Determination of Free Electron Density in Sequentially Doped Inx Ga1 x As by Raman Spectroscopy Kenneth R. Kort,a P.Y. Hung,b Wei-Yip Loh,b Gennadi Bersuker,b Sarbajit Banerjeec,* a b c
Department of Chemistry, University at Buffalo, The State University of New York, Buffalo, New York 14260 USA Front-End Processes Division, SEMATECH, 257 Fuller Road, Suite 2200, Albany, New York 12203 USA Department of Chemistry, Texas A&M University, College Station, TX 77842 USA
The advent and exponential growth of mobile computing has spurred greater emphasis on the adoption of III–V compound semiconductors in device architectures. The introduction of high charge carrier densities within InxGa1 xAs and the development of metrologies to quantitate the extent of doping have thus emerged as an urgent imperative. As an amphoteric dopant, Si begins to occupy anionic sites at high concentrations, thereby limiting the maximum carrier density that can be obtained upon Si doping of III– V semiconductors. Here, we present Raman results on sequentially doped In0.53Ga0.47As wherein sulfur monolayer doping is used to introduce additional carrier density to Si-doped samples. The sequential doping of Si and S allows for high carrier concentrations of up to 1.3 6 1019 cm 3 to be achieved without damaging the III–V lattice. The coupling of the plasmon in the doped samples to the longitudinal optic phonons allows Raman spectroscopy to serve as an excellent probe of the extent of dopant activation, charge carrier density, and the surface depletion region. In particular, the energy position of a high-frequency coupled mode (HFCM) that is detected above 400 cm 1 is used to extract the free electron density in these samples. The extracted free electron densities are well correlated with measured sheet resistance values and the carrier densities deduced from Hall measurements. Index Headings: Doping; Electron-phonon coupling; InGaAs; Monolayers; Raman spectroscopy; III–V semiconductors.
INTRODUCTION The exponential rise of mobile computing has brought new urgency to the design of device architectures that provide reduced power consumption while being compatible with the accelerated scaling of device dimensions. 1 III–V semiconductors have emerged as frontrunners for the active channels of proposed device geometries that are amenable to aggressive scaling such as field-effect transistors (FETs) characterized by Si/III-V interfaces across trenches (FinFETs; recently demonstrated on the 300 mm wafer scale by IMEC) or multiple gate FETs (MuGFETs) that provide better spatial control of potential in different regions of the chip.2 In comparison to state-of-the-art strained Si, III–V semiconductors (with InxGa1 xAs being the most prominent example) offer electron mobility and significantly enhanced injection velocities that can be as much as sixfold higher. However, of the impediments to scaling these geometries to ever smaller dimensions, development of controllable doping methods and verification of Received 20 May 2014; accepted 10 July 2014. * Author to whom correspondence should be sent. E-mail: banerjee@ chem.tamu.edu. DOI: 10.1366/14-07602
Volume 69, Number 2, 2015
activated dopant concentrations through non-destructive metrologies represent two of the most formidable challenges.3,4 Here, we demonstrate the use of Raman spectroscopy as a sensitive probe of plasmon–phonon coupling in InxGa1 xAs samples doped sequentially with silicon and sulfur and provide a measure of the free electron density based on the energy divergence of a high-frequency coupled mode (HFCM). It is well established that Si can serve as an amphoteric dopant for III–V semiconductors and that, beyond a certain dopant concentration (greater than 6 3 1018 cm 3), incorporated Si dopants begin to occupy both cation and anion sites, thereby negating the benefits of additional dopant incorporation from the perspective of increasing the free carrier density.5–7 During molecular beam epitaxy (MBE) growth, incorporated tetravalent Si atoms preferentially occupy trivalent Ga–In cationic sites (generating a free charge carrier for each In–Ga site occupied by Si), but upon annealing or at higher dopant concentrations, ‘‘site switching’’ phenomena are manifested wherein doped Si atoms migrate to vacancies on the As sublattice (surface segregation has also been noted), thereby setting essentially immutable limits to the electron density that can be realized by dopant incorporation during crystal growth.6 To avoid manifestation of such charge compensation, multiple doping strategies are being investigated. All such strategies would greatly benefit from the development of suitable analytical methods for quantitating the efficacy of dopant activation, given that it is not enough to simply incorporate dopants within the semiconductor lattice; the dopant atoms must occupy the appropriate lattice sites wherein they can increase the charge carrier density.3,4,8 The monolayer doping method developed by Javey et al.8 is notable for being agnostic to the channel geometry (thereby ensuring better homogeneity of carrier density across irregularly shaped channels, particularly for nanometer-sized features), selflimiting (the extent of dopant incorporation is tunable based on the areal concentration of the dopant precursor species in the monolayer, the annealing conditions, and the capping layer), provides access to ultra-shallow junctions, and does not damage the III–V zinc blende lattice.3,8 Consequently, this method is being extensively investigated as a means to overcome the limitations of ion implantation and solid source diffusion processes. Raman spectroscopy serves as a sensitive contactless probe of electron–phonon coupling in InxGa1 xAs ternary semiconductors and has been used to evaluate the stoichiometry, free carrier density, band bending, and damage to the crystalline lattice induced by dopant incorporation.9–11 We have previously shown that charge
0003-7028/15/6902-0239/0 Q 2015 Society for Applied Spectroscopy
APPLIED SPECTROSCOPY
239
TABLE I. Target Si doping levels for In0.53Ga0.47As samples (during MBE growth), peak positions of the HFCM bands (Fig. 2), activated dopant concentrations deduced from the peak positions of the HFCM modes using the Lindhard–Mermin formalism (nRS), additional charge carrier density introduced as a result of sulfur monolayer doping, sheet resistance values measured using the van der Pauw method, Neff values determined from Hall measurements, and estimated Neff values calculated by dividing the determined Neff values by the thickness of ˚ ). the In0.53Ga0.47As epilayer (200 A
Sample
Si Target Doping (cm 3)
1 1-S 2 2-S 3 3-S
5 5 5 5 3 3
a
3 3 3 3 3 3
S Monolayer Doping
HFCM Peak Position (cm 1)
No Yes No Yes No Yes
n.d.a 835 881 1043 1132 1198
1017 1017 1018 1018 1019 1019
Estimated nRS (cm 3) ,1.4 4.6 5.3 8.9 1.1 1.3
3 3 3 3 3 3
1017 1018 1018 1018 1019 1019
Contribution from SMLD (cm 3)
Sheet Resistance X sq. 1
.4.4 3 1018
2199 562 380 196 153 143
3.6 3 1018 2 3 1018
Hall Neff (cm 2) 1.2 4.7 9.1 1.7 2.4 2.5
3 3 3 3 3 3
Estimated Neff (cm 3)
1012 1012 1012 1013 1013 1013
6 2.4 4.6 8.4 1.2 1.3
3 3 3 3 3 3
1017 1018 1018 1018 1019 1019
Mobilities (cm2V 1s 1) 2390 2340 2110 1890 1720 1720
n.d.: not detectable.
carrier densities created by the facile method of sulfur monolayer doping of In0.53Ga0.47As can be measured non-destructively by Raman spectroscopy (analogous to the more extensively studied Si-doped III–V materials).12 As a measure of the plasmon–phonon coupling, the energy divergence of the HFCM mode as a function of dopant concentration ought to be agnostic to the nature of the dopants and should vary only as a function of the charge carrier density within the electron gas. Therefore, it stands to reason that the extent of activation of the incorporated dopants and the generated free electron density should be accessible to monitoring by Raman spectroscopy, both for the initially doped samples as well as upon subsequent sulfur doping by the monolayer deposition method. To verify this contention, we have acquired Raman spectra for Si-doped In0.53Ga0.47As (with Si incorporated during growth via MBE) before and after being further subjected to sulfur monolayer doping (Table I).
EXPERIMENTAL In0.53Ga0.47As (100) samples were epitaxially grown as a 20 nm layer on lattice matched (100) InP substrates with the intermediation of an In0.52Al0.48As spacer layer (40 nm). Si was incorporated during the MBE growth process. The samples denoted 1, 2, and 3 correspond to intended Si concentrations of 5 3 1017, 5 3 1018, and 3 3 1019 cm 3, respectively (Table I). Samples 1-S, 2-S, and 3-S correspond to precisely the same Si concentrations as their singly doped counterparts, but have been additionally subjected to sulfur monolayer doping using a process described previously in the literature.3,8,13 Briefly, after deposition of the (NH4)2S layer from aqueous solution, an amorphous SiO2 capping layer was deposited onto the doped samples (as well as onto the exclusively Si-doped samples as a control), which were then annealed as described previously in the literature.3,8,13 Figure 1 illustrates the stack structure examined in this study. Raman measurements were performed at room temperature in backscattering geometry (selection rules permit the observation of only longitudinal optic modes for scattering from (100) In0.53Ga0.47As) using a Jobin Yvon Horiba LabRAM HR (Villeneuve d’Ascq, France) instrument coupled to an Olympus BX41 microscope with a 503 objective using 514.5 nm laser excitation from an
240
Volume 69, Number 2, 2015
Ar-ion laser, which provides a probe depth estimated to be 22.4 nm given the high optical absorption coefficient of InxGa1 xAs at 514.5 nm.14 An 1800 lines/mm grating spectrometer equipped with a Peltier-cooled chargecoupled device (CCD) detector (Andor) was used to acquire spectra yielding a spectral resolution better than 2 cm 1 and spatial resolution of about 1 lm. The laser power was maintained below 10 mW to minimize local heating and photoexcitation of charge carriers in the highly doped samples. Sheet resistance, mobilities, and Hall Neff values were derived from Hall measurements. The Neff (cm 3) values (Table I) were derived by assuming uniform distribution of dopants in the 20 nm InxGa1 xAs epilayer, which is somewhat of an approximation given the stabilization of ultra-shallow junctions, but notably at 514.5 nm laser excitation, the entire epilayer is expected to be probed as verified by observation of the AlAs longitudinal optic mode in the Raman spectra (see Results and Discussion).
RESULTS AND DISCUSSION In polar semiconductors characterized by an appreciable electronegativity difference between the cation and the anion, the longitudinal optic phonons establish a periodically varying macroscopic electric field that can strongly couple to the electron gas derived from free charge carriers subject to the plasmon frequency being comparable to the optical phonon frequencies.9,10,14 For doped ternary semiconductors such as InxGa1 xAs,
FIG. 1. Schematic depiction of epilayer In0.53Ga0.47As samples doped conjointly, first with Si during MBE growth of the III–V semiconductor, and subsequently by sulfur monolayer deposition.
FIG. 2. Evolution of the HFCM region of the Raman spectra of Si-doped In0.53Ga0.47As epilayers upon doping via the sulfur monolayer deposition method. Sample 1 does not have an adequate concentration of free carriers to give rise to a plasmon that can couple with the longitudinal optic phonons. The activated dopant concentrations extracted using the peak positions of the HFCM modes based on the Lindhard–Mermin formalism are listed in Table I. * indicates CCD artifact.
coupling of the longitudinal optic modes with the plasmon gives rise to three distinct phonon–plasmon coupled modes: a HFCM, which diverges over a broad energy range as a function of the dopant concentration, an intermediate frequency coupled mode (IFCM), and a low-frequency coupled mode (LFCM); the latter two spectral features are relatively close in energy to the optical modes of the undoped alloy. Owing to its strongly plasmon–phonon coupled nature, the energy of the HFCM band is closely correlated to the free carrier density in InxGa1 xAs, thereby providing a sensitive contactless optical measure of the carrier concentration.9,10,14 This allows Raman spectroscopy to serve as a direct measure of carrier concentration unlike techniques such as secondary ion mass spectrometry or glow discharge methods that provide an indication of the dopant concentration but do not discriminate between active (corresponding to dopants that have been incorporated in the desired lattice sites and have generated free carriers) and inactive dopants. Figure 2 shows the Raman spectra for Samples 1, 2, and 3 in the HFCM region with the corresponding spectra for their sequentially doped counterparts overlaid to illustrate the pronounced modification of these spectral features upon sulfur monolayer doping. For Sample 1, the intended Si concentration is 5 3 1017 cm 3. The concentration of activated Si dopants is likely lower, and the electron density in this sample is not adequate to give rise to a HFCM feature (,1.43 1017 cm 3). However, upon doping with sulfur through the monolayer deposition process, a HFCM band appears at 835 cm 1, indicating a pronounced increase in the free charge carrier density to 4.6 3 1018 cm 3. The free charge carrier density has been estimated for all the samples using the previously described Lindhard–Mermin formalism.10,11 Consistent with the increase in the charge carrier density deduced from the Raman spectra, Hall measurements show a significant decrease in sheet resistance from 2199 to 562 X sq. 1 (Table I). Samples 2 and 3 show HFCM features at 881 and 1132 cm 1, respectively, which correspond to free electron concentrations of 5.3 3 1018 and 1.1 3 1019, respectively. A substantial Raman shift to
FIG. 3. Evolution of the longitudinal optic region of the Raman spectra of In0.53Ga0.47As epilayers for Sample 1 before and after sulfur monolayer deposition. The spectra are contrasted to the spectrum obtained for an undoped In0.53Ga0.47As epilayer. The additional creation of charge carrier density by sulfur doping causes a decrease of the surface depletion layer resulting in a dampening of the intensity of the GaAs-like longitudinal optic mode.
higher frequencies is evidenced upon sulfur monolayer doping for both these samples. The increase in carrier density surmised from the Raman measurements (and listed in Table I) is further corroborated by the measured decrease of the sheet resistance values. Additionally, Table I list Neff values for the three sample sets, which have been obtained using the carrier densities calculated from Hall measurements and extrapolated assuming relatively homogeneous doping across the 20 nm InxGa1 xAs epilayers. The Neff values are seen to be in good accord with the charge carrier densities extrapolated using the Lindhard–Mirmin formalism.10,11 For all three samples, which were treated with the same sulfur monolayer doping protocol, the increase in free electron density appears to be comparable, on the order of 2 3 1018 to 5 3 1018 cm 3. Sequential co-doping on the In–Ga and As lattices is thus observed to an effective means to obtain high free electron densities in InxGa1 xAs. Figure 3 shows the optical phonon region for Sample 1 before and after S monolayer deposition contrasted with the Raman spectrum of an undoped In0.53Ga0.47As sample. For the undoped sample, the two longitudinal optic modes corresponding to the GaAs-like phonon at 266 cm 1 and the InAs centered at 230 cm 1 are clearly discernible. The former grows in intensity at the expense of the latter due to the intrinsically weaker oscillator strength of InAs and the increased polarization of the GaAs longitudinal optic branch.15,16 Upon Si doping, the GaAs-like LO mode at 266 cm 1 is dampened, and instead, the IFCM band at 259 cm 1 grows in intensity. The IFCM is observed to further grow in intensity and becomes the most prominent feature above 250 cm 1 upon sulfur monolayer deposition. In the Raman spectrum of Sample 1, the remnant feature at 266 cm 1 arises from the surface depletion region, where the plasmon is screened.10 The surface depletion region is thinned with increased doping, resulting in a significant diminution in the spectral contribution from the GaAs-like feature upon sulfur monolayer deposition. Both the IFCM and LFCM
APPLIED SPECTROSCOPY
241
phonon region provide a measure of the surface depletion layer. ACKNOWLEDGMENTS We gratefully acknowledge support of this work from the New York State Center of Excellence in Materials Informatics.
FIG. 4. Raman spectra acquired in the 330–380 cm 1 region for Sidoped In0.53Ga0.47As epilayers upon doping via the sulfur monolayer deposition method. The InP and AlAs-like longitudinal optic modes are clearly discernible indicating that the incident laser probes the entire sample down to the substrate. The band at 353 cm 1 (denoted by *) is likely an As2S3 feature that arises due to S reacting with As in the In0.53Ga0.47As epilayer.17
are shifted towards lower frequencies with increased doping. A shift to low frequencies for the coupled modes with increasing carrier density is atypical of polar semiconductors (increased doping usually results in phonons coupling to higher energy plasmon bands) and occurs in In0.53Ga0.47As due to the unique circumstances of its low phonon energy and small effective mass, which results in these modes residing within the Landau damping region of In0.53Ga0.47As.10,12 Figure 4 shows that, for both InP and AlAs-like longitudinal optic modes at 347 and 367 cm 1, respectively, are observed for all six samples suggesting that the incident laser is probing well pass the In0.53Ga0.47As and In0.52Al0.48As epilayers through to the InP substrate. For 2-S and 3-S, the slight dampening of the InP longitudinal optic mode upon sulfur monolayer deposition is likely a result of the reduced penetration depth resulting from the increased carrier density.
CONCLUSIONS In summary, we have shown that Raman spectroscopy serves as a sensitive probe of free charge density in sequentially doped InxGa1 xAs epilayers incorporating both Si and S dopants. This method of determining dopant activation and free carrier density is observed to be agnostic of the nature of the individual dopants or the mode of dopant incorporation. Sulfur monolayer deposition is observed to be a viable means to further increase the electron density of doped InxGa1 xAs samples without damage to the crystalline lattice. The HFCM Raman band is observed to be a sensitive probe of the free electron density and concentration of activated dopants in InxGa1 xAs. The relative intensities of the GaAs-like feature and the IFCM mode in the optical
242
Volume 69, Number 2, 2015
1. R. Stevenson. ‘‘Compound Semiconductors Join the Race to Sustain Moore’s Law’’. IEEE Spectrum. December 2013. http:// spectrum.ieee.org/semiconductors/materials/compoundsemiconductors-join-the-race-to-sustain-moores-law [accessed Dec 1 2014]. 2. C. Hock-Chun Chin, G. Xiao, W. Lanxiang, L. Hock-Koon, S. Luping, Y. Yee-Chia. ‘‘III–V Multiple-Gate Field-Effect Transistors With HighMobility In0.7Ga0.3As Channel and Epi-Controlled RetrogradeDoped Fin’’. IEEE Electron Device Lett. 2011. 32(2): 146-148. 3. J.H. Yum, H.S. Shin, R. Hill, J. Oh, H.D. Lee, R.M. Mushinski, T.W. Hudnall, C.W. Bielawski, S.K. Banerjee, W.Y. Loh, W.-E. Wang, P. Kirsch. ‘‘A Study of Capping Layers for Sulfur Monolayer Doping on III–V Junctions’’. Appl. Phys. Lett. 2012. 101(25): 253514-253513. 4. K.-W. Ang, J. Barnett, W.Y. Loh, J. Huang, B.G. Min, P.Y. Hung, I. Ok, J.H. Yum, G. Bersuker, M. Rodgers, V. Kaushik, S. Gausepohl, C. Hobbs, P.D. Kirsch, R. Jammy. ‘‘300 mm FinFET Results Utilizing Conformal, Damage Free, Ultra Shallow Junctions (Xj5nm) Formed with Molecular Monolayer Doping Technique’’. International Electron Devices Meeting, 1988. IEDM ’88. Technical Digest., International 01/2011; doi: 10.1109/IEDM.2011.6131679. 5. M.E. Greiner, J.F. Gibbons. ‘‘Diffusion and Electrical Properties of Silicon-Doped Gallium Arsenide’’. J. Appl. Phys. 1985. 57(12): 51815187. 6. R.C. Newman. ‘‘The Lattice Locations of Silicon Impurities in GaAs: Effects Due to Stoichiometry, the Fermi Energy, the Solubility Limit and DX Behaviour’’. Semicond. Sci. Technol. 1994. 9(10): 1749-1762. 7. A.A. Quivy, A.L. Sperandio, E.C.F. da Silva, J.R. Leite. ‘‘p-Type Doping of GaAs (001) Layers Grown by MBE Using Silicon as a Dopant’’. J. Cryst. Growth. 1999. 206(3): 171-176. 8. J.C. Ho, R. Yerushalmi, Z.A. Jacobson, Z. Fan, R.L. Alley, A. Javey. ‘‘Controlled Nanoscale Doping of Semiconductors via Molecular Monolayers’’. Nat. Mater. 2008. 7(1): 62-67. 9. J.E. Maslar, J.F. Dorsten, P.W. Bohn, S. Agarwala, I. Adesida, C. Caneau, R. Bhat. ‘‘Electron-Phonon Interactions in n-Type In0.53Ga0.47As and In0.52Al0.48As Studied by Inelastic Light Scattering’’. Phys. Rev. B: Condens. Matter Mater. Phys. 1994. 50(23): 1714317150. ˜ M. Hopkinson. ‘‘Raman 10. R. Cusco´, L. Artu´s, S. Herna´ndez, J. Iba´nez, Scattering by LO Phonon-Plasmon Coupled Modes in n-Type In0.53Ga0.47As’’. Phys. Rev. B: Condens. Matter Mater. Phys. 2001. 65(3): 035210. 11. S. Hernandez, R. Cusco, N. Blanco, G. Gonzalez-Diaz, L. Artus. ‘‘Study of the Electrical Activation of Si-Implanted InGaAs by Means of Raman Scattering’’. J. Appl. Phys. 2003. 93(5): 2659-2662. 12. K. Kort, P.Y. Hung, P.D. Lysaght, W.-Y. Loh, G. Bersuker, S. Banerjee. ‘‘Raman Spectroscopy Studies of Dopant Activation and Free Electron Density of In0.53Ga0.47As with Sulfur Monolayer Doping’’. Phys. Chem. Chem. Phys. 2014. 16(14): 6539-6543. 13. J.C. Ho, A.C. Ford, Y.-L. Chueh, P.W. Leu, O. Ergen, K. Takei, G. Smith, P. Majhi, J. Bennett, A. Javey. ‘‘Nanoscale Doping of InAs via Sulfur Monolayers’’. Appl. Phys. Lett. 2009. 95(7): 072108-072103. 14. S. Herna´ndez, N. Blanco, I. Ma´rtil, G. Gonza´lez-Dı´ az, R. Cusco´, L. Artu´s. ‘‘Evidence of Phosphorus Incorporation into InGaAs/InP Epilayers after Thermal Annealing’’. J. Appl. Phys. 2003. 93(11): 9019-9023. 15. C.S. Jung, H.S. Kim, G.B. Jung, K.J. Gong, Y.J. Cho, S.Y. Jang, C.H. Kim, C.-W. Lee, J. Park. ‘‘Composition and Phase Tuned InGaAs Alloy Nanowires’’. J. Phys. Chem. C. 2011. 115(16): 7843-7850. 16. J. Groenen, R. Carles, G. Landa, C. Guerret-Pie´court, C. Fontaine, M. Gendry. ‘‘Optical-Phonon Behavior in Ga1-xInxAs: The Role of Microscopic Strains and Ionic Plasmon Coupling’’. Phys. Rev. B. 1998. 58(16): 10452-10462. 17. S. Mamedov, A. Kisliuk, D. Quitmann. ‘‘Effect of Preparation Conditions on the Low Frequency Raman Spectrum of Glassy As2S3’’. J. Mater. Sci. 1998. 33(1): 41.
Department of Chemistry, University at Buffalo, The State University of New York, Buffalo, New York 14260 USA Front-End Processes Division, SEMATECH, 257 Fuller Road, Suite 2200, Albany, New York 12203 USA Department of Chemistry, Texas A&M University, College Station, TX 77842 USA
The advent and exponential growth of mobile computing has spurred greater emphasis on the adoption of III–V compound semiconductors in device architectures. The introduction of high charge carrier densities within InxGa1 xAs and the development of metrologies to quantitate the extent of doping have thus emerged as an urgent imperative. As an amphoteric dopant, Si begins to occupy anionic sites at high concentrations, thereby limiting the maximum carrier density that can be obtained upon Si doping of III– V semiconductors. Here, we present Raman results on sequentially doped In0.53Ga0.47As wherein sulfur monolayer doping is used to introduce additional carrier density to Si-doped samples. The sequential doping of Si and S allows for high carrier concentrations of up to 1.3 6 1019 cm 3 to be achieved without damaging the III–V lattice. The coupling of the plasmon in the doped samples to the longitudinal optic phonons allows Raman spectroscopy to serve as an excellent probe of the extent of dopant activation, charge carrier density, and the surface depletion region. In particular, the energy position of a high-frequency coupled mode (HFCM) that is detected above 400 cm 1 is used to extract the free electron density in these samples. The extracted free electron densities are well correlated with measured sheet resistance values and the carrier densities deduced from Hall measurements. Index Headings: Doping; Electron-phonon coupling; InGaAs; Monolayers; Raman spectroscopy; III–V semiconductors.
INTRODUCTION The exponential rise of mobile computing has brought new urgency to the design of device architectures that provide reduced power consumption while being compatible with the accelerated scaling of device dimensions. 1 III–V semiconductors have emerged as frontrunners for the active channels of proposed device geometries that are amenable to aggressive scaling such as field-effect transistors (FETs) characterized by Si/III-V interfaces across trenches (FinFETs; recently demonstrated on the 300 mm wafer scale by IMEC) or multiple gate FETs (MuGFETs) that provide better spatial control of potential in different regions of the chip.2 In comparison to state-of-the-art strained Si, III–V semiconductors (with InxGa1 xAs being the most prominent example) offer electron mobility and significantly enhanced injection velocities that can be as much as sixfold higher. However, of the impediments to scaling these geometries to ever smaller dimensions, development of controllable doping methods and verification of Received 20 May 2014; accepted 10 July 2014. * Author to whom correspondence should be sent. E-mail: banerjee@ chem.tamu.edu. DOI: 10.1366/14-07602
Volume 69, Number 2, 2015
activated dopant concentrations through non-destructive metrologies represent two of the most formidable challenges.3,4 Here, we demonstrate the use of Raman spectroscopy as a sensitive probe of plasmon–phonon coupling in InxGa1 xAs samples doped sequentially with silicon and sulfur and provide a measure of the free electron density based on the energy divergence of a high-frequency coupled mode (HFCM). It is well established that Si can serve as an amphoteric dopant for III–V semiconductors and that, beyond a certain dopant concentration (greater than 6 3 1018 cm 3), incorporated Si dopants begin to occupy both cation and anion sites, thereby negating the benefits of additional dopant incorporation from the perspective of increasing the free carrier density.5–7 During molecular beam epitaxy (MBE) growth, incorporated tetravalent Si atoms preferentially occupy trivalent Ga–In cationic sites (generating a free charge carrier for each In–Ga site occupied by Si), but upon annealing or at higher dopant concentrations, ‘‘site switching’’ phenomena are manifested wherein doped Si atoms migrate to vacancies on the As sublattice (surface segregation has also been noted), thereby setting essentially immutable limits to the electron density that can be realized by dopant incorporation during crystal growth.6 To avoid manifestation of such charge compensation, multiple doping strategies are being investigated. All such strategies would greatly benefit from the development of suitable analytical methods for quantitating the efficacy of dopant activation, given that it is not enough to simply incorporate dopants within the semiconductor lattice; the dopant atoms must occupy the appropriate lattice sites wherein they can increase the charge carrier density.3,4,8 The monolayer doping method developed by Javey et al.8 is notable for being agnostic to the channel geometry (thereby ensuring better homogeneity of carrier density across irregularly shaped channels, particularly for nanometer-sized features), selflimiting (the extent of dopant incorporation is tunable based on the areal concentration of the dopant precursor species in the monolayer, the annealing conditions, and the capping layer), provides access to ultra-shallow junctions, and does not damage the III–V zinc blende lattice.3,8 Consequently, this method is being extensively investigated as a means to overcome the limitations of ion implantation and solid source diffusion processes. Raman spectroscopy serves as a sensitive contactless probe of electron–phonon coupling in InxGa1 xAs ternary semiconductors and has been used to evaluate the stoichiometry, free carrier density, band bending, and damage to the crystalline lattice induced by dopant incorporation.9–11 We have previously shown that charge
0003-7028/15/6902-0239/0 Q 2015 Society for Applied Spectroscopy
APPLIED SPECTROSCOPY
239
TABLE I. Target Si doping levels for In0.53Ga0.47As samples (during MBE growth), peak positions of the HFCM bands (Fig. 2), activated dopant concentrations deduced from the peak positions of the HFCM modes using the Lindhard–Mermin formalism (nRS), additional charge carrier density introduced as a result of sulfur monolayer doping, sheet resistance values measured using the van der Pauw method, Neff values determined from Hall measurements, and estimated Neff values calculated by dividing the determined Neff values by the thickness of ˚ ). the In0.53Ga0.47As epilayer (200 A
Sample
Si Target Doping (cm 3)
1 1-S 2 2-S 3 3-S
5 5 5 5 3 3
a
3 3 3 3 3 3
S Monolayer Doping
HFCM Peak Position (cm 1)
No Yes No Yes No Yes
n.d.a 835 881 1043 1132 1198
1017 1017 1018 1018 1019 1019
Estimated nRS (cm 3) ,1.4 4.6 5.3 8.9 1.1 1.3
3 3 3 3 3 3
1017 1018 1018 1018 1019 1019
Contribution from SMLD (cm 3)
Sheet Resistance X sq. 1
.4.4 3 1018
2199 562 380 196 153 143
3.6 3 1018 2 3 1018
Hall Neff (cm 2) 1.2 4.7 9.1 1.7 2.4 2.5
3 3 3 3 3 3
Estimated Neff (cm 3)
1012 1012 1012 1013 1013 1013
6 2.4 4.6 8.4 1.2 1.3
3 3 3 3 3 3
1017 1018 1018 1018 1019 1019
Mobilities (cm2V 1s 1) 2390 2340 2110 1890 1720 1720
n.d.: not detectable.
carrier densities created by the facile method of sulfur monolayer doping of In0.53Ga0.47As can be measured non-destructively by Raman spectroscopy (analogous to the more extensively studied Si-doped III–V materials).12 As a measure of the plasmon–phonon coupling, the energy divergence of the HFCM mode as a function of dopant concentration ought to be agnostic to the nature of the dopants and should vary only as a function of the charge carrier density within the electron gas. Therefore, it stands to reason that the extent of activation of the incorporated dopants and the generated free electron density should be accessible to monitoring by Raman spectroscopy, both for the initially doped samples as well as upon subsequent sulfur doping by the monolayer deposition method. To verify this contention, we have acquired Raman spectra for Si-doped In0.53Ga0.47As (with Si incorporated during growth via MBE) before and after being further subjected to sulfur monolayer doping (Table I).
EXPERIMENTAL In0.53Ga0.47As (100) samples were epitaxially grown as a 20 nm layer on lattice matched (100) InP substrates with the intermediation of an In0.52Al0.48As spacer layer (40 nm). Si was incorporated during the MBE growth process. The samples denoted 1, 2, and 3 correspond to intended Si concentrations of 5 3 1017, 5 3 1018, and 3 3 1019 cm 3, respectively (Table I). Samples 1-S, 2-S, and 3-S correspond to precisely the same Si concentrations as their singly doped counterparts, but have been additionally subjected to sulfur monolayer doping using a process described previously in the literature.3,8,13 Briefly, after deposition of the (NH4)2S layer from aqueous solution, an amorphous SiO2 capping layer was deposited onto the doped samples (as well as onto the exclusively Si-doped samples as a control), which were then annealed as described previously in the literature.3,8,13 Figure 1 illustrates the stack structure examined in this study. Raman measurements were performed at room temperature in backscattering geometry (selection rules permit the observation of only longitudinal optic modes for scattering from (100) In0.53Ga0.47As) using a Jobin Yvon Horiba LabRAM HR (Villeneuve d’Ascq, France) instrument coupled to an Olympus BX41 microscope with a 503 objective using 514.5 nm laser excitation from an
240
Volume 69, Number 2, 2015
Ar-ion laser, which provides a probe depth estimated to be 22.4 nm given the high optical absorption coefficient of InxGa1 xAs at 514.5 nm.14 An 1800 lines/mm grating spectrometer equipped with a Peltier-cooled chargecoupled device (CCD) detector (Andor) was used to acquire spectra yielding a spectral resolution better than 2 cm 1 and spatial resolution of about 1 lm. The laser power was maintained below 10 mW to minimize local heating and photoexcitation of charge carriers in the highly doped samples. Sheet resistance, mobilities, and Hall Neff values were derived from Hall measurements. The Neff (cm 3) values (Table I) were derived by assuming uniform distribution of dopants in the 20 nm InxGa1 xAs epilayer, which is somewhat of an approximation given the stabilization of ultra-shallow junctions, but notably at 514.5 nm laser excitation, the entire epilayer is expected to be probed as verified by observation of the AlAs longitudinal optic mode in the Raman spectra (see Results and Discussion).
RESULTS AND DISCUSSION In polar semiconductors characterized by an appreciable electronegativity difference between the cation and the anion, the longitudinal optic phonons establish a periodically varying macroscopic electric field that can strongly couple to the electron gas derived from free charge carriers subject to the plasmon frequency being comparable to the optical phonon frequencies.9,10,14 For doped ternary semiconductors such as InxGa1 xAs,
FIG. 1. Schematic depiction of epilayer In0.53Ga0.47As samples doped conjointly, first with Si during MBE growth of the III–V semiconductor, and subsequently by sulfur monolayer deposition.
FIG. 2. Evolution of the HFCM region of the Raman spectra of Si-doped In0.53Ga0.47As epilayers upon doping via the sulfur monolayer deposition method. Sample 1 does not have an adequate concentration of free carriers to give rise to a plasmon that can couple with the longitudinal optic phonons. The activated dopant concentrations extracted using the peak positions of the HFCM modes based on the Lindhard–Mermin formalism are listed in Table I. * indicates CCD artifact.
coupling of the longitudinal optic modes with the plasmon gives rise to three distinct phonon–plasmon coupled modes: a HFCM, which diverges over a broad energy range as a function of the dopant concentration, an intermediate frequency coupled mode (IFCM), and a low-frequency coupled mode (LFCM); the latter two spectral features are relatively close in energy to the optical modes of the undoped alloy. Owing to its strongly plasmon–phonon coupled nature, the energy of the HFCM band is closely correlated to the free carrier density in InxGa1 xAs, thereby providing a sensitive contactless optical measure of the carrier concentration.9,10,14 This allows Raman spectroscopy to serve as a direct measure of carrier concentration unlike techniques such as secondary ion mass spectrometry or glow discharge methods that provide an indication of the dopant concentration but do not discriminate between active (corresponding to dopants that have been incorporated in the desired lattice sites and have generated free carriers) and inactive dopants. Figure 2 shows the Raman spectra for Samples 1, 2, and 3 in the HFCM region with the corresponding spectra for their sequentially doped counterparts overlaid to illustrate the pronounced modification of these spectral features upon sulfur monolayer doping. For Sample 1, the intended Si concentration is 5 3 1017 cm 3. The concentration of activated Si dopants is likely lower, and the electron density in this sample is not adequate to give rise to a HFCM feature (,1.43 1017 cm 3). However, upon doping with sulfur through the monolayer deposition process, a HFCM band appears at 835 cm 1, indicating a pronounced increase in the free charge carrier density to 4.6 3 1018 cm 3. The free charge carrier density has been estimated for all the samples using the previously described Lindhard–Mermin formalism.10,11 Consistent with the increase in the charge carrier density deduced from the Raman spectra, Hall measurements show a significant decrease in sheet resistance from 2199 to 562 X sq. 1 (Table I). Samples 2 and 3 show HFCM features at 881 and 1132 cm 1, respectively, which correspond to free electron concentrations of 5.3 3 1018 and 1.1 3 1019, respectively. A substantial Raman shift to
FIG. 3. Evolution of the longitudinal optic region of the Raman spectra of In0.53Ga0.47As epilayers for Sample 1 before and after sulfur monolayer deposition. The spectra are contrasted to the spectrum obtained for an undoped In0.53Ga0.47As epilayer. The additional creation of charge carrier density by sulfur doping causes a decrease of the surface depletion layer resulting in a dampening of the intensity of the GaAs-like longitudinal optic mode.
higher frequencies is evidenced upon sulfur monolayer doping for both these samples. The increase in carrier density surmised from the Raman measurements (and listed in Table I) is further corroborated by the measured decrease of the sheet resistance values. Additionally, Table I list Neff values for the three sample sets, which have been obtained using the carrier densities calculated from Hall measurements and extrapolated assuming relatively homogeneous doping across the 20 nm InxGa1 xAs epilayers. The Neff values are seen to be in good accord with the charge carrier densities extrapolated using the Lindhard–Mirmin formalism.10,11 For all three samples, which were treated with the same sulfur monolayer doping protocol, the increase in free electron density appears to be comparable, on the order of 2 3 1018 to 5 3 1018 cm 3. Sequential co-doping on the In–Ga and As lattices is thus observed to an effective means to obtain high free electron densities in InxGa1 xAs. Figure 3 shows the optical phonon region for Sample 1 before and after S monolayer deposition contrasted with the Raman spectrum of an undoped In0.53Ga0.47As sample. For the undoped sample, the two longitudinal optic modes corresponding to the GaAs-like phonon at 266 cm 1 and the InAs centered at 230 cm 1 are clearly discernible. The former grows in intensity at the expense of the latter due to the intrinsically weaker oscillator strength of InAs and the increased polarization of the GaAs longitudinal optic branch.15,16 Upon Si doping, the GaAs-like LO mode at 266 cm 1 is dampened, and instead, the IFCM band at 259 cm 1 grows in intensity. The IFCM is observed to further grow in intensity and becomes the most prominent feature above 250 cm 1 upon sulfur monolayer deposition. In the Raman spectrum of Sample 1, the remnant feature at 266 cm 1 arises from the surface depletion region, where the plasmon is screened.10 The surface depletion region is thinned with increased doping, resulting in a significant diminution in the spectral contribution from the GaAs-like feature upon sulfur monolayer deposition. Both the IFCM and LFCM
APPLIED SPECTROSCOPY
241
phonon region provide a measure of the surface depletion layer. ACKNOWLEDGMENTS We gratefully acknowledge support of this work from the New York State Center of Excellence in Materials Informatics.
FIG. 4. Raman spectra acquired in the 330–380 cm 1 region for Sidoped In0.53Ga0.47As epilayers upon doping via the sulfur monolayer deposition method. The InP and AlAs-like longitudinal optic modes are clearly discernible indicating that the incident laser probes the entire sample down to the substrate. The band at 353 cm 1 (denoted by *) is likely an As2S3 feature that arises due to S reacting with As in the In0.53Ga0.47As epilayer.17
are shifted towards lower frequencies with increased doping. A shift to low frequencies for the coupled modes with increasing carrier density is atypical of polar semiconductors (increased doping usually results in phonons coupling to higher energy plasmon bands) and occurs in In0.53Ga0.47As due to the unique circumstances of its low phonon energy and small effective mass, which results in these modes residing within the Landau damping region of In0.53Ga0.47As.10,12 Figure 4 shows that, for both InP and AlAs-like longitudinal optic modes at 347 and 367 cm 1, respectively, are observed for all six samples suggesting that the incident laser is probing well pass the In0.53Ga0.47As and In0.52Al0.48As epilayers through to the InP substrate. For 2-S and 3-S, the slight dampening of the InP longitudinal optic mode upon sulfur monolayer deposition is likely a result of the reduced penetration depth resulting from the increased carrier density.
CONCLUSIONS In summary, we have shown that Raman spectroscopy serves as a sensitive probe of free charge density in sequentially doped InxGa1 xAs epilayers incorporating both Si and S dopants. This method of determining dopant activation and free carrier density is observed to be agnostic of the nature of the individual dopants or the mode of dopant incorporation. Sulfur monolayer deposition is observed to be a viable means to further increase the electron density of doped InxGa1 xAs samples without damage to the crystalline lattice. The HFCM Raman band is observed to be a sensitive probe of the free electron density and concentration of activated dopants in InxGa1 xAs. The relative intensities of the GaAs-like feature and the IFCM mode in the optical
242
Volume 69, Number 2, 2015
1. R. Stevenson. ‘‘Compound Semiconductors Join the Race to Sustain Moore’s Law’’. IEEE Spectrum. December 2013. http:// spectrum.ieee.org/semiconductors/materials/compoundsemiconductors-join-the-race-to-sustain-moores-law [accessed Dec 1 2014]. 2. C. Hock-Chun Chin, G. Xiao, W. Lanxiang, L. Hock-Koon, S. Luping, Y. Yee-Chia. ‘‘III–V Multiple-Gate Field-Effect Transistors With HighMobility In0.7Ga0.3As Channel and Epi-Controlled RetrogradeDoped Fin’’. IEEE Electron Device Lett. 2011. 32(2): 146-148. 3. J.H. Yum, H.S. Shin, R. Hill, J. Oh, H.D. Lee, R.M. Mushinski, T.W. Hudnall, C.W. Bielawski, S.K. Banerjee, W.Y. Loh, W.-E. Wang, P. Kirsch. ‘‘A Study of Capping Layers for Sulfur Monolayer Doping on III–V Junctions’’. Appl. Phys. Lett. 2012. 101(25): 253514-253513. 4. K.-W. Ang, J. Barnett, W.Y. Loh, J. Huang, B.G. Min, P.Y. Hung, I. Ok, J.H. Yum, G. Bersuker, M. Rodgers, V. Kaushik, S. Gausepohl, C. Hobbs, P.D. Kirsch, R. Jammy. ‘‘300 mm FinFET Results Utilizing Conformal, Damage Free, Ultra Shallow Junctions (Xj5nm) Formed with Molecular Monolayer Doping Technique’’. International Electron Devices Meeting, 1988. IEDM ’88. Technical Digest., International 01/2011; doi: 10.1109/IEDM.2011.6131679. 5. M.E. Greiner, J.F. Gibbons. ‘‘Diffusion and Electrical Properties of Silicon-Doped Gallium Arsenide’’. J. Appl. Phys. 1985. 57(12): 51815187. 6. R.C. Newman. ‘‘The Lattice Locations of Silicon Impurities in GaAs: Effects Due to Stoichiometry, the Fermi Energy, the Solubility Limit and DX Behaviour’’. Semicond. Sci. Technol. 1994. 9(10): 1749-1762. 7. A.A. Quivy, A.L. Sperandio, E.C.F. da Silva, J.R. Leite. ‘‘p-Type Doping of GaAs (001) Layers Grown by MBE Using Silicon as a Dopant’’. J. Cryst. Growth. 1999. 206(3): 171-176. 8. J.C. Ho, R. Yerushalmi, Z.A. Jacobson, Z. Fan, R.L. Alley, A. Javey. ‘‘Controlled Nanoscale Doping of Semiconductors via Molecular Monolayers’’. Nat. Mater. 2008. 7(1): 62-67. 9. J.E. Maslar, J.F. Dorsten, P.W. Bohn, S. Agarwala, I. Adesida, C. Caneau, R. Bhat. ‘‘Electron-Phonon Interactions in n-Type In0.53Ga0.47As and In0.52Al0.48As Studied by Inelastic Light Scattering’’. Phys. Rev. B: Condens. Matter Mater. Phys. 1994. 50(23): 1714317150. ˜ M. Hopkinson. ‘‘Raman 10. R. Cusco´, L. Artu´s, S. Herna´ndez, J. Iba´nez, Scattering by LO Phonon-Plasmon Coupled Modes in n-Type In0.53Ga0.47As’’. Phys. Rev. B: Condens. Matter Mater. Phys. 2001. 65(3): 035210. 11. S. Hernandez, R. Cusco, N. Blanco, G. Gonzalez-Diaz, L. Artus. ‘‘Study of the Electrical Activation of Si-Implanted InGaAs by Means of Raman Scattering’’. J. Appl. Phys. 2003. 93(5): 2659-2662. 12. K. Kort, P.Y. Hung, P.D. Lysaght, W.-Y. Loh, G. Bersuker, S. Banerjee. ‘‘Raman Spectroscopy Studies of Dopant Activation and Free Electron Density of In0.53Ga0.47As with Sulfur Monolayer Doping’’. Phys. Chem. Chem. Phys. 2014. 16(14): 6539-6543. 13. J.C. Ho, A.C. Ford, Y.-L. Chueh, P.W. Leu, O. Ergen, K. Takei, G. Smith, P. Majhi, J. Bennett, A. Javey. ‘‘Nanoscale Doping of InAs via Sulfur Monolayers’’. Appl. Phys. Lett. 2009. 95(7): 072108-072103. 14. S. Herna´ndez, N. Blanco, I. Ma´rtil, G. Gonza´lez-Dı´ az, R. Cusco´, L. Artu´s. ‘‘Evidence of Phosphorus Incorporation into InGaAs/InP Epilayers after Thermal Annealing’’. J. Appl. Phys. 2003. 93(11): 9019-9023. 15. C.S. Jung, H.S. Kim, G.B. Jung, K.J. Gong, Y.J. Cho, S.Y. Jang, C.H. Kim, C.-W. Lee, J. Park. ‘‘Composition and Phase Tuned InGaAs Alloy Nanowires’’. J. Phys. Chem. C. 2011. 115(16): 7843-7850. 16. J. Groenen, R. Carles, G. Landa, C. Guerret-Pie´court, C. Fontaine, M. Gendry. ‘‘Optical-Phonon Behavior in Ga1-xInxAs: The Role of Microscopic Strains and Ionic Plasmon Coupling’’. Phys. Rev. B. 1998. 58(16): 10452-10462. 17. S. Mamedov, A. Kisliuk, D. Quitmann. ‘‘Effect of Preparation Conditions on the Low Frequency Raman Spectrum of Glassy As2S3’’. J. Mater. Sci. 1998. 33(1): 41.
