2.1

Micron-scale simulations for photonics data visualization

Many photonics engineering programs encourage new students to use commercial simulation software such as the Ansys Lumerical or Synopsys tool suites. While powerful and complex optical modeling tools have many advantages, there are also major drawbacks to using commercial software, which is not designed for education, to train undergraduate and early graduate students. The most significant is the cognitive overload students experience when first using these tools. There is a large barrier to entry for novice users who are not familiar with the long list of dropdown menu options and adjustable settings. Students are not focused on the intended photonics learning goals of these activities if they are busy defining finite difference time domain (FDTD) mesh regions, correctly setting up boundary conditions, navigating the software GUI, learning how to write code in custom scripting languages, waiting minutes to hours for each simulation sweep to run, and finally assembling and plotting the results. The feedback loop between initial setup and viewing results is too prolonged (and the parameter space too large) for students to make fast progress in understanding photonics fundamentals. Additionally, many students will be hesitant when interpreting the results from a commercial simulation tool, knowing a small mistake could lead to incorrect outputs, which a more experienced user would easily be able to catch and correct.

A more effective training environment can use expert-vetted simulation results to provide a simpler, intermediate steppingstone between textbook examples and exercises using commercial software. For training in integrated photonics, VM-Lab has created a series of 20 introductory modules featuring scaffolded simulation sequences which allow users to explore waveguide modes and photonic circuit components using sliding bars, graphs, and simple user interfaces.

As shown in Figure 2, the VM-Lab team has designed web-based photonics training simulations with novel 2D and 3D data visualization tools. Students can view real-time vector field profiles for Transverse Electric (TE) and Transverse Magnetic (TM) waveguide mode polarizations (Figure 2a), colorized mode cross sections (Figure 2b), observe a linear interpretation of mode superposition at different positions along a waveguide structure (Figure 2c), and uncover data across parameter sweeps through an interactive graph (Figure 2d). The final simulation GUI shown in Figure 2e/2f is the result of an iterative multi-year development process used to create and deploy these training simulations, including formative and summative learning science research, described in Sections 3-6 below.

Figure 2.

Photonics training simulations showing waveguide modes in a rectangular Si/SiO₂ dielectric waveguide, which use novel visualization training tools including (a) dynamic 3D vector field representation of the waveguide mode, (b) dynamic colorized cross-sectional profiles of field components, (c) linear visualizations of modes to demonstrate mode superposition, and (d) explorable graphs for each simulation parameter space. The lower screenshots show examples of the final GUI layout of the training simulations including (e) Waveguide Mode Explorer (WME) and (f) Waveguide Fundamentals.

2.2

Educational web game: building photonic circuits for sensing applications

Digital games can be a great way for students to discover the applications of STEM fields and explore complex dynamic systems. For students studying remotely or taking online courses without access to classroom activities, interactive learning games can fill a gap in their studies. Engaging games also increase student motivation and prevent students from becoming disillusioned with poor performance on difficult exercises and tasks.¹³ In games, failing before achieving a goal is not only natural to the medium, but can be a rewarding and satisfying experience. When stuck, many players will methodically explore the rules of their environment, get feedback from the game, and try different approaches to achieve an in-game objective.¹⁴ Progressing through the levels of a learning game and seeing measurable success can quickly build student confidence in their STEM skills.^15,16 Games have been successfully used for a wide variety of teaching, including introductions to electromagnetism,¹⁷ demonstrate the intricacies of supply chain management,¹⁸ and to encourage underrepresented youth to pursue careers in STEM fields by creating identity-building narratives.¹⁹

The Virtual Manufacturing Lab at MIT created four educational web games demonstrating the main application areas of integrated photonics. In addition to the datacenter management game “Databytes, Inc.” shown in Figure 1 (right), the team created a drone-building game to teach students about RF communication, an infinite runner where players update the LiDAR imaging capabilities of their vehicle, and a circuit-building web game “Illuminator” shown in Figure 3. In “Illuminator,” learners use electro-optic circuit components to design and build a photonic sensor which can detect methane gas. In a sequence of 17 levels, players choose wavelengths of operation, lay out the sensing and reference paths required for the optical circuit design, select the appropriate sensing circuit elements such as spirals or ring resonators, use functionalized coatings to increase signal-to-noise ratio, and select photodetector material systems. The screenshot in Figure 3a shows the chip view where players place circuit components on the grid next to fiber input/output ports. Figure 3b shows the printed circuit board view where players place lasers and detectors. Players are challenged to be efficient in their designs by minimizing chip footprint and foundry costs. Each level has a series of objectives and hints (Figure 3c), available components (Figure 3d), and the ability to expose and coat sections of the circuit to the gas or liquid analyte for chemical sensing (Figure 3e). Students can also test their incomplete circuits with and without the presence of methane and other chemicals and, when they are ready, submit a design to pass the level. Each level of the game introduces a new learning goal for the player, and the levels get progressively more difficult.

Figure 3.

The integrated photonics web game “Illuminator”, created by MIT and Fire Hose Games studio. Players build (a) photonic integrated circuits laying out electro-optic circuit components for on-chip chemical sensing, and (b) printed circuit boards with lasers, photonic chips, and detectors. Each level includes (c) detailed level objectives, (d) a variety of circuit components which can be placed on the chip grid, and (e) the ability to expose and add a functionalized coatings to elements on the chip, and then test and submit your chemical sensing circuit to complete the level.

The VM-Lab team tested the web games “Illuminator” and “Databytes, Inc.” during multiple workforce training programs including residential learning at MIT, the AIM Photonics Academy training bootcamp, and a 15-month technician training program. The games were well received, and participants showed increased motivation during these sessions reporting high levels of satisfaction with the interactive experience.

3.1

Simulation prototyping and formative assessment research

We developed and tested our interactive simulations, as well as full online-learning modules, complete with videos and assessment exercises, in three phases. During each phase we recruited volunteers from a pool of students and workers who were likely to engage with simulations as part of their education, reskilling, or upskilling.

Phase 1 – Testing the first iteration of simulations with undergraduate physics students

In Phase 1, the subject matter experts (SMEs) and the learning sciences research team discussed and developed learning outcomes for the simulations. By the end of the simulation sequence, learners were expected to be able to:

Additionally, the learning sciences team conducted a cognitive task analysis with SMEs to understand the goals of the simulations and how learners should benefit from them.

Participants: An initial playtest was conducted with one industry expert and four undergraduate students from a university in Massachusetts. The students were undergraduates majoring in physics who had a basic background in optics and electromagnetic wave theory. They did not have a background in photonics or waveguides. The playtest was included as part of a field trip to MIT to learn more about photonics work currently being conducted at MIT.

Procedure: After signing a consent to participate in the study, participants heard a brief introduction to the simulations shown in Figure 4. The agenda was planned to include three simulations: (1) Waveguide Fundamentals, (2) Directional Coupler, and (3) The Multimode Interferometer (MMI). Although they were not held strictly to a time limit, participants were allotted approximately 15 minutes to complete each simulation, along with answering assessment questions they received in a separate handout. Participants were asked to ‘think out loud,’ or articulate their thoughts, as they interacted with the simulations. A researcher recorded their comments or questions and observed their interactions with the simulations. At the conclusion of the session, participants were asked to complete a questionnaire regarding the usability of the simulations. Due to limitations of time, participants did not complete the MMI simulation.

Figure 4.

First iteration of photonics simulations for (left) Waveguide Fundamentals, and (right) Directional Coupler.

Instruments: The SMEs developed assessment questions associated with each simulation. Participants were free to respond to the questions either during their interaction with the simulation or immediately afterward. The questions were intended to assess participants’ understanding of basic concepts related to waveguides and their use. Researchers also utilized an observation tool to record participants’ behaviors and articulations as they engaged with the simulation. A standard usability survey was utilized for user feedback on the simulation.

Data Analysis and Results: Data from the assessment questions, researchers’ observations, and the usability survey were analyzed to explore participants’ experience during the session. For the Waveguide Fundamentals simulation, the participants’ mean score was 73% for the five assessment questions, with the largest number of incorrect responses related to the concept of “confinement factor”, or how confined the mode is inside the waveguide core. We also observed confusion about mode naming conventions, when an optical mode was supported, and when it was considered excited. Researcher observations conducted during the Waveguide Fundamentals simulation revealed that participants asked questions about the meaning of effective index, confinement, mode order, and mode polarization. When engaging with the simulation, researchers observed that participants did not always explore the entirety of each simulation, thus missing the sequencing and incremental building of knowledge as planned. For the Directional Coupler simulation, the participants’ mean score for the six questions was 70%, with the largest number of errors relating to understanding the effect of coupler gap on effective index and oscillation of the field between a top and bottom wave guide. As they progressed through the simulation, participants commented and asked more questions about the representations of simulation elements, e.g., one participant said, “Having [a] red arrow and red signal was confusing.” This participant expressed that they would like more detailed explanation of what was on the screen. Students were confused about the red/blue colorized representation of cross-sectional electromagnetic field profiles (e.g., E_x, E_y, and E_z) shown in the top-left of the simulations as in Figure 4 and expressed a desire for a more basic introduction to waveguide modes.

Actions: Analysis of the data collected supported the need for creation of a new teaching simulation that contained a basic introduction to wave propagation. This module should explain the red/blue colorized cross-sectional mode profiles. The SMEs also concluded that an orienting depiction of light propagation through free space would provide a helpful contrast to subsequent simulations showing propagation through a rectangular waveguide. In addition, the design and development team decided to include a tutorial sequence for every simulation to orient the user to each GUI element.

Phase 2 – Photonics expert feedback and review

Following student feedback from Phase 1, the learning sciences team observed photonics faculty and experts in the field who attended an MIT blended learning bootcamp. Only selected sessions which included the simulations were observed. In these sessions, bootcamp participants engaged with a series of photonics simulations. No formal data collection was conducted; however, the research team gathered feedback from participants after their simulation experiences.

Observations: When exploring the simulations, bootcamp participants tended to use the plane view rather than the line or point view of wave propagation. Additionally, they articulated additional questions about waveguide behavior as they engaged in the simulations. They expressed lack of understanding of the following areas: 1) Definition of effective index and its relationship to refractive index; 2) Mode coupling from one device to another, and distinguishing between coupling into a standard taper and an inverse tapered structure for edge coupling (in this case, participants articulated a specific misconception around the method of coupling⎯that inverse tapers are relying on expanding the mode into the cladding, while regular tapers are making the core dimensions larger); and 3) Explanation of device behavior when mode theory was more appropriate than a ray optics interpretation, specifically with the Y-branch splitter/combiner shown in Figure 5 (left), Multimode Interferometer (MMI) shown in Figure 5 (right), and ring resonator (not shown); for instance, many could not explain a 50% drop in power when only one input path is excited in a Y-combiner, as this requires understanding of modal interference.

Figure 5.

First iteration of the (left) Y-branch Splitter/Combiner, and initial prototype of the (right) Multimode Interferometer MMI simulation.

Actions: Further discussion of bootcamp observations led to a move away from using all three views of electric field profiles (point, line, and plane) in each waveguide simulation. Developers made the decision to rely on a linear “free space” simulation at the beginning of the module and added camera controls and additional animations/illustrations shown in Figure 6 below to orient learners to the wave properties of the light in the direction of propagation. From suggestions from the bootcamp attendees, the developers made significant changes to the simulations which featured more advanced photonic circuit component, including the Directional Coupler, MMI, and Y-branch simulations, including updated UI and additional visualizations to make the circuit components more intuitive.

Figure 6.

Prototype of Waveguide Mode Explorer simulation sequence with: (left) an introduction to free space EM wave propagation, and (right) new 3D visualizations for guided modes in a dielectric waveguide.

Phase 3 – Testing new introductory simulations with students enrolled in a 2-year optics program

Based on a strong need for additional introductory content, the development team created a new “back-to-basics” introduction to waveguides, the Waveguide Mode Explorer (WME) module shown in Figure 6. The WME module included an introductory free space simulation, as well as videos and simulations that depicted the x, y, and z components of the fundamental waveguide modes for each polarization (TE₀ and TM₀). The intended learning outcomes for this module were retained from the earlier version. As part of the development process, two SMEs participated in think aloud sessions in which they engaged with module simulations in their current stage of development. During the sessions, they articulated the various ideas and concepts that the simulations were designed to illustrate.

Participants: The second formal playtest was conducted with 15 students from a technical community college in Massachusetts. All students were enrolled in the community college’s Optics and Photonics Program, which focuses on the fundamental properties of light and using those properties in practical applications. The playtest was included as part of a field trip to MIT to learn more about photonics work currently being conducted at MIT.

Procedure: After signing a consent to participate in the study, participants completed a pretest. They used an identification number on their pre- and posttest, also used to log into the simulations. They then heard a brief introduction to integrated photonics and its applications and were subsequently introduced to three simulations in the WME module: Free Space Optics, Transverse Electric (TE) Guided Modes, and Transverse Magnetic (TM) Guided Modes. They were allocated approximately 20 minutes to complete each simulation and answer assessment questions they received in a separate handout. At the conclusion of the session, participants were asked to complete a posttest, as well as a questionnaire regarding their perception of the experience.

Instruments: The SMEs developed 13 assessment questions associated with the introductory material and simulations. These questions were designed to help participants focus on important concepts of the experience, and participants were asked to respond to the assessment questions during their interaction with the simulation. The pre- and posttest contained five identical questions and a sixth question was included on the posttest to assess participants’ understanding of electric field components. The pre- and posttest was designed to measure the learning outcomes of the WME module. The post experience survey contained four items that asked participants to reflect on their learning from the experience, and to describe any difficulties they had with the simulations.

Data Analysis and Results: Analysis of the pre- and posttest scores showed only small gains in participants’ understanding of integrated photonics concepts. The mean score for the pre-test was 41%, whereas the mean score for the same five questions on the posttest was 45%, and 46% when scores for the additional sixth question were added. The normalized gain for questions 1 - 5 ranged from -2.00 to .40, with a mean <g> = -0.06. The negative gain score occurred for a participant that did not have a laptop during the experience; however, this does not explain why that individual answered several questions correctly on the pretest and either omitted them or answered incorrectly on the posttest. As evidenced by the number of participants who answered incorrectly, some test items pointed to areas in which participants’ understanding of the concepts did not change after engaging with the simulations. When asked to draw a sketch of light traveling down a fiber or silicon waveguide (question 3), one participant correctly drew a modal view on both the pre- and posttest, whereas the remaining participants continued to draw the ray optics view on the posttest as they had done on the pretest. Identification of a valid guided mode (question 4) continued to be problematic from pre- to posttest, as only 3 of the 15 participants recognized a mode profile in which the electric field was totally confined within the core as being a valid guided mode. A greater number of participants (11) correctly identified a valid guided mode when the electric field was less confined to the core, but 5 of those participants were either unable to explain their reasoning or described light as ‘escaping’ or ‘being lost’ when it was outside of the core. Participants also had difficulty depicting the electric field in TE and TM polarizations (question 5); only 4 of the 15 participants drew the TE electric field correctly, and 3 participants were able to correctly depict the TM electric field.

Actions: After analysis of data and synthesis of results, the team determined that undergraduate students and early graduate students would be a more appropriate target group for information at this level of detail. Within our 2-year student sample, participants did not have the prerequisite knowledge about E required to benefit from the WME module.

Important Takeaways: Many 2-year students demonstrated a misconception regarding loss of energy in a waveguide. While these students were familiar with the concept of total internal reflection in the context of fiber optics, they were very confused when presented with a waveguide mode profile which showed an evanescent electromagnetic field extending into the cladding material. Many students, holding to a ray optics interpretation of total internal reflection, described this as loss, with light escaping from the core into the cladding. In reality, this evanescent field is part of the guided mode. We then decided to make this evanescent field misconception an object for further study in Phase 4.

Phase 4 – Final “think-aloud” formative research study with undergraduate and early graduate students

Having identified the correct education level which would most benefit from these simulations (undergraduate and early graduate engineering students), the development team assembled the five-part Waveguide Mode Explorer simulation sequence (Figure 7) into a full online learning module with videos and exercises. One of the learning sciences research team members then conducted a Phase 4 “think-aloud” study to test the full module with the target audience and make final recommendations for changes to the simulations, videos, and exercises.

Figure 7.

Final version of the Waveguide Mode Explorer simulation sequence with (left) introduction to free space EM field point/wave view, followed by (right) four planar vector field component breakdowns for a dielectric waveguide.

Participants: The sample was composed of 22 volunteer student participants recruited from university engineering programs in the southwestern, southern, and northeastern United States. Eight participants were female and 14 were male. Ten of the participants were graduate students, and the remaining 12 were in undergraduate programs. All undergraduate participants had taken introductory courses on electricity and magnetism, and most of the graduate student participants were enrolled in PhD programs in Optical Physics or Optical Engineering. After the first two participants demonstrated some confusion during the exercises and assessment, the team reduced the number of embedded assessment items and modified them to improve clarity for the remaining 20 students.

Procedure: University faculty contacts were asked to distribute study recruitment announcements to their undergraduate and graduate students. Students responded to researchers via email, signed a consent for participation, and scheduled a teleconference appointment for a think-aloud session. Sessions were conducted via Zoom and the researchers asked for permission from participants to record audio and the student’s screen as they used the simulations. All recordings were anonymized; students did not share their video feed and were not referred to by name during the recorded session. Participants took a pretest, went through the module, and then took a posttest. Pretest and posttest questions were answered either verbally or by using Zoom’s annotation tools. Students progressed through the full WME training module including instructional videos, exercises, and interactive simulations. Researchers also collected analytics data of participant activity in each simulation through Unity Analytics.

Instruments: A revised pre- and posttest based on the questions from Phase 3 included five multi-part questions that measured learners’ understanding of the following: (1) vector field component breakdown along the x, y, and z axes, (2) interpreting colorized cross-sectional field profiles, (3) ray optics vs. modal analysis perspective, (4) optical loss in a waveguide mode with prominent evanescent fields, and (5) waveguide mode order and polarization. The question about evanescent fields was intended to shed light on the misconception that was observed in Phase 3 and verify the presence of this misconception in an undergraduate and early graduate audience.

Data Analysis and Results: All participants except one showed overall learning gains from pre- to posttest due to the WME module. In general, undergraduates made the most progress on the more basic pre/posttest questions and did not make many gains on more conceptual questions. In the pretest, most of the graduate students were able to give satisfactory answers to items 1 and 2 above, and instead showed larger gains than the undergraduates on the more conceptual questions in items 3 and 4 about modal analysis and optical loss. All students made significant progress identifying waveguide modes by order and polarization. Both graduate and undergraduate students showed signs of the misconception about evanescent fields leading to optical loss, with some students describing the mode profile demonstrating power “leaking” out of the core, the mode “losing some of the field” due to the prominent evanescence.

Actions: Data from the think-aloud study provided the development team with student rationale for their responses to assessment questions as well as their pre- and posttest answers. The development team then made changes to the order and wording of the videos, simulations, and exercises. Additional text descriptions were added to explain the presence of evanescent fields in waveguide modes and highlight the fact that this mode did not inherently lead to optical loss.

3.2

Creating the full library of photonic circuit simulations

Following feedback and lessons learned from the formative assessment in Phases 1-4, the team finalized the following modules: Waveguide Mode Explorer (Figure 7), Waveguide Fundamentals (Figure 2f), Directional Coupler (Figure 8a), Ring Resonator (Figure 8b), Y-Branch Splitter (Figure 8c), and Multimode Interferometer (Figure 8d).

Figure 8.

Representative screenshots of the completed simulation sequences for the following photonic circuit components: (a) Directional Coupler, (b) Ring Resonator, (c) Y-branch Splitter/Combiner, and (d) Multimode Interferometer (MMI).

The team also completed simulation sequences exploring the following photonic circuit components: radial waveguide bends, Mach-Zehnder interferometers, edge and grating couplers, waveguide escalators, 1D Bragg gratings, on-chip photodetectors, waveguide crossings, asymmetrical directional couplers, and polarization splitter/rotators. An optical fiber mode explorer (similar to the 3D vector field profiles in the Waveguide Mode Explorer for cylindrical dielectric waveguides) as well as a desktop training simulation for fiber-to-chip coupling, for both fiber optic and integrated photonic training curricula. All content has been or will be made available in MIT-led online courses and bootcamps.

5.1

Student feedback from Photonic Integrated Circuits 1 advanced online course

The 2021 offering of the PIC1 course included a single “full explorer” simulation from the Waveguide Mode Explorer module with the largest parameter space (right image of Figure 7) which allowed students to explore E_x, E_y, E_z, and |E| for both TE and TM polarizations. The research team embedded a series of survey questions shown in Tables 1-3 below.

Table 1.

Survey response from the 2021 Photonic Integrated Circuits 1 online course.

Of the students who responded to our survey, Table 1 shows a large majority (88%) of students had never seen a 3D vector field representation of a waveguide mode. From our discussions with experts and graduate students, this perspective is seldom shown in static textbooks or as simulation output from commercial photonics simulation software.

Table 2.

Survey of student experience during the 2021 Photonic Integrated Circuits 1 online course.

From the survey results in Table 2, 82% of students agreed or strongly agreed that the simulation gave them a new perspective of waveguide modes, and 84% agreed or strongly agreed that they would like to explore similar simulations for higher-order modes (e.g., TE₁, TE₂, TM₁, etc.) and that it would be beneficial for future students to use the Waveguide Mode Explorer simulations. In general, the photonics simulations included in all PIC1 course offerings have been very well received and students are excited to use them.

Table 3.

Survey of user satisfaction during the 2021 Photonic Integrated Circuits 1 online course.

The development team also included user satisfaction surveys shown in Table 3 to verify ease of use and performance of the simulations. A large majority of users were satisfied or very satisfied with the quality of the simulations.

5.2

Probing student misconceptions through the Photonic Integrated Circuits 1 advanced online course

To further explore the misconception regarding loss of power due to the presence of an evanescent EM field that we observed in our earlier work, we added a revised version of our question to an optional survey in the 2021 PIC1 online course. Even though this course did not directly address photonics theory, we reasoned that it was a difficult course and those who participated would likely have some expertise in the field.

The visual shown in Figure 10 was provided with a short description and a two-part question so that participants could provide their rationale for each response as shown below.

Figure 10.

Image provided to the PIC1 course offering as an optional survey question. The core and cladding materials are clearly delineated with a strong evanescent field extending into the cladding.

For light in the guided mode above, if the waveguide continues straight for an additional 100 μm:

Of the 41 students in this advanced photonics course who responded to the questions, 39% responded incorrectly, stating ‘yes’ that power would be lost. The rationale provided for the two questions revealed their thinking about evanescent fields. Their responses included statements such as “There are evanescent leakage fields;” “There is a field in the oxide cladding, therefore the field in the waveguide has lost power compared to the input;” and “Cladding radiates light (not guides), leading to radiative modes.” This provided additional evidence to support the presence of this misconception, even in an audience that was likely to have prior knowledge in photonics and were taking a rigorous online course. All module materials were subsequently revised to emphasize information about evanescent fields and power loss.