Infrared Monitoring of Diamond Metalization

ROBERT LOVRINCIC, SEBASTIAN NOEBEL, and ANNEMARIE PUCCI*Kirchhoff-Institut fur Physik der Universitat Heidelberg, Im Neuenheimer Feld 227, 69120 Heidelberg, Germany

The ultra-high-vacuum growth of thin chromium and aluminum films on

single diamond crystals at various temperatures was monitored by means

of infrared spectroscopy, which provides instant information on electronic

and morphological properties of the produced metal films. In accordance

with atomic force microscopy results, spectra for Al indicate a clear

island-like growth in the beginning followed by a transition to a

continuous layer. Cr growth results in much smoother films because of

a special growth mechanism. Analysis of infrared spectra of both kinds of

ultrathin metal layers yields a high DC conductivity well above 106 S/m

starting from only a few nanometers thickness for sufficiently high

substrate temperatures.

Index Headings: Infrared spectroscopy; Thin-film growth; Metal–diamond

interfaces; Electronic properties of thin films.

INTRODUCTION

The availability of high-quality synthetic diamond singlecrystals has greatly increased in recent years, and more andmore applications are being realized for this wide-gapsemiconductor, e.g., in high energy particle detectors.1,2 Oneimportant step towards satisfying performance of diamond-based electronics is the controlled fabrication of good metalliccontacts. Whereas the metal–diamond interface has beenstudied intensely by photoelectron spectroscopy,3–6 only alittle is known about the in-plane conductivity of thin metalfilms on diamond. To improve this situation, we studied thegrowth of aluminum and chromium, two commonly usedelectrode materials in diamond electronics, on single crystallinediamond-(1 0 0) at different growth temperatures by means ofinfrared spectroscopy (IRS) and atomic force microscopy(AFM). In this study we will show the surprising result thatunder ultra-high-vacuum conditions physical vapor depositionat room temperature and above leads to well-conductingnanofilms.

Infrared spectroscopy directly measures the dynamic con-ductivity of thin films in the inspected spectral range.7–10

Moreover, with the validity of the Drude model in this range,the two Drude parameters of plasma frequency, xp, andrelaxation rate, xs, can be separately derived from themeasured spectra,10,11 which enables the extrapolation to thestatic conductivity rDC, given by

rDC ¼ e0x2px�1

s ð1Þ

The dynamic (optical) conductivity is the most importantquantity for the plasmonic behavior of metal nanostructures.For strong plasmonic signals it is important to reduce electronicdamping from structural defects in lithographic nanostructureproduction. The relaxation rate in nanofilms can be signifi-

cantly higher than the bulk value due to three main effects.First, scattering at defects inside the film enhances xs by xd

s ,which depends on the crystalline quality of the layer. Second,the conduction electrons may scatter at surfaces and interfaces,which increases xs by xs

s } d�1 (for constant surfaceroughness),8,12 giving the total relaxation rate

xsðd; TÞ ¼ xbulks ðTÞ þ xs

sðdÞ þ xdsðdÞ ð2Þ

for a thin film. Then, from a certain thickness on, if filmstructure and morphology do not change anymore (xd

s ¼const.), a linear behavior for xs(d) over d�1 is observed. Anextrapolation to d! ‘ (xs

s ! 0) can then yield information onxd

s . Its difference to the bulk value gives the residual resistivityincrease due to defects.8 Third, for thicknesses of the order ofthe Fermi wavelength, thickness-dependent quantum-sizeeffects (QSE) with oscillations of xp and of xs may occurvisibly in IR spectra under certain circumstances.8,10

EXPERIMENTAL

The experiments were carried out in an ultra-high-vacuum(UHV) chamber (base pressure ;1 3 10�10 mbar). Thesubstrate was a CVD diamond single crystal (Element6 (4.5 34.5 3 0.5) mm3) with (1 0 0) surface orientation. Prior totransfer into vacuum the diamond was chemically cleaned,13

resulting in an oxygen-terminated surface. Al and Cr wereevaporated from boron nitride and tungsten crucibles, respec-tively. The deposition rate (about 0.1 nm/min) was calibratedwith a quartz micro balance. All IR spectra were measured innormal transmittance with a Bruker IFS66v/S Fourier trans-form spectrometer coupled to the UHV chamber via KBrwindows. The transmittance of the bare substrate was taken asthe reference spectrum prior to film deposition.

Conductivity and Relaxation Rates. Figure 1 shows thechange of relative transmittance with thickness for variousselected wavenumbers for Cr (T ¼ 400 K; for T ¼ 300 K seeRef. 14) and for Al (at T¼ 300 K) growth on diamond. For thecase of Al, the intersection of the curves at one certain averagethickness (about 6 nm) indicates percolation of Al islands asobserved for noble metals on various substrates.9,15 To obtainthe Drude parameters for continuous films from measured IRtransmittance spectra, data analysis was performed using thecommercial software package Scout16 with the Drude-likedielectric function e(x),

eðxÞ ¼ 1� x2pðx2 þ ixxsÞ�1 ð3Þ

with xp and xs as thickness-dependent parameters for the metalfilms. The maximum fit range was 500 cm�1 to 1500 cm�1 forCr and 1000 cm�1 to 4000 cm�1 for Al, respectively, and hencebelow the onset of interband transitions in both cases.17,18 It isimportant to note that for an island-like (granular) morphologythe area (volume) filling is not complete and thus the IRspectrum of such a layer in general cannot be described with

Received 28 July 2010; accepted 4 October 2010.* Author to whom correspondence should be sent. E-mail: pucci@kip.uni-heidelberg.de.DOI: 10.1366/10-06081

Volume 65, Number 1, 2011 APPLIED SPECTROSCOPY 1050003-7028/11/6501-0105$2.00/0

� 2011 Society for Applied Spectroscopy

the Drude-type dielectric function, except below a certainmaximum frequency.9,19 This maximum frequency roughlycorresponds to the mathematical inflection point of thetransmittance spectra of island films after coalescence and is

shifting to higher values with increasing filling factor (usuallywith increasing average thickness). With the fit rangeextending only up to such maximum frequency, the correlationcoefficient for the fit stays above 0.99. For 9.5 nm average Althickness, a good accordance between measurement and fit upto rather high frequencies is possible; however, the obtained xp

¼ 77 700 cm�1 is still below the bulk value (101 000 cm�1),which indicates a filling factor below 100% (see Ref. 19 Eq.12) and thus a granular morphology. Nevertheless, the DCconductivity already reaches 5 3 106 S/m (see Fig. 2). Belowpercolation, the strongly frequency dependent extinction ofevaporated Al is due to particle plasmon-polariton excitation inAl islands that with increasing coverage change their size,shape, and interaction. For Cr deposition no such behavior isobserved, but instead, up to a critical thickness dc, a nearlyfrequency independent transmittance appears that cannot bedescribed with a simple Drude model and that is related to avery special fcc nano-crystalline phase of the ultrathin Crlayers.14 For Cr layers above this critical thickness, spectral fitsto the Drude model fully reveal the bulk-like xp of a denselayer, and the deviation from the bulk-like DC conductivity isdue to defect scattering (see Fig. 3).

Figure 3 shows the development of xs(d) versus the inversefilm thickness for Cr and Al at different substrate temperaturesduring metal film growth. The increase of xs(d) towards smallthicknesses for Al at T ¼ 300 K indicates the high filmroughness after passing the percolation threshold dc.

9,20 For Alat T ¼ 120 K the growth mode is also island-like, but dc is at1.5 nm only (spectra not shown). A strong temperaturedependence of dc is related to atomic diffusion and nucle-

FIG. 2. DC conductivity with average film thickness for (top) Al and(bottom) Cr as extrapolated from IR spectra. The substrate temperatures fordeposition and IR measurement are given in the figure. The insets (relativetransmittance vs. wavenumbers in units of cm�1) show a selection of measuredspectra (open symbols) for T ¼ 300 K (Al) and T ¼ 120 K (Cr), and best-fitcalculations (continuous lines) from which the Drude-parameters were derived.For Al, the high-frequency limit of the fit range was set to keep the correlationcoefficient above 0.99 (see text).

FIG. 1. Relative transmittance at different wavenumbers versus average filmthickness d, (left) Cr and (right) Al, for substrate temperatures as indicated.The constant transmittance before opening and after closing the shutter(indicated by vertical lines) demonstrates the measurement stability.

FIG. 3. xs(d) vs. inverse film thickness as calculated from IR spectra for (top)Al (inset shows experimental part of linear range for 300 K), and (bottom) Cr,on diamond at different substrate temperatures, as given. The continuous linesare linear extrapolations to d! ‘. Bulk values for xs are indicated by brokenlines.

106 Volume 65, Number 1, 2011

ation.9,21 Lower temperatures lead to a higher island density(lower average island distance) and hence to a coalescence ofthe film at lower average thickness (smaller dc). Highersubstrate temperatures would shift percolation to considerablyhigher average thickness,9,19 which contradicts a high DCconductivity of nanofilms. In contrast, dc of Cr only exhibits amarginal temperature dependence. It is always around 2.5 nm,which is related to the peculiar growth behavior at very lowthicknesses where energetic effects are dominating.14 For Algrowth at 300 K, an extrapolation from the range of the thickesttransparent films between 8.1 nm and 9.6 nm of xs towards d! ‘ results in a relaxation rate that corresponds (within errorbars) to twice the bulk value (780 cm�1).18 This result reflectsthe granular film morphology. At 120 K, xs exceeds 5000cm�1 throughout all Al thicknesses. Because percolationoccurred at only 1.5 nm, the formation of only very smallcrystallites is likely and their random mutual arrangementhampers electronic transport. In the case of Cr, the relaxationrate for films prepared at 300 K also does not approach the bulkvalue (see Fig. 2). Instead, a substrate temperature of 400 K isneeded to sufficiently decrease the number of defectsincorporated in the film. The reference value xbulk

s at 400 Kwas calculated from literature data assuming that thetemperature coefficient of the electric resistivity22 is valid forthe relaxation rate. Beyond dc, Cr growth at T¼ 120 K starts todeviate from growth at higher temperatures: the relaxation rateincreases with thickness towards typical values for cold-condensed metal layers. It is interesting to note that in the rangeof a few nanometers thickness beyond the critical thickness(;5 nm) the initial nearly temperature-independent growthmechanism of Cr leads to a low electronic relaxation rate closeto the values at higher temperatures. In contrast, for aluminumgrowth at lower temperature, mainly suppressed diffusion isrelevant.

Ex situ AFM measurements (see examples in Fig. 4) fullycorroborate the morphology conclusions from IR spectroscopy.The 8 nm Al film of Fig. 4 (prepared at 300 K) has an averagegrain size of 40 nm and exhibits a mesoscopically roughsurface. The 5 nm Cr film, on the other hand, appears rather

smooth, but defects within the film may decrease the electricalconductivity additionally to surface scattering.

CONCLUSION

In conclusion, we want to emphasize the benefit of IRS forthe contact-less conductivity measurement of metal-contactformation. Our study reveals the importance of the substratetemperature for producing highly conductive metalizations onsingle crystal diamonds. For Cr growth, a temperature of 400 Kavoids a high defect density inside the film and gives highlyconductive films at thickness of as little as 3 nm. At 300 K, Alforms conductive polycrystalline layers with a defect-relatedrelaxation rate of only twice the bulk value from 6 nm on. Aldeposition at lower temperature hampers extended islandgrowth but produces defective films with very low metallicconductivity.

ACKNOWLEDGMENTS

The authors gratefully acknowledge a fruitful collaboration with EleniBerdermann from the GSI, Darmstadt. Work supported by EU, project RII3-CT-2004-506078.

1. E. Berdermann, M. Ciobanu, S. H. Connell, A. M. O. D. da Costa, L.Fernandez-Hernando, A. Oh, and J. P. F. Sellschop, Phys. Stat. Solidi A201, 2521 (2004).

2. E. Berdermann, M. Pomorski, W. d. Boer, M. Ciobanu, S. Dunst, C. Grah,M. Kis, W. Koenig, W. Lange, W. Lohmann, R. Lovrincic, P. Moritz, J.Morse, S. Mueller, A. Pucci, M. Schreck, M. D. S. Rahman, and M.Trager, Diamond Relat. Mater. 19, 358 (2010).

3. D. A. Evans, O. R. Roberts, A. R. Vearey-Roberts, D. P. Langstaff, D. J.Twitchen, and M. Schwitters, Appl. Phys. Lett. 91, 132114 (2007).

4. M. Werner, Semicond. Sci. Technol. 18, 41 (2003).5. T. Tachibana, B. E. Williams, and J. T. Glass, Phys. Rev. B 45, 11968

(1992).6. T. Tachibana, B. E. Williams, and J. T. Glass, Phys. Rev. B 45, 11975

(1992).7. G. Fahsold and A. Pucci, ‘‘Non-contact Measurement of Thin-Film

Conductivity by IR Spectroscopy’’, in Advances in Solid State Physics(Springer Verlag, Berlin, Heidelberg, 2003), pp. 833–847.

8. A. Pucci, F. Kost, G. Fahsold, and M. Jalochowski, Phys. Rev. B 74,125428 (2006).

9. F. Meng and A. Pucci, Phys. Stat. Solidi B 244, 3739 (2007).10. D. V. Fedorov, G. Fahsold, A. Pucci, P. Zahn, and I. Mertig, Phys. Rev. B

75, 245427 (2007).11. G. Fahsold, M. Sinther, A. Priebe, S. Diez, and A. Pucci, Phys. Rev. B 65,

235408 (2002).12. E. H. Sondheimer, Adv. Phys. 1, 1 (1952).13. M. Yokoba, Y. Koide, A. Otsuki, F. Ako, T. Oku, and M. Murakami, J.

Appl. Phys. 81, 6815 (1997).14. R. Lovrincic and A. Pucci, Phys. Rev. B 80, 205404 (2009).15. S. Berthier and J. Peiro, J. Phys.: Condens. Matter 10, 3679 (1998).16. W. T. Hard, ‘‘Scout’’ (Aachen, Germany).17. S. S. Fouad, A. H. Ammar, and M. H. El-Fazary, Phys. Stat. Solidi B 187,

99 (1995).18. M. A. Ordal, R. J. Bell, J. R. W. Alexander, L. L. Long, and M. R. Querry,

Appl. Opt. 24, 4493 (1985).19. G. Fahsold, M. Sinther, A. Priebe, S. Diez, and A. Pucci, Phys. Rev. B 70,

115406 (2004).20. G. Fahsold, A. Bartel, O. Krauth, N. Magg, and A. Pucci, Phys. Rev. B 61,

14108 (2000).21. H. Brune, Surf. Sci. Reports 31, 125 (1998).22. H. Landolt and R. Bornstein, Eds., Zahlenwerte und Funktionen aus

Physik, Chemie, Astronomie, Geophysik, Technik (Springer, Berlin, 1959),6th ed., vol. II.

FIG. 4. AFM measurement of (left) a Cr layer with an average thickness of 5nm (contact mode) and (right) an Al layer with an average thickness of 8 nm(tapping mode), both deposited on diamond at T ¼ 300 K. Height and lateralscales were chosen differently in order to give a better impression of thedifferent morphologies.

APPLIED SPECTROSCOPY 107