Review Papers Claim by Claim

Concise critical summary

The paper reports an on chip inverse designed active THz wavelength division multiplexer integrated with a quantum cascade laser comb, demonstrating three broadband output ports across roughly 2.2 to 3.2 THz with measured interport crosstalk near -6 dB, on chip amplification up to ~3.5x (5.4 dB) for one port, and preserved comb coherence (SWIFTS) across overlapping channels; fabrication used a planarized double metal GaAs/AlGaAs QCL platform and inverse design via SPINS with 105 iterations on a 200 micrometer design region Primary Nature Communications Paper

Paper review and critique On chip inverse designed active wavelength division multiplexer at THz frequencies

This review summarizes methods and results, evaluates strengths and limitations, reproducibility, and suggests concrete next steps and experiments to falsify or extend the work. All factual statements below are cited to the paper using inline journal style citations.

1 Major claims and what the data actually show

• Claim The authors demonstrate an on chip active WDM at THz frequencies routing 2.2 3.2 THz into three broadband ports with compact 200 micrometer footprint and low crosstalk approx -6 dB. Device footprint channels and crosstalk
• Claim The design used gradient based adjoint inverse design (SPINS) with a discrete GaAs BCB permittivity parameterization and ran 105 iterations to converge. Inverse design method details
• Claim The WDM can act as an active amplifier when electrically biased independently of the laser, showing threshold at ~6 V and saturation at ~9 V and a maximum amplification factor approx 3.5x (5.4 dB) on port 2. Measured amplification behaviour
• Claim Comb coherence is preserved after demultiplexing: SWIFTS shows coherent operation over approx 250 GHz covering channels 2 and 3 and beatnote maps are largely robust to WDM bias. SWIFTS coherence evidence

2 Strengths

1. Clear demonstration of transferring inverse design paradigms from telecom to THz on a planarized double metal platform, including active device integration and on chip amplification, which is conceptually novel and practically relevant for integrated THz systems Novelty statement from paper
2. Multi modality characterization: FTIR spectra, spatially resolved port measurements, electrical power curves, amplification curves, beatnote mapping and SWIFTS were used to support claims, showing coherent and amplifying behavior under controlled cryogenic conditions Experimental methods summary
3. Practical fabrication demonstration on a GaAs AlGaAs QCL heterostructure with wafer bonding BCB planarization and patch array antennas integrated to achieve surface emission, a non trivial engineering feat in THz photonics Fabrication process details

3 Limitations and blindspots

• Single device demonstration The results are reported for fabricated devices but characterization appears limited to one or a few chips and the authors state primary data are available on request; this reduces statistical confidence and hides device to device variability Data availability statement
• Antenna modeling omission The CST time domain simulations used to predict port routing explicitly exclude the frequency dependent antenna response and reflectivity; therefore simulation predictions compared to measured antenna outcoupled spectra require caution CST simulation limitation
• Laser emission not matching design at low freq The active region spectrum was slightly blueshifted and did not extend to frequencies below about 2.6 THz so port 4 (design center 2.37 THz) is attenuated in the experiment, meaning measured system performance is partly source limited rather than WDM limited Measured spectral blueshift and port 4 attenuation
• Propagation loss magnitude The authors note MIM waveguide losses can be large (approx 40 dB cm) which the active WDM partly compensates for, but high passive propagation loss is an important constraint for scaling and for passive THz interconnects Propagation losses and role of gain

4 Reproducibility assessment

The methods are described with reasonable engineering detail: inverse design algorithm and objective function, materials and refractive indices, fabrication flow summary, measurement methods (FTIR, SWIFTS, beatnote, spatial filtering). However: raw datasets and simulation input files are not posted publicly; measured demonstration appears limited to one device; CST simulation omitted antenna spectral response. For these reasons I score reproducibility at 7 of 10. Key missing reproducibility items are the SPINS run inputs and constraints, CST project files including boundary conditions, and full measurement raw data. Reproducibility data availability note

5 Concrete suggestions to strengthen claims and next experiments

1. Publish SPINS and CST project files plus parameterized layout and permittivity maps in a repository so others can reproduce inverse design and simulate the exact antenna inclusive device response.
2. Fabricate and measure multiple devices across wafers to quantify device to device variability (yield), and measure port routing and crosstalk statistics vs temperature and bias. This addresses single device limitation.
3. Integrate antenna frequency dependent models into the CST and validate simulated vs measured antenna spectral far field to decouple WDM routing performance from antenna spectral filtering.
4. Demonstrate routing with a QCL whose emission better covers the designed low frequency channel (or retune WDM to match the actual QCL spectrum) to show full three port performance not limited by source blueshift.
5. Provide SPINS objective function variants that explicitly penalize crosstalk or optimize narrower channels and show tradeoffs between footprint efficiency and crosstalk; include designs and measured results.

6 How to falsify the main claims

Attempt independent fabrication of the same topology and show that either

• the device cannot route 200 GHz wide bands to separate ports with crosstalk near -6 dB, or
• the comb coherence after demultiplexing measured by SWIFTS and beatnote mapping is not preserved under the same experimental conditions, or
• amplification cannot be reproduced when the WDM is electrically biased independently of the laser.

Failure of any of these would falsify key claims.

7 Quantitative reproduction figure and suggested visualization

8 Short critical appraisal and final verdict

Verdict The paper presents a significant technical advance: a compact inverse designed active WDM integrated with a THz QCL comb, combining inverse design, on chip amplification, and preserved comb coherence. The work is novel and useful for integrated THz photonics but currently limited by partial demonstrations (port 4 attenuation due to source spectrum), a lack of public raw data and simulation projects, and single device statistics. Reproducibility would be greatly improved by publishing SPINS and CST inputs and measured raw datasets. Confidence level in the core experimental observations as reported: high, because multi modal experimental evidence is provided, but overall generalizability pending multi device statistics and open data. Paper concluding remarks

9 Tools replication checklist for a reproducer

• Request SPINS design files and discrete permittivity maps from authors or reproduce using provided objective function f obj equals sum over j 1 minus overlap squared and run adjoint optimization with same bounds (GaAs n approx 3.6 BCB n approx 1.57) on a 200 micrometer design region; keep 40 micrometer waveguides.
• Perform 2D in plane electromagnetic forward and adjoint simulations matching boundary conditions and mode injection (TM 00), then validate with CST full 3D including antennas and BCB planarization layer and metal stacks.
• Fabricate on MBE grown GaAs AlGaAs heterostructure with thermocompression wafer bonding BCB planarization and patch array antenna metallization; recreate V shaped top metallization gap to electrically isolate sections.
• Measure at 20 K with FTIR and SWIFTS beatnote setup and spatial filtering as described; collect full raw spectra and beatnote maps for multiple devices.

10 Buttons and useful next queries

Author review quick links

If you want I can request the SPINS and CST parameter files from the corresponding authors summarize the supplementary materials, or produce a synthetic SPINS objective file and starter CST template for reproduction attempts.