Extended electrical tuning of quantum cascade lasers with digitalconcatenated gratings

S. Slivken, N. Bandyopadhyay, Y. Bai, Q. Y. Lu, and M. Razeghia)

Center for Quantum Devices, Department of Electrical Engineering and Computer Science,Northwestern University, Evanston, Illinois 60208, USA

(Received 17 September 2013; accepted 19 November 2013; published online 6 December 2013)

In this report, the sampled grating distributed feedback laser architecture is modified with digital

concatenated gratings to partially compensate for the wavelength dependence of optical gain in a

standard high efficiency quantum cascade laser core. This allows equalization of laser threshold

over a wide wavelength range and demonstration of wide electrical tuning. With only two

control currents, a full tuning range of 500 nm (236 cm�1) has been demonstrated. Emission is

single mode, with a side mode suppression of >20 dB. VC 2013 AIP Publishing LLC.

[http://dx.doi.org/10.1063/1.4841635]

One of the primary applications for quantum cascade

lasers (QCLs) is chemical spectroscopy. QCLs are compact,

with high power output and the ability to access a wide por-

tion of the electromagnetic spectrum (3< k< 16 lm) at

room temperature. InP-based QCLs also benefit from a

mature epitaxial growth technology developed over many

years.1 Dynamic tuning of the laser output wavelength

remains a challenge, however, and various approaches have

been utilized to maximize the tuning range for a given laser.

External cavity lasers have great versatility, but they are

mechanically tuned and require additional optical compo-

nents.2 This increases size and limits tuning speed.

Distributed feedback (DFB) laser arrays represent a mono-

lithic (compact) tuning alternative, but quickly increase in

complexity as the number of elements (and tuning range)

increases. This complexity includes both the control algo-

rithm and the difficulty in efficiently combining many lasers

of different wavelengths into a common output beam.3,4

To reduce the complexity of the laser array, it is neces-

sary to realize the widest tuning possible for discrete array

elements. One approach is to sampled grating technology, as

previously proposed.5 Towards this end, our group applied

sampled grating distributed feedback (SGDFB) technology

to a standard laser core in order to demonstrate continuous

wave (CW) electrically tunable SGDFB QCLs.6 This initial

attempt showed 50 cm�1 of continuous tuning was possible

near k¼ 4.8 lm with a single device and two control cur-

rents. With anti-reflective (AR) coating of the mirror facets,

we also showed the ability to access a wide spectral range

(361 cm�1) using a SGDFB laser array with a standard laser

core.7

The final goal, however, is to achieve similar functional-

ity from a single SGDFB laser. At telecommunication wave-

lengths, related technology has demonstrated tuning of

170 cm�1 and is still in development.8 This broad tuning

takes advantage of the high sensitivity of an interband laser’s

refractive index on current and temperature. The tuning was

limited primarily by available gain bandwidth and applica-

tion needs, such as a high side mode suppression ratio

(SMSR) and cavity length restriction. The QCL can have

much wider bandwidth, however, either through heterogene-

ous emitter integration2 or by distributing the dipole strength

across multiple transitions.9 Bandwidths up to Dk/k� 40%

have been achieved for both techniques. Unfortunately, the

QCL also has a much more stable refractive index, which

requires larger changes in current density to realize wide

tuning.

In addition, a typical QCL exhibits a much lower modal

differential gain. As the emission wavelength deviates from

the peak gain, the modal differential gain decreases, leading

to a larger threshold current density and a reduced range of

operation. A reduced current density range in turn limits the

tuning, defeating the original purpose.

This can be overcome by modifying the diffractive feed-

back within the laser cavity to partially compensate for the

curvature of the gain spectrum. This type of manipulation

was previously attempted with continuous gratings in a QCL

in an attempt to produce simultaneous emission at multiple

wavelengths.10 Due to the sampled grating basis of a

SGDFB, however, a different approach is necessary to real-

ize wide tuning, such as the digital concatenated grating

(DCG).11

A DCG consists of multiple sampled gratings with dif-

ferent Bragg wavelengths and the same sampling period. In

the low grating contrast regime, the reflectivity of the grating

superstructure can be seen as a linear sum of its substruc-

tures.9 Using this property, the reflectivity profile can be

engineered by selecting the proper number of grating periods

at each Bragg wavelength.

To confirm this behavior, a two grating basis for the

DCG structure was chosen with an engineered overlap. A

schematic of the DCG-SGDFB laser and sampled grating

layout is shown in Figure 1. The front and back sections are

identical except for a small change in sampling period, Z.

The two grating periods are notated K1 and K2. The gratings

are located immediately adjacent to each other, with the K1

grating being oriented closer to the center of the cavity in

both sections.

The reflectivity was simulated using a transfer matrix

approach described in Ref. 6. Shown in Figure 2(a) is thea)Electronic mail: razeghi@eecs.northwestern.edu

0003-6951/2013/103(23)/231110/4/$30.00 VC 2013 AIP Publishing LLC103, 231110-1

APPLIED PHYSICS LETTERS 103, 231110 (2013)

reflectivity of one section of the DCG structure with 25 sam-

pling periods (Ns¼ 25) and no net gain. The values for K1

and K2 are 670 and 920 nm, respectively. Instead of two well

defined reflectivity envelopes, however, we see that the

shape is much more complex. Clearly there is interference

between the two gratings, which is a result of the strong cou-

pling coefficient (>80 cm�1) and small number of grating

periods per sampling period (Ng¼ 4) used in this experi-

ment. A small Ng and large Ns are used, compared to our

previous experiments, to enhance the overall tuning range

while preserving single mode emission. Also shown in

Figure 2(a) is the simulated modal gain for the laser structure

at a current density of 2.5 kA/cm2. The shape is assumed

Gaussian with a FWHM of 500 cm�1, as determined from

electroluminescence measurements. The grating periods and

positions have been chosen to enhance reflectivity away

from the gain peak.

Unlike its use as a passive reflector, however, the DCG-

SGDFB is actively pumped, which alters the effective reflec-

tivity significantly. Shown in Figure 2(b) is the calculated

reflectivity for the same DCG structure with the modal gain

included, which changes with the current density. Though

sufficient for our initial experiment, the shape is not quite

balanced, with a tendency towards longer wavelengths. It is

possible with more effort to further refine the DCG design to

optimize the reflectivity curve.

Laser threshold and primary emitting wavelength are

determined by solving for the effective mirror loss (am) as

seen from between the two sections, given as

am ¼1

2Lln

1

Rf Jf ; Sf½ �Rb Jb; Sb½ �

� �; (1)

where L is the total cavity length and Rf(Rb) represents the

power reflectivity from the front (back) section. In general,

reflectivity is sensitive to both current density and photon

density. Current density is the sum of pulsed and CW contri-

butions, with induced heating effects calculated accordingly.

CW current refers to constant current applied to a specific

section. As a simplification, due to the weak overall grating

coupling per sampling period, the photon density in each

section is assumed constant. Above laser threshold, the mir-

ror loss is solved iteratively until only a single emission line

reaches oscillation.

The mirror loss is also a strong function of the mirror

facet reflectivity. Mirror reflection causes a virtual inversion

in the grating order, which, due to interference, produces no

net increase in diffractive feedback. As the facet reflectivity

increases, the normal material gain starts to dominate, as

shown in Figure 3(a). This leads to a reduced side mode sup-

pression and reduction in tuning range compared to the ideal

cavity, as can be seen in Figure 3(b). For the tuning simula-

tion, the pulsed current density (Jf,p and Jb,p) is kept constant,

and tuning is achieved by ramping the CW current density

(Jf,CW) in the front section. For a good AR coating, however,

several hundred nanometers of tuning is expected.

To experimentally demonstrate wide electrical tuning, a

laser was fabricated using the DCG-SGDFB design

FIG. 1. (a) Schematic of DCG-SGDFB

device structure. (b) Oblique scanning

electron micrograph of etched dual

grating structure.

FIG. 2. (a) Calculated reflectivity of dual grating passive DCG and simu-

lated modal gain spectrum of quantum cascade laser at 2.5 kA/cm2. (b)

Calculated reflectivity of active DCG at the same current density.

231110-2 Slivken et al. Appl. Phys. Lett. 103, 231110 (2013)

principle. The laser wafer chosen is the same used for an

earlier publication on ring lasers and has a peak pulsed

Fabry-Perot emission wavelength of �4.8 lm.6 Dual

sampled gratings with periods of K1¼ 670 and K2¼ 920 nm,

respectively, were patterned into a 300 nm GaInAs grating

layer placed 100 nm above the laser core using e-beam li-

thography and plasma etching. For this experiment, the sam-

pling periods Z1 and Z2 were chosen to be 132.2 and

128.3 lm, respectively, which gives an estimating step tun-

ing of 12 cm�1. The ideal tuning range for this design,

according to simulation, is �800 nm (350 cm�1).

After patterning, laser fabrication followed the proce-

dure described in Ref. 5. The only exception is the modifica-

tion of the electrical isolation channels etched between the

adjacent laser sections. These channels were etched 2 lm

deep through a 120 lm wide mask into the 4 lm thick upper

InP layers. Experimentally, this provided >1500 X isolation

between sections measured near zero bias.

Laser cavities of 9 mm length were cleaved (containing

two sections) and both mirror facets were AR coated with

670 nm of Y2O3. The individual lasers were bonded

epilayer-side up on a copper submount with indium solder.

Laser sections were driven with independent 100 ns current

pulses at 5% duty cycle. A CW (DC) current was added to

the front or back section to observe tuning behavior. Power

measurements were carried out on a thermoelectric cooler

stage using a calibrated thermopile placed directly in front of

the laser facet. All measurements were performed at a stage

temperature of 298 K. Laser spectra were collected using a

vacuum interferometer (Bruker IFS 66v) at 0.125 cm�1

resolution.

In pure pulsed operation, the laser DCG-SGDFB QCL

threshold current density is measured to be 2.3 kA/cm2, com-

pared to a value of 3 kA/cm2 measured for an unpatterned

laser. Spectral measurements are shown in Figure 4 and were

performed using a pulsed current density of 2.5 kA/cm2 for

both sections. Figure 4(a) shows the tuning behavior for the

laser as a function of the additional CW current density

applied to either the front or back section. Though the exper-

imental tuning range falls short of the theoretical expecta-

tion, 500 nm (236 cm�1) of electrical tuning for a single

QCL was demonstrated, which is significantly higher than

realized previously. Maximum power was not investigated,

but several hundred milliwatts of peak power were observed

at this low current density at most wavelengths. In addition,

Figure 4(b) shows the SMSR, with nearly 30 dB demon-

strated for most wavelengths, with a minimum SMSR of

20 dB.

The next step in this technology will be to understand

better the discrepancy between ideal and experimental tuning

characteristics. There are several primary factors which need

to be considered. First, the DCG reflectivity function is not

FIG. 3. (a) Calculated mirror loss for DCG-SGDFB with different mirror

facet reflectivities. (b) Simulated tuning behavior for DCG-DGDFB QCL

with different mirror facet reflectivities.

FIG. 4. (a) Experimental tuning behavior for DCG-SGDFB QCL with single

layer AR coating. (b) Overlay of emission spectra at various CW current

densities showing wide tuning (500 nm) and good SMSR (>20 dB).

231110-3 Slivken et al. Appl. Phys. Lett. 103, 231110 (2013)

ideal for suppressing laser emission near the peak gain, as

indicated in Figure 2(b). In addition, as the simulation cur-

rently uses a uniform photon density, it completely neglects

spatial hole burning effects. Variations in local reflectivity

which will result, and may also impact SMSR and tuning

range. Finally, it is clear that residual reflections still exist

within the cavity. The simple evidence is that the unpat-

terned laser reference still reaches threshold, as mentioned

above. This can be improved with better fabrication and

anti-reflection coating technology. Once this work is com-

pleted, it should be possible to extend the tuning range sig-

nificantly, up to 350 cm�1 for a standard high efficiency

QCL core design.

In conclusion, this paper has presented the DCG-

SGDFB structure as a means to significantly expand the elec-

trical tuning range over a standard SGDFB QCL.

Simulations were performed to analyze the modified reflec-

tivity function in the presence of optical gain. Mirror loss

and tuning behavior were studied as a function of residual

facet reflectivity. An initial fabrication attempt yielded elec-

trical tuning of 500 nm (236 cm�1) centered around a wave-

length of 4.65 lm. Emission is single mode, with a side

mode suppression of >20 dB.

The authors would like to acknowledge the support, in-

terest, and encouragement of Dr. K. K. Law from the Naval

Air Warfare Center.

1M. Razeghi, The MOCVD Challenge: A survey of GaInAsP-InP andGaInAsP-GaAs for Photonic and Electronic Device Applications,Electronic Materials and Devices, 2nd ed. (CRC Press, 2010).

2A. Hugi, R. Terazzi, Y. Bonetti, A. Wittmann, M. Fischer, M. Beck, J.

Faist, and E. Gini, Appl. Phys. Lett. 95, 061103 (2009).3M. Carras, G. Maisons, B. Simozrag, V. Trinite, M. Brun, G. Grand,

P. Labeye, and S. Nicoletti, Proc. SPIE 8631, 863113 (2013).4B. G. Lee, J. Kansky, A. K. Goyal, C. Pfl€ugl, L. Diehl, M. A. Belkin, A.

Sanchez, and F. Capasso, Opt. Express 17, 16216 (2009).5K. K. Law, R. Shori, J. K. Miller, and S. Sharma, Proc. SPIE 8031,

80312E (2011).6S. Slivken, N. Bandyopadhyay, S. Tsao, S. Nida, Y. Bai, Q. Y. Lu, and M.

Razeghi, Appl. Phys. Lett. 100, 261112 (2012).7S. Slivken, N. Bandyopadhyay, S. Tsao, S. Nida, Y. Bai, Q. Y. Lu, and M.

Razeghi, Proc. SPIE 8631, 86310P (2013).8S. Zhang, J. Meng, S. Guo, L. Wang, and J.-J. He, Opt. Express 21, 13564

(2013).9K. Fujita, S. Furuta, T. Dougakiuchi, A. Sugiyama, T. Edamura, and M.

Yamanishi, Opt. Express 19, 2694 (2011).10Q. Y. Lu, N. Bandyopadhyay, S. Slivken, Y. Bai, and M. Razeghi, Appl.

Phys. Lett. 99, 131106 (2011).11J. Zhao, N. Zhou, Y. Tang, J. Zhao, L. Wang, X. Chen, X. Huang, Y. Yu,

and W. Liu, Semicond. Sci. Technol. 28, 035001 (2013).

231110-4 Slivken et al. Appl. Phys. Lett. 103, 231110 (2013)