Non-destructive determination of the boron concentration of heavily doped metallic diamond thin films from Raman spectroscopy

Non-destructive determination of the boron concentration of heavily doped metallic diamond thin films from Raman spectroscopy

Diamond and Related Materials 13 (2004) 282–286 Non-destructive determination of the boron concentration of heavily doped metallic diamond thin films...

164KB Sizes 7 Downloads 64 Views

Recommend Documents

No documents
Diamond and Related Materials 13 (2004) 282–286

Non-destructive determination of the boron concentration of heavily doped metallic diamond thin films from Raman spectroscopy M. Bernard, A. Deneuville*, P. Muret ´ ´ Electroniques des Solides, CNRS and University of Grenoble, BP 166, 38042 Grenoble Cedex ´ Laboratoire d’Etudes des Proprietes 9, France

Abstract Raman spectra of heavily boron doped polycrystalline diamond films prepared by MPCVD exhibit ‘500’ and ‘1220 cmy1’ structures on a wider boron range than the ‘Fano’ deformation of the ‘1332 cmy1’ peak of diamond induced by the metallic conductivity of these films. Therefore, these peaks were ascribed to the boron concentration rather than directly to the metallic conductivity of the films. The ‘500 cmy1’ wide peak down-shifts as the boron content in the film increases. It can be fitted satisfactorily by the sum of a Gaussian and a Lorentzian component. The wavenumber of the Lorentzian component exhibits a simple and accurate one-to-one relationship with the boron concentration. Therefore, analysis of the Raman spectra can be used as a non-destructive technique to deduce easily and without contact the boron concentration of diamond films in the range 2=1020 to 1022 cmy3. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Diamond; Thin polycrystalline films; Boron doping; Metallic conductivity; Raman

1. Introduction Metallic conductivity of diamond thin films is needed for all the electrical devices to achieve ohmic contact. This requirement is especially stringent in the devices where the contact resistance of the ohmic contact on monocrystalline films has to be as low as possible, therefore, in the conducting state of power Schottky diodes, and in the ‘on’ state of MESFET used for RF electronic. It is also required for polycrystalline films in (temperature, pressure) sensors, MEMS, and electrode to remove contaminant from water, by electrochemical oxidation of organic species as well as reduction of nitrates. Low resistance ohmic contact is obtained when the current crosses a very thin potential barrier between the metal and the diamond. This barrier thickness decreases as the concentration of dopant increases w1x. Therefore, more precisely than the conductivity, the knowledge of the dopant concentration is needed to optimize the ohmic contact. This requirement is also stringent for the diamond electrodes as their efficiency increases by three orders of magnitudes when the boron concentration increases above the concentration corre*Corresponding author. Tel.: q33-476881009; fax: q33476887988. E-mail address: [email protected] (A. Deneuville).

sponding to the threshold of metallic conductivity w2x. At that time, the boron concentration of metallic films cannot be derived from cathodoluminescence because there are unknown non-radiative processes. It can only be obtained from destructive SIMS measurements or from a very accurate measurement the IR absorption of thin (-1 mm) films deposited on transparent substrates. For polycrystalline metallic films, we will show here that it can be easily derived from Raman measurements. We showed that polycrystalline metallic films exhibit a very special Raman spectra—with the disappearance of the classical ‘1332 cmy1’ peak of diamond, whose intensity decreases drastically while the peak widens with a downshift in its wavenumber—and two wide new bands appear at approximately 500 and 1220 cmy1 w3x. This behavior was confirmed by the later works on polycrystalline films w4–7x and was seen to hold also for (100) w8,9x and (111) monocrystalline films w6x and various faces of crystallites w10x. The origin of the 500 and 1220 cmy1 structures still remains unclear. We showed previously in monocrystalline films that the 1220 cmy1 structure exhibits only a very small shift while the 500 cmy1 has a significant shift with the boron concentration in the films w8x. Disorder w3,4x and intraband optical transitions w10x were proposed as the origin of the 500 cmy1 band. The metallic conductivity

0925-9635/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.diamond.2003.10.051

M. Bernard et al. / Diamond and Related Materials 13 (2004) 282–286

283

Fig. 1. Raman spectra of heavily boron doped polycrystalline films vs. wavenumber for ByC in the gas phase varying from 2240 to 14 000 ppm.

is obtained on a somewhat narrow boron concentration range (3–8=1020 cmy3) in monocrystalline films, but on a larger range on polycrystalline films as shown hereafter. We will show that the sums of a Gaussian and a Lorentzian component satisfactorily reproduce the shapes of the ‘500 cmy1’ band of these polycrystalline films. Moreover, there is an accurate one-to-one relationship between the wavenumber of the Lorentzian component and the boron concentration in these films. Therefore, the wavenumber of this component can be used to check the boron concentration in these films. 2. Experimental techniques The polycrystalline films were deposited during 24 h at a pressure of 30 Torr by Microwave plasma chemical vapor deposition on silicon at 880 8C with gas mixtures with 99.5%H2 y0.5%CH4, and parts per million of B2H6, with ByC ratio in the gas phase varying from 2000 to 14 000 ppm. The films thicknesses decreases

from 8 to 4 mm as the ByC ratio in the gas phase increases from 2240 to 14 000 ppm. The micro-Raman spectra were recorded at room temperature with a ‘LabRam infinity’ apparatus from Dilor, excited by a He–Ne laser at 632.8 nm, detected by a liquid nitrogen cooled CCD detector. The film conductivity was measured from 300 to 660 K by a KEITHLEY 236 measurement unit. 3. Experimental results The Fig. 1 shows the Raman intensity vs. wavenumber when the ByC ratio in the gas phase increases from 2240 to 14 000 ppm. The classical ‘1332 cmy1’ diamond peak still appears for 2240 ppm, with two other broad structures at approximately 500 (peak) and 1220 (band) cmy1. There is also a narrow peak at 520 cmy1 from the silicon substrate, because even with a thickness of 8 mm and a boron concentration of approximately 1020 cmy3, the

284

M. Bernard et al. / Diamond and Related Materials 13 (2004) 282–286

calibration by SIMS for several B concentrations in the films w11x. Then, Fig. 3 gives the variation of the boron concentration within the film vs. the wavenumber of the Lorentzian component of the ‘500 cmy1’ wide peak. From 1022 to 3=1020 cmy3 the logarithm of the boron concentration varies exponentially with y0.048 W when W increases from 432 to 500 cmy1. 4. Discussion

Fig. 2. Example for ByCs8000 ppm in the gas phase of the fit of the ‘500 cmy1’ peak by a Lorentzian and a Gaussian component.

films still remain transparent until they exhibit metallic conductivity. The 520 cmy1 disappears completely for the slightly thinner film with ByCs2800 ppm, while the ‘1332 cmy1’ peak widens and shifts to lower wavenumber and the relative intensities of the 1220 and 500 cmy1 peaks increase significantly. The ‘1332 cmy1’ peak decreases and down-shifts as the ByC ratio increases. This peak disappears for ByCs4800 ppm, where it remains only a shoulder in this wavenumber range. In the same time the relative intensities of the 500 and of the 1220 cmy1 structures increase. From By Cs2240 to 12 000 ppm, there is a progressive and significant down shift of the 500 cmy1 wide peak, but a very small downshift of the 1220 cmy1 band. There is a striking evolution of the whole spectrum for ByCs 14 000 ppm: two new peaks appear at approximately 1350 and 1580 cmy1, the ‘500 cmy1’ peak remains, while the ‘1220 cmy1’ band appears only as a shoulder at approximately 1200 cmy1 on the 1350 cmy1 peak. The 500 cmy1 peak cannot be fitted by a simple elementary (Gaussian or Lorentzian) shape. After subtraction of the background, its purely empirical fit as the sum of a Gaussian and a Lorentzian component is found satisfying. An example of the fit for ByCs8000 ppm, with a Lorentzian component of approximately 456.8 cmy1 and a Gaussian component of approximately 507 cmy1 is given on the Fig. 2. The quality of the fit is similar for the other ByC ratios. There is no clear correlation between the wavenumber of the Gaussian component and the ByC ratio in the gas phase, while that of the Lorentzian decreases systematically as the ByC ratio increases. Moreover, there is a one-to-one relationship between the wavenumber W of the Lorentzian component and the ByC ratio, in the ranges 500)W)432 cmy1 and 2800-ByC-14 000 ppm. Quite generally, the B content in the solid phase was determined for each ByC ratio in the gas phase from the infrared absorption of the Local Vibrational Mode of the boron at approximately 1290 cmy1, after

The Fig. 1 shows for ByCs2240 ppm, wide structures of approximately 500 cmy1 and a shoulder of approximately 1220 cmy1, with the usual Raman diamond peak of approximately 1332 cmy1. The relative intensities of the ‘500 cmy1’ peak and of the ‘1220 cmy1’ band increase with increasing boron content in the gas phase, up to ByCs2800 ppm (3=1020 Bcmy3 according to SIMS). In the same time, the intensity of the ‘1332 cmy1’ diamond peak decreases. It widens and shifts to a lower wavenumber. Such variations are kept up to ByCs12 000 ppm where the ‘1332 cmy1’ peak appears only as a shoulder. For ByCs 14 000 ppm, two new broad peaks appear at approximately 1350 and 1580 cmy1, which originate from micro crystalline graphite w12x. However, the X-ray diffraction of this film shows only diamond, i.e. less than approximately 1% of the graphite. But, its specific Raman peaks appear because their Raman cross section are approximately 100 times higher than that of the ‘1332 cmy1’ diamond peak w12x. From Fig. 4, the low concentration of graphite induces deformations of both the ‘500 cmy1’ peak and of the ‘1220 cmy1’ band, while the shoulder corresponding to the ‘1332 cmy1’ peak has completely disappeared. The striking variation of the ‘1332 cmy1’ diamond peak from ByCs2800 to 12 000 ppm originates from a

Fig. 3. Boron concentration in the heavily doped polycrystalline films vs. wavenumber of the Lorentzian component of the ‘500 cmy1’ peak.

M. Bernard et al. / Diamond and Related Materials 13 (2004) 282–286

Fig. 4. Conductivity vs. reciprocal temperature of the heavily boron doped polycrystalline diamond films.

quantum interference between the phonon and a continuum of electronic state in a degenerate semiconductor (‘Fano’ effect w8,9x), i.e. only when the diamond films exhibit metallic conductivity). The variation of the conductivity with the reciprocal temperature distinguishes clearly the ByC range corresponding to the metallic conductivity of the diamond films. From Fig. 4, only the diamond films with 2800FByCF12 000 ppm exhibit metallic conductivity, while from the Fig. 1, the 500 and 1220 cmy1 structures appear in the wider 2240y14 000 ppm range of ByC. This might suggest that these signals originate from a second phase responsible for conductivity, which would give metallic conductivity from percolation at a higher ByC value. However, the phase distribution and, therefore, the percolation threshold would be different in polycrystalline and in homoepitaxial films while we obtain the same boron concentration (wBxf3=1020 cmy3) for the metallic conductivity threshold. Therefore, this hypothesis can be ruled out. Thus, the ‘500 cmy1’ and ‘1200 cmy1’ bands are not directly connected to the metallic conductivity in the diamond, but rather to the boron concentration in the films. We previously found the Fano effect in monocrystalline diamond films w8x with B concentration (wBx) from 3 to 8=1020 cmy3 according to SIMS. It appears here for the same threshold, but up to the higher boron concentration of 5–6=1021 cmy3. Such high values were also found by the other groups w1,7x. As the films are polycrystalline, a significant concentration of boron

285

atoms might be trapped within parasitic phases at the grain boundaries. However, the boron concentration is calculated here from the infra-red absorption from the Local Vibration Mode of boron atoms in the diamond lattice. Therefore, IR checks only the boron concentration within the diamond phase, while the SIMS measures the total concentration of boron. This is a mean value, as the boron incorporation rate in polycrystalline films varies according to the growth sector. As the threshold obtained here for metallic conductivity in polycrystalline films is identical to that obtained for the monocrystalline films w8x, there is a negligible concentration of boron trapped in the grain boundaries of these polycrystalline films. The higher boron concentration found in diamond polycrystalline films might originate from the higher deformation allowed in the grain of a polycrystalline film than in a monocrystalline film. As the top of ‘500 cmy1’ peak is somewhat flat (Fig. 2), a fit has to be used to check with a sufficient reproducibility, the wavenumber corresponding to this broad band. A purely empiric satisfying fit of this broad peak is obtained as the sum of a Gaussian and a Lorentzian component. Several origins (disorder, intra band optical transitions,...) connected to the boron atoms have been proposed for this broad peak, which might give several physical components to this broad peak (additional work is in progress in our group to get a better knowledge about the origin of this ‘500 cmy1’ peak). We find (Fig. 4) an especially simple and accurate relationship between the wavenumber in cmy1 of the Lorentzian component of the 500 cmy1 broad peak and wBx in cmy3. wBx cmy3s8.44=1030 expy0.048W (cmy1). From the numerous points of approximately 1.6=1021 B-cmy3 (ByCs6000 ppm), the Fig. 4 gives an idea of the accuracy of the determination of the B content in the film from the wavenumber of the Lorentzian component. For a maximum error on the wavenumber of "8 cmy1, the maximum inaccuracy on the B content is within a factor of f2. This factor of two is similar to the accuracy of the SIMS. 5. Conclusion The Raman spectra at room temperature of polycrystalline diamond films exhibit two broad peaks of approximately 500 and 1220 cmy1 for a boron concentration range from 2=1020 to 1022 cmy3, which is wider than that (3=1020 –6=1021 cmy3) corresponding to the metallic conductivity of the diamond films. Therefore, these structures appear to be connected to the boron concentration rather directly to the metallic conductivity in these films.

286

M. Bernard et al. / Diamond and Related Materials 13 (2004) 282–286

In agreement with other works, this range is also wider than that (3–8=1020 cmy3) we obtained for similar effects in monocrystalline films. This is ascribed to the higher deformation allowed in the grain of a polycrystalline film than in a monocrystalline film. The ‘500 cmy1’ peak down-shifts systematically with increasing boron concentration in these films. It can be fitted by a Lorentzian and a Gaussian component. The boron concentration (wBx in cmy3) can be easily deduced from the wavenumber (W in cmy1) of the Lorentzian component according to the empiric law wBx (cmy3)s8.44=1030 expy0.048W (cmy1) from a non-destructive and contact less measurement. References w1x M. Werner, O. Dorsch, H.U. Baerwind, A. Ersoy, E. Obermeier, C. Johnston, et al., Diamond Relat. Mater. 3 (1994) 983.

w2x A.N. Ndao, F. Zenia, A. Deneuville, M. Bernard, C. Levy´ ´ Clement, Diamond Relat. Mater. 9 (2000) 1175. w3x E. Gheeraert, P. Gonon, A. Deneuville, L. Abello, G. Lucazeau, Diamond Relat. Mater. 2 (1993) 742. w4x P. Gonon, E. Gheeraert, A. Deneuville, L. Abello, G. Lucazeau, J. Appl. Phys. 78 (1995) 7059. w5x J.W. Ager, W. Walukiewicz, M. McCluskey, M.A. Plano, M.I. Landstrass, Appl. Phys. Lett. 66 (1995) 616. w6x R. Locher, J. Wagner, F. Fuchs, M. Maier, P. Gonon, P. Koidl, Diamond Relat. Mater. 4 (1995) 678. w7x H. Spicka, M. Griesser, H. Hutter, M. Grasserbauer, S. Bohr, R. Haubner, et al., Diamond Relat. Mater. 5 (1996) 383. w8x F. Pruvost, E. Bustarret, A. Deneuville, Diamond Relat. Mater. 9 (2000) 295. w9x F. Pruvost, A. Deneuville, Diamond Relat. Mater. 10 (2001) 531. w10x K. Ushizawa, K. Watanabe, T. Ando, I. Sakaguchi, M. Nishitani-Gamo, Y. Sato, et al., Diamond Relat. Mater. 7 (1998) 1719. w11x E. Gheeraert, A. Deneuville, J. Mambou, Diamond Relat. Mater. 7 (1998) 1509. w12x D.S. Knight, W.B. White, J. Mater. Res. 4 (1989) 385.