Article

# Integrated all-photonic non-volatile multi-level memory

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

## Abstract

Implementing on-chip non-volatile photonic memories has been a long-term, yet elusive goal. Photonic data storage would dramatically improve performance in existing computing architectures by reducing the latencies associated with electrical memories and potentially eliminating optoelectronic conversions. Furthermore, multi-level photonic memories with random access would allow for leveraging even greater computational capability. However, photonic memories have thus far been volatile. Here, we demonstrate a robust, non-volatile, all-photonic memory based on phase-change materials. By using optical near-field effects, we realize bit storage of up to eight levels in a single device that readily switches between intermediate states. Our on-chip memory cells feature single-shot readout and switching energies as low as 13.4 pJ at speeds approaching 1 GHz. We show that individual memory elements can be addressed using a wavelength multiplexing scheme. Our multi-level, multi-bit devices provide a pathway towards eliminating the von Neumann bottleneck and portend a new paradigm in all-photonic memory and non-conventional computing.

## No full-text available

... Devices using phase change materials allow for the storage of up to eight levels of data in a single unit that can be adjusted by light pulses [14]. Chalcogenide Ge 2 Sb 2 Te 5 (GST) is a well-studied phase-change material that has been shown to enable photonic synapses in spiking neurons. ...
... In order to reconfigure phase-change photonic devices, it is frequently necessary adjust the intensity [14] and pulse shape [60] of an incident light wave. Resonant str tures [14] are frequently utilized to enable the wavelength-selective operation and i proved modulation depths. ...
... In order to reconfigure phase-change photonic devices, it is frequently necessary adjust the intensity [14] and pulse shape [60] of an incident light wave. Resonant str tures [14] are frequently utilized to enable the wavelength-selective operation and i proved modulation depths. However, the comparable feature is not present ...
Article
Full-text available
The rapid development of neural networks has led to tremendous applications in image segmentation, speech recognition, and medical image diagnosis, etc. Among various hardware implementations of neural networks, silicon photonics is considered one of the most promising approaches due to its CMOS compatibility, accessible integration platforms, mature fabrication techniques, and abundant optical components. In addition, neuromorphic computing based on silicon photonics can provide massively parallel processing and high-speed operations with low power consumption, thus enabling further exploration of neural networks. Here, we focused on the development of neuromorphic computing based on silicon photonics, introducing this field from the perspective of electronic–photonic co-design and presenting the architecture and algorithm theory. Finally, we discussed the prospects and challenges of neuromorphic silicon photonics.
... The drastic change of these properties between these two states is used as a "bit" for storing information in devices. This amorphous to crystalline phase transition (and vice versa) is by now well controlled with electrical pulses and continuous light excitation (rewritable disc for example), even at the industrial level, but new research directions have appeared in the last ten years towards the manipulation of these states with light pulses [56,57], unveiling the possibility, with short light pulses, to envision applications in the GHz and THz frequency range. As a consequence, the understanding of the light-matter interaction is a necessary step. ...
... As mentioned in the beginning of chapter, in the application of phase change material devices, both the amorphization and crystallization are achieved by applying ultrashort laser pulses [56,57]. As we want to avoid such photoinduced transition during our experiments, we performed in the first place, a series of pump-probe experiments as a function of incident pump fluence, in order to reveal the evolution of transient optical reflectivity signals. ...
... On (1 1 1)-oriented cubic substrates (for example SrTiO 3 -STO), rhombohedral 'bulk-like' BFO is typically obtained (figure 3(d)) [43,[53][54][55][56], consistent with predictions of thermodynamic calculations [57]. Even for very small thicknesses (e.g. ...
Thesis
Full-text available
In this thesis work, we study the ultra-fast dynamics of electrons and phonons in ferroic materials, such as GeTe and BiFeO3. This experimental work is carried out using ultra-fast optics and time-resolved X-ray diffraction techniques.In the first part of this research, we studied these dynamics in the GeTe material, which is the reference system for materials known as “phase transition” or “optical memory”. In particular, the change from the amorphous state to the crystalline state is already used on an industrial scale as an optical memory because these two states can be considered as high and low "bit". Nevertheless, with the aim of developing very high-speed (THz) technologies, i.e. beyond the current frequencies of electronics, it is necessary to understand how this material (optical memory) can also be used at these very high frequencies, i.e it is necessary to know the response of the material on the scale of the picosecond. In order to "write" and "read" with light and in an ultra-fast way, it is necessary to understand how the absorbed optical energy is transferred to electrons and then to phonons. We were able to show by optical pump-probe method that when the crystalline material GeTe was optically excited, the "hot" electrons could diffuse very quickly outside the optical absorption zone, that is to say that the hot electrons , within a few picoseconds, non-thermally transport energy into the solid at a distance approximately seven time greater than the optical penetration of the pump beam. This has been demonstrated by analyzing the spectrum of acoustic phonons emitted after optical excitation. This non-thermal transport effect was however not observed in the amorphous phase where the “hot” electrons remain localized in the optical absorption zone. These results show that at the picosecond scale the transformation of optical energy into electronic and phononic energy is very different in the two phases, which may impact the properties of future THz components.In the second part, we discussed how it is possible to manipulate with light the structure of multiferroic materials at the picosecond scale. The archetype material is BiFeO3 which presents a ferroelectric and magnetic order at room temperature. The ferroelectric materials are systems at the base of many applications such as piezoelectric sensors and actuators (NEMS, MEMS). This work, as in the case of GeTe, therefore aims to explore the potential of these materials in the very high frequency domain (GHz-THz). To understand how ultra-fast light pulses act on the structure, we first used optical pump-probe experiments to analyze how ultra-fast optical energy is transformed into acoustic phonons (ultrafast deformation). In particular, we have studied time-resolved Brillouin scattering to show the richness of the phonon emitted in BiFeO3 thin films with the coexistence of longitudinal and transverse modes (LA, TA). In a second step, in order to quantify the amplitude of the deformation (not accessible by Brillouin), we carried out optical pump experiments with RX probe (time-resolved X-ray diffraction). We were thus able to describe how the BiFeO3 lattice distorts and in particular, we showed and quantified the longitudinal (LA) and transverse (TA) deformation fields at the picosecond scale.
... This type of materials can be switched between their amorphous and crystalline phases through thermal or laser irradiation stimuli while entailing a modulation in their refractive index at ultra-short times (of the order of ps) [21,22]. This capability of PCMs has already been exploited in a wide range of reconfigurable photonic platforms that expand from programmable photonics [23,24], neuromorphic computing [25], non-volatile and rewritable data storage [21,26], to tunable metasurfaces and flat optics with amplitude/phase control [27], cloaking [28], and reflective displays [29]. Therefore, in the proposed hot-electron photodetector configuration, the photodetection band can be tuned through the amorphous-to-crystalline phase change in the Sb 2 S 3 that produces a modulation in its refractive index and, consequently, a change in the resonant frequency of the SP generated at the Au grating. ...
... From the experimental point of view, phase change materials not only present pure amorphous and crystalline phases. A mix of both can be controllably induced in the PCM film as already demonstrated for GST [26,40]. The effective refractive index of these samples can be estimated through a Bruggeman model [41] by considering the optical properties of both phases. ...
Article
Full-text available
Hot-carrier based photodetectors and enhanced by surface plasmons (SPs) hot-electron injection into semiconductors, are drawing significant attention. This photodetecting strategy yields to narrowband photoresponse while enabling photodetection at sub-bandgap energies of the semiconductor materials. In this work, we analyze the design of a reconfigurable photodetector based on a metal-semiconductor (MS) configuration with interdigitated dual-comb Au electrodes deposited on the semiconducting Sb 2 S 3 phase-change material. The reconfigurability of the device relies on the changes of refractive index between the amorphous and crystalline phases of Sb 2 S 3 that entail a modulation of the properties of the SPs generated at the dual-comb Au electrodes. An exhaustive numerical study has been realized on the Au grating parameters formed by the dual-comb electrodes, and on the SP order with the purpose of optimizing the absorption of the device, and thus, the responsivity of the photodetector. The optimized photodetector layout proposed here enables tunable narrowband photodetection from the O telecom band ( λ = 1310 nm) to the C telecom band ( λ = 1550 nm).
... However, for the photonics field, they mainly depend on both the technologies used for modulators (for input vectors) or the photodiodes (for output vectors) used in each implementation, which follow the possibilities given by the foundries and rarely are due to architecture choices [60]- [62]. Following that, it is more interesting to focus on common limitations, such as the number of controllers that each circuit requires, the footprint scaling, and the possibility to implement nonvolatile memory elements, such as Photonic RAM (P-RAM) components using Phase Change Materials (PCMs) [63], [64], to further reduce energy consumption. Those figure-of-merits better describe the differences between different circuits, showing that trade-offs must be addressed to evolve into this field. ...
... Moreover, techniques such as coherent detection have been proposed [91], capable to reach 9-bit resolution with WDM MRR architecture. The last piece of confrontation is regarding the possibility to implement P-RAM on the circuits [63], [64], [93], [94], by using PCMs for example [95]. In a larger view, as more and more MVM circuits will be used to implement NNs, having the possibility to integrate photonic memory elements would have a crucial benefit in terms of energy efficiency, as it reduces the power needed to tune the weight as well as the energy required to access external memory elements in DRAM [28]. ...
Preprint
Full-text available
The explosion of artificial intelligence and machine-learning algorithms, connected to the exponential growth of the exchanged data, is driving a search for novel application-specific hardware accelerators. Among the many, the photonics field appears to be in the perfect spotlight for this global data explosion, thanks to its almost infinite bandwidth capacity associated with limited energy consumption. In this review, we will overview the major advantages that photonics has over electronics for hardware accelerators, followed by a comparison between the major architectures implemented on Photonics Integrated Circuits (PIC) for both the linear and nonlinear parts of Neural Networks. By the end, we will highlight the main driving forces for the next generation of photonic accelerators, as well as the main limits that must be overcome.
... Phase change materials (PCMs) have been investigated in the context of PICs due to their non-volatile phase transition and for the high refractive index contrast between their amorphous and crystalline phases [30][31][32] . The ability to reconfigure PICs has been implemented in applications ranging from memories 33 , wavelength division multiplexing 34 , and switches 30 to neuromorphic devices [35][36][37] . Amongst all of the ternary phase change material in non-volatile integrated photonic systems, the most used is Ge 2 Sb 2 Te 5 (GST). ...
Article
Full-text available
A new family of phase change material based on antimony has recently been explored for applications in near-IR tunable photonics due to its wide bandgap, manifested as broadband transparency from visible to NIR wavelengths. Here, we characterize $$\hbox {Sb}_{2} \hbox {S}_{3}$$ Sb 2 S 3 optically and demonstrate the integration of this phase change material in a silicon nitride platform using a microring resonator that can be thermally tuned using the amorphous and crystalline states of the phase change material, achieving extinction ratios of up to 18 dB in the C-band. We extract the thermo-optic coefficient of the amorphous and crystalline states of the $$\hbox {Sb}_{2}\hbox {S}_{3}$$ Sb 2 S 3 to be 3.4 x $$10^{-4}\hbox {K}^{-1}$$ 10 - 4 K - 1 and 0.1 x 10 $$^{-4}\hbox {K}^{-1}$$ - 4 K - 1 , respectively. Additionally, we detail the first observation of bi-directional shifting for permanent trimming of a non-volatile switch using continuous wave (CW) laser exposure ( $$-5.9$$ - 5.9 to 5.1 dBm) with a modulation in effective refractive index ranging from +5.23 x $$10^{-5}$$ 10 - 5 to $$-1.20$$ - 1.20 x 10 $$^{-4}$$ - 4 . This work experimentally verifies optical phase modifications and permanent trimming of $$\hbox {Sb}_{2}\hbox {S}_{3}$$ Sb 2 S 3 , enabling potential applications such as optically controlled memories and weights for neuromorphic architecture and high density switch matrix using a multi-layer PECVD based photonic integrated circuit.
... Compared with those traditional phase-change materials (VO 2 ), the chalcogenide PCMs offer a nonvolatile tuning mechanism, in which no additional energy is needed to maintain the functional state of the device. In particular, Ge 2 Sb 2 Te 5 (GST), as one of the most popular alloys, has led to various photonic applications, such as photonic memory (Cheng et al., 2018;Rios et al., 2014;Ríos et al., 2015), optoelectronic color display , modulators (Wang et al., 2016), switches (Gholipour et al., 2013, absorbers (Tittl et al., 2015), and thermal emitters Qu et al., 2017), and so forth. However, these applications are limited to the visible and infrared spectral range, leaving the extensive potential for further exploration in the THz band (Makino et al., 2019;Pitchappa et al., 2019), especially for computing, communications, information storage, and encryption. ...
Article
Full-text available
Fast and efficient information processing and encryption, including writing, reading, and encryption memory, is essential for upcoming terahertz (THz) communications and information encryption. Here, we demonstrate a THz multi-level, nonvolatile, optically rewritable memory and encryption memory based on chalcogenide phase-change materials, Ge2Sb2Te5 (GST). By tuning the laser fluence irradiated on GST, we experimentally achieve multiple intermediate states and large-area amorphization with a diameter of centimeter-level in the THz regime. Our memory unit features a high operating speed up to 4 ns, excellent reproducibility, and long-term stability. Utilizing this approach, hexadecimal coding information memories are implemented, and multiple writing-erasing tests are successfully carried out in the same active area. Finally, terahertz photoprint memory is demonstrated, verifying the feasibility of lithography-free devices. The demonstration suggests a practical way to protect and store information, and paves a new avenue towards nonvolatile active THz devices.
... This review of the current trends with regards to silicon photonics is by no means a comprehensive overview of the field. Silicon photonics is the subject of many exciting new directions, instances include quantum key distribution [90], compact optical clocks [91] and nonvolatile photonic memories [92]. ...
Thesis
Full-text available
As a corollary of silicon manufacturing, silicon photonics has emerged as a viable photonic platform that has attracted the attention of many. The commercialization efforts of this technology, however, have remain somewhat limited due to several obstacles, technical and cost-related. As silicon is a poor emitter of light, the realization of an electrically-pumped monolithic laser source is unlikely. However, one may argue that the above have been satisfactorily resolved with the hybrid/heterogenous Ⅲ-Ⅴ/silicon photonic platform. In fact, the poor electro-optic conversion of silicon is one of the main factor that enables high performance hybrid/heterogenous Ⅲ-Ⅴ/silicon photonic laser diodes, resulting in significant improvement in performance over its Ⅲ-Ⅴ counterparts. While silicon photonics promises low cost, the premise is that the economies of silicon manufacturing is exploited. The inception of silicon photonics is mainly driven by the “interconnect bottleneck” in telecom and datacom. The data center transceiver market is attractive. However, there is a lack of a singular solution to all requirements in terms of reach, multisource agreement and standards. This implies that the cost of developing silicon photonics technology will be high unless the optical industry makes a concerted effort for standardization. As of now, the volumes required by silicon photonics are too low to draw commitment from large chip-making foundries. This work posits that for silicon photonics to be commercially viable, its range of applications must be widespread. The greater the adoption of silicon photonics in industry, the lower its cost. The condition is that firms must make the first step towards choosing silicon photonics for their applications. This work focuses on the development of silicon photonics technology beyond the traditional O and C bands. As a proof of concept to the broadband properties of the silicon-on-insulator platform, a high-performance arbitrary power splitter is realized at the longer transparency edge. In regard to the 2 μm waveband, which has been touted as a potential window for optical communications, the active Si-SiN multilayer platform, silicon switching as well as hybrid Ⅲ- Ⅴ/silicon photonic tunable lasers operating from 1881-1947, 1955-1992 nm has been demonstrated for the first time. In addition, at the application-rich wavelength region near 1.65 μm, a sub-kHz linewidth, hybrid Ⅲ-Ⅴ/silicon photonic tunable laser with a range of 1647-1690 nm is reported.
... During the write operation, a nanosecond pulse signal carrying certain energy is injected into the optical waveguide and then coupled to the GST after passing through the MR, the energy absorbed by GST reaches its crystallized energy threshold, causing its phase state to change. The switching time for different phase states is subnanosecond [34]. For example, the authors in [32] use a rectangular programming pulse of 50 ns to store 34 unique transmission levels in a single cell of GST, and the programming power is between 68 pJ and 135 pJ. ...
Article
Full-text available
This paper proposes StarLight, a low-power consumption and high inference throughput photonic artificial neural network (ANN) accelerator featuring the photonic ‘in-memory’ computing and hybrid mode-wavelength division multiplexing (MDM-WDM) technologies. Specifically, StarLight uses nanophotonic non-volatile memory and passive microring resonators (MRs) to form a photonic dot-produce engine, achieving optical ‘in-memory’ multiplication operation with near-zero power consumption during the inference phase. Furthermore, we design an on-chip wavelength and mode hybrid multiplexing module and scheme to increase the computational parallelism. As a proof of concept, a 4×4×4 optical computing unit featuring 4-wavelength and 4-mode is simulated with 10 Gbps, 15 Gbps and 20 Gbps data rates. We also implemented a simulation on the Iris dataset classification and achieved an inference accuracy of 96%, which is entirely consistent with the classification accuracy on a 64-bit computer. Therefore, StarLight holds promise for realizing low energy consumption hardware accelerators to address the incoming challenges of data-intensive artificial intelligence (AI) applications.
... Figure 5d-e show the extracted heating and cooling time constants across the silicon heater area using the exponential fitting equation(3). While there is variation across the experimental time constants, it appears to be random and can be attributed to the quality of the fit. ...
Preprint
Full-text available
Optical phase-change materials are highly promising for emerging applications such as tunable metasurfaces, reconfigurable photonic circuits, and non-von Neumann computing. However, these materials typically require both high melting temperatures and fast quenching rates to reversibly switch between their crystalline and amorphous phases, a significant challenge for large-scale integration. Here, we present an experimental technique which leverages the thermo-optic effect in GST to enable both spatial and temporal thermal measurements of two common electro-thermal microheater designs currently used by the phase-change community. Our approach shows excellent agreement between experimental results and numerical simulations and provides a non-invasive method for rapid characterization of electrically programmable phase-change devices.
... Phase-change materials (PCMs) are expected to be used for non-volatile optical devices on a Si photonics platform using abnormally large contrasts of optical properties between amorphous and crystalline phases [12]. In particular, Ge 2 Sb 2 Te 5 (GST), one of the most widely used chalcogenide PCMs [13,14], has been used in tunable metamaterials at visible to infrared wavelengths [15], and integrated Si photonics at near-infrared (NIR) wavelengths [16,17]; for example, optically and electrically controlled photonic switches [18][19][20][21][22][23], all-photonic memory [24,25], in-memory computing [26], and photonic tensor cores [27,28] have been reported. However, the large optical absorption of amorphous and crystalline GST in NIR wavelengths has hindered the realization of optical devices without optical attenuation including a low-loss optical phase shifter. ...
Article
Full-text available
We investigate an optical phase shifter based on Ge 2 Sb 2 Te 5 (GST) integrated with a Si waveguide at mid-infrared (MIR) wavelengths. Since the optical absorption of both amorphous and crystalline GST can be reduced at a longer wavelength, we demonstrate that the optical loss of the phase shifter can be reduced at MIR wavelengths. The measured optical loss per π phase shift of a phase-change material (PCM) phase shifter at 2.32 µm wavelength is 2.6 dB/π, which is more than 80 times smaller than that at 1.55 µm wavelength (21.7 dB/π) and more than 5 times smaller than that at 1.92 µm wavelength (9.7 dB/π). Moreover, resonance wavelength tuning of an add-drop micro-ring resonator using a PCM phase shifter at 2.32 µm wavelength is demonstrated owing to the low-loss optical phase shift. These findings reinforce the applicability of the approach toward a low-loss optical phase shifter based on PCMs operating at MIR wavelengths on a Si photonic platform for quantum computing, sensing, and optical communication.
... To implement integrated photonic non-volatile states, devices based on floating-gate geometries [7], micro-electromechanical systems [8], phase-change materials (PCMs) [9], and ferroelectric materials [10] have been proposed [11,12]. While all these examples can be electrically programmed and optically readout, only PCMs have demonstrated simultaneously optically-written, multilevel, and non-volatile actuation, enabling applications in on-chip all-optical learning [13,14] and processing [15]. ...
Preprint
Full-text available
Controlling changes in the optical properties of photonic devices allows photonic integrated circuits (PICs) to perform useful functions, leading to a breadth of applications in communications, computing, and sensing. Many mechanisms to change optical properties exist, but few allow doing so in a reversible, non-volatile manner. This leads to power inefficiency in reconfigurable circuits and requires external memory elements. In this work, we propose and experimentally demonstrate reversible, non-volatile phase actuation of a silicon nitride PIC with thermally-stable photochromic organic molecules. The use of a high-core-index platform allows, for the first time, the photochemical actuation of a planar-resonator-based photonic memory unit, which enables high performance and permits integrated spectroscopic analysis. We show novel properties of this all-optical memory for a silicon photonics platform, including complete transparency in the optical C-band, as well as first-order photokinetics of the photoconversion that lead to bidirectional scalable switching rates and continuous tuning. Such features are critical for memories in analog applications, such as quantum, microwave, and neuromorphic photonics, where low loss and precision are paramount.
... Multilevel memory systems are technologically appealing, for instance, due to an increase in storage density [1][2][3]. In past years, alternative non-semiconducting materials were proposed as non-volatile multilevel memories in photonic [4] and antiferromagnetic [5] systems. In superconductors, a yTron current combiner can be used for non-destructive current readout and is capable of differentiating between discrete magnetic flux values trapped in superconducting loops [6], finding applications in proposed binary [7,8] and multilevel [9] memory elements. ...
Preprint
Full-text available
With the development of novel computing schemes working at cryogenic temperatures, superconducting memory elements have become essential. In this context, superconducting quantum interference devices (SQUIDs) are promising candidates, as they may trap different discrete amounts of magnetic flux. We demonstrate that a field-assisted writing scheme allows such a device to operate as a multilevel memory by the readout of eight distinct vorticity states at zero magnetic field. We present an alternative mechanism based on single phase slips which allows to switch the vorticity state while preserving superconductivity. This mechanism provides a possibly deterministic channel for flux control in SQUID-based memories, under the condition that the field-dependent energy of different vorticity states are nearby.
... Chalcogenide phase change materials (PCMs) offer a promising solution to nonvolatile programmable PICs [17][18][19]. Traditional PCM devices, however, cannot provide phase-only modulation due to high optical losses associated with classical PCMs exemplified by Ge 2 Sb 2 Te 5 [20,21]. Moreover, most PCM device prototypes to date still rely upon furnace annealing or an external laser stimulus to trigger the structural transition. ...
Article
Full-text available
Optical phase shifters constitute the fundamental building blocks that enable programmable photonic integrated circuits (PICs)—the cornerstone of on-chip classical and quantum optical technologies [1, 2]. Thus far, carrier modulation and thermo-optical effect are the chosen phenomena for ultrafast and low-loss phase shifters, respectively; however, the state and information they carry are lost once the power is turned off—they are volatile. The volatility not only compromises energy efficiency due to their demand for constant power supply, but also precludes them from emerging applications such as in-memory computing. To circumvent this limitation, we introduce a phase shifting mechanism that exploits the nonvolatile refractive index modulation upon structural phase transition of Sb 2 Se 3 , a bi-state transparent phase change material (PCM). A zero-static power and electrically-driven phase shifter is realized on a CMOS-backend silicon-on-insulator platform, featuring record phase modulation up to 0.09 π/µm and a low insertion loss of 0.3 dB/π, which can be further improved upon streamlined design. Furthermore, we demonstrate phase and extinction ratio trimming of ring resonators and pioneer a one-step partial amorphization scheme to enhance speed and energy efficiency of PCM devices. A diverse cohort of programmable photonic devices is demonstrated based on the ultra-compact PCM phase shifter.
... Typically, the response of conventional nanophotonic devices is fixed at the fabrication step. Nevertheless, reconfigurable photonic devices will enable the applicability of photonic circuits in optical neural networks [1,2] and optical computing [3], the design tunable ultrafast tunable metasurfaces and flat optics with amplitude and/or phase control [4], reflective displays [5] and non-volatile and rewritable data storage devices [6,7]. Even though refractive index tuning can be achieved using thermo-optical effects, liquid crystals or Pockels cells, the tunability in these cases is relatively small (of the order of 10 − 2 ) and requires a constant energy supply to hold the state (volatile states). ...
Article
Full-text available
This paper discusses the fundamentals, applications, potential and limitations of polarized light reflection techniques for the characterization of phase-change materials (PCMs). These techniques include spectroscopic ellipsometry, time-resolved ellipsometry and imaging ellipsometry as well as polarimetry. We explore the capabilities of spectroscopic ellipsometry in the determination of the extinction coefficient of PCMs and the capabilities of imaging ellipsometry to characterize PCMs. We show that ellipsometry is capable of more than the determination of thickness and optical properties, and it can be exploited to gain information about crystallization/amorphization kinetics and mapping anisotropies.
... There are both hardware and software approaches to achieve that. On the hardware level, photonic memory [1], [53], [54] and interconnects [12], [14], [32] should be further studied since photonics has no RC delay or I 2 R loss. Besides, 3D integration can also reduce the data movement cost. ...
Preprint
Full-text available
The last few years have seen a lot of work to address the challenge of low-latency and high-throughput convolutional neural network inference. Integrated photonics has the potential to dramatically accelerate neural networks because of its low-latency nature. Combined with the concept of Joint Transform Correlator (JTC), the computationally expensive convolution functions can be computed instantaneously (time of flight of light) with almost no cost. This 'free' convolution computation provides the theoretical basis of the proposed PhotoFourier JTC-based CNN accelerator. PhotoFourier addresses a myriad of challenges posed by on-chip photonic computing in the Fourier domain including 1D lenses and high-cost optoelectronic conversions. The proposed PhotoFourier accelerator achieves more than 28X better energy-delay product compared to state-of-art photonic neural network accelerators.
... The XML identifier of the "iter2-3" potential model discussed in the main text is GAP_2022_4_7_480_18_6_12_970. Fig. 3a is based on ref. 72 , the structural drawings in Fig. 1 were created using VESTA, 73 ...
Preprint
Full-text available
Quantum-accurate computer simulations have played a central role in understanding phase-change materials (PCMs) for advanced memory technologies. However, the drastic growth in computational cost with model system size has precluded simulations on the length scales of real devices. Here we show that a single, compositionally flexible machine-learning (ML) interatomic potential model can describe the flagship Ge-Sb-Te PCMs under practical device conditions, including fully atomistic simulations of non-isothermal heating, and taking chemical disorder into account. The superior computing efficiency of the new approach enables simulations of multiple thermal cycles and delicate operations for neuro-inspired computing, namely, cumulative SET and iterative RESET. A device-scale capability demonstration (40 x 20 x 20 nm3) shows that the new ML potential can directly describe technologically relevant processes in PCM-based memory products. Our work demonstrates how atomistic ML-driven simulations are now entering a stage where they can guide architecture design for high-performance devices.
Article
Tunable thermal emitters have attracted much attention due to their applications in communications, sensing, and control system. To date, the emission performance of many reported thermal emitters require continuous external energy for the purpose of continuous modulation. Here, a Ge2Sb2Te5 (GST) array-based metasurface thermal emitter is proposed and measured. The maximum emissivity (crystalline) is 95.3%, while the minimum emission (amorphous) amplitude is 15.1%. Based on the phase transition states (amorphous and crystalline) of the GST array, the emission amplitude and resonance wavelength of this metasurface thermal emitter are clearly switchable. Moreover, the thickness of the GST arrays is set to be 1.5μm, 2.0μm, and 2.5μm, which results in the emissivity peak shifting to longer resonant wavelengths. The thermal emitter reveals continuous tunability, broad wavelength selectivity, and switchability.
Conference Paper
We demonstrate binary and multilevel electrical programming of the phase change material Ge2Sb2Te5 (GST) memory cells based on ion-implanted silicon-on-insulator (SOI) waveguide microheaters. GST cells can be reversibly switched by using low-amplitude electric pulses.
Conference Paper
We introduce a novel design for integrated phase-change material devices. A figure-of-merit is introduced to quantify the improvement of the proposed design over previous implementations. Additionally, the significance of non-linear effects is discussed.
Article
Photonic systems have seen a recent explosion of investigation for neuromorphic engineering and the execution of machine learning models, owing to the advantages of optics for communications and performing linear operations. Here, we present a novel architecture for executing reconfigurable multiwavelength photonic multiply-accumulate (MAC) operations based on all-solid-state vertically-integrated tunable cavities. We simulate this weighting action through full transfer-matrix calculations of a realistic candidate filter stack. By efficiently leveraging the out-of-plane dimension, we form a conceptual ‘smart pixel’ device that can achieve higher spectral channel count per area, or equivalently computational density, than comparable subpixel architectures and planar integrated photonic devices. While this technology is of use for neuromorphic photonic information processing, its usefulness extends to applications in sensing and communications by covering, for instance, high spatial resolution hyperspectral filters and reconfigurable routing.
Article
Functional devices based on electromagnetic (EM) tunable metasurfaces provide a feasible strategy for controlling of communication signals. Here, an EM metasurfaces based on Ge2Sb2Te5 (GST) layer is proposed and verified. At room temperature, this metasurface sample obtains a transmission window (resonant frequency 8.07 GHz, amplitude 0.92) based on the amorphous state of the GST layer. This transmission window is frequency selectable, which is achieved by changing the structural parameters (diameter of hole array, or thickness of dielectric layer). By changing the annealing temperature, this transmission window is always covered by the transmission valley, thus confirming the switchability of the metasurface sample. Finally, the sensitivity of the transmission window of this metasurface sample to liquid (water, acetone, ethanol, and PEG300) is measured and verified. The metasurface can be applied in communication, sensing, control and other fields.
Conference Paper
Here we demonstrate an ultra-low-loss multi-state photonic memory with phase change materials (GeSbSe), which can be efficiently reprogrammed on-chip. 4-bit memory and over 100,000 cycle tests are shown for this material.
Conference Paper
We proposed an integrated non-volatile optical memory cell using novel analog linear photo-response resistance switch based on 2.5D MoS 2 . It enables a thermally stable and scalable solution for integrated optical memory components. © 2022 The Author(s)
Article
Full-text available
Possessing the merits of high efficiency, low consumption, and versatility, emerging photonic memristive and memristive‐like devices exhibit an attractive future in constructing novel neuromorphic computing and miniaturized bionic electronic system. Recently, the potential of various emerging materials and structures for photonic memristive and memristive‐like devices has attracted tremendous research efforts, generating various novel theories, mechanisms, and applications. Limited by the ambiguity of the mechanism and the reliability of the material, the development and commercialization of such devices are still rare and in their infancy. Therefore, a detailed and systematic review of photonic memristive and memristive‐like devices is needed to further promote its development. In this review, the resistive switching mechanisms of photonic memristive and memristive‐like devices are first elaborated. Then, a systematic investigation of the active materials, which induce a pivotal influence in the overall performance of photonic memristive and memristive‐like devices, is highlighted and evaluated in various indicators. Finally, the recent advanced applications are summarized and discussed. In a word, it is believed that this review provides an extensive impact on many fields of photonic memristive and memristive‐like devices, and lay a foundation for academic research and commercial applications.
Conference Paper
Here we demonstrate low-loss multi-state photonic memory using broadband transparent phase change materials Ge 2 Sb 2 Se 3 and Sb 2 Se 3 , which can be efficiently reprogrammed on-chip.
Article
Full-text available
We present a hybrid device platform for creating an electrically reconfigurable metasurface formed by the integration of plasmonic nanostructures with phase-change material germanium antimony telluride (GST). By changing the phase of GST from amorphous to crystalline through Joule heating, a large range of responses from the metasurface can be achieved. Furthermore, by using the intermediate phases of GST, the metasurface can interact with the incident light in both over-coupling and under-coupling regimes, leading to an inherently broadband response. Through a detailed investigation of the nature of the fundamental modes, we demonstrate that changing the crystalline phase of the GST at the pixel-level enables an effective control over the key properties (i.e., amplitude, phase, and polarization) of incident light. This leads to the realization of a broadband electrically tunable multifunctional metadevice enabling beam switching, focusing, steering, and polarization conversion. Such a hybrid structure offers a high-speed, broadband, and nonvolatile reconfigurable paradigm for electrically programmable optical devices such as switches, holograms, and polarimeters.
Article
Phase-change optical device has recently gained tremendous interest due to its ultra-fast transmitting speed, multiplexing and large bandwidth. However, majority of phase-change optical devices are only devoted to on-chip components such as optical tensor core and optical main memory, while developing a secondary storage memory in an optical manner is rarely reported. To address this issue, we propose a novel phase-change optical memory based on plasmonic resonance effects for secondary storage applications. Such design makes use of the plasmonic dimer nanoantenna to generate plasmonic resonance inside the chalcogenide alloy, and thus enables the performance improvements in terms of energy consumption and switching speed. It is found that choosing height, radius, and separation of the plasmonic nanoantenna as 10 nm, 150 nm, and 10 nm, respectively, allows for a write/erase energies of 100 pJ and 240 pJ and a write/erase speed of 10 ns for crystallization and amorphization processes, respectively. Such performance merits encouragingly prevail conventional secondary storage memories and thus pave a route towards the advent of all-optical computer in near future.
Article
Recently, integrated optical devices working at 2-μm wavelengths have attracted considerable attention. Among such devices, a polarization beam splitter is an essential building block for constructing the on-chip circuit. Here, we have proposed an ultra-broadband polarization beam splitter with tunable transmissions working at 2-μm waveband. Coupled-mode based directional coupler is utilized for the device design. Subwavelength grating structures provide efficient dispersion control, which endows broad optical bandwidths. In addition, a rear-bent coupler is cascaded at the output of the device to furtherly suppress the crosstalk. The silicon-Ge2Sb2Se4Te1 hybrid materials provide tunable transmissions. When the Ge2Sb2Se4Te1 is in the amorphous state, with high extinction ratios more than 20 dB, the bandwidths exceed 200 nm (1900∼2100 nm) and 130 nm (1925∼2055 nm) for TE0 and TM0, respectively. Furthermore, the insertion losses are less than 0.24 dB (1900∼2100 nm) and 0.56 dB (1925∼2055 nm) for TE0 and TM0, respectively. When being in the crystalline state, the device works as an absorber that fully attenuates the input energy. In different intermediate amorphous–crystalline hybrid states, it is also possible to achieve tunable transmissions.
Conference Paper
We experimentally demonstrate for the first time an all-optical fully-integrated InP CAM cell within a complete CAM Matchline architecture with RAM table Encoding and Decoding functionalities. Error-free operation has been evaluated at 5 Gb/s.
Article
The biological neuromorphic system exhibits a high degree of connectivity to process information. Inspired by function, optoelectronic synapses are expected to pave a way to overcome the von Neumann bottleneck for nonconventional computing, which integrates synaptic and optical-sensing functions for visual information processing and complex learning and memory in an energy-efficient way. Herein, this review summarizes the working mechanisms of light-stimulated artificial synapses, including ionization and dissociation of oxygen vacancies, capture and release of carriers through barriers formed by heterojunctions, capture and release of carriers at the semiconductor and dielectric interface, and phase transition. Then, we present a comprehensive overview of the advanced progress in different material systems, including two-dimensional materials, organic material, metal halide, and metal oxide. The existing application scenarios of various synaptic devices are outlined. Finally, the current challenges and perspective toward the development of optoelectronic artificial synapses are briefly discussed for future applications.
Article
Photonic switches have attractive application prospects in optical communication data networks that require dynamic reconfiguration. Integrating optical switching devices with optical fiber, the most widely deployed photonic technology platform, can realize signal transmission and processing in practical applications. Here, we demonstrate the multilevel optical switching using the phase-change material Ge 2 Sb 2 Te 5 (GST) integrated on a graded-index multimode fiber. This switching process works by exploiting the significant difference in extinction coefficient between the crystalline state and the amorphous state of the GST. Using GST to achieve the switch function, no external energy source is needed to maintain the existing state of the switch, and the device is nonvolatile. This multi-level optical switch is an all-fiber integrated device. We apply GST to the end facets of the graded-index multimode fiber by magnetron sputtering, which is a reflective structure. A pulsing scheme is used to control the optical propagation state of the optical modulation signal to realize the switching function. It can store up to 11 non-volatile reliable and repeatable levels encoded by the pump source laser with a wavelength of 1550 nm. At the same time, the switching process between states is on the order of hundreds of nanoseconds. The present experimental results demonstrate the feasibility of 11 multilevel states in the field of optical fibers commonly used in communications. It can be well coupled with the all-fiber terminal device. It also shows that the device is still applicable in the 1525 nm∼1610 nm broadband range, promising for designing future multilevel photonic switches and memory devices.
Conference Paper
Here we demonstrate a low-loss multi-state photonic memory using broadband transparent phase change materials (GeSbSe), which can be efficiently reprogrammed on-chip.
Article
The low optical loss of G e 2 S b 2 S e 4 T e 1 (GSST) makes it a potential functional material for all-optical multilevel photonics memory devices that can operate in the optical telecommunication wavelength band. However, the same characteristic also restricted the tolerance of GSST phase change conditions using 1550 nm as an excitation light source. This work reports on the enhancement of GSST phase change condition tolerance using a graphene oxide (GO) intermediate layer on a polymer waveguide platform. The hybrid waveguide exhibits an insertion loss of around 1 dB and a maximum readout contrast of 25% between amorphous and crystalline states, with a step increase in readout contrast of around 5% per step. This work serves as a proof of concept for the implementation of a GSST–GO hybrid structure as an optical functional material in all-optical photonics memory applications.
Article
There has been growing interest in using photonic processors for performing neural network inference operations; however, these networks are currently trained using standard digital electronics. Here, we propose on-chip training of neural networks enabled by a CMOS-compatible silicon photonic architecture to harness the potential for massively parallel, efficient, and fast data operations. Our scheme employs the direct feedback alignment training algorithm, which trains neural networks using error feedback rather than error backpropagation, and can operate at speeds of trillions of multiply–accumulate (MAC) operations per second while consuming less than one picojoule per MAC operation. The photonic architecture exploits parallelized matrix–vector multiplications using arrays of microring resonators for processing multi-channel analog signals along single waveguide buses to calculate the gradient vector for each neural network layer in situ . We also experimentally demonstrate training deep neural networks with the MNIST dataset using on-chip MAC operation results. Our approach for efficient, ultra-fast neural network training showcases photonics as a promising platform for executing artificial intelligence applications.
Chapter
Silicon photonics is an emerging technology allowing to take the advantage of high-speed light propagation to accelerate computing kernels in integrated systems. Micrometer-scale optical devices call for reconfigurable architectures to maximize resources utilization. Typical reconfigurable optical computing architectures involve micro-ring resonators for electro-optic modulation. However, such devices require voltage and thermal tuning to compensate for fabrication process variability and thermal sensitivity. This power-hungry calibration leads to significant static power overhead, thus limiting the scalability of optical architectures. In this chapter, we propose to use non-volatile Phase Change Materials (PCM) elements to route optical signals only through the required resonators, hence saving calibration energy of bypassed resonators. The non-volatility of PCM elements allows maintaining the optical path. We investigate the efficiency of the PCM elements on the Reconfigurable Directed Logic (RDL) architecture. We also evaluate the static power saving induced by the use of couplers instead of microring to redirect WDM signals into a single waveguide. Finally, we show that the couplers can be efficiently used to cascade the architectures, allowing to increase the number of inputs to be processed without opto-electronic conversions. Compared to a ring-based implementation of RDL architecture, results show that the proposed implementation allows reducing the static power by 53% on average.KeywordsNanophotonicsPhase Change Material (PCM)Reconfigurable computing architectures
Article
A programmable hardware implementation of all-optical nonlinear activation functions for different scenarios and applications in all-optical neural networks is essential. We demonstrate a programmable, low-loss all-optical activation function device based on a silicon micro-ring resonator loaded with phase change materials. Four different nonlinear activation functions of Relu, ELU, Softplus and radial basis functions are implemented for incident signal light of the same wavelength. The maximum power consumption required to switch between the four different nonlinear activation functions in calculation is only 1.748 nJ. The simulation of classification of hand-written digit images also shows that they can perform well as alternative nonlinear activation functions. The device we design can serve as nonlinear units in photonic neural networks, while its nonlinear transfer function can be flexibly programmed to optimize the performance of different neuromorphic tasks.
Article
Phase Change Memory (PCM) is an attractive candidate for main memory as it offers non-volatility and zero leakage power, while providing higher cell densities, longer data retention time, and higher capacity scaling compared to DRAM. In PCM, data is stored in the crystalline or amorphous state of the phase change material. The typical electrically-controlled PCM (EPCM), however, suffers from longer write latency and higher write energy compared to DRAM and limited multi-level cell (MLC) capacities. These challenges limit the performance of data-intensive applications running on computing systems with EPCMs. Recently, researchers demonstrated optically-controlled PCM (OPCM) cells, with support for 5 bits / cell in contrast to 2 bits / cell in EPCM. These OPCM cells can be accessed directly with optical signals that are multiplexed in high-bandwidth-density silicon-photonic links. The higher MLC capacity in OPCM and the direct cell access using optical signals enable an increased read/write throughput and lower energy per access than EPCM. However, due to the direct cell access using optical signals, OPCM systems cannot be designed using conventional memory architecture. We need a complete redesign of the memory architecture that is tailored to the properties of OPCM technology. This paper presents the design of a unified network and main memory system called COSMOS that combines OPCM and silicon-photonic links to achieve high memory throughput. COSMOS is composed of a hierarchical multi-banked OPCM array with novel read and write access protocols. COSMOS uses an Electrical-Optical-Electrical (E-O-E) control unit to map standard DRAM read/write commands (sent in electrical domain) from the memory controller on to optical signals that access the OPCM cells. Our evaluation of a 2.5D-integrated system containing a processor and COSMOS demonstrates 2.14 × average speedup across graph and HPC workloads compared to an EPCM system. COSMOS consumes 3.8 × lower read energy-per-bit and 5.97 × lower write energy-per-bit compared to EPCM. COSMOS is the first non-volatile memory that provides comparable performance and energy consumption as DDR5 in addition to increased bit density, higher area efficiency and improved scalability.
Article
We present a device driving testing platform based on vertically integrated nano light emitting diodes (nano-LEDs). The nano-LEDs with a peak wavelength emission centered at ∼ 445 nm were arranged in arrays and conditioned using a laser-micro-annealing process to individually tune their intensity. They were coupled with freestanding monocrystalline Ge1Sb2Te4 nano-membranes with three different thicknesses (∼ 40, ∼ 60 and ∼ 90 nm) with the aim of initializing ultrafast switching processes and of observing phase changed states simultaneously by Raman spectroscopy. Raman spectroscopy studies reveal that the optical pulses emitted from the nano-LEDs induce substantial, local changes in the nano-membranes‘ states of the Ge1Sb2Te4 layered material. Beside the crystalline state in non-exposed areas (as-grown material), amorphous and different intermediate states were identified in exposed areas as island-like structures with diameters ranging from ∼ 300 nm up to ∼ 1.5 µm. The latter confirms the nano-LEDs‘ emission role in both near- and far-field regimes, depending on the distance between nano-LED and nano-membrane, for driving i.e. inducing the phase change process. The results presented demonstrate the suitability and potential of the vertically integrated nano-LEDs as the key components for a testing platform/for electro-optical convertors driving phase change processes in active optical media. They could also play an important role in the development of future, e.g., non-volatile data storage as well as in optical and neuromorphic computing architectures based on transmistor devices.
Article
Full-text available
The switchable optical and electrical properties of phase change materials (PCMs) are finding new applications beyond data storage in reconfigurable photonic devices. However, high power heat pulses are needed to melt-quench the material from crystalline to amorphous. This is especially true in silicon photonics, where the high thermal conductivity of the waveguide material makes heating the PCM energy inefficient. Here, we improve the energy efficiency of the laser-induced phase transitions by inserting a layer of two-dimensional (2D) material, either MoS2 or WS2, between the silica or silicon substrate and the PCM. The 2D material reduces the required laser power by at least 40% during the amorphization (RESET) process, depending on the substrate. Thermal simulations confirm that both MoS2 and WS2 2D layers act as a thermal barrier, which efficiently confines energy within the PCM layer. Remarkably, the thermal insulation effect of the 2D layer is equivalent to a ∼100 nm layer of SiO2. The high thermal boundary resistance induced by the van der Waals (vdW)-bonded layers limits the thermal diffusion through the layer interface. Hence, 2D materials with stable vdW interfaces can be used to improve the thermal efficiency of PCM-tuned Si photonic devices. Furthermore, our waveguide simulations show that the 2D layer does not affect the propagating mode in the Si waveguide; thus, this simple additional thin film produces a substantial energy efficiency improvement without degrading the optical performance of the waveguide. Our findings pave the way for energy-efficient laser-induced structural phase transitions in PCM-based reconfigurable photonic devices.
Article
In electronic circuits, memristors have been defined as resistors whose resistance depends on past signals. Such elements have promise as low-power weighting and self-learning elements in electronic neuromorphic circuits. Photonic memristors, elements whose transmission depends on past optical signals, are experiencing renewed study alongside the rise in interest in neuromorphic photonics. One potential route to creating photonic memristors involves incorporating photochromic materials, whose optical properties continuously change with optical illumination, into photonic integrated circuits (PICs). In this manuscript we lay out a theoretical model to study the transmission dynamics of devices incorporating photochromic compounds into SiN planar microring resonators which utilize slot waveguide structure, and show that such devices fulfill the criteria for memristive behaviour. This represents a practical path towards incorporating photonic memristors into a technologically mature material platform with minimal additional fabrication processes.
Chapter
Full-text available
In this chapter, we will focus on plasmonic materials and their applications. Section 4.1 briefly introduces the research history of plasmon resonance. Section 4.2 presents some fundamental physics of plasmon resonance, including classification and excitation conditions. Section 4.3 describes the available plasmonic materials. Section 4.4 presents the fabrication method of plasmonic nanostructures. Section 4.5 presents related applications based on plasmon resonances. Finally, a summary and outlook are given in Section 4.6.
Article
In the new computing world, phase-change memory (PCM) has recently evolved as a non-volatile key-enabling technology that has been used as memory storage. PCM has also been explored as non-von Neumann computing for neuromorphic computing applications. It was discovered in the 1960s, still, there are many questions related to its thermal stability, endurance, electrical and structural dynamics. To enhance thermal stability and operation speed, many materials have been explored yet, some rare-earth elements prove to be suitable materials for improving the performance of the device. The article describes the applications of the PCM and the basic processes involved in the working of the device. READ and WRITE processes, material exploration, and cell designs have been discussed, concluding that mushroom-type cell design is good for fabricating PCM devices. Various performance-related properties of the device have been discussed including scalability, reliability, and variability. Finally, an outlook and future scope have been discussed.
Article
Full-text available
Despite their importance in applications such as nonvolatile memory, integrated photonics, and compact optics, the crystalline‐to‐amorphous transition in chalcogenide phase‐change materials (PCMs) is not understood. Herein, this transition in a technologically relevant infrared (IR) transparent chalcogenide material, Ge2Sb2Se4Te1 (GSST), is examined. Thin films of GSST using fully depleted silicon on insulator (FDSOI) microheaters are discussed and the phase transitions by polarized and unpolarized Raman spectroscopy is studied. It is confirmed that the crystalline‐to‐amorphous transition is driven by conversion of Ge–6Se octahedra to Ge–4Se tetrahedra with the extra Se being incorporated into an Se—Se network. This is similar to the mechanism reported in earlier work for Ge2Sb2Te5 (GST). Recrystallization requires disrupting the Se—Se network and the crystallization activation energy is consistent with the Se—Se bond energy. Across 1000 crystallization–amorphization cycles, GSST exhibits no qualitative change in the Raman spectrum, suggesting limited film oxidation or chemical decomposition. After several hundred cycles, recrystallization is less complete, likely due to dewetting of GSST during the high‐temperature amorphization step leading to compromise of the capping layer and loss of GSST. The utility of GSST as a photonic material through fabrication and testing of a GSST‐coated, integrated silicon photonic Mach–Zender interferometer, is discussed. The crystalline‐to‐amorphous transition of Ge2Sb2Se4Te1 (GSST) is investigated by in situ Raman analysis of GSST thin films which are rapidly heated and quenched by silicon‐on‐insulator microheaters. The transition is consistent with the conversion of Ge–6Se octahedral units to Ge–4Se tetrahedral units with the release of Se atoms forming an amorphous Se network.
Chapter
Traditional artificial visual system consisting of separated photodetector, memory unit, and processing unit is facing the problems of high energy consumption and high delay, not conducive to the development of real-time processing. It is in an urgent need to develop sensing-memory-computing electronics for high-efficiency information processing, breaking the bottleneck of separated functional units in artificial visual system. Emerging neuromorphic computing memristors are considered as the most attractive candidate for next-generation sensing-memory-computing electronics owing to excellent characteristics including low power consumption, high speed, and low operation voltage. Various materials and device structures were developed to fabricate neuromorphic computing memristors, such as metal oxide, two-dimensional materials, organic material, phase change material, and ferroelectric materials. Binary oxide material, ternary oxide material, and oxide heterojunctions could be fabricated as active layers in oxide-based memristor with the advantages of CMOS compatibility, uniform distribution operating voltage, and excellent endurance characteristics. Two-dimensional materials have shown great advantages in mechanical flexibility, dangling-bond-free lattice, tunable bandgap, and diverse heterostructures. Organic materials have advantages of low Young’s modulus and easily changing properties by chemical design. Phase change materials show advantages in low power consumption, high speed, multi-bit storage, and optical sensing. Ferroelectric materials have advantages in dielectric, piezoelectric, pyroelectric, electro-optic, and acousto-optic effect. However, there is still a long way to go for industrial applications of sensing-memory-computing devices with these materials. The CMOS compatibility, high-density integration capability, low cost, and stable performance of memristors are key factors of sensing-memory-computing devices for industrial applications. For future sensing-memory-computing chips, novel heterogeneous integration of different material systems and structure by combining the advantages of specific material system may be the possible direction of the next-generation real-time processing neuromorphic system.KeywordsNeuromorphic systemSensing-memory-computingMemristorLow power consumption
Conference Paper
Here we demonstrate a low-loss multi-state photonic memory using broadband transparent phase change materials (GeSbSe), which can be efficiently reprogrammed on-chip.
Article
Full-text available
Phase change memory (PCM) is an emerging technology that combines the unique properties of phase change materials with the potential for novel memory devices, which can help lead to new computer architectures. Phase change materials store information in their amorphous and crystalline phases, which can be reversibly switched by the application of an external voltage. This article describes the advantages and challenges of PCM. The physical properties of phase change materials that enable data storage are described, and our current knowledge of the phase change processes is summarized. Various designs of PCM devices with their respective advantages and integration challenges are presented. The scaling limits of PCM are addressed, and its performance is compared to competing existing and emerging memory technologies. Finally, potential new applications of phase change devices such as neuromorphic computing and phase change logic are outlined.
Article
Full-text available
Conventional flash memory devices are voltage driven and found to be unsafe for confidential data storage. To ensure the security of the stored data, there is a strong demand for developing novel nonvolatile memory technology for data encryption. Here we show a photonic flash memory device, based on upconversion nanocrystals, which is light driven with a particular narrow width of wavelength in addition to voltage bias. With the help of near-infrared light, we successfully manipulate the multilevel data storage of the flash memory device. These upconverted photonic flash memory devices exhibit high ON/OFF ratio, long retention time and excellent rewritable characteristics.
Article
Full-text available
An optical switch operating at a wavelength of 1.55 μm and showing a 12 dB modulation depth is introduced. The device is implemented in a silicon racetrack resonator using an overcladding layer of the phase change data storage material Ge2Sb2Te5, which exhibits high contrast in its optical properties upon transitions between its crystalline and amorphous structural phases. These transitions are triggered using a pulsed laser diode at λ = 975 nm and used to tune the resonant frequency of the resonator and the resultant modulation depth of the 1.55 μm transmitted light.
Article
Full-text available
The phase transformation dynamics induced in Ge2Sb2Te5 films by picosecond laser pulses were studied using real-time reflectivity measurements. With subnanosecond resolution. Evidence was found that the thermal diffusivity of the substrate plays a crucial role in, determining the ability of. the films to crystallize and amorphize. A film/substrate configuration with optimized heat flow. conditions for ultrafast phase cycling with picosecond laser pulses was designed and produced. In this system; we achieved reversible phase transformations with large optical contrast (>20%) using single laser pulses with a duration of 30 ps within well-defined fluence windows. The amorphization (writing) process is completed within less than 1 ns, whereas, crystallization (erasing) needs approximately 13 ns to be completed. (C) 2004 American Institute of Physics.
Article
Full-text available
Three-dimensional integration may allow for continued improvements in the speed, density and cost of non-volatile memory.
Article
Full-text available
Light is intrinsically very difficult to store in a small space. The ability to trap photons for a long time (photon lifetime, τph) and to slow the propagation of light plays a significant role in quantum information1, 2, 3 and optical processing4, 5, 6. Photonic-crystal cavities with an ultrahigh quality factor (Q) are attracting attention7, 8 because of their extremely small volume; however, high-Q demonstrations have been accomplished only with spectral measurements9, 10, 11. Here we describe time-domain measurements on photonic-crystal cavities with the highest Q among wavelength-scale cavities, and show directly that photons are trapped for one nanosecond. These techniques constitute clear and accurate ways of investigating ultrasmall and long τph systems. We also show that optical pulses are delayed for ~1.45 ns, corresponding to light propagation at ~2×10−5 c the speed of light in a vacuum, which is the slowest for any dielectric slow-light medium. Furthermore, we succeeded in dynamically changing the Q within the τph, which is key to realizing the dynamic control of light12, 13 and photon-trapping memory14.
Article
Full-text available
We propose an all-photonic, non-volatile memory and processing element based on phase-change thin-films deposited onto nanophotonic waveguides. Using photonic microring resonators partially covered with Ge2Sb2Te5 (GST) multi-level memory operation in integrated photonic circuits can be achieved. GST provides a dramatic change in refractive index upon transition from the amorphous to crystalline state, which is exploited to reversibly control both the extinction ratio and resonance wavelength of the microcavity with an additional gating port in analogy to optical transistors. Our analysis shows excellent sensitivity to the degree of crystallization inside the GST, thus providing the basis for non-von Neuman neuromorphic computing.
Article
Full-text available
A comprehensive and thorough review of PCM technologies, including a discussion of material and device issues, is provided in this paper. ABSTRACT | In this paper, recent progress of phase change memory (PCM) is reviewed. The electrical and thermal proper-ties of phase change materials are surveyed with a focus on the scalability of the materials and their impact on device design. Innovations in the device structure, memory cell selector, and strategies for achieving multibit operation and 3-D, multilayer high-density memory arrays are described. The scaling prop-erties of PCM are illustrated with recent experimental results using special device test structures and novel material synthe-sis. Factors affecting the reliability of PCM are discussed.
Conference Paper
Full-text available
Phase-change memory (PCM) has emerged as one among the most promising technologies for next-generation non-volatile solid-state memory. Multilevel storage, namely storage of non-binary information in a memory cell, is a key factor for reducing the total cost-per-bit and thus increasing the competiveness of PCM technology in the nonvolatile memory market. In this paper, we present a family of advanced programming schemes for multilevel storage in PCM. The proposed schemes are based on iterative write-and-verify algorithms that exploit the unique programming characteristics of PCM in order to achieve significant improvements in resistance-level packing density, robustness to cell variability, programming latency, energy- per-bit and cell storage capacity. Experimental results from PCM test-arrays are presented to validate the proposed programming schemes. In addition, the reliability issues of multilevel PCM in terms of resistance drift and read noise are discussed.
Conference Paper
Full-text available
Ultra-small, low-power, all-optical switching and memory elements, such as all-optical flip-flops, as well as photonic integrated circuits of many such elements, are in great demand for all-optical signal buffering, switching and processing. Silicon-on-insulator is considered to be a promising platform to accommodate such photonic circuits in large-scale configurations. Through heterogeneous integration of InP membranes onto silicon-on-insulator, a single microdisk laser with a diameter of 7.5 µm, coupled to a silicon-on-insulator wire waveguide, is demonstrated here as an all-optical flip-flop working in a continuous-wave regime with an electrical power consumption of a few milliwatts, allowing switching in 60 ps with 1.8 fJ optical energy. The total power consumption and the device size are, to the best of our knowledge, the smallest reported to date at telecom wavelengths. This is also the only electrically pumped, all-optical flip-flop on silicon built upon complementary metal-oxide semiconductor technology.
Article
Full-text available
Phase-change memory technology relies on the electrical and optical properties of certain materials changing substantially when the atomic structure of the material is altered by heating or some other excitation process. For example, switching the composite Ge(2)Sb(2)Te(5) (GST) alloy from its covalently bonded amorphous phase to its resonantly bonded metastable cubic crystalline phase decreases the resistivity by three orders of magnitude, and also increases reflectivity across the visible spectrum. Moreover, phase-change memory based on GST is scalable, and is therefore a candidate to replace Flash memory for non-volatile data storage applications. The energy needed to switch between the two phases depends on the intrinsic properties of the phase-change material and the device architecture; this energy is usually supplied by laser or electrical pulses. The switching energy for GST can be reduced by limiting the movement of the atoms to a single dimension, thus substantially reducing the entropic losses associated with the phase-change process. In particular, aligning the c-axis of a hexagonal Sb(2)Te(3) layer and the 〈111〉 direction of a cubic GeTe layer in a superlattice structure creates a material in which Ge atoms can switch between octahedral sites and lower-coordination sites at the interface of the superlattice layers. Here we demonstrate GeTe/Sb(2)Te(3) interfacial phase-change memory (IPCM) data storage devices with reduced switching energies, improved write-erase cycle lifetimes and faster switching speeds.
Article
Full-text available
The search for a universal memory storage device that combines rapid read and write speeds, high storage density and non-volatility is driving the exploration of new materials in nanostructured form. Phase-change materials, which can be reversibly switched between amorphous and crystalline states, are promising in this respect, but top-down processing of these materials into nanostructures often damages their useful properties. Self-assembled nanowire-based phase-change material memory devices offer an attractive solution owing to their sub-lithographic sizes and unique geometry, coupled with the facile etch-free processes with which they can be fabricated. Here, we explore the effects of nanoscaling on the memory-storage capability of self-assembled Ge2Sb2Te5 nanowires, an important phase-change material. Our measurements of write-current amplitude, switching speed, endurance and data retention time in these devices show that such nanowires are promising building blocks for non-volatile scalable memory and may represent the ultimate size limit in exploring current-induced phase transition in nanoscale systems.
Article
Full-text available
Phase-change materials offer a promising route for the practical realisation of new forms of general-purpose and 'brain-like' computers. An experimental proof-of-principle of such remakable capabilities is presented that includes (i) the reliable execution by a phase-change 'processor' of the four basic arithmetic functions of addition, subtraction, multiplication and division, (ii) the demonstration of an 'integrate and fire' hardware neuron using a single phase-change cell and (iii) the expostion of synaptic-like functionality via the 'memflector', an optical analogue of the memristor.
Article
Full-text available
We survey the current state of phase change memory (PCM), a non-volatile solid-state memory technology built around the large electrical contrast between the highly-resistive amorphous and highly-conductive crystalline states in so-called phase change materials. PCM technology has made rapid progress in a short time, having passed older technologies in terms of both sophisticated demonstrations of scaling to small device dimensions, as well as integrated large-array demonstrators with impressive retention, endurance, performance and yield characteristics. We introduce the physics behind PCM technology, assess how its characteristics match up with various potential applications across the memory-storage hierarchy, and discuss its strengths including scalability and rapid switching speed. We then address challenges for the technology, including the design of PCM cells for low RESET current, the need to control device-to-device variability, and undesirable changes in the phase change material that can be induced by the fabrication procedure. We then turn to issues related to operation of PCM devices, including retention, device-to-device thermal crosstalk, endurance, and bias-polarity effects. Several factors that can be expected to enhance PCM in the future are addressed, including Multi-Level Cell technology for PCM (which offers higher density through the use of intermediate resistance states), the role of coding, and possible routes to an ultra-high density PCM technology. Comment: Review article
Article
Full-text available
Could optical technology offer a solution to the heat generation and bandwidth limitations that the computing industry is starting to face? The benefits of energy-efficient passive components, low crosstalk and parallel processing suggest that the answer may be yes.
Article
Full-text available
Phase-change materials are some of the most promising materials for data-storage applications. They are already used in rewriteable optical data storage and offer great potential as an emerging non-volatile electronic memory. This review looks at the unique property combination that characterizes phase-change materials. The crystalline state often shows an octahedral-like atomic arrangement, frequently accompanied by pronounced lattice distortions and huge vacancy concentrations. This can be attributed to the chemical bonding in phase-change alloys, which is promoted by p-orbitals. From this insight, phase-change alloys with desired properties can be designed. This is demonstrated for the optical properties of phase-change alloys, in particular the contrast between the amorphous and crystalline states. The origin of the fast crystallization kinetics is also discussed.
Article
Photonic signals were efficiently stored in a semiconductor-based memory cell. The incident photons were converted to electron-hole pairs that were locally stored in a quantum well that was laterally modulated by a field-effect tunable electrostatic superlattice. At large superlattice potential amplitudes, these pairs were stored for a time that was at least five orders of magnitude longer than their natural lifetime. At an arbitrarily chosen time, they were released in a short and intense flash of incoherent light, which was triggered by flattening the superlattice amplitude.
Article
Photonic integration has long been pursued, but remains immature compared with electronics. Nanophotonics is expected to change this situation. However, despite the recent success of nanophotonic devices, there has been no demonstration of large-scale integration. Here, we describe the large-scale and dense integration of optical memories in a photonic crystal chip. To achieve this, we introduce a wavelength-addressable serial integration scheme using a simple cavity-optimization rule. We fully exploit the wavelength-division-multiplexing capability, which is the most important advantage of photonics over electronics, and achieve an extremely large wavelength-channel density. This is the first demonstration of the large-scale photonic integration of nanophotonic devices coupled to waveguides in a single chip, and also the first dense wavelength-division-multiplexing nanophotonic devices other than filters. This work paves the way for optical random-access memories and for a large-scale wavelength-division-multiplexing photonic network-on-chip.
Article
An effective solution to enhance the capacity of an optical-interconnect link is utilizing advanced multiplexing technologies, like wavelength-division-multiplexing (WDM), polarization-division multiplexing (PDM), spatial-division multiplexing (SDM), bi-directional multiplexing, etc. On-chip (de)multiplexers are necessary as key components for realizing these multiplexing systems and they are desired to have small footprints due to the limited physical space for on-chip optical interconnects. As silicon photonics has provided a very attractive platform to build ultrasmall photonic integrated devices with CMOS-compatible processes, in this paper we focus on the discussion of silicon-based (de)multiplexers, including WDM filters, PDM devices, and SDM devices. The demand of devices to realize a hybrid multiplexing technology (combining WDM, PDM and SDM) as well as a bidirectional multiplexing technologies are also discussed to achieve Peta-bit optical interconnects.
Article
Inspired by the brain’s structure, we have developed an efficient, scalable, and flexible non–von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.
Article
The development of materials whose refractive index can be optically transformed as desired, such as chalcogenide-based phase-change materials, has revolutionized the media and data storage industries by providing inexpensive, high-speed, portable and reliable platforms able to store vast quantities of data. Phase-change materials switch between two solid states--amorphous and crystalline--in response to a stimulus, such as heat, with an associated change in the physical properties of the material, including optical absorption, electrical conductance and Young's modulus. The initial applications of these materials (particularly the germanium antimony tellurium alloy Ge2Sb2Te5) exploited the reversible change in their optical properties in rewritable optical data storage technologies. More recently, the change in their electrical conductivity has also been extensively studied in the development of non-volatile phase-change memories. Here we show that by combining the optical and electronic property modulation of such materials, display and data visualization applications that go beyond data storage can be created. Using extremely thin phase-change materials and transparent conductors, we demonstrate electrically induced stable colour changes in both reflective and semi-transparent modes. Further, we show how a pixelated approach can be used in displays on both rigid and flexible films. This optoelectronic framework using low-dimensional phase-change materials has many likely applications, such as ultrafast, entirely solid-state displays with nanometre-scale pixels, semi-transparent 'smart' glasses, 'smart' contact lenses and artificial retina devices.
Article
Phase change materials are widely considered for application in non-volatile memories due to their ability to achieve phase transformation in nano-second time scale. However the knowledge of fast crystallization dynamics in these materials are limited due to the lack of fast and accurate temperature control methods. In this work we have developed an experimental methodology that enables ultra-fast characterization of phase-change dynamics on a more technologically relevant melt-quenched amorphous phase using practical PCM device structures. We have extracted the crystallization growth velocity (U) in a functional capped PCM device over eight orders of magnitude (10(-10) m/s < U < 10(-1) m/s) spanning a wide temperature range (415 K < T < 580 K). We also observed direct evidence of non-Arrhenius crystallization behavior in programmed PCM devices at very high heating rates (>10(8) K/s), which reveals the extreme fragility of Ge2Sb2Te5 in its super-cooled liquid phase. Furthermore, these crystallization properties were studied as a function of device programming cycles and the results show degradation in the cell retention properties due to elemental segregation. The above experiments are enabled by the use of an on-chip fast heater and thermometer called as Micro Thermal Stage (MTS) integrated with a vertical phase change memory (PCM) cell. The temperature at the PCM layer can be controlled up to 600 K using MTS and with a thermal time constant of 800 ns leading to heating rates ~ 10(8) K/s that are close to the typical device operating conditions during PCM programming. The MTS allows us to independently control the electrical and thermal aspects of phase transformation (inseparable in a conventional PCM cell) and extract the temperature dependence of key material properties in real PCM devices.
Article
Phase-change materials integrated into nanophotonic circuits provide a flexible way to realize tunable optical components. Relying on the enormous refractive-index contrast between the amorphous and crystalline states, such materials are promising candidates for on-chip photonic memories. Non-volatile memory operation employing arrays of microring resonators is demonstrated as a route toward all-photonic chipscale information processing.
Article
A class of two-terminal passive circuit elements that can also act as memories could be the building blocks of a form of massively parallel computation known as memcomputing.
Article
Optical computers will be more interesting if they take advantage of phenomena that are unique to optics. In this respect, telecommunications hardware might have something to offer.
Article
Historically, the application of phase-change materials and devices has been limited to the provision of non-volatile memories. Recently, however, the potential has been demonstrated for using phase-change devices as the basis for new forms of brain-like computing, by exploiting their multilevel resistance capability to provide electronic mimics of biological synapses. Here, a different and previously under-explored property that is also intrinsic to phase-change materials and devices, namely accumulation, is exploited to demonstrate that nanometer-scale electronic phase-change devices can also provide a powerful form of arithmetic computing. Complicated arithmetic operations are carried out, including parallel factorization and fractional division, using simple nanoscale phase-change cells that process and store data simultaneously and at the same physical location, promising a most efficient and effective means for implementing beyond von-Neumann computing. This same accumulation property can be used to provide a particularly simple form phase-change integrate-and-fire “neuron”, which, by combining both phase-change synapse and neuron electronic mimics, potentially opens up a route to the realization of all-phase-change neuromorphic processing.
Article
Ultra-compact four-channel wavelength division multiplexing (WDM) devices in add/drop filter and multi-mode interferometer configurations were demonstrated on silicon-on-insulator for future on-chip optical interconnects. Both devices show a crosstalk level ≤-13dB and a footprint ≤0.006mm2.
Article
A rapid and reversible transition between a highly resistive and conductive state effected by an electric field, which we have observed in various types of disordered semiconducting material, is described in detail. The switching parameters and chemical composition of a typical material are presented, and microscopic mechanisms for the conduction phenomena are suggested.
Article
Silicon photonics is currently a very active and progressive area of research, as silicon optical circuits have emerged as the replacement technology for copper-based circuits in communication and broadband networks. The demand for ever improving communications and computing performance continues, and this in turn means that photonic circuits are finding ever increasing application areas. This text provides an important and timely overview of the hot topics in the field, covering the various aspects of the technology that form the research area of silicon photonics. With contributions from some of the worlds leading researchers in silicon photonics, this book collates the latest advances in the technology. Silicon Photonics: the State of the Art opens with a highly informative foreword, and continues to feature: the integrated photonic circuit; silicon photonic waveguides; photonic bandgap waveguides; mechanisms for optical modulation in silicon; silicon based light sources; optical detection technologies for silicon photonics; passive silicon photonic devices; photonic and electronic integration approaches; applications in communications and sensors. Silicon Photonics: the State of the Art covers the essential elements of the entire field that is silicon photonics and is therefore an invaluable text for photonics engineers and professionals working in the fields of optical networks, optical communications, and semiconductor electronics. It is also an informative reference for graduate students studying for PhD in fibre optics, integrated optics, optical networking, microelectronics, or telecommunications.
Article
Phase-change random-access memory (PCRAM) is one of the leading candidates for next-generation data-storage devices, but the trade-off between crystallization (writing) speed and amorphous-phase stability (data retention) presents a key challenge. We control the crystallization kinetics of a phase-change material by applying a constant low voltage via prestructural ordering (incubation) effects. A crystallization speed of 500 picoseconds was achieved, as well as high-speed reversible switching using 500-picosecond pulses. Ab initio molecular dynamics simulations reveal the phase-change kinetics in PCRAM devices and the structural origin of the incubation-assisted increase in crystallization speed. This paves the way for achieving a broadly applicable memory device, capable of nonvolatile operations beyond gigahertz data-transfer rates.
Article
The structure of laser-crystallized thin films of Ge <sub> 2 </sub> Sb <sub>2+x</sub> Te <sub> 5 </sub> (0.0≪x≤1.0) formed by the sputtering method were identified by x-ray diffraction studies to be composed of two phases: one phase is chiefly NaCl type crystal with a lattice constant of about 6 Å and a composition corresponding to Ge <sub> 2 </sub> Sb <sub> 2 </sub> Te <sub> 5 </sub>; the other phase comprises a small amount of an amorphous component such as Sb metal. Results of the Rietveld and the whole-powder-pattern fitting analyses show good agreement when assuming that (i) the 4(a) site is wholly occupied by only Te, (ii) the 4(b) site is randomly occupied by Ge or Sb atoms, and (iii) a little less than 20% of the 4(b) site is always vacant independent of the x value. The above results and the fact that halo noise rises with x increasing from 0.0 to 1.0 indicate a more precise model of crystal structure as follows. That is, Sb atoms added beyond the stoichiometric ratio, Ge <sub> 2 </sub> Sb <sub> 2 </sub> Te <sub> 5 </sub>, never fill up the vacancies of the 4(b) site in the NaCl type structure; the excess Sb atoms will remain in the amorphous state and concentrate, for example, at the grain boundary. The authors conclude that the amounts of the amorphous component produced through the crystallization process predominantly determine the crystallization rates and the critical temperatures of Ge–Sb–Te amorphous films, reportedly that they show a gentle and continuous dependence on the compositional deviation from the GeTe–Sb <sub> 2 </sub> Te <sub> 3 </sub> pse- udobinary line. © 2000 American Institute of Physics.
Article
An optical gate switch using Ge<sub>2</sub>Sb<sub>2</sub>Te<sub>5</sub> phase-change material integrated with a silicon waveguide is reported. The switch is very small (~2 Â¿m) owing to the large difference in absorption coefficient between the crystalline state and the amorphous state. The prototype switch has been fabricated and successfully switched from the transparent on-state to the opaque off-state by laser pulse irradiation. An extinction ratio of more than 12.5 dB was achieved over a wavelength range of 75 nm.
Article
The increasing speed of fibre-optic-based telecommunications has focused attention on high-speed optical processing of digital information. Complex optical processing requires a high-density, high-speed, low-power optical memory that can be integrated with planar semiconductor technology for buffering of decisions and telecommunication data. Recently, ring lasers with extremely small size and low operating power have been made, and we demonstrate here a memory element constructed by interconnecting these microscopic lasers. Our device occupies an area of 18 x 40 microm2 on an InP/InGaAsP photonic integrated circuit, and switches within 20 ps with 5.5 fJ optical switching energy. Simulations show that the element has the potential for much smaller dimensions and switching times. Large numbers of such memory elements can be densely integrated and interconnected on a photonic integrated circuit: fast digital optical information processing systems employing large-scale integration should now be viable.
Article
Non-volatile 'flash' memories are key components of integrated circuits because they retain their data when power is interrupted. Despite their great commercial success, the semiconductor industry is searching for alternative non-volatile memories with improved performance and better opportunities for scaling down the size of memory cells. Here we demonstrate the feasibility of a new semiconductor memory concept. The individual memory cell is based on a narrow line of phase-change material. By sending low-power current pulses through the line, the phase-change material can be programmed reversibly between two distinguishable resistive states on a timescale of nanoseconds. Reducing the dimensions of the phase-change line to the nanometre scale improves the performance in terms of speed and power consumption. These advantages are achieved by the use of a doped-SbTe phase-change material. The simplicity of the concept promises that integration into a logic complementary metal oxide semiconductor (CMOS) process flow might be possible with only a few additional lithographic steps.
The authors also acknowledge support from the DFG and the State of Baden-Württemberg through the DFG-Center for Functional Nanostructures (CFN) within subproject A6.4
• Wirtschaft
Wirtschaft (sdw). H.B. acknowledges support from the John Fell Fund and the EPSRC (EP/J00541X/2 and EP/J018694/1).The authors also acknowledge support from the DFG and the State of Baden-Württemberg through the DFG-Center for Functional Nanostructures (CFN) within subproject A6.4. This work was partly carried out with the support of the Karlsruhe Nano Micro Facility (KNMF, http://www.knmf.kit.edu), a Helmholtz Research Infrastructure at Karlsruhe Institute of Technology (KIT, http://www. kit.edu). The authors thank S. Diewald for assistance with device fabrication and M. Blaicher for technical assistance with device design.