Electronic-photonic circuit crossings

arXiv:2204.14257 · physics.optics, physics.app-ph · Submitted 2022-04-29 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Electronic-photonic circuit crossings".

Title and authors: Kai: So, we're looking at this paper now called "Electronic-photonic circuit crossings". It’s about using topology optimization to design a single layer of circuit that lets you cross electrical wires and photonic waveguides without introducing a lot of electrical resistance or optical losses.

Mira: Yeah, so the main thing here is how they managed to put these two fundamentally different things together in one layer, which is really challenging because electronics rely on electron conduction while photonics relies on light propagation <ref:two thousand two hundred four point one four two five seven#pg1. It’s about solving that integration problem directly.

Lev: From my side, I see the big hurdle as adding more layers or processes every time you need to route an electrical signal past a photonic component, which the paper says gets way worse if you have many crossings because the complexity scales superlinearly with those avoided electronic-photonic crossings <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: Exactly. So instead of adding more layers or vias for every crossing, they introduced this carefully engineered and inversely designed monolithic device that lets you cross them using the same layer without big resistance or optical losses <ref:two thousand two hundred four point one four two five seven#pg3. It's a system-level complexity fix, essentially <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: And the methodology they use is topology optimization to find the best shape for this crossing, measuring quality with a figure of merit Φ which they optimize iteratively <ref:two thousand two hundred four point one four two five seven#pg2. The problem is set up so they maximize the minimum of that figure of merit over different wavelengths, like maximizing alpha times lambda i Phi over E lambda i <ref:two thousand two hundred four point one four two five seven#pg2.

Lev: And they solved that using the finite element method in COMSOL Multiphysics to model the structure and find the optimal geometry for this crossing <ref:two thousand two hundred four point one four two five seven#pg2. It’s a continuous constrained optimization problem where they change the geometry systematically to find that best structure <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: What's really impressive is the performance numbers they hit with this design, achieving up to ninety-nine point eight percent optical transmission across just a twenty nanometer electrical isolation trench <ref:two thousand two hundred four point one four two five seven#pg2. That’s a significant jump from what they had before topology optimization <ref:two thousand two hundred four point one four two five seven#pg3.

Title and authors: Mira: They also managed to get a peak transmission of ninety-three point six percent over one hundred nanometer bandwidth, which is pretty dramatic compared to previous work <ref:two thousand two hundred four point one four two five seven#pg3. They even showed that a power transmission of ninety-nine point eight percent means that hundreds of these crossings in series would result in insertion losses below three decibels <ref:two thousand two hundred four point one four two five seven#pg3.

Lev: That scaling property is what matters for real hardware, because if you need to link many components together, you don't want each crossing to kill the signal that much <ref:two thousand two hundred four point one four two five seven#pg3. This suggests a path toward denser integrated systems where this level of transmission is achievable <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: So, they can integrate these crossings into a monolithic silicon nanoelectromechanical add-drop switch where the flow of photons, electrons, and mechanical motions are all fully integrated within the same layer <ref:two thousand two hundred four point one four two five seven#pg3. That’s a very specific application that shows how versatile this design is <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: The experimental verification also looks strong, they fabricated circuits with zero to six crossings using electron-beam lithography and cryogenic deep reactive ion etching on silicon on insulated wafers <ref:two thousand two hundred four point one four two five seven#pg3. They measured an average transmission of ninety-two point nine percent over one hundred nanometer bandwidth, which they found agrees really well with their theory <ref:two thousand two hundred four point one four two five seven#pg3.

Lev: They also showed the curve-fitted design has a broadband figure of merit Φ equal to T over R plus alpha greater than zero point eight nine two for that one hundred nanometer wavelength, which is good data for characterizing the robustness of the structure <ref:two thousand two hundred four point one four two five seven#pg20.

Kai: It’s interesting how robust they claim this design is to small variations in device thickness; they found that a ten nanometer thinner device layer doesn't have a significant influence on the performance at all <ref:two thousand two hundred four point one four two five seven#pg18. That suggests the broadband optimization procedure really made it stable <ref:two thousand two hundred four point one four two five seven#pg18.

Mira: They also mentioned that replacing features that expand modes with a tapering section results in an overall better performance when you average across the full one hundred nanometer bandwidth, which is a good practical design rule <ref:two thousand two hundred four point one four two five seven#pg18. It shows they're thinking about the whole band, not just one point <ref:two thousand two hundred four point one four two five seven#pg18.

Title and authors: Lev: And they also mentioned that for larger networks, where optical transmission becomes a limiting factor, they design additional crossings while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps <ref:two thousand two hundred four point one four two five seven#pg3. That shows they are looking ahead to scaling it up further <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: So, what does this mean for the next steps? They’re looking at how these isolation trenches can be used for thermal isolation or mechanical isolation in nanophotonic devices or even quantum optomechanics <ref:two thousand two hundred four point one four two five seven#pg2. That opens up some new physical possibilities beyond just routing signals <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: And they also mention that these concepts apply to other semiconductor platforms, like gallium arsenide or indium phosphide, which could allow for independent electrical control of quantum emitters in integrated photonic circuits <ref:two thousand two hundred four point one four two five seven#pg3. That tackles a major scaling problem in integrated quantum photonics <ref:two thousand two hundred four point one four two five seven#pg3.

Lev: If you can do that, it could eventually enable things like phonon-photon crossings, which is a really interesting physical interaction to engineer <ref:two thousand two hundred four point one four two five seven#pg3. That’s moving beyond just simple signal routing <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: So, to wrap up this paper "Electronic-photonic circuit crossings", topology optimization allows for crossing electrical connections and photonic waveguides using the same device layer without introducing significant electrical resistance or optical losses <ref:two thousand two hundred four point one four two five seven#pg3. The authors successfully devised a monolithic silicon nanoelectromechanical add-drop switch in which the flow of photons, electrons, and mechanical motions are fully integrated within the same layer <ref:two thousand two hundred four point one four two five seven#pg3. It demonstrates that near-unity transmission is attainable for smaller gaps through topology optimization <ref:two thousand two hundred four point one four two five seven#pg5.

Mira: That's the core finding, and it opens up a lot of possibilities for new research in nano-opto-electro-mechanical systems, optomechanics, and integrated quantum photonics <ref:two thousand two hundred four point one four two five seven#pg3. It’s not just about better routing anymore <ref:two thousand two hundred four point one four two five seven#pg3.

Lev: And I think the real impact is showing that you can achieve these high transmission numbers, like ninety-nine point eight percent, which means hundreds of EPCCs in series would result in insertion losses below three decibels <ref:two thousand two hundred four point one four two five seven#pg5. That's a big step for realizing practical integrated circuits <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: So, the whole point is that this paper proves we can build these integrated systems using topology optimization and it’s a great piece of work to keep an eye on <ref:two thousand two hundred four point one four two five seven#pg3. We’ll see what they do next <ref:two thousand two hundred four point one four two five seven#pg3.

The paper's summary: Kai: So, to recap, they used topology optimization to create a single layer device that lets you cross electrical wires and photonic waveguides without introducing significant electrical resistance or optical losses <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: Right, it's about solving the problem of routing electrons and photons inside the same microchip by designing a monolithic device where the crossing happens on one layer <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: And they achieved up to ninety-nine point eight percent optical transmission across a twenty nanometer electrical isolation trench <ref:two thousand two hundred four point one four two five seven#pg2. That’s the big number, and it’s really impressive for this kind of integration.

Mira: It is, but the real impact is how they're using that figure of merit, that quality measure, to design a structure that works across a whole range of wavelengths <ref:two thousand two hundred four point one four two five seven#pg2. They aren't just optimizing for one color of light; they’re making sure it works well for the entire band.

Kai: Exactly. And they showed this design is robust; even if you change the device layer thickness by ten nanometers, the performance doesn't really dip much <ref:two thousand two hundred four point one four two five seven#pg18. That stability across parameters is something you need when you’re building actual hardware.

Mira: I think what this means for us is that we can start thinking about much denser integrated systems, where we don't have to stack layers on top of each other just to get the signal through <ref:two thousand two hundred four point one four two five seven#pg3. The power transmission scaling they showed is what really matters there too, because a ninety-nine point eight percent transmission means you can string hundreds of these crossings together and still keep the insertion losses below three decibels <ref:two thousand two hundred four point one four two five seven#pg5.

Kai: So it’s not just about a single crossing; it’s about building these complex networks that handle high data rates, which is what we need for better interconnects in quantum systems <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: And they even connected this to a practical application, showing how you can put this topology-optimized crossing into a monolithic silicon nanoelectromechanical add-drop switch where photons, electrons, and mechanical motions are all fully integrated within the same layer <ref:two thousand two hundred four point one four two five seven#pg3. That’s the kind of holistic system we’re aiming for.

Kai: It opens up new physics too, though; they mentioned using these isolation trenches not just for electrical isolation but also for thermal isolation or mechanical isolation in nanophotonic devices <ref:two thousand two hundred four point one four two five seven#pg2.

Mira: And that's where things get interesting, because if you can use these trenches to isolate things thermally or mechanically, it could lead directly into the area of nano-opto-electro-mechanical systems and quantum optomechanics <ref:two thousand two hundred four point one four two five seven#pg3. That’s a whole new set of possibilities we can explore with this kind of fabrication control.

The paper's improvements: Kai: So, to sum up what they found, the main thing is that topology optimization lets you cross electrical connections and photonic waveguides using the same device layer without adding much electrical resistance or optical losses <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: Right, it’s about solving that integration problem by making a monolithic device where the crossing happens on one layer <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: And they showed this design is robust; even if you change the device layer thickness by ten nanometers, the performance doesn't really dip much <ref:two thousand two hundred four point one four two five seven#pg18. That stability across parameters is something you need when you’re building actual hardware.

Mira: I think what they suggest for improvement is focusing on that broadband figure of merit, that quality measure, and making sure it works well for the entire range of wavelengths <ref:two thousand two hundred four point one four two five seven#pg2. They aren't just optimizing for one color of light; they’re making sure it works well for the entire band.

Kai: They also mentioned that replacing features that expand modes with a tapering section results in an overall better performance when you average across the full one hundred nanometer bandwidth <ref:two thousand two hundred four point one four two five seven#pg18. That’s a practical design rule for making these things work better over time.

Mira: And they also showed that for larger networks, where optical transmission becomes a limiting factor, you should design additional crossings while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps <ref:two thousand two hundred four point one four two five seven#pg3. That tells us how to scale this idea up beyond just one small crossing.

Kai: So they are pushing toward building larger networks that maintain high performance by adapting the gap size based on what the network needs, rather than just optimizing for one perfect number <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: And on a bigger scale, they’re looking at applying these isolation trenches to thermal or mechanical isolation in nanophotonic devices or even quantum optomechanics <ref:two thousand two hundred four point one four two five seven#pg2. That shows the potential for using this geometry beyond just signal routing <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: And they also noted that these concepts apply to other semiconductor platforms, like gallium arsenide or indium phosphide, which could allow for independent electrical control of quantum emitters in integrated photonic circuits <ref:two thousand two hundred four point one four two five seven#pg3. That tackles a major scaling problem in integrated quantum photonics.

Mira: Yes, that’s significant because it means we don't have to be locked into silicon when designing these complex interfaces for quantum systems <ref:two thousand two hundred four point one four two five seven#pg3. This really expands the toolkit for building new kinds of hybrid circuits.

Conclusion: Kai: So to wrap up, we’ve seen how topology optimization allowed them to devise this single layer electronic photonic circuit crossing without introducing significant electrical resistance or optical losses <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: Exactly, it's about solving the integration problem by making a monolithic device where the crossing happens on one layer <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: And they achieved up to ninety-nine point eight percent optical transmission across a twenty nanometer electrical isolation trench <ref:two thousand two hundred four point one four two five seven#pg2. That’s the big number, and it’s really impressive for this kind of integration.

Mira: It is, but the real impact is how they're using that figure of merit, that quality measure, to design a structure that works across a whole range of wavelengths <ref:two thousand two hundred four point one four two five seven#pg2. They aren't just optimizing for one color of light; they’re making sure it works well for the entire band.

Kai: They also showed this design is robust; even if you change the device layer thickness by ten nanometers, the performance doesn't really dip much <ref:two thousand two hundred four point one four two five seven#pg18. That stability across parameters is something you need when you’re building actual hardware.

Mira: I think what they suggest for improvement is focusing on that broadband figure of merit, that quality measure, and making sure it works well for the entire range of wavelengths <ref:two thousand two hundred four point one four two five seven#pg2. They aren't just optimizing for one color of light; they’re making sure it works well for the entire band.

Kai: They also mentioned that replacing features that expand modes with a tapering section results in an overall better performance when you average across the full one hundred nanometer bandwidth <ref:two thousand two hundred four point one four two five seven#pg18. That’s a practical design rule for making these things work better over time.

Mira: And they also showed that for larger networks, where optical transmission becomes a limiting factor, you should design additional crossings while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps <ref:two thousand two hundred four point one four two five seven#pg3. That tells us how to scale this idea up beyond just one small crossing.

Kai: So they are pushing toward building larger networks that maintain high performance by adapting the gap size based on what the network needs, rather than just optimizing for one perfect number <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: And on a bigger scale, they’re looking at applying these isolation trenches to thermal or mechanical isolation in nanophotonic devices or even quantum optomechanics <ref:two thousand two hundred four point one four two five seven#pg2. That shows the potential for using this geometry beyond just signal routing <ref:two thousand two hundred four point one four two five seven#pg3.

Kai: And they also noted that these concepts apply to other semiconductor platforms, like gallium arsenide or indium phosphide, which could allow for independent electrical control of quantum emitters in integrated photonic circuits <ref:two thousand two hundred four point one four two five seven#pg3. That tackles a major scaling problem in integrated quantum photonics.

Mira: Yes, that’s significant because it means we don't have to be locked into silicon when designing these complex interfaces for quantum systems <ref:two thousand two hundred four point one four two five seven#pg3. This really expands the toolkit for building new kinds of hybrid circuits.

Lev: From my side, I gotta say that achieving ninety-nine point eight percent transmission with those fabrication steps sounds like a tough thing to get reliably on real hardware without some huge overhead <ref:two thousand two hundred four point one four two five seven#pg5.

Kai: I hear you on the overhead, Lev, but the way they designed it to be robust against thickness variations suggests that if we can control those fabrication steps well enough, we might get closer to that number <ref:two thousand two hundred four point one four two five seven#pg18.

Mira: The real question is whether they truly accounted for all the noise sources in that measurement record when they calculated the final broadband figure of merit <ref:two thousand two hundred four point one four two five seven#pg20. That is a big assumption there, and it's something we need to check against our theory.

Lev: And for us building error correction hardware, if this scaling holds up in practice across hundreds of crossings, that means the overhead for managing the connections stays manageable <ref:two thousand two hundred four point one four two five seven#pg3. That’s where the real engineering challenge lies.

Kai: So, to wrap up, the paper "Electronic-photonic circuit crossings" shows that topology optimization can build monolithic devices where photons and electrons cross without major loss or resistance <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: It shows we can achieve near unity transmission for small gaps through this method, opening up possibilities for nano-opto-electro-mechanical systems and quantum optomechanics <ref:two thousand two hundred four point one four two five seven#pg3.

Lev: And the scaling property they demonstrated, where hundreds of crossings still keep insertion losses below three decibels, is what makes this interesting for building larger integrated systems <ref:two thousand two hundred four point one four two five seven#pg5.

Kai: It’s a solid piece of work on how to handle that complex integration challenge in silicon chips <ref:two thousand two hundred four point one four two five seven#pg3.

Mira: Next time, we'll be looking at how they apply similar design principles to other semiconductor materials, like gallium arsenide or indium phosphide <ref:two thousand two hundred four point one four two five seven#pg3. That’s where the next big test of this concept will be.

Babak Vosoughi Lahijani, * Marcus Albrechtsen, Rasmus E. Christiansen, Christian A. Rosiek, Konstantinos Tsoukalas, Mathias T. Sutherland, Søren Stobbe

Technical University of Denmark

physics.optics, physics.app-ph

Submitted: 2022-04-29

Updated: 2025-10-27

Comments: 34 pages, 19 figures

Journal ref: Optics Express, Vol. 34, No. 23, pp. 44895-44906 (2026)

DOI: 10.1364/OE.608644

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 89/100

The gist: Electronic-photonic circuit crossings use topology optimization to devise a single-layer electronic-photonic circuit crossing with up to 99.8 % optical transmission across a 20 nm electrical…

Key concepts

Topology Optimization
This is a design method where the shape of a structure is optimized mathematically to meet specific performance goals. In this case, it was used to design the exact geometry of the circuit crossing that maximizes optical transmission while minimizing electrical resistance.
Electronic-Photonic Circuit Crossing (EPCC)
An EPCC is a critical junction in integrated circuits where electrical signals (electrons) must cross photonic waveguides (light). The challenge is routing these two fundamentally different particles together on the same chip without causing major signal loss or interference.
Monolithic Device Design
Instead of adding multiple layers or complex structures to route signals, this work created a single, carefully engineered device layer. This allows electrical connections and light paths to cross within that one layer, solving the problem of scaling complexity in traditional solutions.

Terminology

Summary

Electronic-photonic circuit crossings use topology optimization to devise a single-layer electronic-photonic circuit crossing with up to 99. p. 0

Where the paper says it

Nanoelectromechanical photonics could resolve these issues, but integrating this technology with conventional multilayer metal architectures is challenging, and conventional approaches do not allow crossings of electrical wires and photonic waveguides. Here, we use topology optimization to devise a single-layer electronic-photonic circuit crossing with up to 99.

8 % optical transmission across a 20 nm electrical isolation trench This work addresses an important challenge in incorporating opto-electro-mechanical topologies into photonic integrated circuits and may lead to new functionalities in nano-opto-electro-mechanical systems, optomechanics, and integrated quantum photonics. p. 0
Where the paper says it

We use these concepts to demonstrate a monolithic silicon nanoelectromechanical add-drop switch in which the flow of photons, electrons, and mechanical motions are fully integrated within the same layer. Our work addresses an important challenge in incorporating opto-electro-mechanical topologies into photonic integrated circuits and may lead to new functionalities in nano-opto-electro-mechanical systems, optomechanics, and integrated quantum photonics.

Motivation for Integration

Information technologies rely on controlling two fundamentally different particles: electrons are strongly interacting massive charged fermions while photons are massless and charge-neutral bosons that do not interact <ref:2204.14257#pg2> First, a sizeable fraction of the energy consumption in modern computing is spent on routing information through lossy electrical leads, which may be significantly reduced by chip-level optical interconnects [1, 2] <ref:2204. p. 0

Where the paper says it

These differences underpin the development of optical communication systems based on low losses and high coherence as well as electronic computing relying on strong interactions. For many reasons, however, integrating electronics and photonics is becoming an increasingly important research frontier: First, a sizeable fraction of the energy consumption in modern computing is spent on routing information through lossy electrical leads, which may be significantly reduced by chip-level optical interconnects [1, 2].

14257#pg3>. Second, the developments in hybrid quantum technologies call for high-fidelity faces between stationary and flying qubits [3, 4] which require unprecedented control and integration of electronics and photonics <ref:2204. p. 0
Where the paper says it

Second, the developments in hybrid quantum technologies call for high-fidelity inter- ∗ bala@dtu. dk faces between stationary and flying qubits [3, 4], which require unprecedented control and integration of electronics and photonics.

14257#pg3>. Third, the integration of photonics and electronics is at the heart of a wide range of new technologies that rely on advanced nanofabrication for new applications beyond electronics [5] ranging from smart sensors [6] over ultrahigh-bandwidth optical networks [7] to chip-scale particle accelerators [8], and nano-opto-electro-mechanical systems (NOEMS) [9–14] <ref:2204. p. 0
Where the paper says it

dk faces between stationary and flying qubits [3, 4], which require unprecedented control and integration of electronics and photonics. Third, the integration of photonics and electronics is at the heart of a wide range of new technologies that rely on advanced nanofabrication for new applications beyond electronics [5], ranging from smart sensors [6] over ultrahigh-bandwidth optical networks [7] to chip-scale particle accelerators [8], and nano-opto-electro-mechanical systems (NOEMS) [9–14].

14257#pg3>. Simultaneously controlling the flow of electrons and photons inside the same microchip is challenging because of the conflicting requirements for electronics and photonics: Electronic wires rely on conduction by free electrons but they in turn lead to significant optical losses <ref:2204. p. 0
Where the paper says it

Third, the integration of photonics and electronics is at the heart of a wide range of new technologies that rely on advanced nanofabrication for new applications beyond electronics [5], ranging from smart sensors [6] over ultrahigh-bandwidth optical networks [7] to chip-scale particle accelerators [8], and nano-opto-electro-mechanical systems (NOEMS) [9–14]. Simultaneously controlling the flow of electrons and photons inside the same microchip is challenging because of the conflicting requirements for electronics and photonics: Electronic wires rely on conduction by free electrons but they in turn lead to significant optical losses.

14257#pg3>.

Overcoming Existing Limitations

Existing solutions include optical layers embedded in transparent insulators with metallic surface electrode layers or numerous patterned conducting layers buried below the optical layers and contacted by deep vias in order to route the electrical signals without impeding the optical performance <ref:2204.14257#pg3>. Common to existing solutions is the need to add more layers or processes as the circuits grow <ref:2204. p. 0

Where the paper says it

Existing solutions include optical layers embedded in transparent insulators with metallic surface electrode layers or numerous patterned conducting layers buried below the optical layers and contacted by deep vias in order to route the electrical signals without impeding the optical performance. Common to existing solutions is the need to add more layers or processes as the circuits grow.

14257#pg3>. In other words, the circuit complexity scales superlinearly with the number of (avoided) electronic-photonic crossings <ref:2204. p. 0
Where the paper says it

Common to existing solutions is the need to add more layers or processes as the circuits grow. In other words, the circuit complexity scales superlinearly with the number of (avoided) electronic-photonic crossings.

14257#pg3>. The solution proposed in this work overcomes the system-level complexity in routing electrons and photons inside the chip by introducing a carefully engineered and inversely designed monolithic device <ref:2204. p. 0
Where the paper says it

While these solutions can be employed to integrate photonic circuits with CMOS chips [15–17], they are now a limitation for a wide range of research including routing of single photons from quantum dots [18], chipscale LiDAR solutions [6, 19], photonic interconnects for 5G networks [20], compact sensors [21], electro-optic and thermo-optic devices [22–24], and nanoelectromechanical photonic switches [25]. The solution proposed in this work overcomes the system-level complexity in routing electrons and photons inside the chip by introducing a carefully engineered and inversely designed monolithic device.

14257#pg3>. It allows crossing electronic connections and photonic waveguides using the same device layer without introducing significant electrical resistance or optical losses <ref:2204. p. 0
Where the paper says it

The solution proposed in this work overcomes the system-level complexity in routing electrons and photons inside the chip by introducing a carefully engineered and inversely designed monolithic device. It allows crossing electronic connections and photonic waveguides using the same device layer without introducing significant electrical resistance or optical losses.

14257#pg3>.

Optimization Methodology

The authors use topology optimization to devise a single-layer electronic-photonic circuit crossing with up to 99.8 % optical transmission across a 20 nm electrical isolation trench <ref:2204. p. 0

Where the paper says it

Here, we use topology optimization to devise a single-layer electronic-photonic circuit crossing with up to 99. 8 % optical transmission across a 20 nm electrical isolation trench.

14257#pg2>. The problem is recast as a continuous constrained optimization problem, where the quality of the structure under design is measured using a figure of merit, Φ, which is optimized by changing the structural geometry iteratively and systematically <ref:2204. p. 8
Where the paper says it

Inverse design of electronic-photonic circuit crossings by topology optimization Topology optimization [1] is an inverse design method, applicable across many areas of physics and engineering [2–7]. When applying topology optimization to solve a structural design problem, the problem is recast as a continuous constrained optimization problem, where the quality of the structure under design is measured using a figure of merit, Φ, which is optimized by changing the structural geometry iteratively and systematically.

14257#pg2>. The model domain is discretized and solved using the finite-element method [16] in COMSOL Multiphysics <ref:2204. p. 2
Where the paper says it

In order to experimentally verify these findings, we fabricate a set of photonic circuits with 0 to 6 EPCCs in series using electron-beam lithography and cryogenic deep reactive ion etching applied to SOI wafers. This enables direct measurement of the insertion loss of a single EPCC using the cut-back method, which separates the coupling losses from propagation losses.

14257#pg2>. The optimization problem is formulated to maximize the minimum of Φ over a set of N targeted wavelengths, which can be written as max ξ(r) min λi Φ(E(λi, ˜ξ)) (2) <ref:2204. p. 2
Where the paper says it

The parameter α = 1 is added to the denominator to avoid numerical issues as R → 0 caused by the singularity at R = 0 as well as to balance the priority to T vs R such that the optimization algorithm will not sacrifice T when R << 1. The EPCC is optimized at the wavelengths corresponding to the beginning and the end of the targeted bandwidth, i.

14257#pg2>.

Performance and Applications

The optimized EPCC achieves excellent performance, with a transmission of T > 90.9 % and a reflection of R < 1. p. 1

Where the paper says it

2 1 0 1 0 -1 Wavelength (nm) (a) North: Electrical connections (b) (c) (d) (e) (f) Electrical link Optical link Optical waveguide Electrical link Silicon bridge Input Output Electrical isolation 1 µm 500 nm x y z South: Electrical connections West: Optical connections East: Optical connections 100 80 60 40 20 0 5 4 3 2 1 0 1500 1520 1540 1560 1580 1600 Reflection (%) Transmission (%) 500 nm Fig. Electronic-photonic integration with electronic-photonic circuit crossings.

4 % over a 100 nm bandwidth as well as a peak transmission of T = 93. p. 2
Where the paper says it

9 % and a reflection of R < 1. 4 % over a 100 nm bandwidth as well as a peak transmission of T = 93.

6 %, which is a dramatic improvement compared to the solutions before topology optimization <ref:2204. p. 2
Where the paper says it

4 % over a 100 nm bandwidth as well as a peak transmission of T = 93. 6 %, which is a dramatic improvement compared to the solutions before topology optimization (figs.

14257#pg3>. For larger networks, the optical transmission becomes a limiting factor and we therefore design additional EPCCs while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps [34] <ref:2204. p. 4
Where the paper says it

This demonstrates one of the central building blocks for programmable [14, 32, 33] or quantum photonic circuits [22] where the full benefit of the EPCC is ultimately harvested in larger networks: Regardless of topology and complexity, advanced electronic-photonic circuits can be fabricated in the same device layer and in a single lithography step. For larger networks, the optical transmission becomes a limiting factor and we therefore design additional EPCCs while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps [34].

14257#pg3>. A power transmission of T = 99. p. 1
Where the paper says it

(e) Magnitude and direction of the time-averaged power flow in the middle plane of the structure crossing a 100 nm isolation gap at 1550 nm. (f) Calculated transmission (red) and reflection (blue) of the optimized EPCC over 100 nm wavelength span centered at 1550 nm.

8 % implies that hundreds of EPCCs in series would result in insertion losses below 3 dB <ref:2204. p. 4
Where the paper says it

Because the overall transmission of circuits with N crossings scales as T N , a power transmission of T = 99. 8 % implies that hundreds of EPCCs in series would result in insertion losses below 3 dB.

p. 4
Where the paper says it

Because the overall transmission of circuits with N crossings scales as T N , a power transmission of T = 99. 8 % implies that hundreds of EPCCs in series would result in insertion losses below 3 dB.

14257#pg3>. The resulting topology-optimized EPCCs can be integrated into a monolithic silicon nanoelectromechanical add-drop switch where the flow of photons, electrons, and mechanical motions are fully integrated within the same layer <ref:2204. p. 0
Where the paper says it

9 % over a 100 nm bandwidth, in excellent agreement with theory. We use these concepts to demonstrate a monolithic silicon nanoelectromechanical add-drop switch in which the flow of photons, electrons, and mechanical motions are fully integrated within the same layer.

14257#pg3>.

Experimental Verification

To experimentally verify these findings, we fabricate a set of photonic circuits with 0 to 6 EPCCs in series using electron-beam lithography and cryogenic deep reactive ion etching applied to SOI wafers <ref:2204.14257#pg3>. The measured average transmission of 92. p. 2

Where the paper says it

Six different sets of each photonic circuit are fabricated on the same chip, which allows us to do ensemble measurements and carefully estimate the standard deviation. Figure 2b demonstrates an excellent agreement between the experimental transmission, measured over a 100 nm bandwidth with an average transmission of 92.

9 % over a 100 nm bandwidth, in excellent agreement with theory, is demonstrated through ensemble measurements <ref:2204. p. 2
Where the paper says it

Six different sets of each photonic circuit are fabricated on the same chip, which allows us to do ensemble measurements and carefully estimate the standard deviation. Figure 2b demonstrates an excellent agreement between the experimental transmission, measured over a 100 nm bandwidth with an average transmission of 92.

14257#pg3>. The experimental results show that the curve-fitted topology-optimized design has a broadband figure of merit Φb = T/(R+α) = T/(R+1) > 0. p. 8
Where the paper says it

Inverse design of electronic-photonic circuit crossings by topology optimization Topology optimization [1] is an inverse design method, applicable across many areas of physics and engineering [2–7]. When applying topology optimization to solve a structural design problem, the problem is recast as a continuous constrained optimization problem, where the quality of the structure under design is measured using a figure of merit, Φ, which is optimized by changing the structural geometry iteratively and systematically.

892 for the 240 nm device layer <ref:2204. p. 1
Where the paper says it

(f) Calculated transmission (red) and reflection (blue) of the optimized EPCC over 100 nm wavelength span centered at 1550 nm. Numerical calculations were performed for a device with a thickness of 240 nm determined by the measured thickness of the device layer of the SOI wafer after fabrication.

14257#pg2>. The fabrication involves a detailed process flow including patterning, etching, and wet release etch of suspended silicon structures <ref:2204.14257#pg3>.

Future Directions

The EPCC opens perspectives for a wide range of new research and technology <ref:2204. p. 4

Where the paper says it

Our results indicate that near-unity transmission is attainable for smaller gaps. The EPCC opens perspectives for a wide range of new research and technology.

14257#pg3>. For example, isolation trenches are not only efficient electrical isolators, they are also ideal for thermal isolation in nanophotonic devices [23] or mechanical isolation for self-assembly of nanophotonic circuits [35] and quantum optomechanics [36] <ref:2204. p. 4
Where the paper says it

The EPCC opens perspectives for a wide range of new research and technology. For example, isolation trenches are not only efficient electrical isolators, they are also ideal for thermal isolation in nanophotonic devices [23] or mechanical isolation for self-assembly of nanophotonic circuits [35] and quantum optomechanics [36], and we envision our technology enabling, e.

14257#pg2>. In addition, our concepts apply directly to other semiconductor platforms such as gallium arsenide or indium phosphide, where they enable independent electrical control of quantum emitters in integrated photonic circuits [18] thus solving a major scaling problem in integrated quantum photonics <ref:2204. p. 4
Where the paper says it

, phonon-photon crossings. In addition, our concepts apply directly to other semiconductor platforms such as gallium arsenide or indium phosphide, where they enable independent electrical control of quantum emitters in integrated photonic circuits [18], thus solving a major scaling problem in integrated quantum photonics.

14257#pg3>. The technology enables, e. p. 2
Where the paper says it

In order to experimentally verify these findings, we fabricate a set of photonic circuits with 0 to 6 EPCCs in series using electron-beam lithography and cryogenic deep reactive ion etching applied to SOI wafers. This enables direct measurement of the insertion loss of a single EPCC using the cut-back method, which separates the coupling losses from propagation losses.

g., phonon-photon crossings <ref:2204. p. 4
Where the paper says it

For example, isolation trenches are not only efficient electrical isolators, they are also ideal for thermal isolation in nanophotonic devices [23] or mechanical isolation for self-assembly of nanophotonic circuits [35] and quantum optomechanics [36], and we envision our technology enabling, e. , phonon-photon crossings.

14257#pg3>.

Fabrication Details

The samples are fabricated from undoped (100) 150 mm SOI wafers (SOITEC) with a 3 µm buried oxide <ref:2204.14257#pg3>. The nominally undoped silicon device layer has a resistivity of ∼ 10 Ω cm, which is sufficiently low for making electrical contacts while imposing minimal optical absorption losses <ref:2204. p. 4

Where the paper says it

The drive contacts are electrically connected to the switching elements through the silicon device layer and isolated from each other by the isolation trenches, the EPCCs, and the supporting buried-oxide layer, as shown in Fig. The switch network is characterized by sequentially applying voltages to the drive contacts while monitoring the optical transmission from the input to the output ports.

14257#pg3>. The device layer has a nominal thickness of 250 nm ±30 nm <ref:2204. p. 1
Where the paper says it

(f) Calculated transmission (red) and reflection (blue) of the optimized EPCC over 100 nm wavelength span centered at 1550 nm. Numerical calculations were performed for a device with a thickness of 240 nm determined by the measured thickness of the device layer of the SOI wafer after fabrication.

14257#pg3>. The thickness of the device layer is measured to be 240 nm using variable-angle spectroscopy ellipsometry after fabrication. p. 1
Where the paper says it

(f) Calculated transmission (red) and reflection (blue) of the optimized EPCC over 100 nm wavelength span centered at 1550 nm. Numerical calculations were performed for a device with a thickness of 240 nm determined by the measured thickness of the device layer of the SOI wafer after fabrication.

The devices are patterned with a single electron-beam lithography step and etched directly from the soft-mask using deep reactive-ion etching (DRIE) in a continuous cryogenic (−19 °C) inductively-coupled plasma (ICP) <ref:2204. p. 2
Where the paper says it

In order to experimentally verify these findings, we fabricate a set of photonic circuits with 0 to 6 EPCCs in series using electron-beam lithography and cryogenic deep reactive ion etching applied to SOI wafers. This enables direct measurement of the insertion loss of a single EPCC using the cut-back method, which separates the coupling losses from propagation losses.

14257#pg3>. The wet release etch of suspended silicon structures involves etching of buried oxide (SiO2) followed by cleaning and drying steps <ref:2204. p. 4
Where the paper says it

3c and includes one ground and four drive contacts for actuating the switching elements. The drive contacts are electrically connected to the switching elements through the silicon device layer and isolated from each other by the isolation trenches, the EPCCs, and the supporting buried-oxide layer, as shown in Fig.

14257#pg3>. The metallization and release of fragile suspended silicon membranes are carried out using specific spin-coating, lithography, development, metal deposition, and lift-off procedures <ref:2204. p. 0
Where the paper says it

We focus our effort on silicon membranes fabricated on the silicon on insulator (SOI) platform but the concepts can be readily applied to all other semiconductor platforms. Figure 1a illustrates a generic circuit topology for routing electrons and photons where photons are guided arXiv:2204.

14257#pg3>.

Switch Network Demonstration

The 2 × 2 add-drop network includes four switching elements each comprised of a directional coupler driven by two electrostatic comb-drive actuators [20] <ref:2204.14257#pg4>. The coupling between them and thus the flow of photons is controlled by an adjustable gap (Fig. p. 2

Where the paper says it

Figure 3a shows a dark-field microscope image of the 2 × 2 switch network. The network is built on four reconfigurable directional couplers controlled by electrostatic comb-drive actuators with folded springs [29, 30] (Fig.

S13b) <ref:2204. p. 0
Where the paper says it

We focus our effort on silicon membranes fabricated on the silicon on insulator (SOI) platform but the concepts can be readily applied to all other semiconductor platforms. Figure 1a illustrates a generic circuit topology for routing electrons and photons where photons are guided arXiv:2204.

14257#pg4>. By driving the comb-drive actuators, the gap and thus the coupling strength is adjusted such that the input light is coupled to output 2 (Fig. p. 4
Where the paper says it

The switch network is characterized by sequentially applying voltages to the drive contacts while monitoring the optical transmission from the input to the output ports. The light coupled to the input I1 (I2) can be transmitted to the corresponding through port T1 (T2) or directed to any of the drop ports D1 and D2 depending on the driving voltage of the activated comb-drive actuator.

S14b) <ref:2204. p. 0
Where the paper says it

We focus our effort on silicon membranes fabricated on the silicon on insulator (SOI) platform but the concepts can be readily applied to all other semiconductor platforms. Figure 1a illustrates a generic circuit topology for routing electrons and photons where photons are guided arXiv:2204.

14257#pg4>. The experimental setup shows no significant electrical crosstalk between different nanoelectromechanical actuators <ref:2204. p. 0
Where the paper says it

It is conventionally achieved with thermo-optic tuning, but this suffers from high energy consumption and crosstalk. Nanoelectromechanical photonics could resolve these issues, but integrating this technology with conventional multilayer metal architectures is challenging, and conventional approaches do not allow crossings of electrical wires and photonic waveguides.

14257#pg4>. When actuator 1 is driven, the input light is routed from T2 to D2 (see Fig. p. 4
Where the paper says it

The light coupled to the input I1 (I2) can be transmitted to the corresponding through port T1 (T2) or directed to any of the drop ports D1 and D2 depending on the driving voltage of the activated comb-drive actuator. The through- and drop-port transmission at 1550 nm is plotted in Fig.

3h of the main text) so T2 depends on the actuation voltage <ref:2204. p. 3
Where the paper says it

(b) 45◦ -tilted scanning electron micrograph of one switching element acquired at a low accelerating voltage of 5 kV in order to illustrate the charging effect, showing that excellent electrical isolation is obtained between the actuator electrodes via multiple EPCCs and isolation trenches. The inset in (b) shows a zoomed top view scanning electron micrograph of one of the EPCCs employed to route the actuation signal to the electrostatic actuators.

14257#pg4>.

Experimental Procedure

Loss measurements of each device can be accurately performed by the cutback method, where the device loss is extracted by comparing the optical transmission through a varying number of devices <ref:2204.14257#pg3>. A linearly polarized tunable laser is coupled to the chip via an input grating coupler while swept from 1500 to 1600 nm with a step of 0.5 nm <ref:2204. p. 0

Where the paper says it

We focus our effort on silicon membranes fabricated on the silicon on insulator (SOI) platform but the concepts can be readily applied to all other semiconductor platforms. Figure 1a illustrates a generic circuit topology for routing electrons and photons where photons are guided arXiv:2204.

14257#pg3>. A grating coupler with 90-degree rotation, and hence orthogonal polarization, is used to couple the output light to a single-mode optical fiber <ref:2204. p. 4
Where the paper says it

The switch network is characterized by sequentially applying voltages to the drive contacts while monitoring the optical transmission from the input to the output ports. The light coupled to the input I1 (I2) can be transmitted to the corresponding through port T1 (T2) or directed to any of the drop ports D1 and D2 depending on the driving voltage of the activated comb-drive actuator.

14257#pg3>. The effect of the Fabry-Perot cavity can be discarded by considering the upper peak envelope of the transmission spectra (bold lines in Fig S8) <ref:2204. p. 2
Where the paper says it

1e) of the transverse-electric mode. The calculated transmission and reflection spectra are shown in Fig.

14257#pg3>. A linear fit is applied to each set of data associated with each nominally identical set of photonic circuit at each collected data point, i. p. 2
Where the paper says it

S1 and S2). In order to experimentally verify these findings, we fabricate a set of photonic circuits with 0 to 6 EPCCs in series using electron-beam lithography and cryogenic deep reactive ion etching applied to SOI wafers.

e., each wavelength of the obtained transmission <ref:2204. p. 4
Where the paper says it

Topology-optimized EPCCs with near-unity transmission. The Magnitude and direction of the timeaveraged power flow in the middle plane of the structure is shown for designs obtained with single-wavelength topology optimization with gaps of (a) 100 nm, (b) 60 nm, and (c) 20 nm.

14257#pg3>. The transmission loss, L, can thereby be determined by directly measuring the gradient of the 10 −L 10 <ref:2204.14257#pg3>.

Characterization Results

The calculated transmission and reflection spectra show a peak transmission of T = 99.8 % for a gap of 20 nm <ref:2204. p. 0

Where the paper says it

We focus our effort on silicon membranes fabricated on the silicon on insulator (SOI) platform but the concepts can be readily applied to all other semiconductor platforms. Figure 1a illustrates a generic circuit topology for routing electrons and photons where photons are guided arXiv:2204.

14257#pg5>. The optimized EPCCs with isolation gaps of 20 nm, 60 nm, and 100 nm are presented in Fig. p. 4
Where the paper says it

For larger networks, the optical transmission becomes a limiting factor and we therefore design additional EPCCs while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps [34]. Figure 4 shows EPCCs for (a) 100 nm, (b) 60 nm, and (c) 20 nm isolation gaps topology-optimized for maximum transmission at 1550 nm.

S6 <ref:2204. p. 0
Where the paper says it

We focus our effort on silicon membranes fabricated on the silicon on insulator (SOI) platform but the concepts can be readily applied to all other semiconductor platforms. Figure 1a illustrates a generic circuit topology for routing electrons and photons where photons are guided arXiv:2204.

14257#pg20>. The curve-fitted topology-optimized design shows a transmission above 7 for the broadband figure of merit Φb = T/(R+1) <ref:2204. p. 4
Where the paper says it

For larger networks, the optical transmission becomes a limiting factor and we therefore design additional EPCCs while relaxing the bandwidth requirement and allowing for smaller but still feasible gaps [34]. Figure 4 shows EPCCs for (a) 100 nm, (b) 60 nm, and (c) 20 nm isolation gaps topology-optimized for maximum transmission at 1550 nm.

14257#pg19>. The overall transmission of circuits with N crossings scales as T N, a power transmission of T = 99. p. 4
Where the paper says it

Furthermore, it remains above 99 % across a 35 nm bandwidth (fig. Because the overall transmission of circuits with N crossings scales as T N , a power transmission of T = 99.

8 % implies that hundreds of EPCCs in series would result in insertion losses below 3 dB <ref:2204.14257#pg5>. The performance is robust to small variations in device thickness, showing the 10 nm thinner device layer does not have a significant influence on the performance <ref:2204. p. 1
Where the paper says it

(f) Calculated transmission (red) and reflection (blue) of the optimized EPCC over 100 nm wavelength span centered at 1550 nm. Numerical calculations were performed for a device with a thickness of 240 nm determined by the measured thickness of the device layer of the SOI wafer after fabrication.

14257#pg18>. The replacement of mode-expanding features with a tapering section results in an overall better performance when averaged across the full 100 nm bandwidth <ref:2204. p. 8
Where the paper says it

To improve the performance of the crossing, one may expand the 400 nm suspended single-mode waveguide to a 1 µm wide silicon region through a parameter-optimized tapering section (Fig S2) which results in a noticeable improvement compared to the naive approach. However, Fig S2 shows that a mode expanding region alone cannot diminish the reflection to sufficiently low values for practical applications.

14257#pg18>. The final design is robust to small variations in device thickness, which is likely due to the broadband optimization procedure <ref:2204. p. 1
Where the paper says it

Due to the coherent nature of the photon scattering and interference, designing such a subwavelength mode-matching EPCC is a complicated and non-intuitive problem. We therefore employ inverse design by topology optimization [26, 27] to develop a compact and scalable device that transmits photons efficiently across the electrical isolation trench.

14257#pg18>.

Summary of Key Findings

The single most informative sentence about the paper's subject and finding is that topology optimization allows for crossing electronic connections and photonic waveguides using the same device layer without introducing significant electrical resistance or optical losses <ref:2204.14257#pg3>. The authors successfully devised a monolithic silicon nanoelectromechanical add-drop switch in which the flow of photons, electrons, and mechanical motions are fully integrated within the same layer <ref:2204. p. 0

Where the paper says it

9 % over a 100 nm bandwidth, in excellent agreement with theory. We use these concepts to demonstrate a monolithic silicon nanoelectromechanical add-drop switch in which the flow of photons, electrons, and mechanical motions are fully integrated within the same layer.

14257#pg2>. This demonstrates that near-unity transmission is attainable for smaller gaps through topology optimization <ref:2204. p. 4
Where the paper says it

This shows that highly complex hybrids of nanoelectronics, nanophotonics, and nanomechanics can be integrated with negligible losses. Our results indicate that near-unity transmission is attainable for smaller gaps.

14257#pg5>. The EPCC opens perspectives for a wide range of new research and technology <ref:2204.14257#pg3>.

References

[1] D. A. Miller, Device requirements for optical interconnects to silicon chips, Proc. p. 4

Where the paper says it

Previous research on this subject has used shallow etching of individual doped layers for individual electrical tuning of quantum dots [37], but this approach would not work for homogeneously doped semiconductor platforms and does not address the more general question of how to provide electrical, thermal, and mechanical isolation across a gap with high optical transmission. Miller, Device requirements for optical interconnects to silicon chips, Proc.

IEEE 97, 1166 (2009)

[2] M. Ayata, Y. Fedoryshyn, W. p. 4

Where the paper says it

IEEE 97, 1166 (2009). Fedoryshyn, W.

Heni, B. Baeuerle, A. p. 4
Where the paper says it

Fedoryshyn, W. Baeuerle, A.

Josten, M. Zahner, U. Koch, Y. Salamin, C. p. 4
Where the paper says it

Baeuerle, A. Salamin, C.

Hoessbacher, C. p. 4
Where the paper says it

Salamin, C. Hoessbacher, C.

Haffner, D. p. 4
Where the paper says it

Hoessbacher, C. Haffner, D.

L.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Topology optimization can be used to design monolithic devices that allow crossing of electrical wires and photonic waveguides without introducing significant electrical resistance or optical losses. This enables a new class of integrated circuits where the flow of photons, electrons, and mechanical motions are fully integrated within the same layer.

  2. The system can achieve near-unity transmission for large networks by utilizing topology-optimized electronic-photonic circuit crossings, with a power transmission of T = 99.8 % implies that hundreds of EPCCs in series would result in insertion losses below 3 dB.

  3. The improved AI system can design and fabricate monolithic electro-mechanical photonic add-drop networks where the full benefit of the EPCC is ultimately harvested in larger networks, enabling programmable or quantum photonic circuits.

Abstract

Electrical control of light in integrated photonics is central to a wide range of research and applications. It is conventionally achieved with thermo-optic tuning, but this suffers from high energy consumption and crosstalk. Nanoelectromechanical photonics could resolve these issues, but integrating this technology with conventional multilayer metal architectures is challenging, and conventional approaches do not allow crossings of electrical wires and photonic waveguides. Here, we use topology optimization to devise a single-layer electronic-photonic circuit crossing with up to 99.8 % optical transmission across a 20 nm electrical isolation trench. We focus our experiments on 100 nm trenches and measure an average transmission of 92.9 % over a 100 nm bandwidth, in excellent agreement with theory. We use these concepts to demonstrate a monolithic silicon nanoelectromechanical add-drop switch in which the flow of photons, electrons, and mechanical motions are fully integrated within the same layer. Our work addresses an important challenge in incorporating opto-electro-mechanical topologies into photonic integrated circuits and may lead to new functionalities in nano-opto-electro-mechanical systems, optomechanics, and integrated quantum photonics.

Related papers