High‐Performance Thin‐Film VCSELs Integrated with a Copper‐Plated Heatsink

High‐performance continuous‐wave (CW) vertical‐cavity surface‐emitting lasers (VCSELs) rely on efficient thermal management for which top‐emitting 930 nm thin‐film VCSELs are integrated with a copper‐plated heatsink by using a double‐transfer technique, exhibiting the low‐power consumption, high‐power, and temperature‐stable VCSEL operation. In this study, the top‐emitting 930 nm thin‐film VCSEL structures, including the highly n‐doped GaAs ohmic and lattice‐matched InGaP etch‐stop layers, are epitaxially grown via a low‐pressure metalorganic chemical vapor deposition (LP‐MOCVD) system. The electrical and optical properties of the substrate‐removal thin‐film VCSELs are investigated under CW operation, compared to those of the bulk‐type VCSELs onto the n‐GaAs substrates. The differential series resistance (85.92 Ω) of the thin‐film VCSEL is 9.16% lower than 94.59 Ω of the bulk‐type VCSEL onto the n‐GaAs substrates. The thermal resistance (607 K W−1) of the thin‐film VCSEL is 46.33% lower than 1131 K W−1 of the bulk‐type VCSEL, by which the maximum peak power (11.70 mW) of the thin‐film VCSEL at 24 mA is 12.07% higher than 10.44 mW of the bulk‐type VCSEL at 22.40 mA under room temperature (25 °C).


Introduction
Vertical-cavity surface-emitting lasers (VCSELs) are profitably exploited for energy saving high-speed optical data communication and 3D sensor applications owing to their favorable lowpower consumption, small divergence beam profile, and single mode operation that enable efficient optical coupling with other systems. [1][2][3][4] Especially, high-power and highly reliable VCSELs DOI: 10.1002/admi. 202300191 have gained much attention in that they can deliver the further improved benefits in terms of low spectral shift with temperature and long-range distance communication, satisfying the tremendous demands on the industrial and automotive markets. A myriad of applications such as data center, time-of-flight (ToF) modules, and photonic systems in optical interconnects (OIs) require the temperaturestable VCSEL operation with tight control over the lasing spectrum and the reliably stable device operation under an elevated temperature induced by temperature rise in OIs up to over 85°C. [5][6][7] More recently, the long-range light detection and ranging (Lidar) systems in combination with the ToF technique have driven the high-power VCSELs with the higher slope efficiency, gain volume, and power density, providing the advanced features such as the fully autonomous vehicles equipped with adaptive cruise control (ACC) capabilities. [8,9] However, the most significant hurdle to the high-power and reliable VCSELs is a self-heating effect around the active region under high-current operation, which deteriorates the achievable maximum device performance with premature rollover current as well as device reliability due to the thermally-induced unstable emission spectrum. [10][11][12][13][14][15] It is crucial to manage the self-heating effect during the device operation in that the temperature rise around the active region induced by the current injection is much higher than the ambient temperatures. [16] The device lifetime could be significantly limited by a factor of two for every temperature rise of 10°C. [14] In order to mitigate the self-heating effect, many efforts have been devoted to thermal management for which the sidewall heatsinks covered with the electroplated copper (Cu) and gold (Au), [17,18] flip-chip bonding onto the heatsinks with good thermal characteristics, [19][20][21] and the lapped-GaAs substrates via chemical mechanical polishing (CMP) processes have been presented. The mechanically thinned 10 μm thick thin-film VCSELs showed the lowered thermal resistance (R th ) up to 40% compared to the 200 μm thick bulk-type VCSELs. [12] But, there is still a room for the further enhanced thermal management in order to maximize an efficient heat passage around the active region during the VCSEL operation.
The substrate-removal thin-film VCSELs can be favorably utilized for the high-power and reliable VCSELs in a way of www.advancedsciencenews.com www.advmatinterfaces.de transferring the thin-film VCSEL structures onto the heatsinks with higher thermal and electrical conductivity than the conventionally used GaAs substrates. There have been numerous attempts directed toward transferring the thin-film VCSELs onto the heatsinks such as AlN, [22,23] diamond, [24] and Al [25] heatsinks at the expense of the GaAs substrates, possibly contributing to not only high-power VCSEL operation with the extended rollover current but also reliable VCSEL operation with the tightly controlled emission spectrum. Moreover, the electrical loss could be alleviated with the highly doped GaAs ohmic layer instead of the considerably thick and moderately doped GaAs substrates, providing the high-performance thin-film VCSELs with the low-power consumption. [23] But, the thin-film VCSELs transferred onto the heatsinks using an adhesive material could generate the irregular surfaces and voids induced by the ill-chosen bonding conditions, making a rather high series resistance and steeper upward tilt to the slope of the I-V curve. [25,26] Furthermore, the existence of the voids at the bonding interfaces could suppress the efficient heat transfer of the thin-film VCSELs to the heatsinks. [27] In the present paper, top-emitting 930 nm thin-film VCSEL structures, including the lattice-matched InGaP etch-stop and highly n-doped GaAs ohmic layers, are epitaxially grown on an exact-oriented n-GaAs substrate (001) via a low-pressure metalorganic chemical vapor deposition (LP-MOCVD) system. The VCSELs, operating at the prominent dip of solar spectrum in the range of 920-940 nm, are favorably exploited for 3D sensing systems and modern lidar with the TOF technique due to the relatively weak background noise caused by the ambient solar light. [28] The lattice-matched InGaP etch-stop layer prevents the ingress of acid etchant into the thin-film VCSEL structures during removal of the GaAs substrates, by which the highperformance (CW) thin-film VCSELs are directly integrated with the copper-plated heatsinks by using a double-transfer technique that transfers the thin-film VCSEL stuctures onto the foreign substrates twice in absence of the adhesive materials. The copperbased heatsinks with thermal conductivity of 3.17 W cm −1 K, 5.87 times higher than that of a GaAs substrate (0.54 W cm −1 K), [20] promises high-power and reliable CW VCSEL operation due to the narrower wavelength change over temperature rise as well as the extended rollover current with the lowered thermal resistance. The light-current-voltage (L-I-V) curves and emission peak spectrum of the fabricated thin-film VCSELs onto the copper-plated heatsinks, compared to those of bulk-type VCSELs onto the n-GaAs substrates, are investigated as a function of the ambient temperatures in the range of 25-85°C under CW operation. Figure 1a shows the top-emitting 930 nm thin-film VCSEL structures on the exact-oriented (001) n-GaAs substrates using the LP-MOCVD system. First, the n-GaAs buffer layer was grown in order to have the following high-quality epi structures in which there were an etch-stop layer, highly doped n, p-ohmic GaAs layers, and multi-quantum wells (MQWs) inserted between the bottom n-type and the top p-type distributed Bragg reflectors (DBRs). The lattice-matched n-InGaP layer played a crucial role of an etch-stop layer that prevented the ingress of NH 4 OH-based etchant into the active region during removal of the 350 μm thick n-GaAs substrates. The highly n-doped GaAs ohmic layer was exposed after removing the InGaP etch-stop layer with the HClbased etchant. The electrical loss originated from the considerably thick and moderately doped n-GaAs substrates could be mitigated due to the highly n-doped GaAs ohmic layer that contributes to the enhanced electrical characteristics of the thin-film VCSELs. On top of the n-GaAs ohmic layer, the bottom 38-pair n-DBRs were grown in order to exhibit the maximal reflectivity, whereas 20-pair p-type DBRs were designed to ensure the light emission of the top-emitting thin-film VCSELs. The three-pair InGaAs/GaAsP MQWs inserted between n, and p-DBRs showed the spectral reflectance spectrum with a sharp reflectance dip near 930 nm. The cavity mode (CM) dip in the reflectance spectrum of thin-film VCSELs is centered at 927.5 nm, as shown in Supporting Information Figure S1(Supporting Information). The 10 μm oxide-confined aperture size of the thin-film VCSELs was defined by employing the selective oxidation of Al x Ga 1−x As with high Al fractions above the MQWs. Figure 1b shows that the top-emitting 930 nm thin-film VCSELs with the p-on-n polarity are directly integrated with the copper-plated heatsinks by using a double-transfer technique that transfers the epi structures onto the foreign substrates twice in absence of an adhesive material, as shown in Supporting Information Figure S2 (Supporting Information) and see experimental section. The double-transfer technique identified in the previous research does not require the inverted growth for the top-emitting thin-film VCSELs that could raise the critical issues of a large number of DBR periods for achieving the necessary high reflectivity. [25,29] Figure 1c shows a photograph of the thin-film VCSEL array and a microscopic image of the thin-film VCSEL onto the copper-plated heatsinks. The ≈25 μm thick copper-plated heatsinks enabled the free-standing thin-film VCSEL array without an additional carrier. The 37 μm thickness of the substrate-removal thin-film VCSELs onto the copper-plated heatsinks is measured using a Digimatic indicator, as shown in Figure S3a (Supporting Information). The relatively flexible and compliant thin-film VCSELs onto the copper-plated heatsinks can be strongly favored in the well-established flexiblecircuit technology. [30,31] Figure 1d shows the 10 × 10 μm 2 atomic force microscopy (AFM) images of thin-film VCSEL wafer, and the aperture area in the thin-film VCSEL, respectively. The thinfilm VCSEL wafer showed smooth surface with a root-meansquare (RMS) value of 0.25 nm. The RMS value in the aperture area of the fabricated thin-film VCSEL was 1.77 nm that was relatively rough due to small pits and ridges. This can be attributed to the acid etchants used for the device fabrication. But the RMS value of 1.77 nm in the aperture of the thin-film VCSEL was still good enough. Figure 1e shows the focused ion beam/scanning electron microscopy (FIB/SEM) top-view image of the thin-film VCSELs onto the copper-plated heatsinks, by which the emission aperture with the ring electrodes, and the p-contact pad are shown. The p-contact pad was utilized in order to measure the fabricated thin-film VCSEL performance with a probe station. The 10 μm oxide-confined aperture size of the thin-film VCSELs was defined in the emission aperture. The cross-sectional SEM image obtained by the FIB milling is shown in Figure 1f where p, n-ohmic GaAs layers, p, n-DBRs, MQWs, and oxide aperture are clearly visible. It is worth noting that the smooth interface   between the n-GaAs ohmic layer and n-ohmic metal layers with a combination of Ni/AuGe/Ni/Au is obtained using the latticematched InGaP etch-stop layer that contributes to the highly selective etching of the n-GaAs substrates. Furthermore, the integrated copper-plated heatsinks removed the adverse effects on the device performance caused by the irregular bonding interfaces and unwanted voids due to the ill-chosen bonding materials and processes. Figure 1g shows the cross-sectional scanning transmission electron microscopy (STEM) image of the multiple top p-DBRs and bottom n-DBRs for exhibiting the appropriate reflectivity as well as the MQWs for engineering the lasing wavelength. The Al x Ga 1−x As DBRs with the alternatively composed high and low Al fractions were clearly distinguished in the top and bottom DBRs, respectively. The active region sandwitched between the well-defined top p-DBRs and the bottom n-DBRs is closely observed with the STEM measurement, as shown in Figure 1h. There was a distinct contrast at the interfaces of threepair MQWs inserted between AlGaAs confinement layers. The bright and dark region of the MQWs corresponded to the InGaAs wells and GaAsP barriers, respectively.  Figure 2b shows a microscopy image and spectroscopy of the emitting light obtained from the top-emitting 930 nm thinfilm VCSELs onto the copper-plated heatsinks. The microscopy image of the top-emitting thin-film VCSEL was obtained with a probe station, by which the invisible infrared emission was observed as a purple beam in the aperture area of thin-film VC-SEL. The peak wavelength of the thin-film VCSEL was 926.3 nm at a bias current of 8 mA, which corresponded to the sharp reflectance dip near 927.5 nm, as shown in Figure S1 (Supporting Information). Figure 2c shows L-I-V characteristics of the thinfilm VCSEL directly integrated with the copper-plated heatsinks, compared to those of the bulk-type VCSEL onto the n-GaAs substrates, under CW operation at room temperature (25°C). The photographs of the bulk-type VCSELs onto the thick n-GaAs substrates are shown in Supporting Information Figure S3b (Supporting Information) where the 357 μm thickness of the bulktype VCSELs are measured using the Digimatic indicator. The red (thin-film VCSEL) and black (bulk-type VCSEL) lines were voltage (left y-axis) and the short dashes (thin-film VCSEL), and dots (bulk-type VCSEL) were optical power (right y-axis). Table  1 summarizes the detailed parameters of the thin-film VCSEL compared to those of the bulk-type VCSEL. The threshold current (I th ) of the thin-film VCSEL was 1.45 mA that was similar This can be attributed to the series-resistance reduction due to the highly n-doped GaAs ohmic layer at the expense of the considerably thick and moderately doped n-GaAs substrates. The differential series resistance of the thin-film VCSEL was ≈85.92 Ω that was lower than that of the bulk-type VCSEL of 94.59 Ω, enabling the thin-film VCSEL operation with the low-power consumption. The maximum peak power (11.70 mW) of the thinfilm VCSEL at 24 mA was markedly improved by 12.07%, compared to 10.44 mW of the bulk-type VCSEL at 22.40 mA. The emission features of the thin-film or bulk-type VCSELs with the 10 μm oxide-confined aperture sizes were uniform. The average peak power of the thin-film and bulk-type VCSELs was 10.84 ± 0.4 and 12.14 ± 0.44 mW, respectively. The extended rollover current can be related to mitigation of the self-heating effect around the active region in the thin-film VCSEL in a way of conveying the heat to the copper-plated heatsinks efficiently. Figure 2d shows the wavelength shift (Δ ) in accordance with the dissipated electrical input power (ΔP diss ) under CW operation, by which the Δ /ΔP diss value of the fabricated thin-film VC-SEL was 0.0422 nm mW −1 . The steeper upward tilt to the slope of the wavelength shift in respect to ΔP diss was obtained with the bulk-type VCSEL of 0.0779 nm mW −1 . The efficient heat passage created by the copper-plated heatsink instead of the n-GaAs substrates enabled the narrower wavelength range of the thin-film VCSELs over temperature rise around the active region induced by the electrical input power.

Thermal Resistance (R th ) of Thin-Film VCSELs
The device performance of the top-emitting 930 nm thin-film VC-SELs was investigated with the varied ambient temperatures in the range of 25-85°C, by which the thermal resistance (R th ) was extracted by using an Equation 1 as followed. The Δ /ΔP diss is the output wavelength shift in accordance with the temperature rise around the active region induced by the current injection. The Δ /ΔT is the output wavelength shift as a function of the ambient temperatures. The thin-film VCSELs were loaded onto the temperature-controlled stage with that the device parameters of the thin-film VCSEL as a function of the ambient temperatures are taken in Table 2. Figure 3a shows the L-I-V curves of the thinfilm VCSEL onto the copper-plated heatsinks under CW operation at the varied ambient temperatures from 25 to 85°C in steps of 15°C. The peak power of the thin-film VCSEL was markedly decreased from 11.70 to 7.13 mW, together with the premature rollover current from 24 to 19.1 mA, in accordance with the increased temperatures from 25 to 85°C. Figure 3b shows the I rollover and I th of the thin-film VCSEL with the ambient temperature rise. The downward trend of the rollover current from 24 to 19.1 mA was observed in accordance with the increased temperatures up to 85°C. There were two different trends in the I th of the thin-film VCSELs as a function of the ambient temperatures.
The I th was gradually decreased at the increased ambient temperatures from 25 to 55°C. This can be attributed to the bandgap reduction of MQWs in the active region at the elevated ambient temperatures, resulting in the reduced I th for lasing action in the thin-film VCSELs. [32] Whereas, the I th was markedly increased above the temperature point of 55°C owing to the excessively increased leakage current caused by the further reduced bandgaps with the ambient temperature rise. [13] In fact, the I th was dramatically increased from 1.42 (70°C) to 1.73 mA (85°C). The output spectrum peak positions of the fabricated thin-film VCSEL are gradually redshifted from 926.3 to 930.25 nm in accordance with the increased ambient temperatures up to 85°C, as shown in Figure 3c. The thermal wavelength shift as a function of the ambient temperatures is shown in Figure 3d where the Δ /ΔT value of the thin-film VCSEL is 0.0695 nm K −1 . This was quite the same as 0.0689 nm K −1 of the bulk-type VCSEL, as shown in Figure S4d (Supporting Information). The difference could be attributed to the temperature-controlled stage with accuracy of ±0.1°C provided by the temperature bath. Their Δ /ΔT values were similar to the typical wavelength shifts of 0.07 nm K −1 . [33] The extracted R th value (607 K W −1 ) of the fabricated thin-film VCSEL was obtained from Equation 1. The experimentally measured thermal resistance of the thin-film VCSEL was 46.33% lower than 1131 K W −1 of the bulk-type VCSEL onto the n-GaAs substrates, as shown in Figure S4 (Supporting Information). The lowered thermal resistance enabled the enhanced peak power of  the thin-film VCSEls onto the copper-plated heatsinks with the extended rollover current. Figure 4a shows the increased T active with respect to the current injection of the thin-film and bulk-type VCSELs at room temperature, respectively. The temperature rise around the active region induced by the current injection is investigated with Equation 2.

Temperature-Stable CW Thin-Film VCSELs
The experimentally extracted T active is the temperature around the active region induced by the current injection, the T ambient is the stage temperature for measuring the thin-film VCSEL performance, and the R th value is extracted by measuring the L-I-V curves of the thin-film VCSELs with the varied temperatures identified above. The T active of the bulk-type VCSEL showed the steeper curve than that of the thin-film VCSEL due to the higher thermal resistance. In fact, the temperature rise around the active region was higher than the ambient temperatures, which can be linked to the adverse effects on the bulk-type VCSEL with the premature rollover current. As the injection current was increased, the T active gap between the thin-film VCSELs and bulk-type VC-SELs was markedly enlarged. The T active value (65.96°C) of the thin-film VCSEL at the rollover current of 24 mA was <97.55°C of the bulk-type VCSEL at rollover current of 22.4 mA. This contributed to the narrower wavelength range of the thin-film VC-SELs over temperature rise induced by the electrical input power, enabling the more reliable CW thin-film VCSELs with tight wavelength control than the bulk-type VCSELs. Figure 4b shows comparison of the downward trends in I rollover of the thin-film and bulk-type VCSELs in accordance with the temperature rise. The dramatically reduced rollover current of the thin-film and bulk-type VCSELs was observed in the increased temperatures from 40 to 70°C, respectively. But the gentle downward slopes of the I rollover in the thin-film VCSEL were obtained compared to those of the bulk-type VCSEL due to the alleviation of self-heating effect in the thin-film VCSEL, being capable of keeping the high-power VCSEL operation at the elevated temperatures. The copper-plated heatsinks ensured the efficient heat passage that managed the self-heating effects of the thin-film VCSELs during the device operation, showing the gentle downward slope of the I rollover as well as the 31.6°C reduction in the active region at the maximum peak power.

Conclusion
The top-emitting 930 nm thin-film VCSEL structures, including the lattice-matched InGaP etch-stop and highly n-doped GaAs layers, were epitaxially grown via the LP-MOCVD system. The highly selective etching of the thick n-GaAs substrates was carried out by employing the lattice-matched InGaP etch-stop layer that played a crucial role of achieving the smooth interfaces between the back metals and n-GaAs ohmic layers. The substrateremoval thin-film VCSELs were integrated directly onto the copper-plated heatsinks by using the double-transfer technique that transferred the epi structures onto the foreign carriers twice in absence of an adhesive material. The top-emitting 930 nm thin-film VCSELs onto the copper-plated heatsinks, compared to the bulk-type VCSELs onto the n-GaAs substrates, showed the low-power consumption, high-power, and temperature-stable operation. The differential series resistance of the substrateremoval thin-film VCSEL was ≈85.92 Ω that was lower than that of the bulk-type VCSEL of 94.59 Ω, enabling the thinfilm CW VCSEL operation with the low-power consumption. The narrower wavelength ranges of the thin-film VCSELs in respect to the dissipated electrical input power were observed, enabling the reliable CW operation of the thin-film VCSEL with the lowered R th . The lowered R th value (607 K W −1 ) of the thinfilm VCSELs by 46.33%, compared to 1131 K W −1 of the bulktype VCSEL, was attributed to the efficient heat transfer from the copper-plated heatsinks, mitigating the self-heating effects around the active region induced by current injection during the device operation. The copper-plated heatsinks showed 31.6°C reduction around the active region at the rollover current, enabling high-power VCSEL operation with the extended rollover current. The maximum peak power (11.70 mW) of the thinfilm VCSEL with the extended rollover current was improved by 12.07% than 10.44 mW of the bulk-type VCSEL at room tem-perature. We believe that the substrate-removal thin-film VC-SELs integrated with the copper-plated heatsinks provide great potential to be favorably utilized for high-power and reliable CW VCSELs, enabling the widespread adoptions in a myriad of applications such as the industrial and automotive markets with the low-power consumption, high-power and temperature-stable operation.

Experimental Section
Epitaxial Growth: The top-emitting 930 nm thin-film VCSEL structures, including the lattice-matched InGaP etch-stop and highly n-doped GaAs layers, were grown on the exact n-GaAs (100) substrates via a LP-MOCVD system where trimethylgallium (TMGa), trimethylaluminum (TMAl), trimethylindium (TMIn), arsine (AsH 3 ), and phosphine (PH 3 ) are utilized. The silane (SiH 4 ) and carbon (CBr 4 ) were adoped as the dopant sources for n-, and p-doped GaAs layers, respectively. The lattice-matched InGaP etch-stop layer inserted between n-GaAs ohmic and buffer layers prevented the NH 4 OH-based etchant from penetrating into the active region during removal of the thick n-GaAs substrates. The selective oxidation of Al x Ga 1−x As layer with high content of Al fractions above the three multiple InGaAs/AlGaAs quantum wells was carried out in order to define the 10 μm oxide-confined aperture size of the thin-film VCSELs.
Device Fabrication: The fabrication procedures of the top-emitting 930 nm thin-film VCSELs integrated with the copper-plated heatsinks are described in detail, as shown in Figure S2 (Supporting Information). The top-emitting 930 nm thin-film VCSEL structures were epitaxially grown on the n-GaAs substrates via the LP-MOCVD system. First, the front processes, such as mesa etching for separating the thin-film VCSELs, selective oxidation of Al x Ga 1−x As layer with high Al contents, and top contact formation with the deposition of Ti/Pt/Au metal layers were carried out. The thin-film VCSELs were bonded onto an intermediate carrier under the controlled temperature and pressure by using the adhesive wax. The highly n-doped GaAs layer was exposed after removing the lattice-matched In-GaP etch-stop layer and the n-GaAs substrates with the HCl-based and NH 4 OH-based etchants, respectively. The etch rates of the GaAs substrate and InGaP etch-stop layers were 25 and 27 nm s −1 , respectively. It was worth noting that the highly selective etching of the n-GaAs substrates with respect to the InGaP etch-stop layer contributes the smooth interfaces between the n-GaAs ohmic layer and back metals with a combination of Ni/AuGe/Ni/Au. The copper-plated heatsink was formed onto the Ni/AuGe/Ni/Au metal layers that were deposited onto the n-GaAs ohmic layer. The copper-plating processes were carried out as follows. The surface of the Ni/AuGe/Ni/Au metal layers was treated by a plasma-cleaner system (YES G1000) with Ar plasma for 200 s at 200 W under 130 mtorr pressure in order to enhance the bonding characteristics at the interfaces. The seed metals of 30/300 nm thick Cr/Ti bilayers were deposited via a magnetron sputtering system, respectively. The 25 μm thick copper-plated heatsink was obtained at 1ASD (ampere per square decimeter) for 2 h 30 min. The intermediate carrier was detached by removing the wax, by which the top-emitting thin-film VCSELs were fabricated onto the copperplated heatsinks.
Device Characterization: The reflectance spectrum of the top-emitting 930 nm thin-film VCSEL wafer was obtained using a reflectance mapping machine (ETaMax Plato-series). The L-I-V characteristics of the thinfilm VCSELs integrated with the copper-plated heatsinks were investigated onto the temperature-controlled vacuum stage with a rotary pump under CW operation using a Keithley 2602B. The low-temperature bath (Tokyo Rikakikai CO., LTD.) provided the temperature-control accuracy of ±0.1°C in the range −30-95°C. The current was injected in order to bias the thinfilm VCSELs using the Keithley current source. The 930 nm light emission above the aperture area of the thin-film VCSELs was collected to the Silicon photodiode (Hamamatsu, 2201 photodiode). The emission spectrum of the thin-film VCSELs was obtained with an Ocean optics Maya 2000 PRO www.advancedsciencenews.com www.advmatinterfaces.de spectrometer.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.