Source-linked AI summary
Self-driving laboratory for accelerated discovery of thin-film materials
Benjamin P. MacLeod, Fraser G. L. Parlane, Thomas D. Morrissey, Florian Häse, Loïc M. Roch, Kevan E. Dettelbach, Raphaell Moreira, Lars P. E. Yunker, Michael B. Rooney, Joseph R. Deeth, Veronica Lai, Gordon J. Ng, Henry Situ, Ray H. Zhang, Michael S. Elliott, Ted H. Haley, David J. Dvorak, Alán Aspuru-Guzik, Jason E. Hein, Curtis P. Berlinguette
TL;DR
The paper addresses measurement and comparison of hole-mobility-related properties in thin films. It extracts μ from four-point-probe sheet-resistance measurements and compares manually measured hole mobility with robotic pseudomobility data, under assumptions about optical scattering and interference.
Problem
The work concerns characterizing electronic-material performance, for which properties are often maximized to optimize performance.
Method
The approach extracts each HTM film’s μ from sheet resistance measured with a four-point probe.
Results
The study compares manually measured hole mobility from hole-only devices with pseudomobility data.
Takeaways & Limitations
The reported comparison relates robotic pseudomobility measurements to manually measured hole mobility data.
Takeaways & Limitations
The spectral measurements assume entirely specular and incoherent reflection and transmission, which is reasonable when scattering and interference fringes are minimal.
Abstract
from arXiv · showhide
Discovering and optimizing commercially viable materials for clean energy applications typically takes over a decade. Self-driving laboratories that iteratively design, execute, and learn from material science experiments in a fully autonomous loop present an opportunity to accelerate this research. We report here a modular robotic platform driven by a model-based optimization algorithm capable of autonomously optimizing the optical and electronic properties of thin-film materials by modifying the film composition and processing conditions. We demonstrate this platform by using it to maximize the hole mobility of organic hole transport materials commonly used in perovskite solar cells and consumer electronics. This demonstration highlights the possibilities of using autonomous laboratories to discover organic and inorganic materials relevant to materials sciences and clean energy technologies.
Autonomous workflow step 1: Robotic preparation of spin-coating inks
The autonomous workflow prepared spin-coating inks by varying spiro-OMeTAD and FK 102 Co(III) TFSI stock amounts, while ChemOS controlled their ratio and tert-butylpyridine was fixed. The resulting solutions were aspirated and used within a minute.
- Autonomous workflow step 1: Robotic preparation of spin-coating inks: Precursor solutions mixed spiro-OMeTAD, FK 102 Co(III) TFSI, and 4-tert-butylpyridine stock solutions under ambient conditions.The combined spiro-OMeTAD and FK 102 Co(III) TFSI solutions became dark purple.
- Autonomous workflow step 1: Robotic preparation of spin-coating inks: ChemOS determined the FK 102 Co(III) TFSI:spiro-OMeTAD ratio, which varied from 0 to 1 (n/n).The ratio was adjusted for each precursor solution.
- Autonomous workflow step 1: Robotic preparation of spin-coating inks: The precursor solutions were mixed through aspiration and used within a minute of preparation.Rapid use followed robotic mixing.
Autonomous workflow step 2: Robotic spin coating of thin-film samples
Thin-film samples were prepared autonomously by spin-coating precursor solution onto microscope slides using a custom-built spin-coater. The slides were spun at 1000 rpm for 60 s after dispensing 0.100 mL of solution at the center.
- A custom-built spin-coater provided by North Robotics prepared the thin-film samples.
- The microscope slides were spun at 1000 rpm while 0.100 mL of precursor solution was dispensed at normal incidence at the slide center.
- Rotation continued for 60 s after precursor-solution dispensing.
Autonomous workflow step 3: Robotic thermal processing of thin-film samples
Freshly spin-coated thin-film samples were robotically transferred into a forced-convection furnace for software-controlled annealing. The samples were then immediately fan-cooled to ambient temperature before characterization.
- The N9 slide gripper transferred freshly spin-coated thin-film samples into the furnace, where orchestration software triggered heating for the requested duration.
- The annealing temperature ramped from ambient temperature to 165 °C over the first 100 s and then remained at 165 °C.
- After annealing, the robotic arm removed each sample and held it 25 mm above a 4500 rpm cooling fan for 3 min.
- The cooling period returned samples to ambient temperature regardless of annealing time before further characterization.
Autonomous workflow step 4: Robotic dark field photography
The workflow used dark-field photography to inspect thin-film quality and robotic UV-Vis-NIR spectroscopy to measure and process transmission, reflection, and absorption spectra. Spectroscopy combined calibrated optical hardware, robotic positioning, substrate baselines, and multilayer corrections to estimate film absorption.
- Autonomous workflow step 4: Robotic dark field photography: Dark-field photography captured three overlapping 4000 × 3000 px images to identify dust, defects, and dewetting in thin-film samples.A ring light illuminated samples positioned 90 mm below the camera to contrast smooth and rough film regions.
- Autonomous workflow step 5: Robotic UV-Vis-NIR spectroscopy: The custom fiber-optic station collected UV-Vis-NIR transmission and reflection spectra across 190–1700 nm using separate visible and NIR spectrometers.The BLACK-Comet covered 190 - 900 nm with < 1 nm resolving resolution, while the DWARF-Star covered 900 - 1700 nm with 2.5 nm resolving resolution.
- Autonomous workflow step 5: Robotic UV-Vis-NIR spectroscopy: A fiber-optic reflection probe and collimating lens connected separate lamps for reflection and transmission measurements, with shutters selecting the measurement mode.The reflection probe was positioned above and normal to the sample, while the collimating lens was positioned below and normal to the sample.
- Robotic spectroscopy measurement: For each film, the robot measured blank and coated substrates at seven positions spaced ~1 mm apart to compute an approximation to thin-film absorbance.Reflection and transmission spectra of the blank glass substrate preceded analogous measurements of the annealed thin film on an identical substrate.
- UV-Vis-NIR data processing: Spectra were measured at normal incidence and assumed entirely specular and incoherent when surface scattering and interference fringes were minimal.Film and substrate refractive indices were expected to be ~1.5, allowing film-glass interface reflection to be ignored without significant distortion.
- UV-Vis-NIR data processing: The processing solved substrate and film multilayer equations to obtain corrected film transmission and absorption spectra at each measured position.The method used blank-substrate reflection and transmission to derive substrate optical parameters, then introduced film reflectivity, transmissivity, and single-pass transmission.
Autonomous workflow step 6: 4-point probe conductance instrumentation and characterization
The workflow measured thin-film conductance with a four-point probe and used UV-Vis data to calculate a thickness-independent pseudomobility for optimization. Measurements were collected at seven substrate positions, and their mean pseudomobility was passed to the optimization algorithm.
- Autonomous workflow step 6: 4-point probe conductance instrumentation and characterization: Film conductance was measured at seven positions spaced ~1 mm apart near the substrate center, matching spectroscopy measurement positions.These measurements linked the electrical characterization locations to the UV-Vis characterization locations.
- Autonomous workflow step 7: pseudomobility calculation: Hole mobility was extracted from sheet resistance 4-point probe and UV-Vis measurements, assuming hole conductivity approximated total conductivity under heavy p-doping.The approximation also relied on the high intrinsic mobility of holes relative to electrons in spiro-OMeTAD.
- Autonomous workflow step 7: pseudomobility calculation: Pseudomobility provided a relative mobility measure equivalent to film conductivity divided by p-type carrier density and independent of film thickness.Each film’s pseudomobility was calculated at seven positions, and the mean of those values was passed to the optimization algorithm.
Autonomous workflow step 8: determining the next experiment with Bayesian optimization
After each experiment, ChemOS transfers the realized parameters and pseudomobility to Phoenics, which updates a surrogate model and proposes the next experiment. Precomputed recommendations run in parallel with robotic execution, shortening a 30-experiment campaign by approximately 30 minutes.
- Autonomous workflow step 8: determining the next experiment with Bayesian optimization: ChemOS sends realized experimental parameters and pseudomobility values to Phoenics through a remote server and Dropbox interface.Phoenics then receives the experimental data for Bayesian optimization.
- Autonomous workflow step 8: determining the next experiment with Bayesian optimization: Phoenics updates its surrogate model of the experimental response surface after each experiment.This update is initiated by the transferred experimental data.
- Autonomous workflow step 8: determining the next experiment with Bayesian optimization: Phoenics immediately provides precomputed parameters for the next experiment, allowing model updates to run in parallel with robotic execution.This parallelization minimizes robot downtime between experiments.
- Autonomous workflow step 8: determining the next experiment with Bayesian optimization: Phoenics alternates exploration and exploitation using an adjustable sampling parameter to enable global optimization of the explored response surfaces.The initial sample is selected randomly because no response-surface model is initially available.
Hole mobility measurement using hole-only devices device fabrication
Hole-only devices were fabricated to measure spiro-OMeTAD hole mobility using steady-state SCLC. Mobilities were extracted from current–voltage measurements by fitting the quadratic current-density region to the Mott–Gurney law, with device-level means reported across doping levels.
- Device fabrication: Hole-only devices measured hole mobility in spiro-OMeTAD films using the steady-state space-charge limited current (SCLC) method.The devices were fabricated with ITO-coated glass substrates, PEDOT:PSS, variably doped spiro-OMeTAD films, and 80 nm-thick gold contacts.
- Measurement protocol: Film thicknesses were measured by stylus profilometry, and current-voltage curves were collected from 0 to 1 volts in air using a Keithley 2400 source-measure unit.The final device area was 3.5 mm^2.
- Mobility extraction: Hole mobilities were extracted by fitting the quadratic region of the current density curve to the Mott-Gurney law.The relative permittivity of the HTM was assumed to have a value of 3.
- Data aggregation: Except for the 84% dopant:HTM molar-ratio device, reported mobilities were means from 3 - 7 working devices per doping level, with standard deviations as error bars.The 84% device was excluded because reliable thickness data were not obtained.
Robotic pseudomobility measurements for comparison to hole mobility measurements
Ada measured pseudomobility across FK 102 Co(III) TFSI:spiro-OMeTAD compositions and compared these values with manually measured hole mobility from hole-only devices. Pseudomobility measurements averaged quadruplicate results, with standard-deviation error bars.
- Measurement statistics: Pseudomobility values were averages of quadruplicate measurements, and error bars represented the standard deviation of those results.The measurements were made on HTM films.
- Composition series: Ada measured pseudomobility for FK 102 Co(III) TFSI:spiro-OMeTAD molar ratios of 0, 14, 42, 56, 84, and 98%.Films were prepared and characterized using the Ada platform.
- Hole-mobility comparison: The study compared manually measured hole mobility from hole-only devices with pseudomobility measured using the Ada platform.The comparison was presented with scaled y-axes whose maxima were equal to facilitate visual comparison.