[email protected] A Study of Structural Damage ...

2 downloads 0 Views 1MB Size Report
a) Email address of the corresponding author: prabhu[email protected]. A Study of Structural Damage & Recovery of Si, Ge and Ga FIB implants in Silicon.
Mater. Res. Soc. Symp. Proc. Vol. 1712 © 2014 Materials Research Society DOI: 10.1557/opl.2014.857

A Study of Structural Damage & Recovery of Si, Ge and Ga FIB implants in Silicon Prabhu Balasubramanian1,a), Jeremy F. Graham2 and Robert Hull1 1 2

Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA. FEI, Hillsboro, OR, 97124, USA

ABSTRACT The focused ion beam (FIB) has the necessary precision, spatial resolution and control over ion delivery for potential nano-scale doping of nanostructures such as semiconductor quantum dots (QDs). The ion current density in a FIB is 0.1-10 A/cm2, which is at least three orders of magnitude higher than that in a commercial broad beam ion implanter. Therefore an understanding of FIB implantation damage and recovery is of substantial interest. In this work we employ Raman probes of wavelengths 514 nm and 405 nm for quantifying ion implantation damage—both before and after annealing—in 30 kV Si2+, Ge2+ and Ga+ implants (fluences: 1x1012-5x1015 ions/cm2) into Si(100), for the purpose of understanding the effect of ion species on damage recovery. INTRODUCTION Focused ion beam (FIB) instruments employ a nano-scale ion beam for controlled delivery of ions for locally modifying specimen surfaces. For example, it has been shown that 2D arrays of spots created on Si(100) surfaces using FIB can template, upon subsequent epitaxial growth of SixGe(1-x), the positions of strain relieving quantum dot structures1. Also, it has been shown that FIB has the necessary precision and spatial resolution for delivering ions (at fluences down to a single ion) with nano-scale precision, thus making nano-scale doping possible2. While only a handful of ion species are available in a conventional FIB, a mass-selecting FIB (MSFIB), that is a FIB incorporating a Wien filter with orthogonal electric and magnetic fields for massseparation can be used to produce ion beams of many of the elements in the periodic table from suitable alloy sources. While the ion implantation damage and recovery of Si substrate caused by broad ion beams in commercial ion implanters has been extensively studied, ion implantation damage & recovery of Si by the FIB is not as well understood, although there are reports in the literature that show that the damage and recovery in FIB implants is different from that in the corresponding broad beam implants (e.g.3). In this work we study the effect of FIB ion species (Ge, Si, Ga) on structural damage recovery of FIB-ion implanted Si using Raman spectroscopy. MATERIALS & METHODS FIB implantation damage and post-annealing recovery of Si(100) substrates was studied by performing implantation (ion fluences: 1x1012-5x1015 ions/cm2) of Si2+, Ge2+ and Ga+ accelerated to 30 kV into Si(100), such that the ion energy of the Si and Ge species is 60 keV and that of the Ga species is 30 keV. The ion range and the associated straggle for the Si2+, Ge2+ and Ga+ are 88±33 nm, 48±17 nm and 28±11 nm respectively, determined by performing a)

Email address of the corresponding author: [email protected]

simulations in SRIM 2013 (www.srim.org) that simulates ion-solid interactions. The FIB implants were performed using AuSiGe and Ga sources in a MS-FIB (Orsay Physics Canion column) and conventional FIB (Zeiss Dual Beam) respectively. For comparing the damage in the Si-FIB implants (ion flux is 0.3 A/cm2 for 22 pA current with an estimated beam diameter of 100 nm, which is a conservative estimate considering that beam diameter of ion beams of interest in our Canion column has been estimated to be 25-35 nm for 5 pA currents4) to that in the Sibroad beam implants, Si(100) specimens implanted to 1x1012-1x1015 ions/cm2 fluences of 60 keV Si+ were obtained commercially (http://www.coresystems.com/HOME.asp). Ion flux for the broad beam implantation is estimated at 6-60 µA/cm2 for 1-10 µA ion currents used for the implants, from the estimates for ion beam diameter provided by CORE. As-implanted and annealed specimens were characterized using Raman probes of wavelength 405 and 514 nm, the former being the default Raman probe. Structural damage, D, was determined by subtracting from unity a ratio of the crystalline Si (c-Si) peak height measured from an ion implanted specimen, H, to that from a reference un-implanted Si(100) specimen, Ho (see equation (1)). Peak shifts in Raman spectra from implanted specimens were calculated by noting the shift in the peak position relative to that in the reference spectrum. D=

(1)

The peak used for the measurements of structural damage and peak shifts is the first order optical band of Si at ~ 520 cm-1. For measuring peak heights and peak positions curve fitting was performed in Renishaw’s Raman analysis software, WiRE 3.3. Raman spectra from nonamorphizing implants were fitted using a Voigt function, while spectra from amorphizing implants were fitted using a Gaussian function, and spectra from partially amorphizing implants were fitted using a mix of Voigt and Gaussian functions. Fully and partially amorphizing implants were identified based on the presence of the a-Si peak at around 480 cm-1. The error bars quoted in this work are standard errors. Following implantation and Raman spectroscopy measurements in the as-implanted state, specimens were annealed in a 99.999% purity nitrogen environment. RTA was used for annealing conditions of 730 ˚C, 10 minutes and 800 ˚C, 30 minutes, and a Bruce furnace was used for annealing to 900 ˚C, 30 minutes. RESULTS Raman measurements from several Si-FIB implants are shown in figure 1 a). A comparison of structural damage between Si-FIB implants and the corresponding Si-broad beam implants for the entire range of ion fluences studied is shown in figure 1b). For most fluences the structural damage in the FIB implants is higher than that in the corresponding broad beam implant. For example, in the fluence ranges 1x1012-1x1014 Si/cm2 and 3x1014-1x1015 Si/cm2 the nominal damage in the FIB implants are respectively 5% and 16% higher. Figure 1 c) shows that when the structural damage in the Si-FIB implants is measured using a 514 nm Raman probe instead of a 405 nm Raman probe the measured structural damage is different. This difference is presumably because of different penetration depths of photons of wavelength 514 nm and 405 nm in ion implanted and reference Si, leading to different measurements of D. This is confirmed by the known differences in penetration depths of these photons in un-implanted and ion implanted Si. First, the penetration depths of 514 nm and 405 nm photons for reference Si,

calculated from the absorption coefficients for these wavelength photons in Si reported in Jellison et al.5 are 818 nm and 125 nm. Secondly, although, the photon penetration depths in our FIB implanted Si specimens are not known, the literature data shown in figure 1d) for 80 keV Ar+ implanted Si clearly shows different penetration depths for photons of wavelength 514 nm and 405 nm for a wide range of ion fluences. Having established that the as-implanted damage in the Si-FIB implants is significantly higher when compared to the broad beam Si implants it is clear that the ion flux effect in FIB implantation can cause significant changes to the as-implanted state, and thus also presumably to the damage state post-annealing. In the following discussion, measurements of structural damage and peak shifts from Si and Ge implants are referenced to those from the corresponding Ga implants.

Figure 1. Raman measurements of either signals or structural damage as a function of Si fluence in a) FIB implants at various fluences, and unimplanted Si, using a 405 nm Raman probe, b) FIB and broad beam implants (405 nm Raman measurements), and c) in FIB implants, using 405 nm and 514 nm Raman probes. d) Variation of absorption depth with fluence of 80 keV Ar+ in Si, for 405 nm and 514 nm wavelength photons. The plotted variations in the absorption depth of the two photons is based on the absorption coefficients in figure 1 of Wilbertz et al.6 Figure 2a) compares structural damage for Si, Ge and Ga implants as a function of ion fluence, from which several important points can be made. First, relative to the Ga implants the structural damage is consistently higher in the Ge implants for the fluence range 3x1012-3x1014 ions/cm2, with an average ΔD of 0.20±0.04. Secondly, in the case of the Si implants the ΔD relative to the Ga reference are 0.07±0.04 and -0.07±0.03 for ion fluence ranges 1x1012-3x1013 ions/cm2 and 5x1013-5x1014 ions/cm2 respectively. Thirdly, the structural damage reaches unity at threshold fluences of 1x1015 Si/cm2, 3x1014 Ge/cm2 and 5x1014 Ga/cm2 signifying full amorphization in the implanted region. Additionally, it should be noted that for 3x1014 Si/cm2, 1x1014 Ge/cm2and 3x1014 Ga/cm2 implants, the Raman peak for amorphous Si (a-Si) could be detected, therefore, fluences in the range from these values up to their respective fullamorphization thresholds represent the fluence regimes where partial amorphization has taken place.

Since the damage recovery mechanism of the implanted samples are presumably different depending on whether or not there is amorphization, fluences in all three types of implants have been classified into “non/slightly”, “partially” and “fully” amorphizing fluences. As Table I summarizes, ΔD for the structural damage difference between non/slightly-, partially- and fully- amorphizing Si implants, and the corresponding Ga reference are 0.09, -0.09 and 0.0 respectively. Similarly, ΔD for the Ge implants, referenced to the Ga values, for each of the three categories are: 0.25, 0.0 and 0.0. The Raman peak shifts for the Si, Ge and Ga as-implanted samples is shown in figure 2 b). For Si implants the Raman peak shift for representative fluence in non-amorphizing fluence category is 0.06 cm-1 higher relative to the Ga reference. For the Ge implants the representative peak shift for non-amorphizing fluence category is 0.40 cm-1 higher than the Ga reference. Here, our quantitative comparisons have been limited to the non/slightly amorphizing category because the peak shift measured for the Ga reference for partially-amorphizing fluence category has a large error bar (see the peak shift for 3x1014 Ga/cm2 in figure 2) associated with it, thus forbidding quantitative comparisons. However, as figure 2 shows the peak shift for the representative Ge fluence for partially amorphizing Ge implants is greater than that for the Ga reference; for partially-amorphizing Si implants the nature of the peak shift—relative to the Ga reference—for the representative Si implant is unclear. These results for structural damage and peak shift observed for the Ge and Ga implants might be expected based on the ion species dependent number of vacancies created per incident ion (1413, 1053 and 746 vacancies created by each 30 kV Ge2+, Si2+ and Ga+ ion, as calculated by SRIM). Figure 2. A comparison of Raman measurements of (a) structural damage (b) and peak shift, from the Ga implants as a function of ion fluence to those from the Ge and Si implants.

TABLE I. As-implanted Structural Damage (D) in Non-amorphizing, Partially-amorphizing and Fully-amorphizing Fluences of 30 kV Ga+, Si2+ and Ge2+ FIB implants into Si(100) Implant Species Ga+ Si2+ Ge2+

Representative D in the Indicated Damage Categories No/little Amorphization Partial-amorphization Full-amorphization (c-Si peak only present in (both c-Si and a-Si peaks (a-Si peak only spectra) present in spectra) present in spectra) 0.11 @ 1x1013 0.99 @ 3 x1014 1.0@ 3x1015 (1x1012-1x1014) (5x1014-5x1015) 13 14 0.20 @ 1x10 0.90 @ 3x10 1.00@ 3x1015 (1x1012-1x1014) (3x1014-5x1014) (1x1015-5x1015) 13 14 0.36 @ 1x10 0.99 @ 1 x10 1.0@ 1x1015 (1x1012-5x1013) (3x1014-5x1015)

Figure 3a) compares Raman measurements of structural damage in the Si and Ge implants to those in the Ga implants, post-annealing to 730 ˚C, 10 minutes. In the Ga implants D for the representative non-/slightly-amorphizing, partially-amorphizing and fully-amorphizing fluence categories are 0.04±0.03, 0.14±0.05 and 0.27±0.04.For the Si implants, ΔD for the representative fluences in non-/slightly-amorphizing, partially-amorphizing and fully-amorphizing fluence categories, relative to the Ga reference, are -0.06, -0.16 and -0.29 (measured D for Si implants is essentially zero). For the Ge implants ΔD for the representative fluences in the three damage categories are -0.02, -0.11 and -0.27, relative to the reference Ga implants. It is thus clear that for partially and fully amorphizing Si and Ge fluences the damage recovery is better relative to the corresponding Ga implants.

Figure 3. A comparison of Raman measurements of structural damage from the Ga implants as a function of ion fluence to those from the Ge and Si implants, after post-annealing to a) 730 ˚C, 10 minutes b) 800 ˚C, 30 minutes, and c) 900 ˚C, 30 minutes. D values for samples annealed to 800 ˚C, 30 minutes are shown in figure 3 b). The representative structural damage in the Ga implants corresponding to no/little-, partial- and fullamorphization are: 0.00±0.02, 0.05±0.01 and 0.18±0.02.The ΔD values—calculated relative to the Ga references—for the representative Si implants in each of the three categories are 0.00 , 0.01 and -0.16. Similarly, in case of the Ge implants the ΔD values for the three categories are 0.01, -0.05 and -0.19. Clearly, in the fluence regimes corresponding to partial- and fullamorphization the structural damage recovery in the Si and Ge implants is generally greater than that in the Ga implants. Secondly, in the fluence regime corresponding to no/little-amorphization the damage recovery in all the three types of implants shows full recovery for this annealing condition. Subsequently, all specimens were annealed to 900 ˚C, 30 minutes, refer to figure 3c). The remaining structural damage in the Ga, Si and Ge implants is now very small for no/slightlyamorphizing and partially-amorphizing fluences. In the amorphizing Ga implants the remnant structural damage is significantly high, D = 0.16±0.02 for representative Ga fluence for this category, thus, damage recovery is far from complete. In contrast, in the amorphizing Si and Ge implants the remnant structural damage is almost zero (D= 0.00±0.01 and 0.02±0.01 for representative Si and Ge fluences). DISCUSSION & CONCLUSION In this work better damage recovery in the Si and Ge implants compared to the Ga implants has been observed for partially and fully amorphizing ion fluences for annealing up to 900 ˚C,

which we correlate to unlimited solubility of Si and Ge species in Si, and limited solubility of Ga in Si at the annealing temperatures used for this study. Solubility limits of Ga in Si at 800 ˚C and 900 ˚C are 3.1x1018 Ga/cm3 and 6.4x1018 Ga/cm3 respectively7. These numbers translate into fluences of 1.6x1013 Ga/cm2 and 3.2x1013 Ga/cm2 respectively, after taking into account the depth of 50 nm within which 95.4% of Ga ions in the as-implanted state reside (this depth corresponds to the mean ion range plus twice the longitudinal straggle for 30 kV Ga+ implantation into Si). The observed structural damage for 1x1013 Ga/cm2 after annealing to 800 ˚C, 30 minutes is 0.00±0.02, and it is -0.01±0.02 in 3x1013 Ga/cm2 after annealing to 900 ˚C,30 minutes; corresponding to essentially full recovery in both implants that form the solubility limits at 800 ˚C and 900 ˚C. Secondly, it is observed that the structural damage in the Ga implants steadily increases beyond its solubility limits at 800 ˚C and 900 ˚C, whereas the structural damage in the Si and Ge implants is largely recovered. From these observations it is clear that the solubility of FIB ion species in Si can play an important role in the damage recovery of FIB implanted Si during subsequent annealing. The observed solubility effect of FIB-Ga species on the damage recovery of Si is consistent with the literature report for broad beam-Ga implantations into Si 8. Finally, it is also clear that the structural damage in the 60 keV Si-FIB and Ge-FIB implants in Si can be largely recovered after annealing to 730 ˚C, 10 minutes. After this annealing the structural recovery for the Si and Ge implants are better than 93% and 88% respectively (a structural recovery of 0% and 100% corresponds to structural damage of 1.00 and 0.00 respectively, as defined by the D value in eqn. 1), across the entire fluence range to 5 x 1015 cm-2. After annealing the Si and Ge implants to 800 ˚C, 30 minutes the damage recovery in these implants is at least 94%, and after annealing to 900 ˚C, 30 minutes the damage recovery improves to at least 95%. We acknowledge the financial support from NYSTAR Focus Center (contract numbers: C060117, C080117 and C100117) at RPI. REFERENCES 1

2

3

4 5

6

7

8

M. Kammler, R. Hull, M. C. Reuter, and F. M. Ross, Applied Physics Letters 82 (7), 1093 (2003); J. L. Gray, R. Hull, and J. A. Floro, Journal of Applied Physics 100 (8), 7 (2006). T. Matsukawa, T. Shinada, T. Fukai, and I. Ohdomari, Journal of Vacuum Science & Technology B 16 (4), 2479 (1998); R. Hull, J. Floro, J. Graham, J. Gray, M. Gherasimova, A. Portavoce, and F. M. Ross, Materials Science in Semiconductor Processing 11 (5-6), 160 (2008). M. Tamura, S. Shukuri, M. Moniwa, and M. Default, Applied Physics a-Materials Science & Processing 39 (3), 183 (1986); M. Tamura, S. Shukuri, T. Ishitani, M. Ichikawa, and T. Doi, Japanese Journal of Applied Physics Part 2-Letters 23 (6), L417 (1984). J. F. Graham, C. D. Kell, J. A. Floro, and R. Hull, Nanotechnology 22 (7), 5 (2011). G. E. Jellison, F. A. Modine, C. W. White, R. F. Wood, and R. T. Young, Physical Review Letters 46 (21), 1414 (1981). C. Wilbertz, K. L. Bhatia, W. Kratschmer, and S. Kalbitzer, Materials Science and Engineering B-Solid State Materials for Advanced Technology 2 (4), 325 (1989). S. Haridoss, F. Beniere, M. Gauneau, and A. Rupert, Journal of Applied Physics 51 (11), 5833 (1980). B. L. Crowder, Journal of the Electrochemical Society 118 (6), 943 (1971).