Controlled Gettering of Implanted Platinum in Silicon Produced by Helium Co-Implantation. Pavel Hazdra and Jan Vobecký

Transcription

1 Solid State Phenomena Online: ISSN: , Vols , pp doi: / Journal Citation (to be inserted by the publisher) Copyright 2004 Trans by Trans Tech Publications, Tech Publications Switzerland Controlled Gettering of Implanted Platinum in Silicon Produced by Helium Co-Implantation Pavel Hazdra and Jan Vobecký Department of Microelectronics, Czech Technical University in Prague, Technická 2, CZ-16627, Prague 6, Czech Republic, Keywords: platinum, helium, ion implantation, gettering, silicon, lifetime control. Abstract. The guided in-diffusion of platinum into a low-doped n-type float zone silicon using radiation damage produced by implantation of 7 MeV He 2+ at a dose of cm -2 was investigated. A finite diffusion source was made by implantation of 1 MeV platinum ions at different doses ranging from to cm -2 and subsequent diffusion was performed at different temperatures ranging from 650 to 750 C. The distribution of in-diffused platinum was studied by monitoring the acceptor level of substitutional platinum (E C -E T =0.23eV) by deep level transient spectroscopy. Results show that, similarly to guided in-diffusion from surface PtSi layer, the distribution follows well the profile of the damage produced by helium ions, but the amount of gettered platinum is one order of magnitude lower and the diffusion temperature must be set to 725 C to receive optimum results. Introduction Local reduction of carrier lifetime by radiation defects introduced by helium irradiation became a powerful tool for optimization of electrical parameters in power silicon devices. However, properties of these defects are not ideal and limit their application in cases when a high density of recombination centers is desired [1]. For this reason, recent investigation turned back to platinum, the well known lifetime killer, to improve the localization and controllability of its introduction. The platinum atom is a hybride solute in silicon which can occupy both interstitial and substitutional sites. Diffusion proceeds via a small fraction of platinum solutes being in the interstitial configuration whereas the majority of the platinum atoms are stationary in substitutional sites exhibiting two levels in the band gap: an acceptor level at E C -0.23eV and a donor level at E V +0.32eV [2]. One possible way for platinum to change from an interstitial position to a substitutional site is via Frank-Turnbull (dissociative) mechanism [3] which assumes the interstitial platinum atom Pt i to recombine with a lattice vacancy V to a substitutional platinum atom ( Pt i +V ). The alternative is the so-called kick-out mechanism [4] where an interstitial platinum atom generates a silicon self-interstitial I when occupying a substitutional site ( Pt i +I ). Since both the reactions are affected by the number of lattice defects (vacancies and interstitials), the platinum diffusion can be significantly influenced by their artificial introduction, e.g. by co-implantation of atoms which are electrically neutral impurities in silicon. The guided in-diffusion of platinum from the platinum silicide (PtSi) surface layer by means of damage created by oxygen, fluorine, or chlorine implantation was first described in PtSi/Si n-type Schottky diodes [5]. Afterwards, protons [6] and especially helium ions [7-8] were successfully applied to create deeper lying damage regions capable to accumulate substitutional platinum and moved this technique, which is now called low-dose proximity gettering, closer to potential applications. Although promising results were recently received by the proximity gettering of platinum on high power devices [9], the controllability of the gettering process is still weak compared to well established ion and electron irradiation techniques. For this reason, this work is primarily focused on finding the optimum conditions for radiation-enhanced platinum in-diffusion and subsequent gettering in the depths of several tens of micrometers. In contrast with previously published studies, All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans Tech Publications, (ID: , Pennsylvania State University, University Park, USA-04/03/16,21:38:55)

2 560 Gettering and Defect Engineering in Semiconductor Technology X 2 Title of Publication (to be inserted by the publisher) DLTS Signal C/C E2 E no Pt not annealed Pt only 725 o C 20 min -08 E1 (Pt ) Pt + He 725 o C 20 min Temperature (K) Fig.1. C-DLTS spectrum of the unprocessed diode (dashed), the diode implanted with 1MeV Pt + at a dose of cm -2 (dashed dotted) and the identically platinum implanted diode which was co-implanted with 7 MeV He 2+ at a dose of cm -2. Both the diodes were subsequently annealed at 725 C for 20 minutes (rate window 260 s -1 ). where the source for platinum diffusion was formed by thin PtSi surface layer [5-9], we used the implantation of platinum at different doses to create a finite and laterally homogeneous diffusion source. Experimental Proximity gettering of implanted platinum was studied on the low-doped (phosphorous concentration below cm -3 ) <100>-oriented float zone (FZ) n-type silicon substrate forming the n-base of the planar p + nn + diodes. The diodes were first implanted with 1 MeV platinum ions in the dose range from to cm -2 using the 5MV Tandem facility in FZ Rossendorf to create a homogeneous and finite source for platinum in-diffusion in the depth of 252 nm corresponding to the projected range of platinum ions. The diodes were subsequently co-implanted with 7 MeV helium ions at a dose of cm -2 to create a damaged layer in the depth of 40 m which served for platinum gettering. The platinum in-diffusion was performed by 20 min furnace annealing in vacuum. Five different diffusion temperatures in the range from 650 to 750 C were selected for each set of samples which also included the reference one without helium co-implantation. The stability of the proximity gettered platinum layer was studied at 700 and 725 C by annealing for different times ranging from 5 to 80 minutes. The samples from the same set (equal dose of platinum) were annealed simultaneously and the annealing time was always measured from the instant when the temperature at the sample position stabilized after the approximately 3 minutes long transient given by the insertion of the samples into the central area of the horizontal furnace. Electrically active defects were subsequently measured by capacitance deep level transient spectroscopy (C-DLTS) and high-voltage current transient spectroscopy (HVCTS) allowing non-destructive inspection in the full-depth of the diode base [10]. The experiment was performed on diode structures with removed anode and cathode metallic contacts. However, a detailed comparison of electrical characteristics of the diodes with and without metallic contacts showed that the contact removal had no influence on the accuracy of DLTS measurement. Results and discussion Fig. 1 shows a typical example of capacitance DLTS spectra measured on all diodes implanted with platinum (with and without the co-implantation) and subsequently annealed at temperatures from 650 to 750 C. The excitation conditions were chosen in that way to fully cover the projected range of helium atoms. The figure also shows a typical spectrum of the diode taken prior to processing (implantation and annealing). Three peaks labeled E1, E2, and E3 corresponding to three electron traps are clearly resolved. The levels E2 and E3 with the ionization enthalpies E C -E T equal to 0.26 and 0.47 ev, respectively, were detected in all unprocessed as well as processed (implanted and annealed) diodes. They always appear with the same concentration and are probably connected with

3 Solid State Phenomena Vols Concentration (10 12 cm -3 ) Pt 5x10 12 cm -2 Pt + He Pt only 725 o C 20 min C-DLTS HV-CTS Vacancies simulation Depth (µ m) Fig.2. Concentration vs. depth profiles of the level measured in the platinum implanted diodes (1MeV cm -2 ) without and with helium co-implantation (7MeV cm -2 ) by C-DLTS (lines) and HV-CTS (boxes). Platinum was in-diffused at 725 C for 20 minutes. Simulated distribution of primary vacancies resulting from helium implantation is also shown for reference. the diode fabrication process. The level E1 with the ionization enthalpy E C -E T = 0.23 ev and the extrapolated electron capture cross section σ n = cm 2 was detected only in platinum implanted samples. The identification parameters of the level E1 are in good agreement with those of the acceptor state of the substitutional platinum [2]. The signal of this level significantly increases with helium co-implantation and increasing diffusion temperature. The co-implantation with helium at a dose of cm -2 did not introduce any additional electrically active defect in the range of investigated diffusion temperatures. All these observations are in good agreement with results of previous investigation using the PtSi layer as a source for platinum diffusion [7-8]. The effect of helium co-implantation is shown in Fig. 2 where concentration versus depth profiles of the level measured in the platinum implanted diodes (1MeV cm -2 ) without and with helium co-implantation are compared. Both diodes were simultaneously annealed at 725 C for 20 minutes. Whereas the diode implanted only with platinum exhibits nearly flat distribution of, in the helium co-implanted diode, the concentration of substitutional platinum is significantly enhanced at the peak of the damage introduced by helium implantation (the simulated distribution of primary damage vacancies is shown for reference). The peak of the profile is nearly symmetrical and, allowing for broadening to a finite Debye length, shows an almost perfect decoration of primary damage distribution. However, an accumulation of in the region between the peak and the implanted surface (the tail region) is higher than might be expected from the simulated distribution of the primary damage. While the concentration ratio of the primary damage in the peak and the flat region of the tail is approximately equal to 50, the comparison of the magnitudes of the concentration gives the ratio of about 2.2. The HV-CTS measurement allowed to disclose the end part of the profile lying behind the peak which was not investigated previously [5-8]. It is clearly shown, that the profile changes its gradient at the depth where the damage ends (~ 43 µm) and slowly decreases up to the depth of 58 µm where it crosses and undergoes the profile measured on the sample without the helium co-implantation. The shape of the profiles measured on the diodes co-implanted with helium is similar to those which were taken on the samples where the surface PtSi layer formed an infinite source for platinum in-diffusion [7]. However, magnitudes of concentration are by order of magnitude lower. Fig. 3. shows the effect of annealing temperature on the concentration vs. depth profiles of the level measured in the platinum implanted diodes (1MeV cm -2 ) without (thin) and with helium(thick) co-implantation (7MeV cm -2 ). In the samples implanted only with platinum, no measurable signal originating from the level was detected for diffusion temperatures below 700 C. If the diffusion temperature achieves 700 C, the magnitude of starts to increase gradually. The resulting concentration profile of decreases from the implanted surface and becomes nearly flat at the depth of 50 µm. The co-implantation of helium significantly increases the platinum in-diffusion already at 650 C and the amount of Primary Vacancies (10 18 cm -3 )

4 562 Gettering and Defect Engineering in Semiconductor Technology X 4 Title of Publication (to be inserted by the publisher) Concentration (10 12 cm -3 ) Pt 5x10 12 cm -2 He 1x10 12 cm -2 Pt + He C C Pt only C C C C C C Depth (µ m) Fig.3. Concentration vs. depth profiles of the level measured in the platinum implanted diodes (1MeV cm -2 ) without (thin) and with helium (thick) co-implantation (7MeV cm -2 ) by C-DLTS. Diodes were annealed for 20 minutes at different temperatures ranging from 650 to 750 C. in-diffused achieves its maximum between 700 and 725 C. Diffusion at 750 C results in decreasing of the concentration and noticeable broadening of the peak on the platinum profile. Fig. 3. shows that, in the co-implanted diodes, the diffusion temperature controls namely the amount of in-diffused platinum while the shape of the profile remains nearly the same. The influence of the platinum dose and diffusion temperature on the platinum in-diffusion in the diodes co-implanted with helium is summarized in Fig. 4. Since nearly all measured profiles exhibited the identical shape with the peak to tail ratio ranging from 1.6 to 2.3, the peak concentration of the level can be taken as a quantitative measure of the platinum atoms gettered in the substitutional position. Fig.4. shows that the gettering efficiency increases for all implantation doses with temperature up to 725 C where it reaches its maximum. Increasing the diffusion temperature to 750 C decreases the amount of gettered in the damage region. It is also evident that, excluding the temperature 725 C, there is no correlation between the platinum dose and the amount of gettered in the peak region. It is also worth pointing out that both the in-diffusions performed at 700 and 750 C showed a very bad reproducibility while the annealing at 725 C gave stable and reproducible results. This is demonstrated in Fig. 5 which shows the concentration versus depth profiles of the substitutional platinum for the different diffusion times at 725 C. It can be clearly observed that the diffusion time influences the concentration of Pts. Diffusion for 10 minutes gives the maximal and best resolved peak of Pts concentration while Peak Concentration (10 12 cm -3 ) 2.5 Pt dose (cm -2 ) 5x x x x Diffusion Temperature ( o C) Fig.4. The magnitude of the level peak concentration measured in the depth of 40µm vs. diffusion temperature shown for different doses of 1MeV platinum implantation. Diodes were co-implanted with 7MeV cm -2 helium, the diffusion time was 20 minutes. longer diffusion times result in gradual lowering of the peak which is further accelerated when the annealing time exceeds 60 minutes. The diffusion time also influences the slope of the tail region extending towards the implanted surface. Its originally positive magnitude decreases and changes to negative, if annealing exceeds 40 minutes. Contrary, the part of the profile extending into the bulk seems not to be strongly affected by the time of diffusion.

5 Solid State Phenomena Vols Concentration (10 12 cm -3 ) 725 o C Pt 5x10 12 cm -2 He 1x10 12 cm min 40 min 20 min 10 min 80 min 80 min 10 min 20 min 60 min 40 min Depth (µm) Fig.5. Concentration vs. depth profiles of the Pt s level measured in the diodes implanted with platinum (1MeV cm -2 ) and co-implantated with helium (7MeV cm -2 ). Diodes were annealed at 725 C for different time ranging from 10 to 80 minutes. It is apparent, that the in-diffusion from the implanted source lacks of platinum interstitials due to their stacking to the defects produced by platinum implantation itself - the number of primary defects created by 1MeV platinum ion is approximately 60 times higher than the number for 7MeV helium. Therefore, the platinum in-diffusion can be accelerated only if platinum interstitials are released from these defects. It is expected that the damage produced both by platinum and helium implantation is similar for the implantation doses used in this experiment. Therefore, the helium co-implantation accelerated platinum in-diffusion is taking place under conditions when both the damaged regions start to be annealed. This effect shifts the optimum diffusion temperature from 700 C (the diffusion from the PtSi layer [5-8]) to 725 C and narrows the temperature window where the proximity gettering can be controlled. The unbalance between the number of defects in the shallow (0.25µm) and deeper (40µm) damaged regions causes that the gettering potential of the deeper region is not fully exploited compared to the diffusion from PtSi. Increasing of the diffusion temperature to 750 C leads to annealing of the defects which were originally responsible for platinum gettering and the amount of starts to decrease. This is indirectly illustrated in Fig.6. showing the concentration versus depth profiles of the level E3 (the defect which is originally presented in all diodes prior to platinum implantation) after platinum diffusion at different temperatures. These profiles are compared to the one measured on the diode without helium E3 Concentration (10 11 cm -3 ) Pt + H e C C C C R Pt only PH e Depth (µ m ) Fig.6. Concentration vs. depth profiles of the level E3 measured in the platinum implanted diodes (1MeV cm-2) without (boxes) and with helium (circles, triangles and diamonds) co-implantation (7MeV cm -2 ). Diodes were annealed for 20 minutes at different temperatures ranging from 650 to 750 C. co-implantation. It is clearly shown that already after the diffusion at 650 C the concentration of this defect, which was originally homogeneously distributed, is enhanced at the maximum of the damage produced by helium ions (R phe ). The shape of the profile remains unchanged up to the temperature of 750 C when the peak dampens and broadens. Congruous broadening and dampening exhibits the profile measured on the same sample.

6 564 Gettering and Defect Engineering in Semiconductor Technology X 6 Title of Publication (to be inserted by the publisher) As mentioned above, we received similar shapes of in-diffused profiles for different doses of implanted platinum which are very close in shape to that received by the diffusion from the PtSi source at the same temperature and the same dose of helium co-implantation [7]. This implies that the shape of the in-diffused profile is given by the distribution of the vacancy-type defects gettering the platinum that were introduced by helium co-implantation. Our experiment with lower concentrations of in-diffused platinum show that the low value of the measured peak to tail ratio is given by the distribution of gettering defects and not by the saturation of platinum substitutionals at the peak position. The shape of the in-diffused distribution can be therefore set by a proper selection of the co-implanted ion (peak skewness), its energy (depth), dose and diffusion temperature (peak height and tail slope), while the amount of the in-diffused platinum can be controlled by engineering of the platinum implantation. Summary The effect of helium co-implantation on the in-diffusion of the implanted platinum into low-doped n-type FZ silicon was investigated. Results show that, similarly to guided in-diffusion from surface PtSi layer, the distribution follows well the profile of the damage produced by helium ions, but the amount of gettered platinum is one order of magnitude lower. To obtain reproducible results, the diffusion temperature must be increased to 725 C which is close to the temperature when the gettering damage anneals out. This effect is explained by the influence of the shallow damage region which is produced by platinum implantation and which stacks platinum interstitials at temperature lower than 700 C. The shape of the profile is primarily set by conditions of helium implantation while the amount of gettered platinum may be controlled by implantation of platinum. Acknowledgement This work was supported by the EC - Access to Research Infrastructure Action of the Improving Human Potential Program project HPRI , by the Grant Agency of the Czech Republic under the grant number 102/03/0456 and the Research Programme no. JE MSM Authors also acknowledge ABB Switzerland Ltd, Semiconductors for diode preparation. References [1] P. Hazdra, J. Vobecký, K. Brand, Nucl. Instr. and Meth. Vol. B 186(2002), p.414. [2] S.D. Brotherton, P. Bradley, J. Bicknell, J. Appl. Phys. Vol. 50(1979), p [3] F.C. Frank, D. Turnbull, Phys. Rev. Vol. 104(1956), p [4] U. Gösele, W. Frank, A.Seeger, Appl. Phys. Vol. 23(1980), p.361. [5] B. Holm and K. Bonde Nielsen, J. Appl. Phys. Vol. 70(1995), p [6] D.C. Schmidt, B.G. Svensson, N. Kestikalo, S. Godey, E. Ntsoenzok, J.F. Barbot and C. Blanchard, J. Appl. Phys. Vol. 84 (1998), p [7] D.C. Schmidt, B.G. Svensson, S. Godey, E. Ntsoenzok, J.F. Barbot and C. Blanchard, Appl. Phys. Lett. Vol. 74 (1999), p [8] D.C.Schmidt, B.G.Svensson, J.F.Barbot and C.Blanchard, Appl.Phys.Lett.Vol.75(1999), p.364. [9] J.Vobecký, P. Hazdra, IEEE Electron Device Letters Vol. 23 (2002), p [10] P. Hazdra, K. Brand, J. Vobecký, Nucl. Instr. and Meth. Vol. B 192(2002), 291.

7 Gettering and Defect Engineering in Semiconductor Technology X / Controlled Gettering of Implanted Platinum in Silicon Produced by Helium Co-Implantation /