Nature of damage in diamond implanted at low temperatures

Nature of damage in diamond implanted at low temperatures

$$AMOND RELATED MATERIALS Diamond and Related Materials 4 (1995) 5699574 Nature of damage in diamond C. Uzan-Saguy, implanted at low temperatures...

668KB Sizes 0 Downloads 42 Views

$$AMOND RELATED MATERIALS Diamond and Related Materials 4 (1995) 5699574


of damage in diamond

C. Uzan-Saguy,


at low temperatures

V. Richter, S. Prawer I, Y. Lifshitz 2, E. Grossman

Solid State Institute,


Institute of Technology,

2, R. Kalish

Haifa, 32000, Israel

Abstract The most effective scheme for electrically activating ion implanted dopants in diamond involves implantation at low temperatures (Ti 2<300 K) followed by rapid thermal annealing. For such implantations, all defects (both vacancies and interstitials) are believed to be “frozen in”, a fact which facilitates subsequent annealing. In the present work? the nature of the defects introduced into diamond by low temperature implantations is determined by combining channelling, electrical conductivity, swelling and Raman measurements on type IIa diamonds irradiated with 320 keV Xe ions at Ti = 1.50 K over the dose range 1 x 1013-2 x 1016 Xe cm-‘. The critical dose for amorphization was found to correspond to an energy deposited in the modified layer of 5.5 eV per C target atom. The carriers were determined to be holes for these cold implantations. The measurements suggest that isolated point defects in diamond behave as acceptors, whereas more complex agglomerated defects behave as donors. Keywords:

Ion implantation;

Defects; Doping;



1. Introduction For effective doping of diamond by ion implantation, it is necessary not only to anneal the damage caused by the ion irradiation, but also to activate the dopant. For the case of B (p-type) implantation, Prins [ 1 ] and Zeidler et al. [2] have shown that the most effective scheme for dopant activation involves cold implantation followed by rapid thermai annealing (CIRA scheme). The basic idea behind the success of this scheme is that point defects are immobile for low temperature implantations (Ti 2( 300 K) and therefore most of the implantation induced damage is “frozen in”. On rapid annealing, the defect mobilities are such that a sufficiently large fraction of the implanted ions recombine with vacancies and thus occupy electrically active sites. There is competition between vacancydopant recombination, which results in dopant activation, and vacancy-self-interstitial recombination, which is necessary for annealing the ion beam induced damage. For higher implantation temperatures (typically Ti 3 300 K), defects can diffuse during the implantation, which can result in the formation of more complex defect structures that are difficult to anneal and prevent the implanted dopants from occupying lattice sites.

’ Permanent address: School of Physics, University of Melbourne, Parkville, Vie 3052, Australia. ’ Soreq NRC, Yavne 81800, Israel. 0925-9635/95/$09.50 0 1995 Elsevier Science S.A. All rights reserved 0925.9635(94)05290-5


For effective use of the CIRA scheme, it is imperative that the dose be kept below that at which diamond is known to graphitize or transform into non-diamond carbon on furnace annealing [3-61. The critical dose for amorphization D, depends on the implantation parameters, including the implantation temperature, species and ion energy [S-S]. Below D,,furnace annealing results in the near-complete recovery of the diamond lattice, but once D, is exceeded, an irreversible structural transformation occurs on furnace annealing leading to the formation of graphite-like layers [ 5,6]. It is clear that, in order to optimize implantation/ annealing schemes, a thorough knowledge of the defect structures present in diamond implanted at low temperatures is necessary. Although the electrical conductivity [ 5,6], swelling [ 5,6] and crystalline quality as assessed by channelling experiments [3,5,6] of cold implanted diamond have been studied previously in separate experiments, there is much which remains unknown about the nature of the electrical and structural properties of diamond implanted at these low temperatures. For example, the carrier type for cold implantations has not been determined, although there are reports of p-type conductivity for implantations performed at “room temperature” [ 91. Also, the relationship between the onset of the electrical conductivity and the amorphization threshold has not been established. In this paper, we report measurements designed to elucidate the nature of the damage induced in diamond


C. U-_ud%p,wy et crl.JDiunwndand

by low temperature Xe (320 keV) ion implantation. Insight into the ion beam transformation was obtained by combining the information gained from swelling, electrical conductivity and channelling measurements on the same set of samples irradiated under the same implantation conditions. Xe was chosen because it is chemically inert, it is possible to introduce very considerable ion beam damage at relatively low doses (and thus minimize any possible effects due to the formation of inert gas bubbles), and extensive work has already been done on Xe implantations in diamond at various implantation temperatures [ 7,8].

2. Experimental


The samples used in these studies were type IIa diamond windows supplied by Drukker International (Amsterdam). The diamonds were implanted with 320 keV Xe ions which have a range (R,) and straggling (AR,) of 56 and 9 nm respectively [lo]. In one set of experiments, the resistance of the diamond was monitored in situ during the implantation, which was carried out at Ti = 150 K using sufficiently low beam currents such that beam heating was negligible. The resistance was measured using a two point technique with an electrometer capable of measuring up to 2 x 1012 Q using an apparatus and method described previously [7]. A second set of implantations (also 320 keV Xe) was performed at Ti = 240 K at varying doses to provide samples for the ex situ microstructural characterization of the implanted diamonds. In order to ascertain the degree of lattice damage, channelling Rutherford backscattering spectroscopy (c-RBS) was employed using





320 keV protons and a scattering angle of 16.5”. The swelling of the implanted layer was measured using a surface profilometer and an atomic force microscope (AFM) scanned across an implantedjunimplanted interface. The conduction mechanism was investigated by measuring the temperature dependence of the resistivity. Finally, the carrier type was determined by the “hotpoint” (Seebeck) technique, and in some cases Hall effect measurements were undertaken.

3. Results Fig. 1 shows the conductivity of a type IIa diamond implanted with 320 keV Xe ions at 1.50 K as a function of the Xe ion dose. The first observable increase in the conductivity is observed for a dose of about 1.5 x 1014 Xe cmP2. Once this dose is exceeded, the conductivity rises rapidly. For doses greater than about 1 x 10” increases approximately as a Xe cmP2, the conductivity linear function of the dose (see Fig. 1). Unlike the case of C implantations at 77 K [5,8], the conductivity does not appear to saturate even at the highest doses. Also shown in Fig. 1 is the swelling of the diamond which accompanies irradiation. Again, there is no noticeable swelling until a dose of about 1.5 x 1014 Xe cmP2 is exceeded. The fact that diamond implanted at 77 K only swells when the dose exceeds a critical value has been established previously [S]. Of particular note in Fig. 1 is the fact that the degree of swelling closely follows the dose dependence of the conductivity. AFM measurements on the implanted and non-implanted parts of the sample verified the swelling data taken by the surface prohlometer and, in addition, revealed an ever increasing



4 [ IY95) 560-574




Fig. 1. The conductivity (a) as a function of dose (D) for type IIa diamond irradiated with 320 keV Xe ions at 150 K (m). The straight line is a of o on D is almost least-squares fit to the conductivity data over the dose range 6 x 10’4P1 x lOI Xe cm-‘. Over this range the dependence linear. The swelling which accompanies Xe ion irradiation (0) is also shown as a function of dose for diamond irradiated at 240 K.

C. Uzan-Saguy

et al./Diamond

and Related Mate&Is

roughness of the surface as a function of the ion dose. No signs of individual ion impacts could be detected even for the lowest doses employed. Fig. 2 shows the c-RBS data for diamond implanted with 2 x 1013, 2 x lOi and 1 x 1Or5 Xe cm-‘. The c-RB spectra prior to implantation (i.e. virgin diamond) and the RB spectrum for non-aligned (i.e. random) cases are also shown for comparison. For the lowest dose, only a very slight increase in backscattering yield is observed. The spectrum for the sample implanted with 2 x 1014 Xe cm-’ shows the formation of an amorphous layer extending from the surface down to a depth of about 50 nm. (This is in sharp contrast with the c-RBS results (not shown) for a diamond implanted at 470 K to a dose of 2 x lOi Xe cmm2 for which most of the diamond remains intact and only a very thin buried graphitic region at the end of the range can be observed.) For the cold implantations (Fig. 2), increasing the dose beyond 2 x 1014 Xe cmm2 appears to have the effect of extending the depth of the amorphous surface region. It should be noted that the conductivity begins to increase at about the same dose at which the c-RBS yield reaches the random level. Similar observations were made for 100 keV C implantations [ 111. In that case, the conductivity begins to increase and the c-RBS yield reaches the random level at a dose of (2.5-3) x 1015 C cmm2. This dose is somewhat higher than the value of 1.5 x 1015 C cme2 determined in Ref. [3] for the critical dose for amorphization under similar implantation conditions with 100 keV C ions. Raman spectra on samples irradiated with 3 x 1Or3 and 1 x 1014 Xe cmm2 (not shown here) displayed evidence of the presence of amorphous sp3 bonded carbon Energy 0.10

0.15 I


4 (1995) 569-574

[ 123. The Raman spectra for these samples did not display any evidence for the formation of amorphous or polycrystalline sp2 bonded carbon, whereas samples irradiated with 2 x 1016 Xe cm-2 displayed a broad Raman peak centred at about 1550 cm-’ which is characteristic of amorphous sp2 bonded carbon. Fig. 3 shows the conductivity (c) vs. temperature (T) for samples implanted with 320 keV Xe at 240 K for various doses plotted as log(c) vs. l/T. Clearly, a single activation energy cannot adequately describe the data over the whole of the temperature range examined. In an attempt to check for the possibility that the conduction mechanism may be dominated by variable range hopping [13], the data were replotted as log(o) vs. ( 1/T)1’4. However, these plots were not linear either over the temperature range examined. Despite the fact that the curves in Fig. 3 are not linear over the whole temperature range, reasonably linear regions can be found over a restricted range of 300,< r,< 500 K. The slopes in this temperature regime have been used to calculate the room temperature “activation energies” (&), and these are plotted as a function of dose in Fig. 4. Also included in Fig. 4 is the dose dependence of the conductivity from Fig. 1, plotted as a resistivity in order to show that the decrease in activation energy


500 I

400 I

300 I


(MeV) 0.20 I

0.25 I











c 2 g

u 400 a, .F

9 8

“, 300 .! 0 E 200 b z

1o-1’3 e


,o-l: Channel








INVERSE Fig. 2. c-RB spectra for type IIa diamond irradiated with 320 keV Xe ions at 240 K at doses of 2 x 10 I3, 2 x 10“’ and 1 x 10’s Xecm-s. The c-RB spectra for virgin (i.e. unimplanted) diamond and for diamond taken in the random (i.e. non-channelled) orientation are shown for comparison. The c-RBS data were taken using a 320 keV proton beam and a scattering angle of 165”.



















Fig. 3. Conductivity e vs. inverse temperature for type IIa diamond implanted with 320 keV Xe ions at 240 K to doses of 2 x 1014, 5 x 10i4, 2 x 10’s and 2 x lOi Xe cm-‘. The straight lines are leastsquares fits to the data in the temperature range 300 < T< 500 K, from which “activation energies” are deduced.



lo4 1o13


I 111,11

I I I lllrl/

I I I1"C







Fig. 4. The activation energy extracted from the straight line fits shown in Fig. 3 vs. dose for type IIa diamond irradiated with 320 keV Xe ions at 240 K as a function of ion dose (C). Also shown for comparison are the conductivity data of Fig. 1 replotted here as the resistivity vs. dose (W).

appears to follow closely the decrease in resistivity of the ion implanted material as the dose is increased. The carrier type for all samples implanted at low temperatures was determined to be p-type by hot-point (Seebeck) measurements. Performing Hall effect measurements on these samples is difficult due to their small size and low conductivity. Such measurements were only possible for a sample irradiated with 2 x lOI6 Xe cm -’ (which shows peaks in the Raman spectrum corresponding to sp2 bonded carbon). The Hall measurements for this sample yielded a mobility ,Mand carrier density y of 20 cm2 V’ s-l and 3 x 10” cmp3 respectively. For all the cold implanted samples, the irradiated layer proved to be resistant to etching in a solution of boiling acids which is known to attack and dissolve graphite-like layers on diamond. This shows that the implantation has not caused graphitization, which occurs when diamond is implanted with similar doses at elevated temperatures [ 71. Annealing the sample implanted with 2 x 1015 Xe cme2 at 700 K in vacuum for 1 h resulted in an increase in o. For this annealed sample, the temperature dependence of (r was very close to that found for the 2 x 10lh Xe cmm2 sample shown in Fig. 3, with E,= 0.06 eV. We stress that, after this annealing, the carriers remained p-type, and that the material could not be etched by boiling acids. However, following annealing at 1100 K for 1 h, the carriers were identified as n-type, and in this case the material could be slowly etched by boiling acids.

4. Discussion The dose D, at which the conductivity begins to increase appears to correspond closely to that at which

swelling of the implanted region is first observed and to the dose at which the c-RBS yield reaches the random level. For the present case of 320 keV Xe implantations, this dose is about 1.5 x 1014 Xe cme2. The computer code TRIM [lo] can be used to determine parameters for the ion-solid interaction, such as the range, straggling and percentage of energy lost by the Xe ion in nuclear collisions. Using these calculated parameters, the above dose can be transformed into the nuclear energy imparted by the Xe ion per target C atom. Following the method outlined in Ref. [ 141, for 320 keV Xe irradiation, D, corresponds to a value of 5.6 eV per C atom. This value is comparable with that for the cohesive energy in diamond of about 7 eV per C atom, i.e on average sufficient energy has been delivered to the lattice to break every CC bond about once. Similar measurements to those presented here for 100 keV C implantations [ 8,111 showed that D,z 2.5 x 10” C cme2 which corresponds to 5.4 eV per target atom, in close agreement with the results for 320 keV Xe irradiation. Other estimates for D, have been given in terms of the fraction of carbon atoms displaced from their lattice positions (50% 75%) [4] or the number of vacancies created per cubic nanometre, i.e. 12 vacancies nmp3 [ 31. However, these estimates rely on a knowledge of the displacement energy from TRIM and ignore dynamic annealing effects. By contrast, the estimate of about 5.5 eV per target C atom given above relies only on a knowledge of the nuclear and electronic stopping powers. It is important to note that, although the c-RBS yield reaches the random level at D,,the conductivity only increases significantly for doses well in excess of D,.It is also important to stress that the Raman spectra for samples implanted at doses close to, but less than, D, show evidence for the presence of amorphous sp3 carbon, but no evidence for sp2 bonded carbon.

C. &an-Saguy

et al.lDirrmond und Related Materials

The above observations lead us to propose that, below D, (i.e. the critical dose for the amorphization of the diamond lattice), ion implantation results in the formation of isolated islands of amorphous material. At D,, a semi-continuous amorphous sp3 bonded structure is created, and at this dose a percolative pathway is formed between these amorphous zones resulting in an increase in the observed conductivity. As the dose is increased further, a gradual conversion from sp3 to sp2 bonded C is observed, which gives rise to a further increase in the conductivity, and also to swelling (see Fig. 1). The Raman results, showing no evidence for amorphous sp2 bonded structures for doses less than D,, also suggest that the gradual transformation from sp3 to sp2 bonded carbon is inhibited until the diamond is amorphized. The reason for this may possibly be that a large decrease in density accompanies the sp3 to sp2 conversion, and that prior to amorphization, the diamond lattice is rigid enough to prevent this from occurring. If this is true, then we should observe increases in the conductivity of amorphous sp3 bonded C (in which case there is no lattice to constrain the sp3 to sp2 bonded conversion) at doses much lower than D,. Indeed, recent measurements [lS] have shown that the conductivity of amorphous sp3 bonded carbon begins to increase at doses as low as 1 x 1Or2 Xecmm2. It is very interesting to note that the carriers induced by cold implantation are p-type over the whole range of doses employed in this study. A similar result was found by Fang et al. [9] for 80 keV Kr irradiation at room temperature. However, in that work, for ri > 325 K, the carrier type was identified to be always n-type. For the cold implantations performed in the present study, the defects are expected to be immobile [5], whereas for implantations performed above room temperature, interstitials are expected to be mobile [ 51, resulting in a high probability of defect clustering. The present results therefore suggest that isolated point defects behave as acceptors in diamond, whereas more complex defects, such as those formed for elevated temperature implantations, behave as donors. If this is true, then we would expect that neutron irradiation, which produces predominantly point defects, should result in p-type conductivity, and indeed this has recently been found to be the case [ 161. Furthermore, we have shown that, when the cold irradiated samples are annealed at 1100 K, the carriers change to n-type. We propose that this change in carrier type is due to defect clustering, which is expected to occur when the cold implanted sample is annealed at these high temperatures. These observations suggest the interesting possibility of being able to construct a p-n junction in diamond without any chemical doping. n-Type conductivity, induced in diamond by residual damage, has been used previously to construct p-n-type devices based on natural p-type (type IIb) diamonds [5,6,17]. The present

4 (1995) 569.-574


results suggest that such devices might be constructed purely by Xe ion beam irradiation of undoped diamond by varying the implantation temperature to produce nand p-type regions on the diamond. The damage related carrier mobilities are expected to be low, as found in this work for the case of heavily Xe irradiated (2 x lOi Xe cm-‘) diamond (y=20 cm2 V-’ s-l). This low mobility is, however, comparable with that obtained by B implantation of type IIa natural diamond followed by suitable annealing [2], for which it was shown that implanted layers with hole mobilities of the order of p= 20-50 cm2 V-i s-r can be used successfully in the construction of field effect transistors. Thus it would appear that p-type layers created by damage alone might be useful in diamond devices. The conductivity is related to the carrier mobility p and carrier concentration p by a=el*p


where e is the electronic charge. Assuming that the carrier mobility is constant (i.e. not a function of the carrier concentration) and that p is proportional to the ion dose D, a plot of log(a) vs. log(D) should yield a straight line with a slope of unity. In Fig. 1, the log(a) vs. log(D) plot has been least-squares fitted with a straight line for the dose range 6 x 1014-2 x 1016 Xecm-‘; the fit in this range has a slope of 1.15fO.l. Thus it appears, at least over this dose range, that the number of carriers increases almost linearly with dose, and this observation is consistent with our previous proposition that, following the production of a continuous amorphous layer, there is a gradual conversion from sp3 to sp2 bonded carbon. Finally, it is clear from Fig. 4 that the increase in conductivity (or equivalently decrease in resistivity) is accompanied by a decrease in the “activation energy”. For the case of B doped CVD diamond films, the activation energy was found to decrease as the doping level was increased [ 181. This was interpreted as being due to the formation of a band by the acceptors whose width increased with acceptor concentration. This increase in width brings the bottom of the impurity band closer to the top of the valence band, thus decreasing the energy difference between the acceptor levels and the valence band, which is experimentally observed as a decrease in EA. A similar mechanism may be operative here, i.e. as the dose is increased, the number of carriers also increases, leading to an increase in the width of the acceptor impurity band and thus a decrease in the activation energy.

5. Conclusions Type IIa diamond was irradiated with 320 keV Xe at temperatures below 240 K. The dose dependence of the

conductivity, c-RBS yield, activation energy, swelling and carrier type were determined. The following conclusions can be drawn. D, is equivalent (1) The critical dose for amorphization to about 5.5 eV per C atom. Below D,,the ion beam creates isolated amorphous sp3 bonded zones within the irradiated region. At D,,the c-RBS yield reaches the random level, and the diamond begins to swell. We propose that, at this dose, the amorphous regions overlap to form a semi-continuous layer. The overlap of these amorphous zones results in the formation of a conducting pathway. At yet higher doses, conversion from sp3 to sp2 bonded C occurs, with a concomitant increase in conductivity. reported here, the (2) For the cold implantations damage induced carriers behave like acceptors, lying between 0.2 and 0.04 eV above the valence band. Over the dose range 6 x 1014-2 x 10lh Xe cm-‘. the number of carriers created appears to increase linearly with ion dose. The “activation energy” decreases with increasing ion dose, possibly due to an increase in the width of the acceptor impurity band as the dose is increased. (3) Samples annealed at sufficiently high temperatures show a change from p- to n-type conductivity. This result, as well as the observation that the ion beam conductivity induced for implantations for Ti > 325 K is always n-type [9], suggests that isolated point defects in diamond act as acceptors, whereas more complex agglomerated defect structures behave as donors.

References 111 J.F. Prins, Nucl. Instrum. Mahods Phys. Rex B, 801X1 (1993) 1433. [21 J.R. Zeidler, CA. Hewett and R.G. Wilson, Phys. Ret:. B. 47 (1993) 2065. E.A. Hill, N.R. Parikh and G. [31 J.D. Hunn, M.L. Swanson, Hudson, Int. Corrf. on the New Diumond Sciewe tmd Technology, 1990, p. 929. 141 R.A. Spits, J.F. Prim and T.E. Derry, NLICI. In.~rurx Methods Phys. Rex B, X5 (1994) 347-351. I51 J.F. Prins, Mater. Sci. Rep.. 7 (1992) 271, and references cited therein. and R. Kalish, Ion Implonfution m Diarnond, [61 MS. Dresselhaus Graphite. und Related Materials, Springer-Verlag, Berlin, 1992. and references cited therein. R. Kalish. Diamond R&u. Murer., -7 (1993) 621. and R. Kalish, Appl. Phy.s. Letf. 57 c71 S. Prawer, A. Hotfman (1990) 2187. CSI S. Prawer and R. Kalish, Phys. Rra. B, to be published. and S.S. Lau, IEEE 191 F. Fang, C.A. Hewett, M.G. Fernandes Tr~m.s. Elrctron Devices, 36 (1989) 1783. The Stopping and Cl01 J.F. Zeigler. J.P. Biersack and U. Littmark, Range of Ions m Solids, Pergamon, New York. 198.5. G.A. Anderson and S. Prawer, unpublished, 1994. [ItI K.A. Nugent, D.N. Jamieson, S. Dooley and R. [I21 S. Prawer. Kalish, to be published. 1131 N.F. Mott and E.A. Davis. h’kcfrorric Processes in Noncry.stn//ine Solids. Clarendon, Oxford, 1979. S. Prawer and R. Kalish, Phys. Rev. B. 45 1141 A. Hoffman, (1992)12(16736. 1151 I.G. Gerstner. D.R. McKenzie. D.G. McCulloch, S. Prawer and R. Kalish. to be published. and R. Kalish, unpublished, 1994. 1161 J.W. Vandersande M.D. McAleese, M. 1171 M. Geis. N.N. Efremow, J.D. Woodhouse, Marchywka. D.G. Sacker and J.F. Hochedez. IEp;E Hectron Device Left., 117 (1991 ) 456. IIS1 K. Okano. H. Naruki. Y. Akiba, T. Kurosu, M. Iida. Y. Hirsoe and T. Nakamura. Jp,l. J. Appl. Phys.. 28 (1989) 1066.