I. Introduction
Nowadays, photonic integration represents the key technology to reduce the costs (in terms of both capex and opex), the energy consumption and the size of the optical components, thus allowing mass production of cost-effective high-performance photonic integrated circuits (PICs) [1]– [4]. Moreover, integration technologies can increase the stability, reliability and robustness of optical transceivers. Silicon photonics provide significant leverage in design and low-cost fabrication of small foot-print device (smaller than alternative platform such as InP or LiNbO3) [3]. Silicon photonics is also strategic in view of monolithic integration with complementary metal–oxide–semiconductor-based electronics. This way, the shortest possible electric interconnects between electronics and photonics could be allowed, and both bondwires and bondpads, usually detrimental for high speed performance, could be avoided. Recent developments also demonstrated the possibility to include laser sources by exploiting heterogeneous integration [4]. Nowadays, a capacity of 100 Gb/s exploiting Si-based modules with dual-polarization quadrature phase shift keying per channel is almost available for metro-regional applications [5], and optical coherent technology offers enormous advantages with respect to systems based on direct detection [6]. In this scenario, monolithic integration with the electronics is also crucial for mass–volume production of transceivers requiring digital signal processing. Another important aspect for the next-generation optical networks is the spectral efficiency, which can be increased by exploiting more complex modulation formats such as 16-quadrature amplitude modulation (QAM). Several architectures to implement 16-QAM signals have been investigated, as in [7]– [10]. Among these, in a previous work we proposed a solution based on a reconfigurable dual-drive in-phase (I) and quadrature (Q) modulator structure, allowing a 16-QAM signal generation as well as QPSK and others formats [10]. In particular, the proposed structure enables 16-QAM generation by simply driving the modulator with equal-amplitude binary electrical signals, thus avoiding complex electronics for the generation of multiple-level signals [12] . Moreover, the presence of tunable optical splitters in [10] allows energy-efficient offset-free 16-QAM generation and minimize the required phase modulators. In [11], the proposed architecture has been monolithically integrated on a InP platform. However, the high spurious amplitude modulation occurring in p-n junction-based phase modulators limited the capabilities of the circuit so much so that only QPSK modulation was successfully demonstrated.