Optical memristive switches
Optical memristive switches are particularly interesting for the use as latching optical switches, as a novel optical memory or as a digital optical switch. The optical memristive effect has recently enabled a miniaturization of optical devices far beyond of what seemed feasible. The smallest optical – or plasmonic – switch has now atomic scale and in fact is switched by moving single atoms. In this review, we summarize the development of optical memristive switches on their path from the micro- to the atomic scale. Three memristive effects that are important to the optical field are discussed in more detail. Among them are the phase transition effect, the valency change effect and the electrochemical metallization.
KeywordsOptical switches Memristors Photonics Plasmonics Atomic scale Optical memory
Optical memristive switches are versatile integrated optical circuit elements [1, 2, 3, 4, 5, 6, 7]. They are digital optical switches with distinct transmission states based on a resistive switching with memory and thus feature characteristics of optical modulators and optical memories in one structure. Typically, they are controlled by electrical signals – but sometimes by optical signals. Normally, the operation speed is moderate and in the MHz range. The application range includes usage as a new kind of memory which can be electrically written and optically read, or usage as a latching switch that only needs to be triggered once and that can keep the state with little or no energy consumption. In addition, they represent a new logical element that complements the toolbox of optical computing.
The optical memristive effect has been discovered only recently . It is of particular interest because of strong electro-optical interaction with distinct transmission states and because of low power consumption and scalability . Such devices rely in part on exploiting the electrical memristive effect to optically probe the material changes. In another aspect, they rely on plasmonics to enhance light-matter interaction and the switching effect.
In electrical resistive switching, a wide range of resistance switching processes are known [8, 9, 10, 11]. Although all of them can be induced electrically, switching happens through mechanical, chemical, thermal or other effects or as a combination of them. Only four mechanisms have also been shown to work for optical devices [1, 3, 4, 5, 6, 7, 12]. These are the phase transition effect, the valency change effect, the electrochemical metallization and the phase change effect.
If the memristive section is in or close to an optical waveguide, then a change in resistance can be detected by an optical signal – which makes the effect an optical memristive effect. Optical transmission measurements showed a clear hysteresis with well distinguishable transmission levels depending on the device’s resistance state [1, 3, 4, 5, 6, 7]. All effects were demonstrated in electro-optic configurations and with the exception of the phase change effect also shown to work as optical memristive switches such that both the electrical resistance and the optical transmission switching could be observed in the same device. Still, the inherent connection of the two switching mechanisms in phase change materials is well known [13, 14, 15]. Phase transition effect based optical memories are volatile due to the thermal activation that is needed to keep the state [5, 16, 17, 18, 19]. Valency change effect or electrochemical metallization based optical memories are generally nonvolatile (or latching) since the conductive path remains open or closed, respectively, even without bias [1, 3, 4]. Likewise, phase change effect based optical memories are nonvolatile also since the disorder in the crystallinity is kept after the heat pulse [13, 20].
Memristive switches at the nanoscale are already widely implemented in electronics [8, 9, 10, 21, 22, 23, 24, 25, 26] while optical memristive device are only explored now. The reason for this is in fact that memristive effects typically are performed in a few nanometers thin layers with areas below 1 μm2. The memristive effect thus remains mostly unnoticed by an optical signal. While a photonic wave has a rather low power density and is too large to detect a sub-micrometer material change, a plasmonic wave can confine the optical intensity to the active region and therefore is a good fit to detect a memristive material change . And indeed, latest research used strong material changes and plasmonics as means to increase light-matter interaction by orders of magnitude [27, 28]. The ultimate scaling limit was reached in 2015 when an atomic scale plasmonic switch was reported . The next challenge now is to increase the integration density of optical switches. This way, optical circuits can be scaled down from the micro- to the nanoscale. The space requirements of optical devices will be reaching electronic scales ultimately enabling the long awaited co-integration of optics and electronics on the same chip.
This review is organized in three parts. In the first part, the optical and electrical concepts of optical memristive switches are discussed with a focus on plasmonics as a mean to enhance light-matter interaction. The second part shows the latest progress in the field and an example of optical switching for each of the electrically induced resistive switching mechanism presented above. The third part covers the progress from the micro- to the atomic scale and presents the atomic origin of resistance switching by electrochemical metallization and its first implementation as an optical switch.
2 Optical & electrical concepts
The successful implementation of optical memristive switches requires an understanding of the underlying concepts of both the optical and electrical dynamics and their interaction. Therefore, this chapter introduces the concept of plasmonics, which exploits strong light confinement on metal surfaces to enhance light-matter interaction and a brief explanation of resistive switching.
The strong field confinement and the high field intensities on the metal surface make plasmonics particularly suitable for modulators and switches since any light-matter interaction in an active material close to the metal is strongly enhanced [27, 28, 29]. For maximum light confinement, also the lateral dimension can be reduced. This is done in plasmonic waveguides by structuring the metal or the dielectric similar to standard photonic waveguides. Since SPPs can only be excited when light has an electric field component perpendicular to the metal surface, different waveguide designs have to be used for the two polarizations of light. Common choices are a hybrid plasmonic waveguide (HPW) for a transverse magnetic (TM) polarization and metal-insulator-metal (MIM) waveguide for a transverse electric (TE) polarization.
Plasmonics offers the possibility to convert light into an electromagnetic wave coupled to free charge oscillations on a conducting surface. The plasmonic wave can be confined to nanoscale dimensions and therefore can overcome the diffraction limit of photonics. The strong field enhancement multiplies light-matter interaction at the price of higher propagation losses along the metals. Fortunately, there is a trade-off since the required propagation distances are reduced from mm’s to μm’s. Hence, the total losses remain reasonable. Typical plasmonic devices have an area of about 10 μm2 or below. The compact size has advantages. First, capacitive device limitations are necessarily low and plasmonic devices can thus in principle be fast. Second, the high confinement on such a compact space allows for lowest power operation .
Currently more and more plasmonic devices are added to the toolbox of plasmonics. This encompasses devices for light generation, modulation, amplification and detection. While first plasmonic circuit elements are emerging , fully conceived all-plasmonic circuits are still to come [34, 35, 36].
2.2 Resistive switching
2.3 Electro-optical interaction
The electro-optical interaction between the light in the waveguide and the electrical resistive switching may rely on different possibilities to interact and the exact mechanisms can be difficult to determine. A first possibility is a refractive index change of the active material by resistive switching which in turn changes the propagation constant of the optical mode. A second possibility is via irregularities or defects in the waveguide caused by the switching operation. They will perturb the optical mode and therefore change the propagation losses. Alternatively, the increased conductivity through switching to the on state could lead to a stronger coupling to charge oscillations of SPPs on the metal walls and influence the plasmonic character of the mode. Lastly, also a resonance effect is probable where the resistive switching alters the resonance condition. These electro-optical interactions are not necessarily independent and combinations are expected to be observed.
3 Optical memristive switches
The investigation of memristors as optical switches evolved only in the last few years. As mentioned in the introduction, four types of memristive effects were shown for electro-optical switching; the phase transition effect, the valency change effect, electrochemical metallization and the phase change effect, c.f. Fig. 1. The phase transition effect changes electrical and optical properties from insulating to metallic through Joule heating. Since the device at room temperature is always in the insulating state, its state is volatile and must be kept actively. Both the valency change effect and the electrochemical metallization are based on building and destroying a conductive path. This path and hence the device state generally endure during ambient conditions which renders the device nonvolatile. Nonvolatility is beneficial since no power is required to keep the current state. This could be especially interesting for optical memory applications. The nonvolatility is also given in devices based on the phase change effect where short heating pulses swap the material phase from crystalline to amorphous and back.
In this chapter, we first introduce the best performing optical memristive device. Then, we show the first three memristive effects on the basis of an exemplary paper each. We omit the phase change effect since its implementation in an optical memristive switch is still ongoing research. Recent work demonstrates electrically controlled displays based on a phase change material  and hereby indicates great potential for the application as electro-optical switches. However, they did not show active optical switching.
It should be noted, that the same device structure might be used for operation of an ultrafast modulator based on the free carrier dispersion effect in indium-tin-oxide (ITO) [38, 39]. A material study in the visible showed a tunability of the refractive index in the order of unity . Later, a waveguide integrated device showed absorption modulation within a ITO-SiO2-Au stack . However, the origin of the modulation remains uncertain as long as no fast response times are demonstrated.
The formation of a conductive path was the key to this switch. However, the underlying physics could not be fully resolved. Literature suggests that the switching happens through the formation of either a conductive path of oxygen vacancies [41, 42] or a metal filament , c.f. section 3.3 or 3.4, respectively, or even a combination of them.
3.2 Phase transition effect
Optical memristive switches exploiting a reversible phase transition from insulating to metallic phase were reported using vanadium dioxide (VO2) as active material [5, 16, 17, 18, 19]. The phase transition can be activated by substrate heating , optical absorption [44, 45, 46, 47] or electrical currents [5, 48, 49, 50]. Here, we present the most recent publication of a VO2 switch where electrically induced Joule heating is exploited to achieve the phase transition . The switch reached an ER of 12 dB on a length of 1 μm and showed operation across at least 100 nm.
The authors investigated the switching dynamics by means of time-resolved transmission measurements . They show in pulsed measurements (period of 1 s) that the transmitting state can be preserved for one or a few microseconds. From the measured phase transition times, sub-GHz switching speeds are estimated but not shown since speed measurements are very challenging due to thermal accumulation. The authors claim that high switching speeds of tens of GHz are potentially possible by activating the phase transition through electron injection, which happens on picosecond timescales. However, this comes most probably at the price of losing the memory effect.
3.3 Valency change effect
Two distinct reflection states were observed with a difference of about 5% in reflection corresponding to an ER of about 0.35 dB. The hysteresis in reflection gives an indication of the memristive nature of the switch, which was also shown electrically. Since the reflection states are preserved without bias, such devices allow use as a nonvolatile optical memory.
The growth and rupture of a conductive path of charge vacancies is governed by the motion of oxygen ions under a voltage bias [8, 51]. The carrier dynamics are key to understand the resistive switching mechanism and the interaction with light. Recently developed measurement techniques are capable to separate different current contribution and hence to access material diffusion parameters . This insight paves the way towards more compact switches – potentially even reaching the atomic scale.
3.4 Electrochemical metallization
The full potential of the memristive effect based on metal filament formation could not be exploited in the current device configuration due to limitations from the design. A deeper understanding of carrier dynamics in electrochemical metallization switches, i.e., the metal ion transport and their redox reactions [8, 52], was opening new possibilities. Recently, the atomic origin of resistive switching  and quantized conductance levels  were demonstrated in electrochemical metallization switches showing that the underlying physics are an atomic process. This fact enabled scaling the device from the micro- down to the nano-scale as shown below.
4 From the micro- to the atomic scale
The need for higher integration density and lower energy consumption drives the development of optical switches to the nanoscale. Electronic components are already at nanoscale dimensions and atomic scale transistors or switches were demonstrated in the last two decades [8, 9, 21, 22, 55, 56, 57, 58]. Optically, this is a very new but quickly growing research field. State-of-the-art research in plasmonics is expected to show efficient and highly compact solutions for light generation and detection on-chip in the near future calling for optical modulators and switches on the nanoscale. With the development of atomic scale CMOS processes, the co-integration of electronic and optical devices on a new energy efficient platform will eventually emerge.
Standard optical switches are still based on photonic waveguides of mm lengths whose size reduction is limited by diffraction and rather weak light-matter interaction. In the last years, plasmonics evolved quickly and reported switches, c.f. chapter 3, could reduce the interaction lengths to several μm. Only recently, quantum plasmonic devices [59, 60, 61, 62, 63, 64] and the first atomic scale plasmonic switch  were reported. The authors showed digital optical switching of 9.2 dB based on electrochemical metallization. The grown filament showed discrete quantum conductance levels proving the atomic scale origin of the switching mechanism.
4.1 Atomic scale resistance switching
The exact process of atomic scale filament formation is difficult to access by experiment. Still, atomic scale resistance switching was already demonstrated [21, 22, 55, 56, 57]. Atomistic simulations [53, 65, 66, 67, 68, 69, 70, 71] are an accurate and promising tool to describe the exact processes happening during the growth and the destruction of a filament in electrochemical metallization cells. Such a model must incorporate and couple electronic transport, nucleation, electrochemical reactions and the diffusion of ions [72, 73]. In 2015, Onofrio, et al., presented a technique based on reactive molecular dynamics augmented with a charge equilibrium method which was able to describe the full dynamics during electrochemical metallization . They demonstrated the atomic origin of resistance switching at the example of a copper filament growing within amorphous silicon dioxide.
4.2 Atomic scale plasmonic switching
The atomic scale plasmonic switch presented in  explores the ultimate limits of atomic scale resistance switching since the quantum conductance of a single silver filament is used to switch light on or off. A similar resistance switching has also been found in memristive plasmonic antennas where resonances are shifted by single atoms [74, 75, 76, 77].
The work in  is the first optical switch implementation on a single atom scale. This work actually scaled the active area of optical on-chip devices to sizes never reported before. Furthermore, the device introduced in Ref.  had an ER of 12 dB, was operated with MHz speed and consumed as little as 12.5 fJ/bit.
This review presented the research on optical memristive switches. These optical switches are based on the phase transition effect, the valency change effect or the electrochemical metallization. State-of-the-art devices and examples for each effect were presented. Finally, it was shown how the concept of resistive switching can be exploited in an atomic scale plasmonic switch.
Overview over waveguide integrated optical memristive switches. Extinction ratio (ER), switch area, modulation speed and power consumption are given as metrics to compare the switching performance. The last column shows if the switch can operate as a nonvolatile optical memory
In the future, it can be expected that the performance of optical memristive switches will be further improved towards higher speed and lower energy consumption. The exploration of other resistive switching mechanisms integrated into optical devices could lead to further improvements. Finally, the use as optical memory might attract more attention on the path towards all-optical computing.
- 2.C. Hoessbacher, Y. Fedoryshyn, A. Emboras, D. Hillerkuss, A. Melikyan, M. Kohl, M. Sommer, C. Hafner, and J. Leuthold, Latching Plasmonic Switch with High Extinction Ratio. in CLEO: 2014, (San Jose, California, 2014), p. FTu3K.6Google Scholar
- 33.C. Haffner, W. Heni, Y. Fedoryshyn, J. Niegemann, A. Melikyan, D.L. Elder, B. Baeuerle, Y. Salamin, A. Josten, U. Koch, C. Hoessbacher, F. Ducry, L. Juchli, A. Emboras, D. Hillerkuss, M. Kohl, L.R. Dalton, C. Hafner, J. Leuthold, All-plasmonic Mach–Zehnder modulator enabling optical high-speed communication at the microscale. Nat. Photonics 9(8), 525–528 (2015)CrossRefGoogle Scholar
- 38.A. Melikyan, T. Vallaitis, N. Lindenmann, T. Schimmel, W. Freude, and J. Leuthold, A Surface Plasmon Polariton Absorption Modulator. in Conference on Lasers and Electro-Optics 2010, (San Jose, California, 2010), p. JThE77Google Scholar
- 50.P. Markov, J. D. Ryckman, R. E. Marvel, K. A. Hallman, R. F. Haglund, and S. M. Weiss, Silicon-VO2 Hybrid Electro-optic Modulator. in CLEO: 2013, (San Jose, California, 2013), p. CTu2F.7Google Scholar
- 66.J. Zhang, H. Zeng, Theory and application of quantum molecular dynamics (World Scientific, Singapore, 1999)Google Scholar
- 70.S. Datta, Electronic transport in mesoscopic systems, (sixth printing) ed (Cambridge University Press, Cambridge, 2005)Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.