Plasmonic nanogap enhanced phase-change devices with dual electrical-optical functionality

Although integrated photonics has gained great traction over the last decade, primarily for its potential to overcome fundamental limitations of today’s electronic circuitry (1), the conversion of electrical and optical signals seamlessly on a chip has remained elusive. The development of compact devices for efficient electro-optic conversion holds great importance as sharing the computing load between the electrical and optical domains shows increasing promise for applications including integrated optical switches, reconfigurable photonic circuits, photonic artificial neural networks, and more (2–4). Phase-change materials (PCMs) are considered outstanding candidates for dual-mode operation as they, in principle, provide both electrical and optical modulation functionality. To this effect, several devices implementing nonvolatile, optical PCMs have been proposed (5, 6), but none have been successfully demonstrated on an integrated platform. This is because the high electrical contrast between the conductive and insulating state in PCMs requires very close spacing between the metal contacts (usually tens of nanometers) to initiate a phase transition (7). In addition, the resulting conductive region formed after electrical switching is, at most, a few hundred nanometers in diameter—thus reducing the total volume of material for light-matter interaction (8).

Combining plasmonics with PCMs is a particularly promising approach for satisfying such stringent requirements, since the dimensions of such devices can be reduced to tens of nanometers and smaller—significantly below the diffraction limit of traditional optical devices (9). The combination of high electrical conductivity and strong plasmonic resonance at optical wavelengths in silver and gold has led to extremely compact electro-optic nanogap devices such as integrated light sources (10), photodetectors (11, 12), and modulators (13, 14). In addition, the extremely high field enhancement possible with subwavelength nanogaps enables very high–sensitivity spectral measurements for applications such as label-free detection of biomolecules (15, 16).

While plasmonics permits very strong light-matter interaction in nanometer-scale devices, the relatively high loss of metals at optical frequencies makes guiding light inefficient. Combining integrated photonics with nanoscale plasmonics, however, allows for both low-loss light delivery and strong light-matter interaction in a compact footprint (17, 18). Here, we combine waveguide-integrated plasmonic nanogaps with a PCM, Ge2Sb2Te5 (GST), to create an electro-optic memory cell that is fully addressable in both electrical and optical domains. Previous demonstrations of such mixed-mode devices either have used nonvolatile PCMs, such as VO₂ (19, 20), which requires significant power consumption for data retention, or have been limited to write/erase operations either electrically or optically but not both (21–24). By exploiting both the nanoscale dimensions and strong field confinement of a plasmonic nanogap, we enable both electrical and optical nonvolatile switching of GST within the gap, allowing for full mixed-mode operation of a PCM memory cell.

A three-dimensional (3D) illustration of our device can be seen in Fig. 1A. We use a partially etched Si₃N₄ rib waveguide to route the optical signal to the plasmonic memory cell, which are coupled via a tapered geometry (25–27). Plasmonic nanogaps are formed between two metal electrodes (3 nm Cr/75 nm Au) fabricated via lift-off using electron-beam lithography (EBL) and thermal evaporation. A thin film (75 nm) of GST with a 5-nm capping layer of SiO₂ bridges the nanogap, as shown by the atomic force microscope (AFM) micrograph in fig. S1 (see Supplementary Materials), controlling both the electrical resistance and optical transmission of the device depending on the state of the material. By sending either electrical or optical pulses, we can reversibly switch the GST within the nanogap between its highly resistive amorphous phase and conductive crystalline phase (28–30). Optical and scanning electron microscope (SEM) micrographs of the completed device are shown in Fig. 1 (B to D).

The gold electrodes form a plasmonic metal-dielectric-metal waveguide in the nanogap region that couples to and from the silicon nitride waveguide, allowing for broadband modulation of the optical transmission. The reduction in mode volume serves to both enhance the electric field in the nanogap and reduce the switching volume of the active PCM. To quantify the field enhancement of the plasmonic nanogap, we performed both 2D eigenmode and 3D finite-difference time-domain (FDTD) simulations using Lumerical Solutions and plot the field profile cross sections of the device when GST is in the amorphous and crystalline phases (see Fig. 1E). The amplitude of the electric field intensity is scaled relative to the field amplitude of the waveguide mode. Figure 1E shows that the electric field intensity is enhanced by more than an order of magnitude in the case of amorphous GST due to strong field confinement within the 50-nm nanogap. This enhancement reduces by a factor of 5 (defined by the ratio of maximum field intensity in the nanogap for the amorphous and crystalline states) when the GST within the nanogap switches from a fully amorphous, nonabsorbing state to a fully crystalline, absorptive state. Figure 1F shows the simulated transmission spectrum for the complete waveguide-nanogap system. Here, we observe that, at longer wavelengths, the transmission actually increases when GST in the nanogap is crystalline. This is a result of the modulation of the coupling between the nanogap and waveguide where an increase in the wavelength-dependent refractive index causes an increased coupling to the plasmonic mode within the nanogap. We see in 3D FDTD simulations that the light that is scattered and reflected between the waveguide/nanogap interface decreases when the GST is in the crystalline state. Although the optical absorption also increases for crystalline GST, the enhanced coupling to the nanogap results in an overall increased transmission when GST is switched from the amorphous to the crystalline state.

We experimentally verify this plasmonic field enhancement by sending pulses of increasing amplitude to the device and measuring the nonvolatile change in transmission of a counter-propagating probe signal (λ = 1570 nm). The measured change in transmission as a function of pulse energy can be seen in Fig. 1G. Here, we observe a change in transmission with a pulse energy of 16 ± 2 pJ using a 5-ns optical pulse. Given the linear relationship between switching energy and the change in optical transmission in Fig. 1G, we can explain the modulated transmission of our waveguide-nanogap device as a volume-dependent effect—the volumetric fraction of crystalline domains inside the nanogap grows with increasing pulse energy. We also note that this energy is significantly lower than previous demonstrations of evanescently coupled phase-change devices (28) due to the strong field enhancement and small mode volume of our plasmonic nanogap.

We then perform optical programming of our phase-change memory cell as illustrated in Fig. 2A. We send optical write and erase pulses to partially amorphize and crystallize the GST within the nanogap while simultaneously monitoring both the optical transmission and electrical resistance of the device. Figure 2A shows a schematic of the experimental setup where the pump laser is used to switch the GST inside the nanogap. Piecewise optical write pulses (7.5 mW for 8 ns followed by 3 mW for 400 ns) and rectangular erase pulses (7.5 mW for 8 ns) are used to switch the GST between crystalline and amorphous states, respectively. We use a constant-power optical probe to monitor the change in transmission, while a source-meter unit (SMU) in constant voltage mode (V_bias = 50 mV) is used to monitor the change in resistance. Time-dependent traces of the simultaneous change in both the transmission and resistance of the device can be seen in Fig. 2 (B and C) during consecutive optical write and erase pulses separated by 1 s. In agreement with our FDTD simulations of Fig. 1F, the resistance and transmission traces change as expected—i.e., an amorphization (erase) pulse results in an increase in electrical resistance and decrease in optical transmission, while a crystallization (write) pulse results in the opposite effect.

We subsequently demonstrate successful operation of the device in the electro-optic domain, wherein a change in optical transmission is observed as a result of electrical switching of the device. Here, we add a bias tee between the device and SMU to monitor the DC resistance of the device while sending write and erase pulses via the RF port of the bias tee (see Fig. 3A). By sending a 10-ns, 350-mV pulse across the device (5-ns rise and fall time), we amorphize the GST in the gap, while a 350-mV triangular pulse (5-ns rise, 500-ns fall time) recrystallizes the GST. As shown in Fig. 2 (B and C), the state of the device can be seen in both the optical transmission and electrical resistance readout. As we observed previously and as expected from our FDTD models, the transmission increases when the GST is switched to the crystalline state (see Fig. 2B) and is repeatable for many cycles, as shown in Fig. 4 (D and E). However, in these measurements, we observe higher contrast in the electrical resistance than during optical switching. This is because we are able to switch only the volume of material needed to create or disrupt a conductive path between the electrodes (31). The enhanced level of optical sensitivity to such a small volume of material switching between its amorphous and crystalline phase is attributable to the strong light-matter interaction within the plasmonic nanogap. Moreover, higher sensitivity is observed in the optical readout when programming the device optically, whereas higher sensitivity is observed in the electrical reading when the device is programmed electrically. This is attributed to the different energy distribution in the active region of the device when pulsed electrically versus optically. Each mode results in switching at locations of high field strength for that mode. However, this is not the case for the reading of the complementary mode. Crucially, the voltage required to switch the state of the GST is aided by reducing the spacing between the metal contacts and thereby increasing the electric field within the nanogap (see fig. S2). Our device therefore benefits both the optical and electrical design by improving light-matter interaction for the former and reducing the volume and separation between the electrical contacts for the latter, resulting in an efficient mixed-mode device.

As the optical transmission and electrical conductance are both dependent on the fractional volume of crystalline domains, both are dependent on the energy of the write and erase pulses. Figure 4 (A to C) shows the dependence of both the optical transmission and electrical resistance of our device for various optical erase pulse energies (fixed 8-ns pulse width). Between each erase pulse, a fixed piecewise write pulse (1.3 nJ total energy and 408 ns in duration) is used to return the device to the crystalline state. We note that due to the stochastic nature of the formation of amorphous and crystalline domains within the memory cell, the resistance trace of Fig. 4A shows more variation than the readout of the optical probe during optical switching cycles. On the other hand, the variation of the optical transmission is largely limited by the signal-to-noise ratio of our optical readout that can be addressed by increasing the coupling efficiency between the waveguide and plasmonic nanogap, which, based on previous work, could be improved by a factor of 3 (27). We attribute this to the different mechanisms between electrical conductance and optical transmission in our device. While the change in optical transmission depends on the overlap between the optical mode and the fractional volume of crystalline versus amorphous domains, a change in electrical conductance requires a continuous path of crystalline domains across the device to be formed—similar to a percolation network (32). This results in a dependence on both the fractional volume of crystalline GST and the position of those domains, which leads to greater variation of the device resistance as seen in Fig. 4A. Because of this, the electrical domain both has a higher switching energy threshold and saturates at a lower erase pulse energy than the optical transmission readout, as shown in Fig. 4 (B and C). The lower saturation threshold is due to the fact that once the circuit is fully broken by an erase pulse, greater pulse energies will not affect the overall conductance of the device.

Our devices show good cyclability in both the electrical resistance and transmission for optically and electrically induced switching between the amorphous and crystalline states (see Fig. 4, D and E). The optical and electrical properties of GST and related phase-change chalcogenides have both been demonstrated commercially to be robust for more than 10¹² write-erase cycles (7, 33). The cyclability of these phase-change chalcogenides, combined with a storage lifetime of more than 10 years at room temperature (34), makes our approach highly promising for future integrated electro-optical storage. Table 1 provides a comparison of our work with other nonvolatile photonic memories published to date. Not only do our results compare favorably with the literature in terms of active area and minimum switching energy, but we also show full optical and electrical programming and readout in an integrated device.

We have demonstrated a nonvolatile nanoscale electro-optic device that enables both electrical and optical programming and readout using the synergetic combination of PCMs and nanoplasmonics. This is an unprecedented demonstration of an integrated, reversible, and nonvolatile phase-change memory cell that fully bridges the gap between electro-optic mixed-mode operations. This was enabled by using a plasmonic design that simultaneously reduces the footprint of the device, enhances light-matter interaction, and reduces the separation between electrical contacts, creating a compact and highly sensitive device. Our approach also enables a direct comparison of both optical and electrical read and write operations in a single device, demonstrating the relative merits and limitations of both. The nonvolatile nature of our platform provides an exciting outlook in the development of switchable and reconfigurable metadevices by means of optical or electrical stimuli, enabling novel approaches to switchable metamaterial-based optical components (35–37). We anticipate that a plethora of novel devices and platforms should arise in the coming years, which will capitalize on the bridge between the electrical and photonic domains that are demonstrated here. These devices potentially herald true device-level integration of hybrid optoelectronic computing platforms with in-memory computing and multilevel data storage, which is readily applicable to this work (38, 39).

H.B. thanks A. Ne for stimulating conversations. All authors thank the collaborative nature of European science for allowing this work to be carried out. Funding: This research was supported by EPSRC via grants EP/J018694/1, EP/M015173/1, and EP/M015130/1 in the United Kingdom; the Deutsche Forschungsgemeinschaft (DFG) grant PE 1832/2-1 in Germany; the European Research Council grant 682675; and the European Union’s Horizon 2020 research and innovation program under grant agreement no. 780848 (Fun-COMP project). Author contributions: All authors contributed substantially. N.F. and N.Y. led the design, fabrication, and testing of the devices. J.T., X.L., J.L.S., and Z.C. aided the modeling, characterization, and the experimentation. C.D.W., W.H.P.P., and H.B. led the research, with N.F., N.Y., and H.B. drafting the manuscript. Competing interests: H.B. and W.H.P.P. are inventors on several patents related to this work filed by Oxford University Innovation [application no. PCT/GB2016/052871 (lapsed); application no. US15/760 945 (pending); application no. EP16770973.2 (pending); application no. GB1607345.4 (lapsed); and application no. PCT/GB2017/051143, publication no. WO 2017/187151 (A1)]. H.B. also serves on the board of Bodle Technologies Ltd. H.B. works for the University of Oxford, an institution that has financial and reputational incentives to publish many papers. All the other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors or Oxford Research Archive for Data (https://ora.ox.ac.uk).