Opto-Electronic Advances, Volume. 4, Issue 10, 210039-1(2021)

Hybrid artificial neural networks and analytical model for prediction of optical constants and bandgap energy of 3D nanonetwork silicon structures

Shreeniket Joshi and Amirkianoosh Kiani*
Author Affiliations
  • Silicon Hall: Micro/Nano Manufacturing Facility, Faculty of Engineering and Applied Science, Ontario Tech University, 2000 Simcoe St N, Oshawa, Ontario L1G 0C5, Canada
  • show less
    Figures & Tables(22)
    Schematic of fabrication set-up. Figure reproduced with permission from ref.6, Elsevier.
    Filtered reflectance data.
    Filtered transmittance data.
    Model validation with experimental transmittance.
    Refractive index as a function of wavelength.
    Extinction coefficient as a function of wavelength.
    PUMA model validation with analytical refractive index (n); R_Puma: Simulation results when only reflectance was input; B_Puma: Simulation results when both reflectance and transmittance were inputs.
    PUMA model validation with analytical extinction coefficient (k).
    PUMA model validation with analytical extinction coefficient (k).
    PUMA model evaluation with experimental transmittance.
    PUMA model evaluation with experimental reflectance.
    Flowchart for deep learning algorithm.
    Deep Learning Model developed.
    Comparison of model-predicted extinction coefficient with analytical values.
    Comparison of model-predicted refractive index with analytical values.
    Absorption regions for fabricated silicon thin film.
    Tauc’s Plot for determining optical bandgap.
    • Table 1. Prediction of extinction coefficient (k).

      View table
      View in Article

      Table 1. Prediction of extinction coefficient (k).

      K_Mean absolute error0.000418969
      K_Mean squared error2.6281E-07
      Sum_K.sq error9.93422E-05
      R20.987741346
    • Table 2. Prediction of refractive index (n)

      View table
      View in Article

      Table 2. Prediction of refractive index (n)

      N_Mean absolute error0.023172839
      N_Mean squared error0.000574442
      Sum_N.sq error0.217139095
      R20.867649736
    • Table 3. Manufacturing parameters.

      View table
      View in Article

      Table 3. Manufacturing parameters.

      Frequency (kHz)Pulse duration (ps)Temperature (°C)
      Sample 1600150RT
      Sample 2900150200
      Sample 31200150RT
      Sample 41200150600
      Sample 512005000RT
    • Table 4. Validation of proposed methodology.

      View table
      View in Article

      Table 4. Validation of proposed methodology.

      SampleMSE of TransmittanceMAE of kMAE of nMSE of kMSE of nR2 value of kR2 value of n
      11.91E–095.47E–048.33E–024.77E–071.39E–020.980.97
      29.65E–117.74E–049.70E–021.00E–061.50E–020.970.97
      31.87E–096.67E–041.04E–016.55E–072.20E–020.970.96
      46.44E–101.04E–032.22E–011.79E–066.86E–020.900.88
      53.97E–107.48E–041.05E–019.58E–071.79E–020.970.97
    • Table 5. Band gap information for various silicon structures7

      View table
      View in Article

      Table 5. Band gap information for various silicon structures7

      SemiconductorsCrystal structureEg (T=300K) Type of band gap
      SiDiamond1.12Indirect
      a-Si:HAmorphous1.7 to 1.8Indirect
      SiC(α) Wurtzite2.9Indirect
    Tools

    Get Citation

    Copy Citation Text

    Shreeniket Joshi, Amirkianoosh Kiani. Hybrid artificial neural networks and analytical model for prediction of optical constants and bandgap energy of 3D nanonetwork silicon structures[J]. Opto-Electronic Advances, 2021, 4(10): 210039-1

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Original Article

    Received: Mar. 19, 2021

    Accepted: May. 4, 2021

    Published Online: Dec. 28, 2021

    The Author Email: Kiani Amirkianoosh (amirkianoosh.kiani@ontariotechu.ca)

    DOI:10.29026/oea.2021.210039

    Topics