Новые мощные фиолетовые лазеры Nuvoton обеспечат рост производительности при производстве чипов

Empowering violet lasers More powerful violet lasers, realised by improving design and introducing superior thermal management, will target laser direct imaging for advanced semiconductor packaging BY SHINJI YOSHIDA AND COLLEAGUES FROM NUVOTON TECHNOLOGY CORPORATION JAPAN Throughout the history of the semiconductor industry, there’s been a relentless drive to improve semiconductor performance. As well as reshaping device architectures, this eternal goal refines the way that chips are packaged and manufactured. Within the silicon industry, conventional transistor scaling is approaching fundamental limits. Due to this, innovation is increasingly shifting towards advanced packaging, where processors, memory devices and specialised chips are assembled three-dimensionally and interconnected by extremely fine patterning. This transition is placing an unprecedented demand on manufacturing equipment, and in particular on the lithographic technologies employed to form redistribution layers and high-density interconnects on large-area substrates. Packaging at the forefront There are a number of requirements associated with advanced semiconductor packages. They include finer features and resolutions, a higher overlay accuracy between multiple layers, large substrates and high productivity. Efforts to excel on these fronts are stretching the capabilities of traditional steppers using photomasks. While the photolithography equipment that relies on ultraviolet radiation from mercury lamps and solid photomasks has served the silicon industry well, there are intrinsic challenges when substrate sizes grow and product variants multiply. Mask preparation is now costly and time consuming, as it is increasingly difficult to ensure a high overlay accuracy across large areas, because patterns on solid photomasks are fixed and fail to compensate for distortion and deformation of the substrate. Due to these constraints, there’s renewed attention in laser direct imaging (LDI). This is a maskless exposure technique, with circuit patterns written directly by scanning a focused laser beam. Thanks to elimination of photomasks, LDI enables rapid design changes, excellent resolution and overlay accuracy, and efficient support for high-mix manufacturing. In addition, this solution complies with a global trend for regulating and phasing out mercury-based light sources. Despite these advantages, today’s adoption of LDI for next-generation semiconductor packaging is limited to pilot-line equipment. Holding back full-scale industrial deployment of LDI are concerns over throughput, which is governed by the power of the laser light source. Existing semiconductor lasers emitting in the UV or violet have enabled early deployment of LDI systems, but their output powers are insufficient for high-volume production of advanced semiconductor packages. The light-source bottleneck In practical LDI systems, laser diodes must deliver a high continuous-wave output power while maintaining a stable emission wavelength and delivering reliable long-term operation. Simply increasing drive current is not a viable solution, as this increases self-heating, in turn increases the junction temperature of the laser diode, and leads to degradation of laser performance and an acceleration of output power degradation during long-term operation. To overcome this bottleneck, there must be a holistic approach, addressing both the laser diode and the thermal path that removes heat from the active region. At Nuvoton Technology Corporation Japan we are addressing these issues during our development of a high-power InGaN-based edge-emitting laser diode that emits at 405 nm – that’s a wavelength well matched to photoresist sensitivity, and widely used in LDI systems. By jointly optimising the device geometry, epitaxial waveguide structure and package thermal design, we have realised a high-output power and an improved wall-plug efficiency. These strengths have enabled improved long-term reliability through suppression of the junction temperature in violet laser diodes. Managing heat at its source A major limitation of the high-power operation of InGaN-based laser diodes is the rise in junction temperature at high current densities. Elevated junction temperatures increase carrier leakage from the quantum well, leading to a higher threshold current and a reduced slope efficiency. In addition, high junction temperatures accelerate device degradation during aging, due to the thermally activated formation of defects, which act as non-radiative recombination centres. Compounding these matters, the unwanted loss mechanisms are amplified in violet and UV lasers, due to reduced band offsets in the III-N quantum well and increased internal optical loss, making high-power operation particularly sensitive to junction temperature. Consequently, it is imperative to suppress the rise in junction temperature under high-power operation, as this holds the key to high performance and long-term reliability. One of our key design choices when working towards this objective is our laser chip geometry. We employ a wide stripe and a long cavity. By expanding stripe width and extending the cavity length, we reduce the optical power density at the facet and the current density in the cavity – and ultimately distribute heat generation over a larger volume. There’s a reduction in the thermal resistance of the laser chip, helping to suppress the rise in junction temperature. For the devices described in this article, stripe widths are up to 45 μm, and cavity lengths up to 1400 μm. In addition to enabling a higher output power, this geometry mitigates degradation mechanisms exacerbated at high optical intensities. As emission heads towards the violet and UV, thermal management is increasingly critical in InGaN-based lasers. Shorter wavelengths are accompanied by higher photon energies, and typically higher operating voltages – both contribute to increased heat generation for a given optical output. At the same time, it is more challenging to ensure carrier confinement in the quantum well, due to reduced band-offsets, making device performance particularly sensitive to junction temperature. There’s a strong interplay between key laser characteristics. As the temperature rises, more carriers spill over from the active region, leading to an increase in threshold current and a rapid deterioration in slope efficiency. In high-power operation, this detrimental feedback loop between temperature and laser performance threatens to dominate device behaviour, and incremental design changes are insufficient for addressing these concerns. This issue is one of the reasons why high-power scaling of violet laser diodes has historically lagged behind that of devices emitting at longer wavelengths. Our approach to improving the performance of violet lasers involves trimming its current density and thermal resistance through a widening of the stripe and an extension of the cavity length. However, just making these geometrical changes does not guarantee high-power performance. In short-wavelength lasers, internal optical loss plays an outsized role in determining wall-plug efficiency. There’s the possibility that free-carrier absorption in heavily doped cladding and guiding layers, as well as scattering losses associated with imperfect optical confinement, offset gains resulting from a reduced current density and a lower thermal resistance of the laser chip. To address this, we treat waveguide optimisation and thermal design as interdependent challenges. We know that reducing internal loss not only improves efficiency directly – it also cuts dissipated power for a given output, thereby alleviating the thermal load on the package. Meanwhile, if we lower the thermal resistance of the entire device, including the package, we will preserve the optical and electrical benefits of the optimised waveguide under practical operating conditions. Due to this strong coupling between optical design and thermal management, it is critical for progress at the device level to be matched by advances in packaging. Optimising the waveguide Thermal management alone is insufficient to ensure success; a high wall-plug efficiency requires low internal optical loss and strong confinement of carriers and photons. We meet these requirements with a laser structure employing a single quantum-well active region to minimise threshold current, surrounded by carefully engineered guiding layers. The heterostructure features high-refractive-index InGaN waveguide layers to confine the optical mode, and AlGaN cladding layers to provide carrier confinement and minimise optical leakage. Our design also addresses free-carrier absorption. Arising from dopants and impurities in the waveguide, this impediment to efficiency is a major source of internal loss in short-wavelength lasers. We mitigate this with a design that minimises the spatial overlap between the optical mode and highly doped regions. It’s a strategy that reduces free-carrier absorption without compromising carrier injection. Through optimisation of dopant concentrations and impurity profiles during growth of the heterostructure, produced by MOCVD, internal losses are reduced without compromising electrical performance. Our emitters also incorporate an electron-blocking layer – this helps suppress carrier overflow from the active region, and supports high efficiency, even at elevated drive currents. Rethinking the package While device-level optimisation improves intrinsic performance, it is the package that governs how effectively heat is extracted. In industrial laser applications, TO-CAN packages, such as TO9, are widely used. They provide a compact form factor and ease of integration. However, these packages have traditionally suffered from relatively high thermal resistance. Moving to much larger packages could address this issue, but it will complicate system design, so we have focused on trimming thermal resistance within the TO9 footprint. Figure 1. Schematic of the low-thermal-resistance TO-9 package, featuring a monolithic copper base, p-side-down mounting, and a diamond submount, designed to enhance heat extraction from the laser chip. To accomplish this, we redesigned the package as a monolithic copper structure, replacing conventional iron-based base plates and minimising internal thermal interfaces (see Figure 1). With this redesign, package thermal resistance is roughly two-thirds of that of a conventional iron-based TO-9 package. Behind this reduction is a substantially higher thermal conductivity, realised by replacing iron-based materials with copper. We have made further gains by mounting the laser chip p-side down, as this shortens the heat path from the active region to the submount. Note that the choice of submount material is equally critical. We have evaluated various high-thermal-conductivity materials, and found that a diamond submount provides a substantial reduction in thermal resistance compared with conventional ceramic materials. Our use of a diamond submount, offering a thermal conductivity several times higher than that of typical ceramic materials, delivered an additional reduction in junction-to-case thermal resistance. Taken together, the redesigned copper package and p-side mounting on diamond have doubled the overall heat extraction capability of the laser, and enabled stable high-power operation within a compact TO-9 package. Power performance and efficiency As one would expect, the combined effect of a tailored device geometry, waveguide optimisation, and a low-thermal-resistance package has produced marked improvement in high-power performance. Operating at a case temperature of 25 °C, our 405 nm lasers deliver a continuous-wave output exceeding 12 W (see Figure 2). When the output is 5.2 W, wall-plug efficiency is 43.9 percent. For industrial applications, high efficiency must be maintained at elevated temperatures. We find that even at a case temperature of 50 °C, our devices exhibit wall-plug efficiencies above 40 percent, when delivering an output power of more than 5 W. Behind this success is the device’s thermal robustness, essential for integration into compact LDI modules, where active cooling capacity is limited and there may be fluctuations in ambient temperature. In addition to enabling high output powers and thermal robustness, our work highlights the flexibility of our device platform, which can be optimised for various performance priorities. If we design devices with a reduced stripe width, we can enhance wall-plug efficiency by decreasing the threshold current – but this goes hand-in-hand with a fall in the maximum output power, due to increased current density and a higher thermal resistance. When making this trade-off, the laser geometry must be tailored to specific application requirements. While a high output power is essential for an LDI system seeking maximum throughput, applications requiring a moderate output power are better served with geometries optimised for peak efficiency. Figure 2. Optical output power and wall-plug efficiency measured under continuous-wave operation at a case temperature of 25 °C for a 405 nm InGaN-based laser diode with a cavity length of 1400 µm and a stripe width of 45 µm. Figure 3. Long-term output stability of the laser diode during aging at 6 W and a case temperature of 25 °C, demonstrating stable operation without catastrophic optical damage. The key message is that our underlying design principles can support both high-power, thermally robust operation and high-efficient operation at lower output powers. What’s required is to optimise the device geometry for the required characteristics. Reliability at high power When laser diodes operate at high output powers, there’s severe stress on their facets and internal structures. This raises concerns related to long-term reliability. To look into this matter, we conducted aging tests under automatic current control, involving an initial output power of 6 W and a case temperature of 25 °C. These conditions are representative of demanding industrial operation beyond typical commercial levels. During extended operation, exceeding 1900 hours, we did not observe any catastrophic optical damage, and recorded minimal output degradation (see Figure 3). Based on the low degradation rate during our investigation, we can conclude that our device design provides a sufficient margin for long-term operation at this power level. These results underscore the importance of reducing optical and thermal stress through appropriate design of the device and its package. Implications for LDI and beyond The performance of our lasers overcomes a key barrier to the wider adoption of LDI in advanced semiconductor packaging. By delivering a substantially higher optical power from a compact semiconductor laser, we are enabling engineers in silicon fabs to shorten exposure times and boost throughput, without sacrificing system stability or lifetime. This leads to a step change in productivity, allowing LDI to provide the throughput required for industrial-scale manufacturing on large substrates. Beyond LDI, there are opportunities for high-power, high-efficiency violet laser diodes in a number of other industrial applications, including additive manufacturing, precision processing and illumination systems – all benefit from short-wavelength, compact light sources. As semiconductor manufacturing continues to evolve, the demands placed on light sources are only going to increase. Meeting them is our next step. We are targeting further improvements in efficiency, extension to other ultraviolet and visible wavelengths, and integration into application-optimised modules. Such efforts will build on successes to date, demonstrating that a coordinated approach – spanning epitaxial design, device geometry and thermal packaging – will enable semiconductor laser diodes to meet the stringent requirements of next-generation manufacturing technologies.