CHAPTER 3. Experimental Setup
3.4 Probe station and spectrum measurement system
Probe station system was essential instrument for basic device characteristics measurement such as I-L (current versus light output), I-V (current versus voltage).
Scheme of probe station system, as illustrated in Figure 3-6, include probe station, current source, and power-meter module. Keithley 238 supplies continuous current for diode laser and receives relative voltage synchronously. The laser was temperature controlled by a precise thermoelectric controller and the output power was measured by Newport power-meter module (model 1835C). An integration sphere was used to pick up whole emitting power from VCSEL to improve the accuracy of power measurement. The integration sphere was calibrated with 850nm, 1300nm, and 1550nm and all the measured values can be traced to NIST standard.
The VCSEL device was placed on a platform of the probe station and was injected bias current with a microprobe. Threshold condition, slope efficiency, turn-on voltage and differential resistance can be obtained from L-I-V information by sweeping bias current. Emission spectrum was measured by optical spectrum analyzer (OSA, Advantest 8381). A multi-mode fiber probe was placed close to the emission aperture to take optical spectra. The OSA had spectrum resolution of 0.1nm which was adequate to measure VCSEL lasing spectra. Scheme of spectrum measurement system was combined with probe station as Figure 3-6.
CCD
Objective Keithley 238 driver
Power supply Power meter
Spectrum analyzer
Computer
Integrating sphere PD
Prober
Fig. 3-6 Probe station measurement instrument setup 3-5 Far-field and near-field
Near-field pattern (NFP) of VCSELs was magnified by high resolution optics
and projected into a B/W CCD then digitized by computer. Beam-view analyzer is useful software we used for the image process of the NFP. We could obtain NFP image under threshold, as spontaneous emission, to define oxide aperture size.
Far-field angle measurement was achieved by scanning the detector cross the laser emission, as shown in Figure 3.7, and recording the intensity with angle simultaneously. A stereo microscope was installed to assistant the wafer-level far-field test.
probe
Scanning photodetector
VCSEL Divergence angle test
machine
probe
Scanning photodetector
VCSEL Divergence angle test
machine
Fig. 3-7 Photograph of far-field measurement setup
3-6 Microwave test system 3.6.1 Setup
The microwave test system was mainly consisted of network analyzer, optical platform and microwave probe, as illustrated with Figure 3-8. Agilent 8720ES network analyzer was a crucial instrument of this microwave measurement.
Transmitter of network analyzer produced -10dBm RF signal. Laser diode drivers (Newport, model 525) provided direct bias current to set the laser above the threshold.
Bias-Tee combined AC and DC signal transmission through the same coaxial cable.
The mixed signal was transmitted through the coplanar microwave probe and was injected into VCSEL, which was bonded on RF submount, as Figure 3-9. The coplanar microwave probe had 700µm pitch (ground (G) to signal (S) tip spacing) and suitable measurement frequency range was up to 40 GHz. Signal probe was higher than ground tips. Fine probing was observed by signal probe skating on contact substrate. Illustration of microwave probe holder in whole was shown in Figure 3-10.
Optical platform contained microscope, beam splitter, objective and fiber coupler etc. We used a long work distance objective (20X, Mitutoyo), fixed in a triple-divide translation stage, to pick up the laser output from VCSEL. Light was separated by
beam splitter and received by CCD and another 10X objective lens(Olympus). One of splitting light was received by a simplified microscope, which was constructed by beam splitter and a CCD to make probing easily. Another light path trough 10X objective coupling into multimode fiber by five-axis fiber aligner (Newport).
The collected light was transmitted into 12 GHz photodetector (Model 1580, NewFocus) and was conversed into electrical signal and fed to network analyzer.
Comparing two channels microwave signal by network analyzer, information of transmission and reflection characteristics could be expressed as vector (magnitude and phase), scalar (magnitude only), or phase-only quantities, that was, S-parameter.
Bias-Tee
VCSEL (Device Under Test) substrate
VCSEL (Device Under Test) substrate
bonding wire
VCSEL (Device Under Test) substrate
Pattern generator
Fig. 3-8 Microwave test system setup
OM vertical view image VCSEL Bonding wire
AlN
1521µm
Au
1050µm 279µm 456µm
Fig. 3-9 Sub-mount substrate design
Photodetector
Laser driver Network analyzer
Optical platform
Microwave probe
Fig. 3-10 Illustration of microwave test system
3.7 Relative Intensity Noise measurement
The experimental setup of RIN is shown in Figure 3-11. The laser was driven by a low noise current source and was temperature controlled by a precise thermoelectric controller. The laser output was collimated and focused by two objective lenses. A beam splitter guided ~ 5% of the laser light into an image-formation lens then the image was recorded by a CCD camera. A multimode fiber was coated with Al thin film on one facet with reflectivity of 10%. The fiber was held by a 5-axises precision fiber positioner and the coated facet was taken as the reflecting mirror. The reflected beam was collimated and re-focused back to the emission aperture of the VCSEL. The tilt of fiber- facet was carefully aligned and formed an external cavity between the top DBR of VCSEL and the fiber-facet. Size of reflected spot can be controlled by the Z-axis of the positioner while the overlapping between the reflected spot and emission aperture was controlled by tilt of the positioner. The external cavity length was about 28 cm. Mirror M1 guide the laser to a fast photodetector (PD, Newfocus 1601) and its output is coupled to a 0.5 GHz oscilloscope (Agilent Infinium) to exam the stability of laser. The output of fiber was connected to a 12 GHz photodetector (NewFocus 1580-A) and its output was coupled
to a RF spectrum analyzer (HP 8563E).
Fig. 3-11 Relative Intensity Noise test system setup
The main problem with the measured noise in the RF spectrum analyzer is that it also represents the shot noise power caused by the quantum nature of the photodetection and the receiver’s total measured thermal noise power. Hence in order
to quantify the shot noise power and the receiver thermal noise power we use, the experimental procedure as described in following.
In order to measure the shot noise power and the receiver thermal noise power, we turn off the transmitter and just have the receiver connected to the RF Spectrum Analyzer. By comparing the noise power level between the ‘Transmitter On &
Receiver On’ and ‘Transmitter Off & Receiver On’ cases, we can find the receiver thermal noise to be ~ -66 dBm. The system relative intensity noise has contributions from the RIN laser, photodiode shot noise power and thermal noise power. Hence
)
The is Pavg,elec found by modulating the transmitter with a 10dBm 200Mhz signal.
The average signal power found at the modulating frequency gives the Pavg,elec. For example, Pavg,elec = -0.587 dBm
Also the Idc current is generated from this Pavg,elec. Hence Idc2
R = Pavg,elec where R
= 50 ohms, resolution bandwidth = 1 MHz.
Hence RIN(Measured) = -66 – 10*log(1000000) – (-0.587) = -125.4 dB/Hz
CHAPTER 4 Strain-Compensated InGaAsP/InGaP MQWs VCSELs
The use of an Al-free InGaAsP based active region is an attractive alternative to the conventional (Al)GaAs active region for IR VCSELs. While edge emitting diode lasers with Al-free active regions have demonstrated performance and reliability surpassing AlGaAs-active devices [8-9]. In addition, theoretical calculations have predicted a lower transparency current density, a high differential gain and better temperature performance in InGaAsP-strain active VCSELs in respect to lattice-matched GaAs quantum-well active devices [10]. These parameters are all very important in high speed and high temperature VCSEL design because the relaxation resonance frequency of the laser depends on the square root of the differential gain as well as the difference of operation current and threshold current [4]. The use of tensile-strained barriers like In0.4Ga0.6P can provide strain compensation and reduce active region carrier leakage. Al-free materials are significantly less reactive to oxide level, compared to AlGaAs materials make them ideal for the reliable manufacture process [8].
Proton implanted VCSELs using compressive strain In0.18Ga0.82As0.8P0.2 active-region has been demonstrated good performance [11]. This chapter present the fabrication and characteristics of high performance 850nm InGaAsP/InGaP SC-MQWs VCSELs. High speed modulation up to 12.5 Gbit/s was demonstrated from 25oC to 85oC utilizing In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P strain-compensated MQWs (SC-MQWs) VCSELs [12]. The DC, small signal, and large signal measurements were performed on SC-MQWs VCSELs.
Finally, the initial reliability data from these VCSELs are presented.
4.1 Theory
To compare the InGaAsP/InGaP SC-MQWs and GaAs/AlGaAs QWs, the gain spectrum and material gain as a function of carrier density were calculated based on the k‧p theory with valence banding mixing effect. An 8×8 Lutttinger-Kohn Hamiltonian matrix is used in this investigation. The method is followed by Chuang [13] and the gain spectral is broadened by Lorentzian's function. The well width and MQWs compositions were both set as 8 nm and In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P to tune the emission wavelength to 842 nm. A scheme of the conduction band diagram was shown in Fig. 4-1. The obtained gain spectrum and material gain were compared with a conventional GaAs/Al0.26Ga0.74As MQWs using same algorithm.
The conduction band offset of In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P and GaAs/Al0.26Ga0.74As are
assumed to be 0.5 and 0.65 respectively. The calculation of bandgap energy formula is adapted from Varshni equation with α = 5.5×10−4 eV/K, β = 225 Κ for both InGaAsP and AlGaAs materials and the bandgap is adapted the value at 298 K. Although the Varshni parameters might not be the same for InGaAsP and AlGaAs, the error incurred is likely tolerable because the small bandgap deviation should not affect the material gain. As shown in Fig. 4-2(a), the material gain of In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P is approximately triple higher than that of GaAs/Al0.26Ga0.74As structure when the input carrier concentration is 2×1018 cm-3. The material gains of In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P and GaAs/Al0.26Ga0.74As as a function of the input carrier concentration are depicted in Fig. 4-2(b). The transparent carrier concentrations of In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P and GaAs/Al0.26Ga0.74As are 1.5×1018 cm-3 and 1.78×1018 cm-3 respectively. Furthermore, In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P provides higher gain than GaAs/Al0.26Ga0.74As in the entire calculation range. The results obtained numerically suggest that the high material gain, low transparency carrier concentration make the InGaAsP SC MQWs a better active layer. In addition, its superior performance has been confirmed in edge emitting lasers. [8-9]
4.2 Fabrication Process of polyimide planarized VCSEL
The schematic structure of a fabricated top emitting VCSEL is shown in Fig. 4-3(a), which has been grown by low pressure metal organic chemical vapor deposition (MOCVD) on a semi-insulating (100) GaAs substrate. The group-V precursors are the hydride sources AsH3 and PH3. The trimethyl alkyls of gallium (Ga), aluminum (Al) and indium (In) are the group-III precursors. The dopant sources are Si2H6 and CBr4 for the n and p dopants, respectively. The bottom n-type distributed Bragg reflector (DBR) consists of 35-period-Al0.15Ga0.85As/Al0.9Ga0.1As. The top p-type DBR consists of 23 pairs of Al0.15Ga0.85As/Al0.9Ga0.1As. The active layer consists of three In0.18Ga0.82As0.8P0.2/ In0.4Ga0.6P (80Å/100Å) SC-MQWs surrounded by Al0.6Ga0.4As cladding layer to 1λ-cavity. A 30nm thick Al0.98Ga0.02As was introduced on the upper cavity spacer layer to form an oxide confinement.
Finally, 1λ thickness of current spreading layer and heavily doped GaAs (p> 2x1019 cm-3) contacting layer was grown. The n-type DBR was grown at 750oC. The quantum well region and p-type DBR were grown at 650oC. Growth interruptions of 5s, 10s, or 15s were introduced before and after In0.18Ga0.82As0.8P0.2 QW growth. Fig. 4-4 shows the comparison of
photoluminescence spectra of In0.18Ga0.82As0.8P0.2/In0.4Ga0.6P with different growth interruption times. The 5s growth interruption is not enough to evacuate residual As in the growth reactor, resulting in the carry-over of As into the In0.4Ga0.6P barrier. The 15s growth interruption is so long that some impurities can be gettered at the interface or indium segregation after strained layer growth, resulting in the degradation of luminescence. The 10s growth interruption seems to give the best luminescence quality. The composition of SC-MQWs is characterized by high-resolution x-ray diffraction. The gain peak position = 835 nm was determined by photoluminescence while the FP-dip resonant wavelength = 842 nm was determined by reflectance measurement. The VCSELs were fabricated utilizing the processing described by Peters et al. to minimize capacitance while keeping reasonably low resistance [3]. The processing sequence included six photomasks to fabricate polyimide-planarized VCSELs with coplanar wave-guide probe pads, as shown in Fig. 4-3(b).
Device fabrication began with the formation of cylindrical mesas of 20 µm in diameter by etching the surrounding semiconductor to the bottom n-type mirror to a depth of 5 µm using a Reactive Ion Etching (RIE) system. The sample was wet-oxidized in a 420 oC steam environment for ~12 min to form the current aperture and provide lateral index guiding to the lasing mode. The oxidation rate was 0.6 µm/ min for the Al0.98Ga0.02As layer, so the oxide extended 7.5 µm from the mesa sidewall. The VCSELs therefore have a 5 µmin diameter emitting aperture defined by lateral oxidation. A 40 µm circular mesa were formed after oxidation using wet chemical etching (H2O:H2SO4:H2O2 = 8:1:8) down to n-buffer layer.
Following Si3N4 was deposited for passivation. Ti/Au was evaporated for the p-type contact ring, and AuGeNi/Au was evaporated onto the etched n–buffer layer to form the n-type contact which is connected to the semi-insulating substrate. Contacts were alloyed for 30 sec at 420 oC using RTA. After contact formation, photosensitive polyimide was spun on the sample for field insulation and planarization. Ti/Au with thicknesses of 200/3000Å were deposited for metal interconnects and coplanar waveguide probe-bond pads. Heat treatment after the metal deposition was utilized to improve metal-to-polyimide adhesion strength. Fig.
4-5 shows the SEM photo of a finished VCSEL.
4.3 Result and discussion
Fig. 4-6 shows the typical light output and voltage versus current (LIV) curves of the SC-MQWs InGaAsP/InGaP VCSEL at room temperature and 85oC under CW opteration.
These VCSELs exhibit kink-free current-light output performance with threshold currents
~0.4 mA, and slope efficiencies ~ 0.6 mW/mA. The threshold current change with temperature is less than 0.2 mA and the slope efficiency drops by less than ~30% when the substrate temperature is raised from room temperature to 85oC. This is superior to the properties of GaAs/AlGaAs VCSELs with similar size [14]. The resistance of our VCSELs is
~95 Ohm and capacitance is ~0.1 pF. As a result, the devices are limited by the parasitics to a frequency response of approximately 15 GHz. The lateral mode characteristics is an important feature since it strongly affects the transmission properties. Fig. 4-7 shows the emission spectrum of a VCSEL at an operating current of 6 mA. This spectrum was recorded with an optical spectrum analyzer (Advantest 8381A) with spectral resolution of 0.1 nm. Two dominant modes were observed at 844.2 nm and 843.7 nm. The root-mean-squared (RMS) spectral linewidths at 2, 6, 8 mA are 0.15, 0.37, and 0.4 nm respectively, which can fulfill the requirement (<= 0.45 nm) of 10Gbps data transmission.[15]
The small signal response of VCSELs as a function of bias current was measured using a calibrated vector network analyzer (Agilent 8720ES) with on-wafer probing and a 50 µm multimode optical fiber connected to a New Focus 25GHz photodetector. Fig. 4-8 shows the measured (dashed lines) and fitted (solid lines) small-signal frequency response of a 5 µm VCSEL at different bias current levels. The modulation frequency is increased with increasing bias current until flattening at a bias of approximately 5mA. With only 3mA (5mA) of bias current, the maximum 3dB modulation frequency response is measured to be ~13 (14.5) GHz at 25 °C and is suitable for 12.5 Gb/s operation. The measured data were fit to a general 3-pole modulation transfer function [16, 17]
where f is modulation frequency, fp is parasitic roll off frequency, fr is resonant frequency and γ is damping rate.
Fig. 4-9 shows the 3dB modulation frequency as a function the square root of the difference in current above threshold. The relaxation resonance frequency is found saturate for driving currents above 3mA. By fitting the lowest current points in Fig. 4-8, we obtain a modulation current efficiency of 11.6 GHz/(mA)1/2[18]. This is higher than GaAs/AlGaAs VCSELs with similar size [6, 19] and is comparable with oxide confined VCSELs with InGaAs based quantum wells [7, 16]. Plotting the damping rate γ versus fr2 reveals a K-factor of 0.3 ns. Neglecting heating effects and external parasitics, the intrinsic bandwidth was found to be 29.6GHz using the relation fmax = . 2(2π/K)
To measure the high-speed VCSEL under large signal modulation, microwave and
lightwave probes were used in conjunction with a 12.5-Gb/s pattern generator and a 12-GHz photoreceiver. The eye diagrams were taken for back-to-back (BTB) transmission on SC-MQWs InGaAsP/InGaP VCSEL. As shown in Fig. 4-10(a), the room temperature eye diagram of our VCSEL biased at 4 mA with data up to 12.5 Gb/s and 6dB extinction ratio has a clear open eye pattern indicating good performance of the VCSELs. The rise time Tr is 28 ps and fall time Tf is 41 ps with jitter (p-p)=20 ps. The VCSELs also show superior performance at high temperature. Fig. 4-10 (b) demonstrated the high speed performance of our VCSELs (biased at 5mA) with reasonably open eye-diagrams at 12.5 Gb/s and 6dB extinction ratio at 85°C. This further confirms the superior performance of our VCSELs.
To guarantee the device reliability is always a tough work but a natural task for the components supplier in the data communication markets. We have accumulated life test data up to 1000 hours at 70oC/8mA with exceptional reliability. As shown in Fig. 4-11, the light output is plotted versus time scale for SC-MQW VCSEL chips under the high temperature operation lifetime (HTOL) test at 70oC/8mA. None of them shows the abnormal behavior.
References
[1] J. A. Tatum, A. Clark, J. K. Guenter, R. A. Hawthorne, and R. H. Johnson,
“Commercialization of Honeywell’s VCSEL technology,” Vertical-Cavity Surface-Emitting Lasers IV, K. D. Choquette and C. Lei, editors, Proceedings of the SPIE, vol. 3946, pp. 2-13, SPIE, Bellingham, WA, 2000..
[2] F. H. Peters and M. H. MacDougal, “High-Speed High-Temperature Operation of Vertical-Cavity Surface-Emitting Lasers” IEEE Photon. Technol. Lett., vol. 13, No 7, p.
645–647, July 2001.
[3] F. H. Peters, D. J. Welch, V. Jayaraman, M. H. MacDougal, J. D. Tagle, T. A. Goodwin, J.
E. Schramm, T. D. Lowes, S. P. Kilcoyne, K. R. Nary, J. S. Bergey, and W. Carpenter, “10 Gb/s VCSEL-based data links,” Photonics West, San Jose, CA, Tech. Rep. OE 3946–26, 2000.
[4] C.W. Wilmsen, H. Temkin, and L.A. Coldren, eds., Vertical-Cavity Surface-Emitting Lasers: Design, Fabrication, Characterization, and Applications, Cambridge University Press, 1999.
[5] K. L. Lear, M. Ochiai, V. M. Hietala, H. Q. Hou, B. E. Hammons, J. J. Banas, and J. A.
Nevers, “High-speed vertical cavity surface emitting lasers,” in Dig. IEEE/LEOS Summer Topical Meetings, 1997, pp. 53–54.
[6] J. A. Lehman, R. A. Morgan, M. K. Hibbs-Brenner, and D. Carlson, “High-frequency modulation characteristics of hybrid dielectric/AlGaAs mirror singlemode VCSELs,”
Electron. Lett., vol. 31, pp. 1251-1252, 1995.
[7] K. L. Lear, A. Mar, K. D. Choquette, S. P. Kilcoyne, R. P. Schneider, Jr., and K. M. Geib,
“High-frequency modulation of oxide confined vertical cavity surface emitting lasers,”
Electron. Lett., vol. 32, pp. 457–458, 1996.
[8] L. J. Mawst, S. Rusli, A. Al-Muhanna, and J. K. Wade, “Short-wavelength (0.7 µm < λ
<0.78 µm) high-power InGaAsP-active diode lasers,” IEEE J. Select. Topics Quantum Electron., vol. 5, pp. 785–791, (1999).
[9] N. Tansu, D. Zhou, and L. J. Mawst, “Low Temperature Sensitive, Compressively-Strained InGaAsP Active (λ=0.78-0.85 µm) Region Diode Lasers,” IEEE Photon. Technol. Lett., Vol.12(6), pp.603-605 , 2000.
[10] T. E. Sale, C. Amamo,Y. Ohiso, and T.Kurokawa, “Using strained lasers (AlGa) In As P system materials to improve the performance of 850 nm surface- and edge-emitting lasers,” Appl. Phys. Lett., vol. 71, p. 1002–1004, (1997).
[11] N. Tansu, L.J Mawst, “Compressively-strained InGaAsP-active (λ=0.85µm) VCSELs”
IEEE Lasers and Electro-Optics Society 2000 Annual Meeting. LEOS 2000. vol. 2, p. 724 -725 (2000).
[12] H. C. Kuo, Y. S. Chang, F. I. Lai, T. H. Hsueh, “High speed modulation of 850 nm InGaAsP/InGaP strain-compensated VCSELs. Electon. Lett. Vol 39, p. 1051-10523 (2003)
[13] S. L. Chuang, "Efficient band-structure calculation of strained quantum-wells", Phys.
Rev. B, 43, 9649-9661 (1991).
[14] David J. Bossert, Doug Collins, Ian Aeby, J. Bridget Clevenger, “Production of
[14] David J. Bossert, Doug Collins, Ian Aeby, J. Bridget Clevenger, “Production of