNmode becomes predominant, and shifts toward a newposition at 1072 cm�1. This value is in the neighbour-hood of the normal mode given for c-BN small

�1 � �monocrystals, at 1065 cm after Gielisse et al. 3 , or�1 � �at 1055 cm after Eremets et al. 4 . Since a linear

dependence between compressive stress and the shiftof the mode position expressed in wave numbers has

� � �1 �1been reported 5 at a rate of 3.39 cm �GPa , thisseems to indicate the occurrence of a residual stressrelease of 2 GPa in our film. We tentatively tried tomeasure residual compressive stress by cantilever mea-surements with an optical interferometric method. Therings were too diffuse and too large to allow propermeasurements, and only an upper limit of 1.9 GPa canbe inferred. This is consistent with the fact that our

Fig. 6. IR spectra at 60� incidence, with light in the S and P states ofpolarisation.

oldest film is stable over a period of 6 months withoutany delamination or cracking. We have determinedthat the relative fraction of c-BN in the film is above

� �95%, according to the calibration formula given in 6 ,based on the IR absorption intensity of the hexagonalA and cubic A phases of BN:h-BN c-BN

1.9Ac-BNc-BN � �98%fraction 1.9A �2.5 Ac-BN h-BN

We have also performed some FTIR spectroscopyexperiments at oblique incidence, and the correspond-ing measurements are presented in Figs. 5 and 6. Thefirst was recorded with natural light at an angle of 60�to the sample normal, and a spectrum at normal inci-dence is also given for comparison. Another peak at1267 cm�1 is then activated, which is attributed to the

Ž .Fig. 7. IR spectrum of c-BN and the theoretical optical parameters deduced see text .

( )A. Soltani et al. � Diamond and Related Materials 10 2001 1369�1374 1373

Ž .longitudinal optical mode LO associated with the TO� �mode of c-BN, according to the Berreman effect 7 .

This later stipulates that LO modes can be observed atnon-normal incidence in layered media with a thicknessshorter than the incoming wavelength. Only the P

Ž .component in the plane of incidence of the electro-magnetic wave is concerned in this effect, whereas TOmodes interact with both P and S states of polarisation.

It is consistent with the previous considerations toassign this new peak to a LO mode, as can be observedin Fig. 6. The two spectra were measured at an oblique

Ž . Žincidence of 60� in the S upper curve and P lower.curve state of polarisation, respectively. We can

observe that the LO mode is enhanced relative to theTO in the P configuration.

The LO position is, however, somewhat away fromthe values of 1340 and 1304 cm�1 given in the previous

� �works 4,5 . In other works on c-BN crystals published� � � �by Chrenko 8 and Gielisse et al. 3 , the LO mode is

at 1232, and the TO at 1000 cm�1. The discrepanciesbetween these results, as well as with ours, are not yetunderstood. However, we observe that the LO-modeposition varies from 1242 to 1267 cm�1 in our samples,apparently with increasing thickness, but this pointneeds to be examined further.

The IR optical parameters of the c-BN layer havebeen deduced by using an iterative method. Firstly, thefilm thickness has been calculated according to a model

� � � �described by Mogab 9 and Jones et al. 11 with the� �data for bulk c-BN given by Gielisse et al. 3 . The

corresponding result is plotted and the parameters ofthe calculation are given on the right of Fig. 7.

This gives a film thickness of 90.7 nm, and the c-BNŽ �1damping constant � �83 cm or ��� �� �1�2 a TO

.0.0774 was calculated for each sample by taking the� �full width at half-maximum of the T �T � 1 �0

Ž .f ��� curve, where T is the transmittance back-TO 0Ž �1 .ground taken above 2000 cm , T is the transmit-

tance value at a wavenumber � and � is theTOwavenumber at maximum absorption. These parame-ters are then kept constant to calculate the opticalparameters in a Kramers�Kronig-like method de-¨

� �scribed by Kozima et al. 10 . It consists of adjusting thedata recorded at normal incidence by varying the re-

Ž .fractive indices n and k through the following equa-tion:

� Ž . ��2� ln �TOŽ .� � � P d��2�� dHTO TO2 2� � ��0 TO

� �with P, the principal part of the integral, and �,which are functions of n and k, respectively, and d thethickness.

The optimal values, which properly adjust the experi-mental spectrum, are labelled on the left of Fig. 7.

They are in good agreement with those given in the� �literature 11 . For the permittivity and the absorp-

� �tion coefficient � , the values reported are 5.5 12int6 �2 � �and 2.6�10 cm 13 , respectively, whereas the cor-

responding values in the present work are 5.26 and2.13�106 cm�2 .

Micro-Raman measurements have also been per-formed and do not reveal noticeable information onthe structures, except for a broad band centred around

�1 Ž .1280 cm LO with weak intensity, and anotherweaker peak corresponding to the TO mode at approxi-mately 1030 cm�1, which is embedded in a broad bandof the silicon substrate.

Current�voltage measurements on a Au�p-Si�c-BN�Ag structure give us a resistivity of approximately1015 ��cm of the BN layer for a voltage bias rangingfrom 0.5 to 50 V.

4. Summary

We have deposited thin films of boron nitride by aPECVD method with self-biasing of the samples with aRF signal at 13.56 MHz. These structures contain avery high fraction of c-BN, at more than 95%. Theposition of the LO mode has been determined atapproximately 1267 cm�1 by infrared spectroscopy intransmittance mode at an oblique incidence. The filmsdo not crack or delaminate, even after a period of afew months in the atmosphere, and are characterisedby a smooth surface.

Acknowledgements

We thank Dr A. Dahoun for performing the AFMmeasurements, Dr M.A. Djouadi for supplying us theNewton interferometric apparatus for bending mea-surements, Mr J.C. Petit for technical support and the

Ž .Metz Institute of Technology IUT for access to theFTIR spectrometer.

References

� �1 O. Mishima, K. Era, J. Tanaka, S. Yamaoka, Appl. Phys. Lett.Ž .53 1988 962.

� � Ž .2 T. Klotzbucher, E.W. Kreutz, Diamond Relat. Mater. 7 1998¨1219.

� � Ž .3 P.J. Gielisse, S.S. Mitra, J.N. Plendl et al., Phys. Rev. 155 19671039.

� �4 M.I. Eremets, M. Gauthier, A. Polian et al., Phys. Rev. B52Ž .1995 8854.

� �5 M. Cardona, J.A. Sanjuro, E. Lopez-Cruz, P. Vogl, Phys. Rev.Ž .B28 1983 4579.

( )A. Soltani et al. � Diamond and Related Materials 10 2001 1369�13741374

� �6 M.A. Djouadi, S. Ilias, D. Bouchier, J. Pascallon, G. Sene, V.´Ž .Stambouli, Diamond Relat. Mater. 7 1998 1657.

� � Ž .7 D.W. Berreman, Phys. Rev. 130 1963 2193.� � Ž .8 R.M. Chrenko, Solid State Commun. 14 1974 511.� � Ž .9 C.J. Mogab, J. Electrochem. Soc. 120 1973 932.

� � Ž .10 K. Kozima, W. Suetaka, N. Schatz, J. Opt. Soc. Am. 56 1966181.

� �11 G.O. Jones, D.H. Martin, P.A. Mawer, C.H. Perry, Proc. R.Ž .Soc. A261 1961 10.

� �12 M. Schubert, B. Rheinlander, E. Franke et al., Appl. Phys. Lett.Ž .70 1819 1997.

� � Ž .13 P. Scheible, A. Lunk, Thin Solid Films 364 2000 40.