Open Access
Add to BibSonomyAbstract
In this paper, we present an InP-based micromechanically tunable VCSEL emitting in the 1.55μm wavelength region with a 26nm tuning range. The laser is based on a two-chip concept, allowing for a separate optimization of the curved top mirror and the amplifying component. Current confinement is achieved by a buried tunnel junction. The design of the microcavity ensures fundamental mode operation with a side mode suppression ratio exceeding 49dB even for large apertures. Simulations indicate that the tuning range is limited by coupled cavity effects and reveal important design criteria like an upper boundary regarding the device thickness.
©2005 Optical Society of America
1. Introduction
Tunable lasers emitting in the 1.55μm wavelength range are important and demanded light sources for fiber-based communication networks and spectroscopic applications like trace gas monitoring or sensor techniques based on fiber Bragg gratings. Beside the well-known advantages of vertical-cavity surface-emitting lasers (VCSELs) like low power consumption and easy fiber coupling, the light emission parallel to the growth direction allows an easy implementation of micro-electro-mechanical systems (MEMS) for wide and mode-hop-free tuning by the control of only one tuning current. Various concepts utilizing optical [1, 2] and electrical [3, 4] pumping schemes have been presented to date. The deflection of the MEMS-mirror for wavelength tuning is typically achieved by applying electrostatic force. The problems related to the fabrication of long-wavelength VCSELs like small index contrast and poor thermal conductivity of potential distributed Bragg reflectors (DBRs) and the difficulty to achieve current confinement for electrical pumping are often addressed by wafer fusion [2], metamorphic growth processes [4] or ion implantation [5].
Here, we present an electrically pumped device utilizing a buried tunnel junction (BTJ) [6] for current confinement and an actuation mechanism that is based on thermal expansion due to Joule heating of the mirror material. Such a mechanism suffers from a reduced tuning speed but the membrane displacement is not restricted to 1/3 of the air gap size and there is no risk of a membrane collapse, as it is the case for an electrostatic approach. In addition, the technology is rather simple and the voltage drop is typically smaller. Dynamic measurements have revealed a 3-dB cut-off frequency of 500MHz and 1/e-time constant of about 1ms [7], sufficient for most applications like gas sensing or as spare lasers in wavelength division multiplexing (WDM) systems. Furthermore, the amplifying and the mechanical part of the laser are fabricated separately and are assembled in a final packaging step. This allows for independent optimization of the components to achieve optimum device performance for a desired application. The schematic structure of the laser is depicted in Fig. 1. Since the lower part, containing the back mirror, the current aperture (BTJ), and the active region, which is also referred to as a “half VCSEL”, is smaller in size compared to the MEMS-mirror, surrounding devices are used as contact surface. Alternatively, a submount technique can be used [7]. Since the packaging can probably be automated in connection with an active alignment similar to what is known from the fabrication of optical interconnects, our device, with the unique feature to tailor the laser performance, can be an interesting alternative to monolithic devices.
2. Device structure
The half VCSEL is fabricated in a two-step growth process using solid-source molecular beam epitaxy (MBE). The device is mounted in an upside down configuration and the substrate, initially located on top of the structure, has been removed completely. The active region consists of 7 compressively strained AlGaInAs quantum wells with a thickness of 6nm [8]. The top n-doped AlGaInAs-cladding is about 1.7μm thick with a modulated doping profile to optimize the trade-off between series resistance and absorption loss. The n ++ doped layer below the top contact has been removed in the light-emitting region by selective wet-chemical etching. The buried tunnel junction, located below the active region, consists of heavily doped n-AlGaInAs and p-GaInAs and is laterally structured prior to the InP-overgrowth provoking a reverse biased pn-junction in the outer parts and therefore efficient current confinement to the BTJ diameter D. The use of n doped InP reduces the electrical and thermal resistance drastically and in connection with a thin dielectric back mirror consisting of only 2.5 pairs of amorphous silicon and calcium fluoride together with a gold coating, the generated heat can efficiently be transferred to the electro-plated gold heat sink that additionally serves as the p-contact. Since lateral current injection is used, a thick overgrowth is desirable to minimize current crowding effects that could favor higher order transverse mode operation. For the present device, a 600nm InP layer was chosen.
Fig. 1. Schematic cross section of a tunable two-chip VCSEL with a buried tunnel junction (BTJ) for current confinement. The sketch at the right of the cross section indicates the three-mirror-configuration with the reflectivity R A/B/C and the length of the cavities L air/SC.
The micro-mechanical top mirror is an MBE-grown distributed Bragg reflector (DBR) consisting of 22.5 pairs of GaAs and Al0.85Ga0.15As leading to a theoretical reflectivity of 99.92%. The displacement of the mirror is achieved by injecting a tuning current into the membrane using via-hole contacts. Due to Joule heating and the resulting thermal expansion of the material, the air gap thickness increases and the wavelength can be shifted. Some of the layers additionally contain 2 to 5% of indium to introduce a certain stress gradient that leads to a rotation symmetric and concave curvature after membrane release. The radius of curvature is typically between 1 and 3mm. Due to this shape the assembly step is rather uncritical and an alignment without tilt-loss can easily be found. In addition, such a plane-concave resonator can be used to improve the side mode suppression ratio even for large current apertures D [3].
It is important to note, that no antireflection (AR) coating is used at the semiconductor-air interface resulting in a three-mirror-resonator (see Fig. 1). For such a configuration, it is desirable to choose the air-gap thickness Lair to be an odd multiple of a quarter wavelength (λ/4), while the optical path length LSC of the half VCSEL is designed to be an even multiple of λ/4 (see Fig. 1). In this case, the air-gap acts as an additional mirror layer and increases the overall reflectivity while keeping the overlap of the optical field with the active region high. This so-called semiconductor-coupled cavity (SCC) design therefore enables low threshold currents but shows a reduced wavelength tuning efficiency [9].
3. Results
The results obtained for a device with a BTJ diameter of D=10μm at room temperature in continuous mode operation are shown in 'Fig. 2.
Fig. 2. Typical laser spectra for three different tuning currents of I 1=0mA, I 2=6mA, and I 3=8.9mA. The envelope indicates the maximum output power during tuning with a constant driving current of 20mA.
Three typical spectra for different tuning currents are depicted together with the envelope function generated by keeping track of the maximum during tuning with a constant laser current of 20mA. As can be seen, a tuning range of 26nm with very constant output power was obtained. The side mode suppression ratio (SMSR), regarding the transverse modes on the high energy side and/or the longitudinal modes on both sides of the lasing fundamental mode, is found to be between 49 and 57dB across the tuning range. Measuring a longitudinal mode-spacing or free spectral range (FSR) of approximately 40nm (see spectrum for I 3 in Fig. 2), we can determine the air gap thickness to be 23 λ/4 which is in good agreement with profile measurements of the structure. Due to the electro-thermal actuation mechanism of the MEMS-mirror, the membrane displacement and, therefore, the wavelength shift is proportional to the dissipated power, i.e. to the square of the tuning current. The experimental data, displayed in Fig. 3(a), is just showing a slight s-shape behavior as it is expected for a SCC design [9]. Performing one-dimensional simulations based on the transfer matrix approach using the given device dimensions and the above mentioned air gap thickness, the resonance wavelength can be calculated for the related membrane displacement. The result is shown in Fig. 3(a) matching the experimental data very well. As can be seen, a tuning current of 8.7mA results in a membrane displacement of 550nm being equivalent to a wavelength shift of 25nm. The slight deviation of some of the measured data points from the theoretical line is due to temperature fluctuations. The wavelength drift of a packaged device can be around 1nm/K. Therefore, for most applications, a wavelength locker is desirable as it is the case for most MEMS-based lasers. The initial deflection of the membrane, i.e. the air gap thickness at zero tuning current depends on the radius of curvature of the membrane and the length of the suspension beams and can differ from device to device. The resistance of the membrane is typically around 800Ω but is also subject to variation. The maximum fiber coupled power and the threshold current versus laser wavelength are shown in Fig. 3(b).
The power is quite constant at a value of 0.25mW slightly increasing at the low energy side of the tuning range due to the reduced effective reflectivity of the top mirror system as a consequence of the detuned air gap. The lowest threshold current is found to be 2mA equivalent to a current density of only 1.8kA/cm2, taking into account a diffusion broadening of about 1μm. Due to a very low differential series resistance of only 17.6Ω, the threshold voltage is 0.87V. The voltage drop at the thermal roll-over at 20mA is 1.24V and very constant across the tuning range. The relative low output power in connection with the very low threshold current density can be attributed to the use of a highly reflecting top mirror. Since the used microcavity enables high SMSR even for large current apertures, high power operation should be possible. A single-mode fiber coupled power of 2.8mW for a similar (D=12μm) but non-tunable device has already been observed in connection with reduced mirror reflectivity (14.5 mirror pairs equivalent to a reflectivity of R=99.6%) [11]. This is the highest output power published for a non-tunable VCSEL to date and should be achievable in connection with a tunable mirror as well.
Fig. 3. Tuning behavior for a device with a BTJ diameter of 10μm. a) Experimental laser wavelength vs. square of the tuning current (red symbols) and theoretical wavelength shift vs. membrane displacement (black line). b) Maximum fiber coupled power and threshold current plotted versus laser wavelength.
4. Simulations
To obtain more insight into the effects that limit the tuning range, one-dimensional simulations based on the transfer matrix approach have been performed to calculate the resonance wavelength and the equivalent threshold gain for SCC devices. The results are plotted in Fig. 4 for four samples of different semiconductor cavity length LSC.
Fig. 4. Simulations of the change in threshold gain and the resonance wavelength for a given membrane displacement compared to the non-detuned case based on the transfer matrix formalism.
Assuming a certain allowed change in the threshold gain gth compared to the non-detuned case, one can find the related maximum tuning range for a given structure. As can be seen from Fig. 5 the tuning range follows a 1/LSC-dependency.
Fig. 5. Experimental and theoretical data for the tuning range plotted versus the inverse length of the half VCSEL. To obtain the theoretical data, a maximum increase of 50% in threshold gain compared to the non-detuned case was assumed. The line is just a guide to the eye.
This relationship is due to the fact that the semiconductor part in a SCC structure acts as a Fabry-Perot filter with a full width at half maximum (FWHM) of the transmission characteristic being inversely proportional to the length of the filter (LSC). The analytic expression for a lossless filter of length LSC and with reflectivity RA and RB is given in Eq. (1).
Therefore, the intensity of the optical field at the active region decreases, i.e. the threshold gain increases stronger with tuning for larger values of LSC . Beside the theoretical data for Δgth/gth=50%, the measured tuning ranges of three different samples are depicted in Fig. 5 [7]. For all devices, a top mirror of the same reflectivity (R=99.92%) was used. In addition, the tuning range of the devices is related to the width of the (filtered) electroluminescence (EL) of the devices without the top mirror. For all three lasers, shown in Fig. 5, the tuning range was about 70–80% of the FWHM of the EL. In comparison, a similar device with an AR coating showed a FWHM of the EL of about 160nm being much larger compared to the observed tuning ranges. As a consequence, simulations and experiments with a broadband dielectric top mirror reveal only a minor influence of the stop band width on the tuning range compared to the dominant filter effect. Obviously, the tuning range of such a complicated structure can depend on many device parameters. Nevertheless, a clear trend is observed and even though long cavities might be desirable to reduce current crowding or to improve single mode operation [10], the limitations regarding the tuning range have to be considered carefully. The above mentioned simulations can only reproduce the expected trend but by comparing this data to the experiment, we can now extract important information regarding the design criteria for these tunable devices an estimate an optimum device thickness for a desired tuning range. Similar considerations hold for the use of partial mirrors at the semiconductor air interface since the increased reflectivity also narrows the transmission characteristic [4].
5. Conclusion
In conclusion, we have shown a MEMS-tunable VCSEL showing a very high SMSR exceeding 49dB due to the use of a curved top mirror in connection with a low threshold current density of 1.8kA/cm2. Simulations for the present SCC layout indicate that the main limiting factor regarding the tuning range of 26nm is the filter-like behavior of the semiconductor part of the device and not the material gain or the spectral characteristic of the DBR. For such a configuration, we have obtained a maximum tuning range of up to 43nm which is the best value published so far for electrically pumped devices and sufficient for most applications since absorption lines of several typical gases can be found within this wavelength interval and since the bandwidth of a typical erbium doped fiber amplifier is of the same order. By the use of an antireflection coating at the semiconductor-air interface we expect to further increase the tuning range.
Acknowledgment
This work was supported by the Federal Ministry of Education and Research, Germany (BMBF).
References and links
1. F. Sugihwo, M. C. Larson, and J. S. Harris jr, “Simultaneous optimization of membrane reflectance and tuning voltage for tunable vertical cavity lasers,” Appl. Phys. Lett. 72, 10–12 (1998). [CrossRef]
2. A. Syrbu, V. Iakovlev, G. Suruceanu, A. Caliman, A. Rudra, A. Mircea, A. Mereuta, S. Tadeoni, C.-A. Berseth, M. Achtenhagen, J. Boucart, and E. Kapon, “1.55-μm optically pumped wafer-fused tunable VCSELs with 32-nm tuning range,” IEEE Photonics Technol. Lett. 16, 1991–1993 (2004). [CrossRef]
3. P. Wang, P. Tayebati, D. Vakhshoori, C.-C. Lu, and R. N. Sacks, “Half-symmetric cavity microelectromechani-cally tunable vertical cavity surface emitting lasers with single spatial mode operating near 950 nm,” Appl. Phys. Lett. 75, 897–898 (1999). [CrossRef]
4. J. Boucart, R. Pathak, D. Zhang, M. Beaudoin, P. Kner, D. Sun, R. J. Stone, R. F. Nabiev, and W. Yuen, “Long wavelength MEMS tunable VCSEL with InP-InAlGaAs bottom DBR,” IEEE Photonics Technol. Lett. 15, 1186–1188 (2003). [CrossRef]
5. A. Bousseksou, M.E. Kurdi, M.D. Salik, I. Sagnes, and S. Bouchoule, “Wavelength tunable InP-based EP-VECSEL operating at room temperature and in CW at 1.55-μm,” IEE Electron. Lett. 40, 1490–1491 (2004). [CrossRef]
6. M. Ortsiefer, R. Shau, G. Böhm, F. Köhler, and M.-C. Amann, “Low-threshold index-guided 1.5 μm long-wavelength vertical-cavity surface-emitting laser with high efficiency,” Appl. Phys. Lett. 76, 2179–2181 (2000). [CrossRef]
7. F. Riemenschneider, M. Maute, H. Halbritter, G. Böhm, M.-C. Amann, and P. Meissner, “Continuously tunable long-wavelength MEMS-VCSEL with over 40-nm tuning range,” IEEE Photonics Technol. Lett. 16, 2212–2214 (2004). [CrossRef]
8. M. Ortsiefer, M. Förfanger, J. Rosskopf, G. Böhm, F. Köhler, C. Lauer, M. Maute, W. Hofmann, and M.-C. Amann, “Singlemode 1.55 μm VCSELs with low threshold and high output power,” IEE Electron. Lett. 39, 1731–1732 (2003). [CrossRef]
9. G. D. Cole, E. S. Bjrlin, Q. Chen, C.-Y. Chan, S. Wu, C. S. Wang, N. C. MacDonald, and J. E. Bowers, “MEMS-tunable vertical-cavity SOAs,” IEEE J. Quantum Electron. 41, 390–407 (2005). [CrossRef]
10. S. Riyopoulos and H. Unold, “Continuously tunable long-wavelength MEMS-VCSEL with over 40-nm tuning range,” J. Lightwave Technol. 20, 1173- (2002). [CrossRef]
11. B. Kogel, M. Maute, H. Halbritter, S. Jatta, G. Bohm, M. -C., and P. Meissner, “High singlemode output power from longwavelength VCSELs using curved micromirrors for mode control” IEE Electron. Lett. 41, 43–44 (2005). [CrossRef]
