Chapter 3 VCSEL fabrication and measurement setup
3.6 Experimental guidelines
Special attention must be paid to the modulation depth of the single applied to the lasers. High modulation amplitude increases the signal to noise ratio but can introduce distortions of the response dues to nonlinerities and cause an artificial compression of the resonance frequency peak, especially with structures operating at very low bias currents such as oxidized VCSEL with sub-milliampere threshold.
For accurate parameter extraction it is also important to have a large signal-to-noise ratio. This can be achieved by using large temporal averaging factors during measurements. Averaging in the frequency domain, so-called smoothing is not recommended; it will reduce the frequency resolution of the signal and often corrupt the phase information.
A good rule is to measure the modulation response at 5 – 10 bias current until a maximum current is reached where the resonance frequency peak saturates or starts
At high frequency the signal to noise ratio is often poor. This noise in the high frequency region is in general not only caused by ordinary amplifier noise but also noise due to calibration problems of the measurement set-up. Even slight movement of the measurement cables will change the phase of distant reflections, which can give a noisy appearance of the measurement curve. Crosstalk can give similar effects. This type of noise is not uncorrelated at different measurement points and is difficult to distinguish from the signal. Hence, it is preferable to omit the highest frequency parts of the measurements where the signal where the signal-to-noise ratio is very poor.
The VCSEL chip has to be mounted on a module for commercial applications.
The package has limitations at high frequencies and the measured bandwidth can be strongly decreased by the package parasitics. Resonance and filtering effects introduced by the inherent package parasitics can be significant sources of error. The inductive and capacitive elements associated with the physical geometry of the package will result in small parasitic LC circuits, which will in turn cause voltage or current peaking at the resonace frequencies and low pass or band rejection filtering effect at other frequencies.
Figure 3-1 Cross section structure of tapered oxide implant VCSEL structure
Al0.98Ga0.02As layer
MQW: GaAs/AlGaAs
Al0.12Ga0.88As/Al0.90Ga0.10As 39.5 pairs (99.8%)
GaAs substrate
Au/Ge/Ni/Au n-contact Ti/Pt/Au p-contact
Al0.12Ga0.88As/Al0.90Ga0.10As 22 pairs (99.6%)
Figure 3-2 Steps of oxide-implant VCSEL process
(1) VCSEL structure epitaxy Substrate
(2) Mesa etching – ring-trench
(3) Passivation process PR
(4) Contact coating – electrode
(5) Oxidation process H2O 420oC
dosage: 1013~1015, 200~420keV
(6) Proton bombing
Figure 3-3 Illustration of oxidation process system setup
Thermal coupler
Furnace
N2
Heater
Boat VCSEL
H2O / H2 Purge
Figure 3-4 Oxidation rate of 98% Al-content layer
Al0.98Ga0.02As layer MQW: GaAs/AlGaAs
Al0.12Ga0.88As/AlAs 30.5 pairs (99.8%)
n+ GaAs substrate Al0.12Ga0.88As/AlAs 22 pairs (99.6%)
Figure 3-5 (a) Cross section structure of oxide photonic crystal VCSEL structure (b) top view image of the PC-VCSEL;(c) scanning electron microscope image of an etched hole.
60µm
(a)
(b)
(c)
Figure 3-6 Steps of oxide photonic crystal VCSEL process
(1) VCSEL structure epitaxy Substrate
(2) Mesa etching – ring-trench
(3) Passivation process PR
(4) Contact coating – electrode
R
Reeaaccttiivvee IIoonn EEttcchhiinngg ((AArr ,, CCll22 ))
(6) Etching air hole (5) Oxidation process
H2O 420oC
Implant region
MQW: GaAs/AlGaAs
Al0.12Ga0.88As/AlAs 30.5 pairs (99.8%)
n+ GaAs substrate Al0.12Ga0.88As/AlAs 22 pairs (99.6%)
Figure 3-7 Cross section structure of implant photonic crystal VCSEL
Probe Station Keithley 238
CW Current Source
Beam view analyzer
Objective nikon 20x NA 0.45 Computer
Objective nikon 20x NA 0.45 Computer
CCD mirror
Objective nikon 20x NA 0.45 Computer
CCD mirror
Computer Computer
CCD mirror
Figure 3-9 Far field pattern measurement system
Figure 3-8 Probe station measurement instrument setup
CCD
Bias-Tee
Photo detector Laser driver
Monitor
Lens CCD
Objective
Objective Fiber coupler
Beam splitter
Microwave probe
bonding wire
VCSEL (Device Under Test) substrate
Pattern generator
Bias-Tee
Photo detector Laser driver
Monitor
Lens CCD
Objective
Objective Fiber coupler
Beam splitter
Microwave probe
bonding wire
VCSEL (Device Under Test) substrate
bonding wire
VCSEL (Device Under Test) substrate
Pattern generator
Figure 3-10 eye diagram measurement instrument setup
Chapter 4
Result and discussion
Oxidation process was regular process for VCSEL fabrication. Most high speed VCSEL was also made by the same method. In our study, we chose different type of oxide implant VCSELs to compare with their high speed characteristics in first section.
In second section, we was according to the static and high speed characteristics of these VCSELs to modeling the equivalent circuit. Equivalent circuit modeling was investigated in section 4-2. With VCSEL modeling, the equivalent components were extracted and observe the modulation limitation caused by structure. In third section, we chose different type of photonic crystal VCSELs to compare with their static characteristics of these VCSELs
4.1 Oxide-implant VCSELs DC and AC characteristics
Tapered oxide VCSEL and blunt oxide VCSEL had same process structure, shown in Figure 4-1. Process structure was described in Chapter 3. Both VCSEL had the same oxide aperture as 5.5 µm and mesa as 30 µm.
The static characteristics of both VCSEL were shown in Figure 4-2. The threshold current is ~ 1 mA (0.9 mA) for tapered (blunt) type VCSELs with the same slope efficiency of ~ 0.35 mW/mA. The similar static performance is not surprising since 1. the oxide apertures here have the same size of 5~6µm and 2. the blunt oxide don’t incur excess scattering except at very small oxide size (< 3µm ). The maximal output power exceeds 3 mW at room temperature and output power rollover occurs as the current increases above 12 mA.
The measured modulation response curves for tapered oxide VCSEL and blunt oxide VCSEL were shown in Figure 4-3(a) and 4-3(b), respectively. The maximum modulation bandwidth of both VCSEL was 13 and 9.5 GHz. At low bias currents, the bandwidth increased in proportion to the square root of the current above threshold, as expected from the rate equation analysis. For blunt oxide VCSEL, low frequency rollover was observed and the modulation response became gradually overdamped
as the bias current increasing. In contrast, no rollover was observed in low frequency for tapered oxide VCSELs and the 3-dB bandwidth reached a maximum value of 13 GHz before fully overdamped. The observations coincide with the simulation results for blunt oxide VCSEL in which a significant overdamping was reported in the relaxation oscillation due to the nonuniformity of the transverse mode and significantly reduced the modulation bandwidth. [11, 12] The bandwidth are similar at low bias current for both tapered oxide VCSELs and blunt oxide VCSELs and saturated at the bias current higher than 8 mA. Modulation current efficiency factor (MCEF), indicated in Figure 4-4, appeared modulation bandwidth as function of (I-Ith)1/2. Tapered oxide VCSEL had 6.5 GHz/mA1/2; As the Figure 4-5 shows, the peak height of modulation response was higher in the tapered oxide VCSEL than that of blunt oxide VCSEL. The higher modulation amplitude in tapered oxide VCSELs implies the lower damping rate which was known to approximately proportion to the ratio between resonant frequency and peak modulation amplitude [13]. The oxide layer of tapered oxide VCSEL had was theoretically investigated to reduce the nonlinear damping effect by making the electrical aperture smaller than the optical aperture, and thereby improves the modulation bandwidth [12].
As the Figure 4-3 shows, the peak height of modulation response was higher in the tapered oxide VCSEL than that of blunt oxide VCSEL. The higher modulation amplitude in tapered oxide VCSELs implies the lower damping rate which was known to approximately proportion to the ratio between resonant frequency and peak modulation amplitude [13]. In depth, the damping rate, resonant frequency, and parasitic roll-off frequency can be obtained by fitting the experimental data to a three-pole approximation of the modulation response equation [13]. The K factor can be found out from the slope of damping rate with the square of resonant frequency, as plotted in Figure 4-5. It was shown the damping rate indeed about two times higher than that in blunt oxide VCSELs and the K factors were 0.15 ns and 0.4 ns respectively. The comparison of the damping rate is also coincident with the theoretical prediction [11, 12]. If parasitic and other heating effects are ignored, the theoretical damping limited 3-dB bandwidth may be over 59 GHz for tapered oxide VCSEL and 22 GHz for blunt oxide VCSEL, from the ration of 2π(2)1/2/K. [13]
For commercial high speed performance testing, tapered oxide and blunt oxide VCSEL were measured as eye diagram shown in Figure 4-7. Eye diagram of blunt
oxide VCSEL shown that jitter was about 25 ps. Tapered VCSEL had small jitter below 20 ps. It still had rise-time/fall-time were 26 and 40 ps respectively. For that, oxide-implant VCSEL passed 10Gbps clarified.
4.2 Equivalent circuit modeling
As previously shown in Figure 4-3(a), the 3-dB bandwidth reached a maximum value of ~13 GHz before fully overdamped. We therefore investigate the extrinsic bandwidth limitation on the tapered oxide VCSELs in the next. An equivalent circuit for the VCSEL impedance is useful for analysis of electrical bandwidth limitations.
Inset of Figure 4-6 shows the equivalent circuit model used to extract the circuit components. The resistance Rm represents the mirror loss while the Ra accounts for active region resistance. Ca represents a combination of capacitance of active area and oxide layer. A shunt resistance Rp is also included to account for pad loss and the pad capacitance is represented by Cp. Using this equivalent circuit, we can also investigate the extrinsic limitations on the modulation speed and determine the influence of the parasitic capacitance and the mirror resistance on the modulation bandwidth. To extract the capacitance of VCSELs, the measured amplitude and the phase of S11 data were fitted from 100 MHz to 20 GHz. Fig. 5 illustrates measured and fitted S11 results of the tapered oxide VCSEL at 6mA. Convergence of the fitting values to physically reasonable values was obtained using the following procedure.
First, the Cp, Rp, Rm, Ca were extracted using zero bias S11 data (where Ra is very large and can be neglected). Second, the Ra and Ca values were extracted by fitting the S11 data for different bias currents. Finally, all the circuit parameters were allowed to vary about these values in order to minimize the squared error. The resulting Cp, Rp, and Rm were 160 fF, 5Ω and 51Ω respectively and the extracted Ra and Ca values for different bias currents were listed in Table 4-1. Based on these extracted values, the electrical bandwidth can be determined from -3dB of the S21 of the equivalent circuit shown in inset of Figure 4-6. In consequence, electrical bandwidth of ~ 12.5 GHz was obtained and showed weak dependence on bias current. The calculated electrical bandwidth is coincident with the maximal measured modulation bandwidth and confirms the parasitic effects as the main limitation on the tapered oxide VCSELs.
4.3 Photonic crystal VCSELs DC characteristics
Oxide VCSEL and oxide photonic crystal VCSEL had same process structure, shown in Figure 4-8. Process structure was described in Chapter 3. Both VCSEL had the same oxide aperture as 12 µm and mesa as 50 µm.
The static characteristics of both VCSEL were shown in Figure 4-9.The threshold current is ~ 0.9 mA (0.5 mA) for oxide (photonic crystal) type VCSELs . The slope efficiency of oxide (photonic crystal) VCSELs ~ 0.239 mW/mA (0.254 mW/mA). The series resistance of oxide ( oxide photonic crystal) VCSELs ~ 145 Ω (191 Ω). The maximal output power exceeds 2.8(1.61) mW at room temperature.
The photonic crystal VCSEL with Λ=5 mm, a/Λ=0.4 exhibits single fundamental mode lasing on the low current range of operation with the side-mode suppression ratio (SMSR) of over 20 dB as shown in Figure 4-10 (a). But as a result of the oxide layer was influence on the index guide of photonic crystal. Therefore the photonic crystal VCSEL becomes multi-mode lasing over the high current range of operation as shown in Figure 4-10 (b). The near field images of the device also shown in the inset of Figure 4-10. Very strong emission can be seen in the central site, where the circular emission pattern is evidence of the PC confined mode as shown in Figure 4-10 (a).
The process structure Implant photonic crystal VCSEL as shown in Figure 4-11.
Process structure was described in Chapter 3. The oxide aperture of Implant photonic crystal VCSEL as 10 µm and mesa as 60 µm
The static characteristics of implant photonic crystal VCSEL were shown in Figure 4-12.The threshold current is ~ 1.25 mA for implant photonic crystal VCSELs .The slope efficiency of implant photonic crystal VCSELs ~ 0.18 W/A.The series resistance of implant photonic crystal VCSELs ~ 1.5K Ω. The maximal output power exceeds 1.24 mW at room temperature.
Lasing spectra of the implant photonic crystal VCSEL is shown in Figure 4-13(a), confirming singlemode operation within overall operation current. The PC-VCSEL reveals a side mode suppression ratio(SMSR) > 40 dB throughout the current range.
For comparison, lasing spectra of a proton-implanted VCSEL without photonic crystal holes hows multiple mode operation as the driving current increased above 4.25 mA Figure 4-13(b). As the driving current increases, even higher-order transverse modes
We present the far-field intensity distribution of implant VCSEL in Figure 4-14(a)~(c), The far-filed image shows a doughnut-liked pattern at injection current of 3.8mA (i.e. 1.1 Ith), which is illustrated in Figure 4-14(a). At higher injection current of 4.9 mA(i.e. 1.3 Ith), some radial structures appear at the periphery of doughnut-liked pattern, which is illustrated in Figure 4-14(b)~(c). It can be seen that multiple transverse modes could be observed when the drive current equals 3.8mA (i.e. 1.1 Ith). On the other hand, the intensity of higher order mode and the beam divergence angle both increase as the drive current is increased to 4.9 mA(i.e. 1.3 Ith).
The far-field intensity distribution of implant photonic crystal VCSEL in Figure 4-14(d)~(f), The far-filed image shows a doughnut-liked pattern at injection current of 2.8mA (i.e. 1.1 Ith), which is illustrated in Figure 4-14(a). At higher injection current of 3.5 mA(i.e. 1.3 Ith), some radial structures appear at the periphery of doughnut-liked pattern, which is illustrated in Figure 4-14(b)~(c). It can be seen that only fundamental modes could be observed when the drive current equals 2.8mA (i.e. 1.1 Ith). On the other hand, the beam divergence angle increases very small as the drive current is increased to 3.5 mA(i.e. 1.3 Ith). Figure 4-15 shows beam divergence angles as functions of VSCELs drive current. It was found that by incorporate with photonic crystal, the divergence angle was decrease. And the divergence angles of single mode VCSEL was not increase with current.
.
0 2 4 6 8 10 12 14
tapered oxide blunt oxide
tapered oxide blunt oxide
Figure 4-2 Typical LIV curve for tapered oxide and blunt oxide VCSELs
Figure 4-1 Tapered oxide and blunt oxide VCSEL structure
Substrate
Oxide layer Implant region Implant region
Tapered oxide layer
2 4 6 8 10 12 14 -15
-10 -5 0 5
Bias current 1.5 mA 2 mA 2.5 mA 3 mA 5 mA 7 mA 9 mA
Frequency response (dB) Frequency (Hz)
2 4 6 8 10 12 14
-10 0 10
Bias current
1.5mA 2.5mA 4mA 6mA 8mA 9mA
Frequency response(dB)
Frequency(GHz)
Figure 4-3 Modulation responses for both tapered oxide and Blunt oxide VCSELs
(a) Tapered oxide VCSEL
0.5 1.0 1.5 2.0 2.5 3.0 oxide and blunt oxide VCSELs
0 10 20 30 40 50 60 70 80 90 100
Damping rate (109 s-1 )
Resonant frequeny2 ( GHz2)
K= 0.15 ns
Figure 4-5 damping rate as a function of square of resonant frequency for tapered oxide and blunt oxide VCSELs
-2 0 2 4 6 8 10 12 14 16 18 20 22
Figure 4-6 Real and imaginary S11 parameter versus frequency from model and measured data (dash line is measured data, solid line is simulated data)
Bl B lu un nt t O Ox x id i de e
Figure 4-7 Eye-diagram for tapered oxide and blunt oxide VCSELs (a) Blunt oxide VCSEL
(b) Oxide-implant VCSEL
T T a a pe p e re r ed d O Ox xi id de e
B B lu l un nt t O Ox xi id de e
Figure 4-8 oxide VCSEL and oxide photonic crystal VCSEL structure
Oxide layer
Substrate n-DBR p-DBR active layer
Oxide layer Photonic crystal region p-contact
Substrate n-DBR p-DBR
active layer
n-contact
Figure 4-9 Typical LIV curve for oxide photonic crystal VCSELs
Figure 4-10 Spectral characteristics and near-field images for the photonic crystal VCSELs
830 835 840 845
0 10 20 30 40 50
dB intensity (a.u)
Wavelength(nm)
1mA
I=1.2mA
(a)
830 835 840 845
0 10 20 30 40 50
dB intensity (a.u)
Wavelength(nm) 1.5 mA
1.75mA
(b)
0 1 2 3 4 5 0
5 10 15 20
Current(mA)
Voltage(V)
Λ=5µm α=2.5µm mesa=60µm loop=3 1 point defect
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Power(mW)
Figure 4-12 Typical LIV curve for implant photonic crystal VCSELs Figure 4-11 Implant photonic crystal VCSEL
Substrate n-DBR p-DBR
active layer p-contact
n-contact
Photonic crystal region
Implant region
Figure 4-13 Spectral characteristics for the implant photonic crystal VCSELs and VCSEL without PC holes (a) PC-VCSEL (b)VCSEL
836 838 840 842 844
-70 -60 -50 -40 -30 -20 -10 0
1.5mA 2.5mA 4mA 5mA
Inte ns ity (dB )
Wavelength(nm) (a)
(b)
1.1 1.2 1.3
8.0 Implant VCSEL
Potonic crytal VCSEL
D ivergen ce an gle(
0)
I/I
thFigure 4-14 Far field patterns of the fimplant VCSELs observed under different drive currents are shown in (a)~(c) and the implant photonic crystal VCSEL are shown in (d)~(f)
Figure 4-15 The divergence angle for implant and photonic crystal VCSELs
0.25
Tapered oxide VCSEL
Table 4-1 Equivalent circuit elements at different bias current for Tapered oxide VCSEL
Chapter 5 Conclusion
We have performed experimental study of small signal modulation behavior of 850nm tapered oxide-implant and blunt oxide-implant VCSELs.
We have found that tapered oxides process gives VCSEL with excellent high speed performance without semi-isolating substrate. The damping rate from the modulation response was found to reduce two times in the tapered oxide VCSEL and therefore enhanced the maximal modulation bandwidth. With same oxide aperture size 5.5 µm, tapered oxide VCSEL shows better modulation bandwidth of 13.2 GHz while blunt oxide VCSEL has 9.5 GHz. A very clean eye was demonstrated from improved VCSEL with rising time of 26 ps, falling time of 40 ps and jitter of less than 20 ps, operating at 10Gb/s with 6mA bias and 6dB extinction ratio. We also build an equivalent circuit model to analyze the bandwidth limitation affected by VCSEL intrinsic impedance. The simulation results could make the modulation limitation clearly and help us to modify the VCSEL process for high speed operation.
In the second part of the thesis, we report a high power (>1 mW) singlemode proton-implanted photonic crystal vertical-cavity surface-emitting laser (PC-VCSEL) with high SMSR (> 40 dB) throughout the whole operation current range. This PC-VCSEL, with an aperture of about 10 µm, has ultra-low threshold current of about 1.25 mA. We analyze the L-I curve, emission spectra, near field pattern, divergence angles of photonic crystal VCSELs fabricated with oxide-confined and implant structure. The present results indicate that a VCSEL using proton implantation for
current confinement and photonic crystal for optical confinement is a reliable approach to achieve high-power singlemode operation of a VCSEL. This concept will be applied to a 1.3µm VCSEL and other commercial applications in the future.
Reference
[1] H. Soda, K. Iga, C. Kitahara, and Y. Suematsu, “GaInAsP/InP surface emitting injection lasers,” Japanese Journal of Applied Physics, vol.18, pp.2329-2330, 1979.
[2] K. Iga, F. Koyama, and S. Kinoshita, “Surface-emitting semiconductor lasers, ”IEEE Journal of Quantum Electronics, vol.24, pp.1845-1855, September 1988.
[3] T. Honda, T. Shirasawa, N. Mochida, A. Inoue, A. Matsutani, T. Sakaguchi, F.
Koyama, and K. Iga, "Design and fabrication processes consideration of GaN-based surface emitting lasers," Trans. IEICE, J81-C-II, pp.97-104,
1998.
[4] “2003 年光電半導體產業及技術動態調查報告,” PIDA, 2004.
[5] R. Brand, “10 gigabit Ethernet interconnection with wide area networks,”
10GEA, March 2003
[6] G. P. Agrawal, Fiber-Optic Communication Systems ~Wiley, New York, 1997.
[7] A. A. Maradudin and A. R. McGurn, J. Mod. Opt. 41, 275 ~1994.
[8] J. D. Joannopoulos, R. D. Meade, and J. N. Winn, Photonic Crystal- Molding the Flow of Light ~Princeton University Press, Princeton, NJ,1995.
[9] E. R. Hegblom, B. J. Thibeault, R. L. Naone, and L. A. Coldren, “Vertical cavity lasers with tapered oxide apertures for low scattering loss,” Electron. Lett., 33, pp.
869–879, 1997.
[10] Yokouchi, N., Danner, A.J., and Choquette, K.D.: ‘Etching depth dependence of the effective refractive index in two-dimensional photonic-crystal-patterned vertical-cavity surface-emitting laser structures’, Appl. Phys. Lett., 2003, 82, (9), pp.
1344–1346
[11] A. Valle, J. Sarma, and K. A. Shore, “Spatial hole burning effects on the dynamics
of vertical cavity surface-emitting laser diodes,” IEEE J.Quantum Electron., 31, pp.
1423–1431, 1995.
[12] Y. Liu, W.-C. Ng, B. Klein, and K. Hess, IEEE J. QUANTUM ELECTRONICS, 39, 99-108 (2003)
[13] G. P. Agrawal and N. K. Dutta, “Semiconductor Lasers,” 2nd 276-280, Van Nostrand Reinhold published.