Figure 7.3: Modulation principle of the EOM VCSEL: The exciton absorption peak is shifted due to the QCSE by applying an electric field; In consequence, the refrac-tive index is altered too; The EOM cavity resonance dip can be shifted out of resonance with the VCSEL cavity; Compared to the resonance position the light transmission is significantly changed. Figure from presentation of ref. [Shc08].
structure to enhance its effect. In consequence, the refractive index change caused by the QCSE shifts the resonance dip of the cavity, and thus modulates the out-coupling efficiency of the laser. The principle of the concept is shown in Figure 7.3. Inserting the EOM part into a cavity results in enhanced optical field strength at the position of the modulator MQW and consequently enhances the modulating strength.
n
n p
constant current modulated voltage
Figure 7.4: Figure shows the schematic cross-section and a drawing of the processed EOM VCSELs. The larger VCSEL section at the bottom is operated in the CW mode while modulated voltage is applied to the top EOM section to modulate the light output.
transport at the numerous interfaces. Doping of the DBRs was performed withSiH4and DM Zn, as detailed in 4.2. A conventional one-λcavity design with a four-fold MQW GaAs/Al0.2Ga0.8Asgain-medium was chosen for the VCSEL section. For the target lasing wavelength of845nmthe active QW-emission was set to≈835nmat RT to account for device heating upon operation. Adjacent to the active cavity, an aluminum-oxide current-aperture was formed by post-growth oxidation ofAlGaAs-gradings bordering anAl0.98Ga0.02Aslayer. This aperture layer was included within the first middle DBR period. The modulator cavity was based on the same layer structure but included a five-fold MQWGaAs/Al0.2Ga0.8As EOM-medium with the GS transition at higher energy to avoid absorption. A schematic of the complete device structure is given in Figure 7.4.
To ensure automatic matching of both cavities within the device, both cavities were designed to have nominally the exact same optical length. This was realized for the four-fold MQW VCSEL cavity and the five-four-fold MQW EOM cavity by adapting QW and barrier thicknesses to form an equal optical length in total. The complete structure totals to about 400 layers and is very demanding in terms of growth accuracy and homogeneity.
7.2.1 Calibrations and growth
Epitaxial growth of this demanding EOM VCSEL design required an extensive amount of calibration and fine tuning samples to align both cavities and DBR properties, as well as to set a large variety of precise material compositions, all combined with a multitude of different doping levels. All parts of the EOM VCSEL were calibrated individually by specially designed test structures e.g. a single cavity within a reduced number of DBR layers to set the resonance dip position within the DBR stop band range. PL investigations for the optically active test samples and XRD measurements for lattice matching and compositional settings were carried out. Doping levels for n- and p-doped layers were
7 5 0 8 0 0 8 5 0 9 0 0 9 5 0
Reflectivity Intensity [a.u.] W a v e l e n g t h [ n m ]
9 . 5 - f o l d D B R
N p 3 8 0 2 n - d o p e d ( S i ) N p 3 8 0 5 p - d o p e d ( Z n )
Figure 7.5:Reflectivity of two nominally identical 9.5-fold DBR test structures. The sole difference is doping. They are either p-doped with Zn, or n-doped with Si. Spectral difference of the stop-band position is measured to 3 nm.
tuned to values between 5·1017 to 4 ·1018 and verified by Hall measurements. All layers could be tuned within this range except for theAl0.9Ga0.1As: Siwhich maxes out aroundn= 7·1017due to the self-compensating behavior of the group IV element silicon. Transfer-matrix method simulations were employed to analyze surface reflectivity measurements of DBR and cavity test samples. For the simulation, an intermediate layer with an averaged constant composition approximated graded layers with good accuracy.
Reflectivity measurements show that growth rates were altered depending on the em-ployed dopant source. Thus stop band positions of the p-DBR and n-DBR varied for identical growth parameters and needed to be tuned separately to the same spectral position (cf. Figure 7.5). In consequence, the resonance dips of the n-i-p VCSEL cavity with adjacent aperture layer and the p-i-n EOM cavity in proximity to theInGaP/GaAs:p+ etch-stop/contact layer did not show the intended automatic matching and were tuned separately to the same spectral position. Figure 7.6 part (a) shows the very good accuracy of less than one nanometer of the spectral resonance dip positioning of EOM (red) and VCSEL (green) cavity resonances. Based on these test samples the final EOM VCSEL structure (black) was grown showing an overall resonance in between the separately cal-ibrated positions. EOM and VCSEL cavity test samples were simulated separately and layer-model parameters were tuned to match dip positions to investigate run-to-run growth stability, cf. 7.6 part (b). Subsequently, the whole EOM VCSEL layer assembly was simulated employing the very same, unchanged layer-model parameters as for the test
8 4 2 8 4 4 8 4 6 8 4 8 8 5 0
0
2 0 4 0 6 0 8 0 1 0 0
N p 3 5 3 1 E O M V C S E L N p 3 5 3 0 E O M c a v i t y N p 3 5 2 9 V C S E L c a v i t y
Simulated Reflectivity [%] W a v e l e n g t h [ n m ]
1 . 4 7 5 1 . 4 7 1 . 4 6 5 1 . 4 6
( a )
( b ) E n e r g y [ e V ]
0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0
Normalized Reflectivity
Figure 7.6: (a)Reflectivity measurements show the resonance dip positions of EOM and VCSEL cavity test samples. Spectral positioning accuracy of both cavities is achieved within 1 nm. The resonance dip of the final complete EOM VCSEL structure based on the shown calibration samples is positioned in between the test sample resonances.
(b) Transfer-matrix simulations of the test samples are tuned to the measured resonance-dip positions. Layer-model parameters of these test sample simulations are used without changes to simulate the whole EOM VCSEL structure. This complete layer model shows perfect matching of the simulated dip position with the measured value, without any additional model parameter modifications.
dashed lines:Guides for the eye are shown at the measured dip positions.
sample simulations. Comparison of the measured and the simulated EOM VCSEL dip positions illustrates a perfect spectral match. The sole difference was based on the depth of the resonance dip, which was attributed to the simplified model employing a constant imaginary part value of the refractive index for all layers with band edge energies beyond the calculated wavelength. These results demonstrate the excellent predictability and accu-racy of the MOVPE growth technology, also for high precision growth runs with several
0 2 4 6 8 1 0 1 2
0123 contact layer
E O M V C S E L
Field Pattern IE2 I [a. u.]
Refractive Index
T h i c k n e s s [µm ] c a l c u l a t e d
f o r 8 4 5 n m
Figure 7.7: black: Refractive index pattern of the grown structure. Graded DBR interfaces are approximated for simulation by intermediate layers with constant refractive index levels. Positions of the VCSEL cavity, EOM cavity, and contact layer are indicated.
red:Optical field-intensity wavepattern of the VCSEL cavity resonance wavelength. Simulation is based on the transfer matrix method with a constant absorption approximation included for all layers beyond the band edge of the calculated wavelength. Nanostructure effects are not included.
hundreds of layers, and evidences the accuracy of the optimization process based on small, individually optimized test structures in combination with transfer matrix simulations.
Optical field-intensity wave-pattern was simulated with the aforementioned transfer matrix method to verify the overall design. The result for 845 nm is shown in combination with the refractive index pattern of the whole grown structure in Figure 7.7. Simulation was carried out without including gain parameters. Maximum field intensity was predicted for the VCSEL cavity, while significant field strength was observed in the EOM part, enabling output power modulation.
The final device was grown by a single arsine-based MOVPE growth run at 700°C for the entire monolithic EOM VCSEL structure. Growth temperature was lowered to 615°C only for the deposition of the lattice matchedInGaP etchant-stop layer, employing the previously described switching procedure (cf. Table 4.5). After switching back from TBP to arsine, theInGaP layer was covered with 5 nmGaAsprior to heating up again to resume DBR growth. Growth was finalized by aGaAs : n+ contact layer serving as oxidation protection. Epitaxy on 2 inchGaAs : Si(1 0 0) substrates yielded good uniformity across the whole wafer and a smooth surface quality.
7.2.2 Processing and characterization
Standard lithography and dry etching techniques were employed to process the EOM VCSELs with varying double-mesa diameters from 25 to 36 µm and 45 to 56 µm for EOM and VCSEL sections, respectively. All processing for this device was performed by my
7 8 0 8 0 0 8 2 0 8 4 0 8 6 0 8 8 0 9 0 0 9 2 0 0 . 0
0 . 2 0 . 4 0 . 6 0 . 8 1 . 0
J o i n e d r e s o n a n c e d i p o f V C S E L + E O M c a v i t i e s .
Normalized Reflectivity W a v e l e n g t h [ n m ]
N p 3 5 3 1 F i n a l E O M V C S E L
1 . 5 5 1 . 5 1 . 4 5 1 . 4 1 . 3 5
E n e r g y [ e V ]
Figure 7.8:Surface reflectivity measurement of the unprocessed EOM VCSEL wafer is shown.
The joined EOM + VCSEL resonance dip is precisely positioned within the DBR stop band. FWHM of the dip is less than 2 nm.
former colleague Alex Mutig as described in [Hop07]. Selective oxidation of the Al-rich aperture layer was done using optimized conditions as described in ref. [Hai02]. Upon oxidation, the volume of the aperture layer is reduced between 6.7% and 13%, depending on the aluminum content [Cho97]. Consequently, the design needed to deal with significant strain at the interfaces to the adjacentGaAslayers. To avoid interlayer delamination or fracture, gradedAlGaAslayers were grown enclosing the aperture layer. Three ohmic contacts were formed for device operation, bottom and middle contact for the VCSEL part, middle and top contact for the EOM part. The bottom contact is formed on the underside of the thinned substrate, while middle and top contacts are circular contacts on the respective mesas (cf. Figure 7.4).
For device operation, the EOM VCSELs were mounted onto a copper heat sink and operated at RT as detailed in [Blo09]. In parallel to the optical output power of the laser, the electric photocurrent generated within the EOM section was recorded. This measured current was due to unintended photo absorption of the lasing wavelength.