近日,美国哈佛大学Lonar, Marko团队提出了集成电光数模链路,用于高效计算和任意波形生成。相关论文于2025年8月25日发表在《自然—光子学》杂志上。
人工智能和现代通信系统的快速发展需要创新的解决方案来提高计算能力和先进的信号能力。集成光子学,在芯片尺度上利用电磁波的模拟性质,为基于数字电子学的方法提供了一个有前景的补充。为了充分释放它们作为模拟处理器的潜力,在传统数字电子学和模拟光子学之间建立一个共同的技术基础对于构建下一代计算和通信系统是必不可少的。然而,由于缺乏有效的接口,模拟光子硬件的优势的全面展示受到了严重的挑战,其中可扩展性、速度和能耗是主要瓶颈。
研究组解决了这一挑战,并展示了一种通用的电光数模链接,使主题基于铸造厂的铌酸锂纳米光子学成为可能。使用纯数字电子输入,他们实现了以高达186 Gb s−1的信息速率按需生成模拟光学和电子波形。光波形解决了光子计算中数模电光转换的挑战,展示了高保真的修改的美国国家标准与技术研究所图像编码,超低功耗为0.058 pJ b−1。电子波形实现了一种无脉冲形状的微波任意波形产生方法,具有超宽带可调延迟和增益。该研究结果为集成光子学实现高效、紧凑的数模转换范式铺平了道路,并强调了模拟光子硬件可能对各种应用(如计算、光互连和高速测距)产生的变革性影响。
附:英文原文
Title: Integrated electro-optic digital-to-analogue link for efficient computing and arbitrary waveform generation
Author: Song, Yunxiang, Hu, Yaowen, Zhu, Xinrui, Powell, Keith, Magalhes, Letcia, Ye, Fan, Warner, Hana K., Lu, Shengyuan, Li, Xudong, Renaud, Dylan, Lippok, Norman, Zhu, Di, Vakoc, Benjamin, Zhang, Mian, Sinclair, Neil, Lonar, Marko
Issue&Volume: 2025-08-25
Abstract: The rapid growth in artificial intelligence and modern communication systems demands innovative solutions for increased computational power and advanced signalling capabilities. Integrated photonics, leveraging the analogue nature of electromagnetic waves at the chip scale, offers a promising complement to approaches based on digital electronics. To fully unlock their potential as analogue processors, establishing a common technological base between conventional digital electronics and analogue photonics is imperative for building next-generation computing and communications systems. However, the absence of an efficient interface has thus far critically challenged a comprehensive demonstration of the advantages of analogue photonic hardware, with the scalability, speed and energy consumption as primary bottlenecks. Here we address this challenge and demonstrate a general electro-optic digital-to-analogue link enabled using foundry-based lithium niobate nanophotonics. Using purely digital electronic inputs, we achieve the on-demand generation of both analogue optical and electronic waveforms at information rates of up to 186Gbs1. The optical waveforms address the digital-to-analogue electro-optic conversion challenge in photonic computing, showcasing high-fidelity Modified National Institute of Standards and Technology image encoding with an ultralow power consumption of 0.058pJb1. The electronic waveforms enable a pulse-shaping-free microwave arbitrary waveform generation method with ultrabroadband tunable delay and gain. Our results pave the way for efficient and compact digital-to-analogue conversion paradigms enabled by integrated photonics, and underscore the transformative impact that analogue photonic hardware may have on various applications, such as computing, optical interconnects and high-speed ranging.
DOI: 10.1038/s41566-025-01719-9
Source: https://www.nature.com/articles/s41566-025-01719-9