Demonstration of III-nitride vertical-cavity surface-emitting lasers with a topside dielectric curved mirror

We report long cavity (65λ) GaN-based vertical-cavity surface-emitting lasers (VCSELs) with a topside dielectric concave mirror, an ion implanted current aperture, and a bottomside nanoporous GaN distributed Bragg reflector. Under pulsed operation, a VCSEL with a 10 μm aperture and a curved mirror with a radius of curvature of 120 μm had a threshold current density of 14 kA cm−2, and a maximum output power of 370 μW for a lasing mode at 404.5 nm. The longitudinal performance has a side-mode suppression ratio of 30 dB up to a current density of approximately 40 kA cm−2. Multiple transverse mode profiles are observed across several devices.


G
aN-based vertical-cavity surface-emitting lasers (GaN VCSELs) have several advantages over conventional edge-emitting laser diodes, including low threshold currents, circular beam profiles, and 2D arraying capabilities, with applications ranging from displays to communications and sensing, among others. 1)[5][6][7][8][9][10][11] However, GaN VCSELs struggle with self-heating due to higher input power requirements and high optical losses from p-GaN and current spreaders, 12) as well as poor heatsinking due to the typically low thermal conductivities of the bottomside distributed Bragg reflectors (DBRs). 13)These issues result in low thermal rollover and low device lifetimes.Additionally, the longitudinal mode spacing for this cavity length regime is approximately 10-20 nm, 14) comparable to the typical width of the gain from the active layer.This means that precise growth control is critical for optimal performance.While there has been recent progress on in situ cavity length monitoring techniques for GaN VCSELs with epitaxial AlInN/GaN DBRs, 15) cavity length precision remains a pressing challenge for other GaN VCSEL architectures.Recently, long VCSEL cavities (L eff > 10 μm) have shown significant promise towards addressing the issues of thermal stability and cavity length control. 16,17)he main benefits of the long cavity are two-fold.First, a larger cavity volume enables generated heat to disperse throughout the structure rather than being concentrated near the active region. 18,19)This consideration is of particular importance for cavity structures that have thermally resistant DBR designs in the thermal path, such as flip-chip bonded dual-dielectric DBR VCSELs and nanoporous DBR (NP DBR) VCSELs. 13,20)The second benefit is that the longitudinal mode spacing is inversely proportional to cavity length, so increasing the cavity length will decrease mode spacing.For example, increasing the short to medium cavity length by a factor of 5 would decrease the mode spacing by an equivalent factor, to 2-4 nm.In this case, cavity control is less important, as the gain region will overlap several longitudinal modes regardless of how the growth rates shift.
For long planar cavities with fixed mirror diameters, diffraction loss increases quickly with cavity length. 21)This effect has been used to suppress higher order transverse modes in other material systems, as higher order modes experience greater rates of diffraction loss. 22)Since the typical gain of GaN QWs is approximately 1% per pass, minimizing this source of loss is critical for long cavity designs.For the recent designs, lateral mode confinement was achieved by introducing a curved concave mirror on one side. 23)Converting one of the planar DBRs into a curved DBR mirror is a recognized method for establishing a stable resonator configuration.This configuration results in the formation of a beam waist on the planar side, effectively reducing diffraction losses.This approach confines the lateral mode, and as a consequence, the dimensions of the beam waist and its propagation within the cavity are chiefly governed by two key parameters, the cavity length (L) and the radius of curvature (ROC) of the mirror.
The initial design demonstrated by Hamaguchi et al. prioritized low threshold conditions by placing the active region at the beam waist (100 nm from the planar mirror) and minimizing aperture size. 14)To form their lens, they polished the substrate down to a thickness of approximately 28 μm and formed lenses with ROC ranging from 37.7 to 56.6 μm on the backside of the polished substrate using thermal reflow of photoresist cylinders. 24)The RMS roughness of the top of the lens was measured to be 0.2 nm, similar to the roughness of a planar single-crystalline GaN substrate.Using this structure, they were able to demonstrate GaN VCSELs with record performances, all while achieving device yields greater than 90%.However, the polished substrate introduces processing complexity, requiring handling of a thin, fragile substrate, and requiring backside alignment of the mirror.It is beneficial to place the lens on top of the VCSEL cavity, thus enabling simple processing techniques without needing to lap or polish the substrate down to the desired cavity thickness.
Recently, we demonstrated a long cavity GaN VCSEL with a topside monolithic GaN lens that lased under CW operation. 25)The cavity length was approximately 11 μm and the GaN mirror had a ROC of 31 μm.The active region was located approximately 6.5 μm from the planar mirror.While the device lased under CW operation, performance was limited by high voltage caused by incomplete activation of the buried tunnel junction (TJ).Additionally, the regrowth of the unintentionally-doped (UID)-GaN/n-GaN that formed the topside lens is suspected to have partially damaged the active region.The total regrowth time was approximately 90 min, and the growth temperature was 900 °C.It has been shown that c-plane In 0.08 Ga 0.92 N layers are resistant to decomposition for annealing up to 15 min at 1000 °C, but experience a 50% decrease in photoluminescence (PL) intensity when annealed for 60 min. 26)We observed a similar phenomenon, where the PL intensity of annealed m-plane In 0.08 Ga 0.92 N QWs dropped by 56% relative to the as-grown sample (Fig. S1, Supporting Information).The mechanisms of thermal degradation of InGaN QWs discussed in the literature suggest that the decomposition of the QWs is initiated by the presence of small voids created by metallic vacancies, and decomposition within these voids is driven by In atoms that diffuse into the relaxed voids. 27)Decomposition of the QWs generates defects that act as non-radiative recombination centers, hurting laser performance. 28)Based on this, it would be advantageous to replace the monolithic GaN mirror with a non-absorbing material that can be deposited and processed under standard temperatures.
Here, we present a long cavity (L eff ~70λ) m-plane GaN VCSEL with a topside SiO 2 concave mirror, an ion implanted current aperture (IIA), and a buried planar nanoporous GaN distributed Bragg reflector (NP DBR).The SiO 2 mirror has a diameter of 26 μm and an ROC of 120 μm.The active region is placed approximately 7 μm from the beam waist.
The epitaxial device structure, shown in Fig. 1(a), was grown and fabricated similarly to Ref. 25, and will be briefly described below.The epitaxial layers were grown using atmospheric metalorganic CVD (MOCVD) on free-standing double-side polished m-plane (101̅ 0) substrates with an intentional 1°miscut in the [0001̅ ] direction.The epitaxial structure consisted of a 21-period UID GaN and n+-GaN superlattice to form the bottomside NP DBR, UID GaN, n-GaN, 2´InGaN quantum wells, and a p-GaN layer.After the first growth, a Ti/Au hard mask was deposited to define the aperture during ion implantation.Al ions were implanted with a dose of 10 15 cm −2 and an acceleration voltage of 20 kV.After removing the hard mask with heated aqua regia, the sample was cleaned with concentrated HF and ozone before regrowing the 8 nm n++-GaN TJ layer and 150 nm n-GaN current spreader by MOCVD. 29)Mesa structures and trenches were formed by reactive ion etching.Following this, 2400 nm SiO 2 was deposited via ion beam deposition.Then, photoresist lenses were formed via photoresist reflow and transferred into the SiO 2 by inductively coupled plasma etching using a CF 4 :O 2 :CHF 3 gas mixture of 3:3:1. 24)fter, the NP DBR was electrochemically etched, followed by deposition of metal contacts and a 16-period dielectric DBR.A cross-section of a completed device, taken using focused ion beam and imaged by scanning electron microscopy (SEM), is shown in Fig. 1(b).A close-up SEM image of the NP DBR is shown in the inset of Fig. 1(b).Electrical characteristics were analyzed under pulsed operation with a 500 ns pulse width and a 0.5% duty cycle at room temperature (20 °C).Optical power measurements were taken by placing the sample 7 mm above a calibrated 3 mm diameter biased silicon photodetector.Spectrum data was acquired with an Ocean Insight HR4Pro spectrometer with a spectral resolution of 0.2 nm.Topside nearfield patterns (NFP) were taken using an optical microscope with a 20x objective lens, and the bottomside farfield pattern (FFP) was taken using a Thorlabs goniometric stage with a 2.54 cm point to rotation.
The RMS roughness at the final regrowth surface was approximately 0.8 nm, measured by atomic force microscopy (AFM).After depositing the SiO 2 the RMS roughness remained near 0.8 nm.The fabricated devices had an effective cavity length of approximately 65λ for a target emission wavelength of 405 nm.A NP DBR porosity of 31% was determined from cross-sectional SEM images, leading to a calculated peak reflectance of 99.673%.
The light-current-voltage (LIV) characteristics of a 10 μm aperture VCSEL were analyzed under pulsed operation with a 500 ns pulse width and a 0.5% duty cycle, shown in Fig. 2(a).The threshold current density was 14 kA cm −2 , and the maximum output power was 370 μW for a lasing mode at 404.5 nm with a spectrometer resolution-limited linewidth of 0.22 nm.The slope efficiency was 0.02 W/A, leading to a differential efficiency (η d ) of 0.64%.As shown in Fig. 2(b), the device exhibited a side-mode suppression ratio (SMSR) of 30 dB down to the detection floor of the spectrometer, up to a current density of approximately 40 kA cm −2 .Above 40 kA cm −2 , a second mode appears at 401.8 nm, with a mode spacing of 2.7 nm.The unexpected SMSR might be influenced by the narrow stopband of the NP DBR, which can be seen in the inset of Fig. 2(b).The calculated stopband width of 21 nm, and fall-off at 410 nm, prohibited half of the active region emission spectrum, which was centered at 405 nm, from experiencing meaningful gain.It is unknown at this time why the threshold for adjacent modes was so high, but similar single mode behavior was also observed by Ito et al. for a 3 μm aperture VCSEL with a cavity length of 25 μm and a lens with an ROC of 33 μm. 30)It has also been established that cavity control structures, such as filtering mirrors, can provide longitudinal mode control in long cavities. 31)ompared to previous work with a monolithic GaN lens, the SiO 2 lens creates an additional planar interface with the GaN epitaxial surface that affects the beam propagation through the cavity.For this initial demonstration, two simple models were constructed using matrix elements from Tables 2-1 in Optical Electronics by Yariv,32) and a Gaussian beam was propagated using ABCD matrix formalism.With this approach, the GaN/SiO 2 interface becomes an interface matrix element that introduces refraction.The beam waist equation, from Eqs. (2.5)-( 8) and (2.5)-( 13), is: where z 0 is the confocal parameter.Propagating the beam modifies z 0 , and the new expression for the beam waist becomes: where A, B, C, and D are matrix elements found by propagating the beam through the cavity structure.It is found with this method that a cavity with a monolithic GaN lens has a beam waist of 1.33 μm, while a cavity with an SiO 2 lens has a beam waist of 1.56 μm, 17% wider than the GaN lens case.Using Eq. ( 2) from Ref. 24 translates to a beam diameter at the active region of 7.88 μm for the GaN lens, and 9.13 μm for the SiO 2 lens, a 15.8% increase.Assuming no other effects, this would improve the coupling efficiency of the 10 μm current aperture into the mode from 62.4% to 83.3%.The coupling is also limited by the alignment between the current aperture and lens, which is approximately ±1 μm.
Topside NFPs for several 10 μm VCSELs can be seen in Figs.3(a)-3(d).The modes are similar to higher order mode profiles that can be calculated using a Laguerre-Gaussian model, 33) which is attributed to the circular symmetry of the lens.Figure 3(e) shows the bottomside FFP for the device in Fig. 3(b).The higher order transverse mode behavior is in line with findings observed by Nakajima et al for a VCSEL with a similar lens ROC. 16)The similarity between the topside NFP and bottomside FFP is promising preliminary evidence that the large irregular voids in the bottomside NP  DBR (Fig. S2, Supporting Information) may not contribute to scattering of the farfield. 34)uring testing, we observed that the modes in Figs.3(b) and 3(c) were off-center from the peak of the lens.SEM imaging revealed the presence of pits across the surface of the lenses, shown in Fig. 4 with a white dotted circle denoting the approximate position of each mode seen in the topside NFP.Measurements on the pitted surface using AFM revealed that the pits were approximately 10-50 nm deep, and 300 nm wide.The increased scattering loss caused by these pits provides one possible reason for the location of the lasing region on the VCSELs and the higher order modes.The evidence for this is that the lasing spots generally occurred in areas with a lower pit density.However, it is possible that the higher order transverse modes are caused by non-uniform current injection into the active region due to non-uniform activation of the IIA TJ, an issue exacerbated by the wider 10 μm current aperture. 35,36)The pits are also the current theory as to why the threshold current densities for the reported VCSEL are higher than what was achieved by the authors with pulsed operation of the GaN lens VCSEL (6.7 kA cm −2 ), 25) despite the reduction in regrowth time.Regardless, the pitting in the lenses is a subject of immediate study, with the goal of total removal from future dielectric lens processing.
In summary, we reported 65λ-cavity GaN VCSELs utilizing a topside dielectric curved mirror.For a device with a 10 μm current aperture and a curved mirror with a 120 μm ROC, single longitudinal mode operation with an SMSR of 30 dB up to 3x J th was observed.Devices lased in a variety of higher order transverse modes, with mode behavior possibly influenced by pits introduced during the lens dry etch.The SiO 2 lens fabrication process offers advantages over prior long cavity curved mirror VCSEL demonstrations, as it can be fabricated using standard techniques and at room temperature.These results show promise for the expanded capabilities of GaN-based VCSELs.

Fig. 2 .
Fig. 2. (a) LIV for pulsed operation of a 10 μm VCSEL.(b) Pulsed emission spectrum of the device above J th showing an SMSR of ~30 dB (black) and the emergence of an additional mode at higher J (red).Mode denoted by black arrow.The inset shows the measured reflectivity of the bottomside NP DBR.

Fig. 3 .
Fig. 3. Optical microscopy images of representative 10 μm VCSEL (a) below J th (b)-(d) above J th .(e) Bottomside FFP of (b) measured at a current density of 27 kA cm −2 .Measurement taken along the c-direction.