Towards Si-based Light Sources Greg Sun University of Massachusetts Boston
UMass System Amherst, Boston, Lowell, Dartmouth Worcester (Medical school)
UMass Boston
UMass Boston Established in 1964 Only public university in the city of Boston Undergraduate Students: 12K Graduate Students: 4K Thomas M. Menino, Bachelor of Arts in Community Planning (1988), UMass Boston, Mayor of Boston and the longestserving mayor in the history of the city
Outline Motivations Semiconductors (direct and indirect bandgap) Semiconductor laser diodes Challenges for light emissions from Si The group-iv (Si, Ge, Sn) band-to-band approach The group-iv intersubband approach Prospects of Si based lasers
What is a semiconductor? When bands are completely filled, they do not conduct current, only partially filled bands are conductive.
Direct and Indirect Bandgaps Eg Eg k k Direct bandgap: GaAs, InP, GaN Good for photonics Indirect bandgap: Si, Ge, AlAs Good for electronics
Photon Emissions from Semiconductors Conduction band Momentum Phonon emission Photon emission u l ħω In-plane k In-plane k Valence band 1) Light emitters require direct band gap 2) Wavelength is determined by the band gap
Band Structures of Real Semiconductors
Semiconductor Quantum Wells d Effective bandgap CB suband Bandgap VB suband GaAs: narrower bandgap AlGaAs: wider bandgap d: quantum well width ~ Electron de Broglie λ
Semiconductor Laser Diodes Wavelength in µm 0.405 0.525 0.635 0.760 0.785 1.064 1.310 1.654 1.877 3.330 GaInAlN AlGaInP GaAlAs InGaAlP GaInAlSb
Why Si Photonics?
Building Blocks for Si Photonics Building blocks: Modulators Waveguides Photodectors Light sources
Advantages of Si Photonics Monolithic integration of photonic device with Si electronics to create the opto-electronics super chip (OEIC) Mature infrastructure of the Si microelectronics for making highly sophisticated Si photonic devices Availability of lowest cost, largest size Si wafers Large contrast in refractive index between Si (3.45) and SiO 2 (1.45) ideal for waveguides basic building block of integrated photonicss
Si OEIC Chip R. A. Soref, Proceedings of IEEE, vol.81, 1687 (1993)
3D Opto-electronic Integration N. Sherwood-Droz and M Lipson, Optics Express, 19(18), 2011
The Missing Piece - Si Light Source Among the many photonic devices, the most difficult challenge is the lack of efficient light sources.
Band Structure of Si
Limitations of Si Momentum Conduction band Phonon emission Photon emission Si has not been the material of choice for efficient light sources because of its indirect bandgap. The emission of a photon requires the participation of a phonon second-order effect. Valence band
Si Laser Efforts Si nanocrystals formed in Si-rich SiOx Optically pumped gain observed (Univ. of Trento, 2003, Univ. of Rochester, 2005) Observations highly dependent on sample preparation Er-doped Si at 1.55µm LEDs with efficiency on par with commercial GaAs LEDs Si is not a good host of Er Optically pumped bulk Ge laser (MIT, 2009) Very heavily doped n-type Ge under tensile strain Optically pumped Si Raman laser (UCLA, Intel, 2003) High pump power (100MW/cm2) and large device size (1 cm) unlikely to be integrated with Si ICs Hybrid of III-V lasers on Si wafer InGaAs QD laser grown on Si (Univ. Michigan, 2005) InP laser bounded on Si (Intel, UCSB, 2006)
Nature, 2000 Observations highly dependent on sample preparation
Appl. Phys. Lett. 64, 2842 (1994)
Opt. Express, 2004 Pump power ~100MW/cm 2 Device size ~1 cm Unlikely to be integrated with Si IC.
The Ge Laser 135 mev Small difference between direct and indirect Eg Possible transition from indirect to direct Eg using a) Tensile strain b) heavy n-type doping
PL and EL from Ge on Si Substrate X. Sun, J. Liu, L. C. Kimerling, J. Michel, Opt. Lett. 34, 1198, (2009)
Optically Pumped Ge Laser Opt. Lett. 2010 30kW/cm 2
Ge Laser Diode Opt. Express, 2012 Threshold current density ~300kA/cm 2
III-V Hybrid Laser on Si Opt. Express, 2006 Active AlGaInAs lasers are attached to Si waveguide via wafer bonding This is not a monolithic process
Group-IV Materials Tin: 1997 at CalTech
The Group-IV (Si, Ge, Sn) Combination The lattice constants spanning over a broad range Bandgap tunable Direct bandgap?
Band Structures of Ge and α-sn Ge α-sn Ge band structure Γ-bandgap 0.798-0.413 L-bandgap 0.664 0.092 α-sn band structure S. Bloom, Solid State Communications, 32 6, 465-467, (1968).
Schematic of Ge1-xSnx band diagram Energy Ge without Sn Ge with 2% Sn ћν Ge with?% Sn Direct transition! Γ L k 1.Ge1-xSnx can become direct-bandgap material-------optical emitter 2.The direct bandgap becomes smaller------- Mid IR detector 33
GeSn: Indirect-to-direct Transition 0% Sn (indirect) L Γ k 3% Sn (quasi-direct) L Γ k D. W. Jenkins and J.D. Dow, Phys. Rev. B 36, 7994 (1987). R. A. Soref and L. Friedman, Superlatt. Microstructur. 14, 189 (1993). L Γ k >6% Sn (direct)
PL and EL Results from Kouvetakis Group (2.2% Sn) J. Mathews, et al, Appl. Phys. Lett. 97, 221912 (2010) R. Roucka, et al, Appl. Phys. Lett. 98, 061109 (2011)
GeSn Laser Design Bandgap Si Ge Lattice constant Sn Added degree of freedom for band gap and strain engineering GeSiSn works as lattice matched confining layer
GeSiSn Double Heterostructure Laser G. Sun, R. A. Soref, and H. H. Cheng, J. Appl. Phys. Vol. 108, 033107 (2010)
Auger Recombination Nonradiative process significant for narrow bandgap materials eeh process eeh+phonon process
DH Carrier Lifetimes
Optical Gain
Threshold and Temperature The DH laser will not lase at room temperature. G. Sun, R. A. Soref, and H. H. Cheng, J. Appl. Phys. Vol. 108, 033107 (2010)
GeSn/GeSiSn QW Design
Multiple QW Structure
QW Carrier Lifetimes
Modal Gain λ=2.3μm Room temperature operation is feasible G. Sun, R. A. Soref, and H. H. Cheng, Opt. Express, Vol. 18, 19957 (2010)
GeSn Material Challenges Very low solubility of Sn in Ge and Si (<1%) Hard to prevent Sn Segregation and precipitations Very large lattice mismatch a Si =5.431Å; a Ge =5.657Å; a Sn =6.489Å Low temperature growth
Experimental Effort NTU-UMass Boston Collaboration Low temperature MBE growth LED with highest Sn content: Ge 0.922 Sn 0.078
Fabrication of p-i-n diode Etch Clean First Second Open Third Deposit down custom metal the SiO2 sample window n-layer mask Si cap P-type doped Ge Ge spacer Ni GeSn layer Ge spacer N-type doped Ge Ge buffer Si buffer Silicon substrate PR PR PR SiO2 PR SiO2 48
49
Next Step for GeSn Laser Diodes Mid-IR double heterostructure GeSn laser on Si substrate (low temperature) Mid-IR quantum well GeSn laser on Si substrate (room temperature) 50
The Group-IV Intersubband Approach Quantum cascade lasers Circumventing the band indirectness completely Mid-IR to THz range With advantages for being nonpolar material system Facing material challenges 51
Band-to-band Transitions in QWs ΔE c u In-plane k Subband ħω ħω ΔE v l In-plane k 1) Light emitters require direct band gap 2) Wavelength is determined by the band gap
Intersubband Transitions in QWs E 2 ΔE c ħω ħω E 1 In-plane k 1) Intersubband transition is always direct in k-space 2) Wavelength is determined by the subband separation, no longer by the bandgap.
Brief History First semiconductor laser demonstrated in 1962 Original idea of IST laser proposed in 1971 by Kazarinov and Suris First QCL based on IST in 1994 (Bell Labs) Why takes so long?
Band-to-band vs. Intersubband i u ħω ħω l g Lasers based on ISTs are far more complex than conventional lasers.
Band Engineering E 2 E 1 Separated QWs Coupled QWs QW width yields the subband level and separation Barrier width and height determine subband overlap
Optically Pumped IST Laser Barrier/QW active region Pumping light (TM polarized) n-type cladding layer n-type cladding buffer Lasing emission (TM polarized) Substrate
Optically Pumped IST Lasing Process p L p L Periodic coupled MQW structure (n-doped) Optical Pumping: 1 3; Lasing: 3 2 Depopulation: 2 1; Population inversion: 32 > 2
Electron Scattering Phonon absorption Subband 2 Subband 1 Intrasubband process Phonon emission Electron-phonon scattering 22 11 electron- Electron scattering Electron-electron scattering
Optical Phonon Scattering γ 21 = 2π ħ 1 H ep 2 2 1 H ep 2 2 δ E 2 E 1 ħω p dn G 12 Q 2 n p, Absorption G 12 Q 2 n p + 1, Emission absorption 2 1 Emission n p = 1 exp ħω p k B T 1 γ 21 [1/ps] ( 0. 1 for GeSn) 1 2 3 4 E 2 E 1 /ħω p
Optical Phonon Emission u l u l g g Q ħω p ħω p ħω p IST above optical phonon Q IST below optical phonon
Population Analysis (Current Injection) N 3 t = ηj e N 3 N 3 τ 3 N 2 t = N 3 N 3 τ 32 N 2 N 2 τ 2 Population inversion: N 3 N 3 = τ 3 1 τ 2 τ 32 ηj e N 2 N 3 τ 32 > τ 2 Thermal backfilling
Current Injection IST Laser Metal contact n-type cladding contact layer V Barrier/QW active Region Metal contact n-type Cladding buffer Lasing n-type Substrate
Quantum Cascade Laser
Features and Advantages Unipolar (n-i-n or p-i-p ) device Fast modulation time Design freedom Lasing wavelength tunable by design: Mid to far-ir to THz range Bandgap directness irrelevant - favorable for Si-based lasers
History and Status 1992, first design of intersubband laser by G. Sun and J. B. Khurgin 1994, first QCL demonstration by Bell Labs Lasers based on III-V (GaAs, InGaAs, AlGaAs, InAlAs on GaAs or InP substrate) material systems Lasing wavelengths from 3μm to ~200μm CW, room temperature operation on selected wavelengths in InP based QCLs THz lasers bridging the gap between far-infrared coherent light source and GHz microwave electronics (100GHz ~ 10THz) operating below 200K
IST Bridging the THz gap For a long time there were no adequate sources operating in this range. Technology Review (MIT) 2004: THz selected as one of 10 emerging technologies that will change your world
THz security imaging Military IR countermeasures
THz medical imaging Clinical image THz image THz molecular sensing
Existing THz Sources Free electron lasers Molecular gas lasers THz Quantum Cascade Lasers
Circumventing the Indirect Bandgap E L-valleys in Ge or Ge-rich SiGe E Δ-valleys in Si or Si-rich SiGe CB Γ Δ or L-valley K Indirect K VB Band indirectness becomes irrelevant!
Si-based Intersubband Efforts Intersubband approach proposed in SiGe/Si QWs (G. Sun et al, 1995) Intersubband EL demonstrated in SiGe/Si Quantum cascade structures: G. Dehlinger et al (Switzerland, 2000) I. Bormann et al (Germany, 2002) S. A. Lynch et al (England, 2002) One scheme in common Intersubband transitions in valence band
SiGe Intersubband Devices Intersubband transitions within the valence band of GeSi/Si quantum Wells Metal Conduction Band p-sige cladding SiGe/Si QW active region p-sige buffer p-type Si substrate Lasing V Lasing Si barrier Eg Valence Band GeSi QW Si barrier SiGe quantum wells with Si barriers Small offset in conduction band Large offset in valence band p-i-p structure Electro-luminescence demonstrated but no lasing so far G. Sun, L. Friedman, and R. A. Soref, APL, vol.66, 3425 (1995)
Challenges and Opportunities Challenges: Unparallel valence subband dispersion due to light hole heavy hole coupling Difficulty in growing multiple QWs of large thickness due to large lattice mismatch (4%) between Si and Ge Low carrier transport and small oscillator strength due to large hole effective mass Opportunities: Lower threshold because of longer subband lifetimes due to weaker scattering of nonpolar optical phonons Strain and band offset engineering by incorporating yet another group-iv element Sn into the system
GeSiSn/Ge Strain Free QCLs Lattice matched structure no limitations on number of layers and thickness Large L-valley offset conduction band intersubband lasers
Ge/GeSiSn Conduction Band Offset Energy Calculation based on Jaros Band offset theory [Phys. Rev. B 37, 7112 (1988)] Conduction Γ, X, and L valleys for lattice matched GeSiSn alloy with Ge L-valleys for both GeSiSn and Ge below all others for Sn < 5% ΔE L =150meV between Ge well and Ge 0.76 Si 0.19 Sn 0.05 barrier
Ge/GeSiSn QCL on Si Substrate Growth of relaxed Ge layers directly on Si substrate Ge layers in 40nm -1µm thickness Edge dislocations formed at interface Low threading dislocation density < 10 5 /cm 2 Strain Free Ge/GeSiSn QCLs can be lattice matched to relaxed Ge buffer layer on Si substrate
L-Valley Ge/GeSiSn QCL L-valley CB offset: 150meV; Effective mass: 0.12m o ; Electric field: 10kV/cm Active region: 3 states with 3 QWs; lasing transition: 3 2 (49µm) Injector region: miniband formed with 4 QWs; upper state 3 in miniband gap; strong overlap between state 2 and miniband.
Lifetimes Carrier scattering due to nonpolar Optical and acoustic phonons Population inversion: 32 2
Population Dynamics and Optical Gain N i )] ( ) (1 [ 2 3 ) ( 2 3 _ 2 _ 32 2 3 2 2 N N e J L n c z e g p o L 2 _ 2 2 32 _ 3 3 2 3 _ 3 3 3 N N N N t N N N e J t N Population rate equation at threshold Current injection 3 2 Carrier scattering Depopulation Optical gain Dipole matrix overlap Lifetime requirement for inversion Thermal population
Plasmon Waveguide Conventional dielectric waveguides unsuitable for far-ir QCLs (d<<λ) Plasmon waveguides with metal cladding layers effective for mode confinement Only TM-polarized modes allowed Waveguide loss associated with metal absorption
Plasmon Waveguide Mode TM mode H y q( x ae bcosh( q( xd ae d /2) kx) e /2) e e j( zt ) j( zt ) j( zt ),,, x d / 2 x d / 2 x d / 2 M cosh( kd / 2) E ˆ ˆ o j q e e x d E E j kx xˆ k kx zˆ e x d cosh( kd / 2) E ˆ ˆ o j q e e x d q( xd /2) j( zt) ( x z), / 2 j( zt ) o[ cosh( ) sinh( ) ], / 2 M q( xd /2) j( zt ) ( x z), / 2 2 P 1 ' j" i 2
Plasmon Waveguide Confinement Electric field Profile: d /2 d /2 E E 2 2 dx dx Optical confinement Γ 1.0 Waveguide loss: α W =110/cm For d=2.1µm (40 periods of QCL)
QCL Threshold Current Optical Gain: g 2e 2 ( L ) 3 z cnl o p 2 2 2 [ (1 3 32 J ) e _ ( N 2 _ N 3 )] 1.0, m 10 / 110 / w cm cm 3 z 2 3. 3nm Confinement factor Waveguide loss FWHM=10meV gth w m Mirror loss In comparison with GaAs-based QCLs, this represents a reduction in threshold current density!
SiGeSn Works from Other Groups Kasper s group, University of Stuttgart Harris group, Stanford University Kolodzey s group, University of Delaware Yu s group, University of Arkansas Imec and Catholic University of Leuven, Belgium Nagoya University, Japan
Summary Group-IV system (Si, Ge, Sn and their alloys) is being investigated for photonic devices on Si substrates Strain and bandgap engineering are being explored. Direct bandgap GeSn is within reach. Group-IV LEDs have been reported as well as IR detectors. Progress is being made, but there remain significant material challenges. GeSn laser diodes are expected in near future. GeSiSn QCL is more futuristic.