1. Introduction

Vertical cavity surface emitting lasers (VCSELs) have many applications that need high-power single optical-mode operation. High-power VCSELs require a large aperture size, allowing several transverse modes to lase. Smaller aperture size leads to single mode operation, but at the expense of reduced output power. To obtain higher single mode output power, different approaches of optical cavity engineering of VCSELs were employed. They include lateral patterning of the top distributed Bragg reflectors (DBRs) by shallow surface etching [1, 2], by metal deposition [3,4], by etching of hole structures [5–9] or by ion indiffusion [10]. Combinations of oxide-defined and implanted apertures [11,12] were also employed in order to position the pattern closer to the active region of the cavity. More sophisticated refractive index patterns have been inserted into the optical cavities by selective etching of an intra-caity layer between two overgrowth steps, realizing antiguided VCSELs [13] and coupled VCSEL arrays [14, 15]. More recently, techniques for patterning the optical gain distribution using buried tunnel junction (TJ) structures have been introduced [16]. This approach has yielded high performance single mode VCSELs, operating at telecommunication wavelengths, fabricated using double wafer fused DBRs mirrors [17] or dielectric mirrors [18]. Combining the buried TJ approach with a separate refractive-index patterning close to the cavity core has not been implemented yet, but would be very powerful for optimally designing the optical mode pattern and gain in such devices.

This is precisely what we report here, using double-wafer-fused, buried TJ long wavelength VCSELs. The gain profile is defined by patterning a buried TJ layer used for carrier injection and the refractive index distribution is tailored by adding sub-wavelength air gaps into the cavity. The combined patterning is employed for improving the mode pattern definition and discrimination, and opens up new ways for implementing new designs of VCSEL optical cavities.

2. Intra-cavity patterning

Figure 1 illustrates schematically the structure of the patterned-cavity VCSEL arrays. The devices were fabricated using organometallic vapor phase epitaxy of the gain In-GaAs/InAlGaAs/InP multiple quantum well (QW) region and the GaAs/AlGaAs DBR mirrors, followed by patterning and double wafer fusion [17]. Optical gain is provided by a stack of six strained InGaAs/InAlGaAs QWs grown on a planar (100)-InP substrate with parameters optimized for emission near 1320 nm wavelength [19]. The tailoring of the gain and refractive index distributions were implemented on the InP-based cavity core using two patterning steps. First, a square buried TJ mesa of side length s_TJ = 10μm was defined using electron beam lithography and wet chemical etching [20], followed by regrowth of another stack of InP/GaInAsP layers. The patterned TJ mesa localizes the injected carriers and thus defines the lateral extent of the gain region. The additional stack of regrown layers provides the level at which the refractive index is patterned within this cavity region. Here it is provided by a set of four squares of side length s_P = 5μm that define four VCSEL apertures in a 2x2 array configuration, located at the center of the TJ mesa. Several otherwise identical structures with different array pixel separations g = 100, 250, 500 and 750 nm were fabricated; a square border of width b = 3μm was also introduced, defining the overall size of the array matrix (see Fig. 1). The pixel separations and array boundary were realized using a SiO₂ mask defined by electron beam lithography, through which 60 nm deep airgaps were etched using inductively coupled plasma (ICP). The TJ pattern and the array matrix were mutually aligned (accuracy of < 100nm) by using common alignment marks in the electron beam lithography steps. Since intra-cavity pattern are supposed to have a strong impact on the cavity mode, the etched patterned was placed close to a node position of the electric field, see Fig. 1(b).

 

Fig. 1 Left panel: 2×2 VCSEL array structure with definition of gain (square tunnel aperture of side length s_TJ = 10μm) and refractive index profile (pixel of side length s_P = 5μm) on separate vertical layers. Right panel: Vertical refractive index profile and electric field distribution inside a plain not patterned cavity.

Download Full Size | PDF

3. Measurement Results and Discussion

Typical light-voltage versus current characteristics (cw operation at room temperature) of a 2x2 array structure with a pixel separation of g = 250nm is depicted in Fig. 2(a). The device exhibits a threshold current I_th = 3.4mA and maximum output power of 1.6 mW at 17.5 mA. The maximum output power of the cavity patterned devices is still lower than comparable devices with only TJ patterning [17,19]. This is probably due to increased scattering introduced by the refractive index pattern. Further optimization is needed to increase output power.

 

Fig. 2 (a) Light-voltage versus current characteristic of a coupled 2x2 VCSEL array with a pixel separation of g = 250 nm. (b) Corresponding near field (left column), far field (middle column) and emission spectra (right column) for injection currents of 1, 3, 5, and 14 mW.

Download Full Size | PDF

The near field- and far field images acquired with a CCD camera and the emission spectra measured with an optical spectrum analyser (HP70950A) are presented in Fig. 2(b). At an injection current of 1 mA (below threshold) the near field shows the amplified spontaneous emission, which already reflects the defined array pixels. This pixelation becomes more pronounced close to lasing threshold. Due to the asymmetric electrical contacting (see Fig. 1), the VCSEL array reveals slightly asymmetric intensity distributions in the near field images below and above threshold. Above the lasing threshold the array is coupled in the out-of-phase supermode [4, 21, 22], for which the electric field of neighboring pixels has a phase shift of π, yielding the four lobes in the far field intensity image [19]. This particular array lases at the same (highest order) supermode from threshold to thermal roll-over, with the other transverse modes being at least 25 dB down in intensity. The red shift of the dominant supermode between threshold and 14 mA current is due to heating of the active region (2.7 nm red shift, which correspond to about 27°C degrees increase in temperature). The devices remained coupled out-of-phase also for pumping currents exceeding thermal roll-over. Since for most applications in-phase emission is desirable, an additional phase shift layer could be used in order to obtain Gaussian-like beam patterns as demonstrated in Ref. [23].

The emission spectra of two nominally identical devices from different wafer locations with a pixel separation of g = 250 nm, measured at a pumping current of 14 mA, are shown in Fig. 3(a) and 3(b). Whereas device (a) shows a side mode suppression ratio of about 26 dB, for device (b) the ratio is less than 6 dB. This difference could stem from device variations due to non uniformities in epitaxial layer thickness / composition as well as inadvertent pattern variations. In order to gain insight into this characteristic, the near field image was spectrally resolved by scanning the imaged near field plane with an optical fiber connected to a high-resolution (100 pm) optical spectrometer. The extracted near field mode patterns thus obtained are displayed in Fig. 3. The intensity pattern of the main mode m₁ exhibits four lobes and corresponds for both devices well with the patterned array pixels, which are indicated by dashed lines superimposed on the patterns. It represents the out-of-phase supermode coupling of the fundamental modes m′₁ of the single pixels. The lower-order supermodes built out of the m′₁ pixel mode, expected to oscillate at longer wavelengths, are suppressed. This is most likely because they suffer larger scattering losses at the etched airgaps between pixels, where their intensity is larger than that of the highest order supermode.

 

Fig. 3 Spectra and near field intensity distributions of the oscillating modes for two nominally identical devices (a) and (b) from different locations on the wafer with a pixel separation of g = 250 nm at an injection current of 14 mA. The near field intensity distributions were plotted in linear scale, using the full colour range shown.

Download Full Size | PDF

The relatively strong pixel confinement induced by the etched airgaps yields higher-order modes confined within each pixel, which we label here m′_l, with l = 2, 3,... Such higher order pixel mode, also coupled in the highest order supermode configuration, is labeled m_h2 in Fig. 3(a), and lases at 1308.8 nm. A similar mode, labeled m_d2, is observed in the spectrum of device (b), near 1310.5 nm wavelength. This mode is composed this time of diagonal sub-peak configuration, as revealed in the corresponding spectrally-resolved near field patterns. For the m_d2 mode, adjacent pixels couple in mirror-symmetrical pixel configuration, yielding a peculiar ring field pattern. Coupled modes made of still higher order pixel modes oscillate at shorter wavelengths, near 1306.2 nm and 1308.4 nm for device (a) and (b), respectively. These modes, labeled m₃ in Fig. 3, consist of coupled pixel modes showing three intensity peaks each, and extend well beyond the outer airgaps definition. For device (a) an additional lower order super-mode at 1311.8 nm appears.

The analysis of Fig. 3 sheds light on the mode selection mechanism. The modal gain is determined by the overlap of the field patterns and the gain region. The modal losses are determined largely by scattering off the airgap structures introduced at the refractive index patterning level. The weaker discrimination against the higher-order pixel modes in device b) may result from the spatial profile of the gain, which better overlaps the higher order pixel mode in this particular case. The higher order modes m₃ are always well suppressed, thanks to the loose confinement of the corresponding pixel mode.

The array structures with gap width of 100 nm showed similar near and far field behavior as for the 250 nm separation, presented in Fig. 2, except that a higher number of modes was observed in the spectrum. This is attributed to a weaker pixel definition. On the other hand, for a pixel separation of 750 nm, the array pixels were not coupled anymore. This can be explained by a too small penetration of the evanescent waves of the pixel modes, which yields localization at the different pixels due to optical disorder [24].

4. Conclusions

In conclusion, the larger mode discrimination and the better conformity of the mode pattern to the pixelation introduced by the air gaps, as compared to solely TJ patterning [19], illustrates the advantage of the cavity patterning approach introduced here. However, further optimization of the structure is necessary in order to increase the single mode power while keeping a large spatial mode discrimination. For this purpose, the contrast of the refractive index patterning should be optimized to achieve the desired pixel definition while keeping the scattering losses low. The gain distribution should also be tailored by the separate tunnel junction definition to favor the desired mode. The same cavity patterning approach could be used for other purposes, e.g., for controlling the spontaneous emission into the desired mode using an appropriate two-dimensional photonic crystal structure at the index patterning level.