1. Introduction

The quantum cascade laser(QCL) [1] has emerged as a practical, compact, room temperature (RT), mid-infrared radiation source with high output power [2]. As a result, a large number of QCL-related applications in homeland security, pollution control, industrial process monitoring, medical diagnosis etc. are gradually being realized. A common theme of all these application is spectroscopy, by which molecules, in trace concentration, are detected due their strong resonant absorption of mid-infrared radiation. The spectral region between 6 and 10 µm is the fingerprint region for identification of a large number of chemicals like NO,CH₄, N₂0, H₂S, SO₂, CO₂ [3], as well as explosives like TNT and RDX.

For unique identification of chemicals, it is desirable to observe multiple absorption features over a broad spectral range. This is used to differentiate between two chemicals which may have similar features at a certain wavelength. Tunable range of 236 cm⁻¹ were previously demonstrated by our group, from a single QCL wafer centered around a wavelength of 4.65 µm [4]. In this paper, we demonstrate a broadband heterogeneous QCL (HQCL) [5] with DFB grating laser array emitting between 5.9 and 10.9 µm (~760 cm⁻¹), from a single wafer, at RT. In HQCL, multiple stacks of discrete wavelength quantum cascade (QC) stages are incorporated within the same quantum cascade laser waveguide core. The summation of the effective gain of the QC stages can give a broad gain over a large wavelength range. This allows selection of the laser emitting wavelength over a wide range using an appropriate feedback mechanism. EC systems are mechanically tuned and require additional optical components, making it bulky and limiting its tuning speed, whereas DFB laser arrays represent a compact, monolithic tuning alternative, with fast tuning speed [6].

2. Design, growth, fabrication of HQCL

The HQCL demonstrations so far, were grown based on the Al_0.48In_0.52As/ Ga_0.47In_0.53As material system lattice matched to InP [5]. Strain balanced Al_0.63In_0.37As/ Ga_0.35In_0.65As/ Ga_0.47In_0.53As QCLs were shown [7] to have good performance across the whole wavelength range 5.2-11 µm, and they can be incorporated into a HQCL active region in a single growth run.

In contrast to 3 well active region designs used in highly efficient QCLs [2], the energy separation between upper laser state and next excited state must be reduced well below 124 meV (~10µm). This may reduce the overall efficiency of shorter wavelength emitter by decreasing injection efficiency to the upper laser level and increasing the thermal escape of electrons from the upper laser level, but is necessary in a HQCL to minimize excess loss at longer wavelengths. For this demonstration, a compromise value of 55-60 meV was chosen for a balance in the efficiency and the range of operation in the broadband HQCL.

Gain in a HQCL core is dependent on all constituent QC stages. Photons emitted by the primary optical transition of a long wavelength QC stage are often cross-absorbed by optical transitions between the upper laser level and the next exited state of other QC stages, due to large oscillator strength between them and large electron population of upper laser level near resonant field. The absorption increases with reduction in detuning of photon energy from the resonant absorption energy.

Modal intensty profile within the waveguide and the modal confinement factors (CF) of the QC stages at different photon energies is shown in Figs. 1(a) and 1(b).

 

Fig. 1 (a) Modal intensity vs distance inside a HQCL (b) Modal confinement factors vs energy of the different QC stages.

Download Full Size | PDF

If the gain of the longer wavelength QC stages are over compensated, without regard for larger differential gain at long wavelength, short wavelength DFBs may fail to lase at lower current than Fabry Perot (FP) mode, which is the natural oscillation mechanism inside the laser cavity, leading to a decrease in spectrum coverage. In ideal scenario, the anti-reflective (AR) coating on the laser facets should suppress the FP mode, by suppressing appreciable feedback for FP oscillation especially at high current injection. However, design of AR coating with minimal residual reflectivity over a large spectral range, is quite challenging and puts a severe constraint on shape of the gain curve for an ultra-broadband HQCL.

The gain is calculated from the first order gain model by the expression,

where $\hslash {\omega }_i$, $N_i$ represent energy and electron concentration of $i$th level respectively; $\hslash \omega$ is the energy of the electromagnetic wave; $g_{ji}(\hslash \omega )$ is the energy dependent gain due to transition of dipole strength $z_{ji}$ between two levels j and i, separated by energy $\hslash {\omega }_{ji}$ and broadened by linewidth ${\Gamma }_{ji}$;$e$, $c$, ${\epsilon }_0$, $n_r$ are the electronic charge, speed of light, vacuum permittivity and refractive index respectively. In comparison, the model [8] generally employed for calculation of first order gain $g_{ji}^{*}$ is

In both of the models, non-parabolicity [8] is ignored. The calculation for electron concentration is based on assumption that when stages of QCL align the injector miniband and the upper laser level share a common quasi Fermi-level, where individual levels are occupied by a Fermi-Dirac distribution [9]. The origin of term $\omega$ in the numerator of (1.2) can be traced to the relation between dipole matrix element and momentum matrix element, under the approximation $\hslash \omega \approx \hslash {\omega }_{ji}$ [10]. In a single core QCL, where the lasing mode is generally near its peak gain corresponding to resonant transition between upper and lower laser level, (1.2) is valid. However, in broadband HQCL, one QC stage in off-resonant condition $\hslash \omega <<\hslash {\omega }_{ji}$ or $\hslash \omega >>\hslash {\omega }_{ji}$ may be a source of loss for photons of energy $\hslash \omega$emitted by another QC stage. The (1.2) in comparison to (1.1), underestimates the gain at longer wavelength. So we need to use (1.1) which is also similar to the relation, derived from the density matrix formalism, under first order approximation [8].

The net gain curve of the HQCL is calculated by the sum of individual QC stage gain $(g_k\left(\hslash \omega \right))$ of the $k$ th stage, weighted by the respective modal confinement factors $({\Gamma }_k(\hslash \omega ))$ at different $\hslash \omega$ . The total modal gain, based on (1.1) and (1.2), at a current density of ~4 kA/cm², are shown in Figs. 2 (a) and 2(b), respectively. Wavefunction engineering ensures that the strong losses corresponding to resonant absorption from upper laser levels to higher states (near 60 meV) does not overlap with the HQCL emission energies of interest. The gain is designed to be very flat throughout the wavelength range. Figure 2 (b) shows a skewed gain profile.

 

Fig. 2 Individual gain and total modal gain of the QC stages at a current density of ~4 kA/cm² following (a) Eq. (1.2) (b) Eq. (1.1)

Download Full Size | PDF

The broadband structure was grown by gas source molecular beam epitaxy on a 4 µm thick low-doped InP buffer (doped to n ~2x10¹⁶ cm⁻³) on top of a low-doped InP substrate (doped to n ~1-2x10¹⁷ cm⁻³). QC stages 1 to 6 (doped similarly to n ~4.7x10¹⁶ cm⁻³), were grown, starting from the buffer region in the configuration shown in Fig. 1 (a). The growth of the active region was followed by a 100 nm thick InP (doped to n ~2x10¹⁶ cm⁻³) and a 750 nm thick lattice matched Ga_0.47In_0.53As grating layer.. Threshold current of the HQCLs are much higher than single core devices due reduced modal confinement factors of each QC stages with respect to a single core device. The devices are doped higher than single core QCLs to ensure that the DFBs reach the threshold current before their maximum current at resonant field.

A distributed feedback (DFB) grating array, with a total of 24 lasers were patterned on the grating layer by e-beam lithography. Etch depth of the gratings were 375 nm, with the grating period varying with the required Bragg wavelength. After standard grating fabrication [11], a 4 µm thick InP (doped to n ~2x10¹⁶ cm⁻³) cladding layer and a 1 µm thick highly doped InP (doped to n ~5x10¹⁸ cm⁻³) cap layer were grown on top of the grating by metal organic chemical vapor deposition. Standard double channel processing was followed to fabricate the DFB laser array, with the exception that Si₃N₄ was substituted for SiO₂ as an insulator [12]. Laser bars of cavity length 3 mm were cleaved and an AR coating of ~1.4µm of Y₂O₃ was deposited on the both the facets. DFB array dies were then bonded epilayer up with indium solder on copper sub mounts. Testing was done on a thermoelectric cooler stage at 298K.

3. Results

Spectra were measured with a Fourier transform infrared spectrometer (FTIR) with a resolution of 0.125 cm⁻¹ in rapid scan mode. Single mode DFB emission from AR coated devices was obtained between 5.9 and 10.9 µm, as shown in Fig. 3. The normalized spectrum of each DFB laser appears as a vertical line. FP emission at 9 µm, which was observed in uncoated devices, was eliminated by AR coating, which had a reflectivity minimum at around 9 µm. At the same time, very uniform threshold current densities of around 4 kA/cm², for most AR coated DFB lasers in the array, are observed. This may be useful for application in a spectrometer, since all wavelengths can be accessed with the same drive current.

 

Fig. 3 Threshold current density as a function of emission energy for the AR coated HQCL DFBs at various wavelengths. Threshold current density of uncoated DFBs along with peak optical output power of the AR coated DFBs are also shown.

Download Full Size | PDF

For testing, average output powers were measured from the front facet of AR coated DFBs using a calibrated thermopile detector placed directly in front of the laser facet, and the peak power was obtained from the measured average power and the known pulse duty cycle, assuming 100% collection efficiency. Pulsed mode testing was done with a pulse width of 200 ns and a duty cycle of 1% (i.e., a repetition rate of 50 kHz). Peak power output of more than 50 mW was obtained over a large spectral range. Some broadening due to thermal chirp was observed at high current, but the emission remained essentially single mode. The maximum power output from the longer wavelength region was considerably higher than shorter wavelengths, presumably due to the higher differential gain at longer wavelengths and larger number of periods of QC stages emitting at longer wavelengths. The lower power output at the short wavelength is also influenced partially by a larger effective $\kappa L$ product, which produces larger effective reflection coefficient, for the DFB grating.

4. Conclusion

In conclusion, a HQCL composed of 6 strain balanced Al_0.63In_0.37As/Ga_0.35In_0.65As/Ga_0.47In_0.53As QC stages was demonstrated to emit at wavelength from 5.9 μm to 10.9 μm (~760 cm⁻¹). Different aspects of device design like active region wavefunction engineering, modal confinement factors by spatial arrangement of QC stages were employed to get a balanced gain, across these wavelength range, which was demonstrated by a flat threshold for DFB laser arrays. Utilizing the DFB grating feedback mechanism in HQCLs opens up the possibility of an ultra-broadband, electrically tunable, fast, compact, lightweight, and monolithically integrated device on a single chip.