Logiciel d’intelligence artificielle pour la fabrication de semi-conducteurs photoniques

*** NEW – May 19, 2020
An attachment has been added. The document contains questions and answers related to the Challenge
*******************************************************
May 8, 2020
New attachment has been added. Please read the document as it provides important information pertaining to the submission of your proposal.
********************************
May 5, 2020

The closing date for this challenge has been extended to May 22, 2020 at 14:00 EDT.
****************************************************
April 14, 2020
Due to the issues surrounding COVID19, we have decided to extend the solicitation closing date until May 7, 2020.
This Challenge Notice is issued under the Innovative Solutions Canada Program (ISC) Call for Proposals 003 (EN578-20ISC3). For general ISC information, Bidders can visit the ISC website.
Please refer to the Solicitation Documents which contain the process for submitting a proposal.
Steps to apply:
Step 1: read this challenge
Step 2: read the Call for Proposals
Step 3: propose your solution here
Challenge title: AI Software for Photonics Semiconductor Fabrication
CHALLENGE SPONSOR: National Research Council of Canada (NRC)
Funding Mechanism: Contract
MAXIMUM CONTRACT VALUE:
Multiple contracts could result from this Challenge.
The maximum funding available for any Phase 1 contract resulting from this Challenge is $150,000.00 CAD excluding applicable taxes, shipping, travel and living expenses, as required, for up to 6 months (excluding submission of the final report).
Estimated number of Phase 1 contracts: 2
The maximum funding available for any Phase 2 contract resulting from this Challenge is $300,000.00 CAD excluding applicable taxes, shipping, travel and living expenses, as required, for up to 12 months (excluding submission of the final report). Only eligible businesses that have successfully completed Phase 1 will be considered for Phase 2.
Estimated number of Phase 2 contracts: 1
This disclosure is made in good faith and does not commit Canada to contract for the total approximate funding. Final decisions on the number of Phase 1 and Phase 2 awards will be made by Canada on the basis of factors such as evaluation results, departmental priorities and availability of funds.
TRAVEL: For Phase 1 it is anticipated that two meetings will require the successful bidder(s) to travel to the location identified below:
Kick-off meeting
Ottawa, ON
Final Review Meeting
Ottawa, ON
All other communication can take place by telephone, videoconference, and WebEx.
Problem Summary Statement
NRC is seeking a software solution that will, through the use of models and data analysis, predict and control the wavelength of a grown semiconductor structure during its fabrication.
Problem Statement
Through the Canadian Photonics Fabrication Centre (CPFC), the NRC offers semiconductor foundry services to our clients for the fabrication of photonic devices. An essential part of those services is the growth by metallo-organic chemical vapour deposition (MOCVD) of heterostructures based on Gallium, Indium, Arsenic, and Phosphorus GaInAsP or Aluminum, Gallium, Indium, and Arsenic (AlGaInAs). The optical properties of these heterostructures are critical for the device performance, and is usually described by the photoluminescence (PL) wavelength, specified as a target (i.e., 1400 nm) and a variation around this target (± 5 nm). Current models used to target this PL wavelength are accurate within ±15 nm. As specifications are often tighter than this, experimental fine-tuning is needed to meet our clients’ needs. This fine-tuning is done using calibration runs, which are costly and time-consuming.
NRC would like to see a modeling software developed that will, using empirical parameters as inputs, accurately predict the PL wavelength of a grown heterostructure, without the need for calibration runs.
Desired Outcomes and Considerations
Essential (Mandatory) Outcomes
NRC is looking for a software that can predict the PhotoLuminescence (PL) wavelength of an epitaxially grown semiconductor heterostructure containing a combination of AlGaInAsP layers within ±5 nm of the actual result, for each wafer.
Proposed solutions must:
Phase 1:
At the end of phase 1, a proof-of-concept software is required. This software must:

Be able to extract data from text files for the various process inputs and outputs, with an error rate less than 1%. The inputs and outputs to read are:PL maps, X-ray Diffraction (XRD) profile, Metalorganic Chemical Vapour Deposition (MOCVD) recipe, temperature profile, reflectivity profile, run log. Approximately 500 datapoints are available for each and will be provided by NRC.
Be able to produce a model based on this data and machine learning principles that can, when given a structure and recipe, predict a PL wavelength.
Demonstrate that PL variations are non-stochastic below ± 5 nm.
Be usable on any windows-based PC, running windows XP or newer versions.

Phase 2:
At the end of phase 2, in addition to phase 1 requirements, the software must:

Be able to predict the PL wavelength within ± 5 nm for any combination of compositions (Al,Ga,In,As,P) and layer structure, within the boundaries of the available data, when using a MOCVD recipe and a structure as input.
Be adaptable by the user; the software must be able to read new data and update its model under the user’s control.
Be able to provide recipe parameters when given a structure, target PL, and recent run results, and have the resulting PL be within ±5 nm of this target. The user must be able to fix parameters. Parameters are, for each layer in a structure: Temperature, zone heating, time, multiple gas flows.
Be applicable to any MOCVD tool of comparable capability.

Note: Bidders are reminded that under Question 1 a (Scope) proposals must describe how solutions clearly meet all eight of the Essential (Mandatory) Outcomes listed in this section. Bidders should focus their Phase 1 project plan on demonstrating the feasibility of Essential Criteria 1-4. The remainder of the Essential Outcomes can be focused on during Phase 2 work.
Background and Context
The CPFC foundry service at the NRC is a low volume, high diversity semiconductor manufacturing environment that provides early-stage development and pilot-scale production of photonic devices for a variety of clients including a number of Canadian SMEs. The semiconductor devices produced, such as amplifiers and lasers, are tailored to meet a specific client’s need, including a very specific PL wavelength. Growth of a new structure must largely rely on an iterative process of calibration and measurement. The difficulty lies in finding the precise combination of input parameters to produce the required output PL wavelength.
Currently available state-of-the-art modeling software are powerful theoretical models that are well suited to predicting qualitative trends. However, when used to make quantitative predictions of output parameters, such as the PL wavelength, these models are often inaccurate, differing from empirical results by a substantial amount. This inaccuracy is further compounded by natural variations associated with the growth process. Such models therefore cannot be used in isolation to target a specific output parameter; an iterative process of calibration and measurement is also necessary.
Semi-empirical models are often used to bridge the gap between the previous models and empirical results. This method encounters two issues:

They are only applicable to a specific semiconductor structure. In a high diversity, low volume foundry, each structure will require its own empirical model.
These semi-empirical models are only applicable to a specific process rather than being of general use.

The goal of this challenge is to leverage the disparate data available from a low volume, high diversity foundry to build an empirical modelling system that is generic enough to be applicable to a wide variety of semiconductor structures.
ENQUIRIES
All enquiries must be submitted in writing to T••••••••@••••••••.••.ca no later than ten calendar days before the Challenge Notice closing date. Enquiries received after that time may not be answered.