I. Introduction
More compact and cost-effective optical transceivers are needed for 100 Gb/s Ethernet (100 GbE) to deal with the exponential growth in data traffic generated in telecommunication networks and data center networks. 100 GbE, standardized in IEEE 802.3ba [1], has a long reach specification for 10 km or 40 km transmission over single-mode fiber (SMF), namely 100 GBASE-LR4 or 100 GBASE-ER4. The specification defines a multi-lane configuration that has four 25.78125 Gb/s lanes with local area network wavelength division multiplexing (LAN-WDM), which is assigned a wavelength channel spacing of 800 GHz in the 1.3 μm wavelength band. A transceiver for 100 GbE is also defined in the multi-source agreement (MSA). To further reduce both the body size and the power consumption of the 100 GbE transceiver, a CFP4 transceiver [2] defined in the centum form-factor pluggable MSA or a QSFP28 transceiver [3] defined in the quad small form-factor pluggable MSA has been enacted. However, with a multi-lane configuration it is complicated to fabricate such a compact transceiver because we have to further miniaturize the optical components installed in the transceiver such as the transmitter optical sub-assembly (TOSA) and the receiver optical sub-assembly (ROSA). Nevertheless, the multi-lane configuration must be a promising approach even for the next generation 400-Gb/s Ethernet (400 GbE), which is currently being discussed by an IEEE task force [4]. Therefore, an integration technique that satisfies the increasing lane number is an important technical issue for a future TOSA and ROSA.