Since the first experimental demonstration of electro-absorption (EA) effect on the metal-oxide-semiconductor capacitor (MOSCAP) structure using indium-tin-oxide (ITO) by J. Leuthold and H. Atwater group in 2010, the interest of developing ultra-efficient photonic modulators with transparent conductive oxide (TCO) has been elevated over the past 14 years. Considering TCO’s strong plasma-dispersion efficiency, epsilon-near-zero (ENZ) effect, and the high electrical conductivity, it seems fairly straightforward and highly promising to realize low-voltage, high-speed photonic modulators with TCO materials --- at least on simulation and PowerPoint.
Despite years of persistent research efforts from several groups as shown in the figure of TCO modulator history, achieving high-speed electro-optic modulation with TCO materials remain a grand challenge. Before this article was published, TCO-driven photonic modulators only achieved 4.5 Gb/s eye diagrams. Although small-signal RF modulation has been claimed to obtain more than 40 GHz bandwidth, it is not convincing that such devices can provide sufficient modulation strength or will be useful for optical communication. There are several hidden challenges that induced such a long-time roadblock. First, the carrier density and mobility of TCO materials strongly depend on the sputtering deposition and post-processing conditions. As results, such complexity presents difficulties in assessing the consistency of material performance. Second, achieving high-speed E-O modulation requires sophisticated doping to both silicon and TCO to reduce the series resistance, optimization of the fabrication process to form good Ohmic contact, and precise design of high-speed electrodes for impedance matching. Especially, TCO materials seem to be quite picky to metal films that can form Ohmic contact with small contact resistance. The third pain point is that no silicon photonics foundry is offering TCO processes in the fabrication and many experimental groups must rely on their own facilities to complete the entire fabrication steps. Researchers in TCO photonics have to reinvent wheels like those in silicon photonics community many years ago.
The work presented in this article addressed existing challenges from three perspectives. First, we engineered our device design based on a microring resonator (MRR) to work in the non-ENZ region of TCO materials, which only requires a small refractive index change. The MRR design is crucial to achieve the low-driving voltage that is high desired by industry. Second, we fabricated the device by combining Intel’s silicon photonics fab with our own TCO patterning process. Specifically, we used Intel’s 300 mm high volume manufacturing (HVM) silicon photonics process to fabricate passive Si-MRRs with p+ and p++ doping, which provided a low optical loss, high qualify platform for post-processing. Third, we used high mobility titanium-doped indium oxide (ITiO) to replace traditional ITO gate, which can significantly improve the Q-factor of the microring resonator. Our work demonstrated a sub-volt, 25Gb/s MOSCAP Si-MRM driven by the high mobility ITiO gate. Driven by slightly higher voltage, it can operate at 35Gb/s. In addition, the success of the combined foundry-university co-fabrication will prove the feasibility of integrating TCO processes into silicon photonics foundry to achieve high-performance hybrid devices and photonic integrated circuits.