Self-driving laboratory for accelerated discovery of thin-film materials

Autonomous workflow step 1: Robotic preparation of spin-coating inks

• Each precursor ink combined spiro-OMeTAD, FK 102 Co(III) TFSI, and 4-tert-butylpyridine stock solutions, with ChemOS setting the FK 102 Co(III) TFSI-to-spiro-OMeTAD ratio between 0 and 1.Preparation took place in ambient conditions.Source: Autonomous workflow step 1: Robotic… · p. 13
• The tert-butylpyridine-to-total-solute mass ratio was fixed at 0.2.The total comprised spiro-OMeTAD and FK 102 Co(III) TFSI.Source: Autonomous workflow step 1: Robotic… · p. 13
• Combining the spiro-OMeTAD and FK 102 Co(III) TFSI solutions turned the precursor dark purple; inks were mixed by aspiration and used within a minute.Source: Autonomous workflow step 1: Robotic… · p. 13

Autonomous workflow step 2: Robotic spin coating of thin-film samples

• Slides were spun at 1000 rpm while 0.100 mL of precursor solution was dispensed at the center at normal incidence, and rotation continued for 60 s.The spin-coater was custom-built and provided by North Robotics.Source: Autonomous workflow step 2: Robotic… · p. 13

Autonomous workflow step 3: Robotic thermal processing of thin-film samples

• The N9 slide gripper moved freshly spin-coated samples into the furnace, where the heat gun was triggered for the heating time requested by the orchestration software.Source: Autonomous workflow step 3: Robotic… · p. 13
• The sample temperature ramped from ambient to 165 °C over the first 100 s, then remained at 165 °C for the rest of the annealing time.Source: Autonomous workflow step 3: Robotic… · p. 13
• After heating, the arm removed each sample and held it 25 mm above a 4500 rpm cooling fan for 3 min before further characterization.This cooling period returned samples to ambient temperature regardless of annealing time.Source: Autonomous workflow step 3: Robotic… · p. 14

Autonomous workflow step 4: Robotic dark field photography

• Dark-field images distinguished smooth from rough film regions and enabled identification of dust, defects, and dewetting.Samples were imaged at three overlapping locations at 4000 × 3000 px resolution, and images were analyzed manually after experiments.Source: Autonomous workflow step 4: Robotic… · p. 14

Autonomous workflow step 5: Robotic UV-Vis-NIR spectroscopy

• The custom fiber-optic station measured UV-Vis-NIR transmission and reflection using spectrometers spanning 190–900 nm and 900–1700 nm.A probe above the sample collected reflection, while a collimating lens below the sample supplied transmission illumination.Source: Autonomous workflow step 4: Robotic… · p. 14

Robotic spectroscopy measurement

• Robotically collected blank-substrate and annealed-film reflection and transmission spectra at seven positions to estimate thin-film absorbance.The seven positions were spaced about 1 mm apart near the substrate center.Source: Autonomous workflow step 4: Robotic… · p. 16

UV-Vis-NIR data processing

• Corrected absorption spectra assumed minimal scattering and interference fringes, and ignored film–glass interface reflection because both refractive indices were expected to be about 1.5.The correction used measured film/substrate reflection and transmission to solve for thin-film transmission and absorption at each measured position.Source: Autonomous workflow step 4: Robotic… · pp. 16–17

Autonomous workflow step 7: pseudomobility calculation

• They inferred HTM hole mobility by combining four-point sheet resistance with UV–Vis-derived hole density, assuming hole conductivity approximates total conductivity under heavy p-doping and dopant:HTM ≤1:1.The ≤1:1 condition limits HTM2+ in the film.Source: Autonomous workflow step 6: 4-point… · pp. 19–20
• Thickness-independent pseudomobility, conductivity divided by p-type carrier density, served as a relative mobility measure; each film’s optimizer input averaged values from its seven measured positions.The thickness independence supported optimization across the compositional and processing variable space used in the work.Source: Autonomous workflow step 6: 4-point… · pp. 20–21

• Realized experimental parameters and pseudomobility were sent via Dropbox to remote ChemOS, which passed the data to Phoenics (15) for global Bayesian optimization.Source: Autonomous workflow step 8:… · p. 21
• Approximately 30 minutes were saved in a 30-experiment campaign by updating Phoenics’ surrogate model while the robot ran the next experiment.Source: Autonomous workflow step 8:… · p. 21
• Phoenics alternated exploration and exploitation using an adjustable sampling parameter, and selected the initial sample randomly because no response-surface model was available.Source: Autonomous workflow step 8:… · p. 21

Hole mobility measurement using hole-only devices device fabrication

• Hole mobility in spiro-OMeTAD films was extracted by fitting the quadratic current-density region to the Mott-Gurney law using SCLC method29,30 and assuming relative permittivity 3.The extraction also used the HTM-layer thickness.Source: Hole mobility measurement using… · pp. 29–30
• ITO-coated glass substrates were etched, sonicated in detergent, water, acetone, and isopropanol, nitrogen-dried, then UV-ozone treated for 30 minutes before use.The substrates were cut into 1.8 cm × 3.5 cm pieces, and 1.5 cm of ITO was removed using Zn powder and 2 M HCl.Source: Hole mobility measurement using… · p. 29
• PEDOT:PSS ink was filtered and mixed with isopropanol, then spin-coated onto ITO at 3000 RPM for 1 minute before baking.The ink used a 1:2 volume ratio of dispersion to isopropanol; excess PEDOT:PSS was wiped from both device sides before baking.Source: Hole mobility measurement using… · p. 29
• The Ada platform robotically pipetted inks, and varying-doping spiro-OMeTAD films were spin-coated onto glass/ITO/PEDOT:PSS devices at 3000 RPM for 1 minute.Each film used 50 µL of ink; acetone-soaked wipes removed film from both sides.Source: Hole mobility measurement using… · pp. 29–30
• Devices were completed with 80 nm gold contacts deposited through a shadow mask, and the final device area was 3.5 mm2.HTM thickness was measured by stylus profilometry; current-voltage curves were collected in air from 0 to 1 volts using a source-measure unit.Source: Hole mobility measurement using… · p. 30
• Mobilities were averaged across 3–7 working devices per doping level, with standard-deviation error bars; the 84% dopant:HTM ratio was excluded for unreliable thickness data.The reported means are shown in figure S6C.Source: Hole mobility measurement using… · p. 30

Robotic pseudomobility measurements for comparison to hole mobility measurements

• Pseudomobility values were averaged over quadruplicate measurements on Ada-prepared films spanning FK 102 Co(III) TFSI:spiro-OMeTAD ratios of 0, 14, 42, 56, 84, and 98%.Error bars represent the standard deviation of the quadruplicate results.Source: Robotic pseudomobility measurements… · pp. 30–31
• Mean film absorption from 495–505 nm was used to calculate pseudomobility, while conductance was calculated from fitted current–voltage slopes.The absorption spectra and current–voltage measurements covered films with varying dopant:HTM ratios in both optimization runs.Source: Robotic pseudomobility measurements… · pp. 34–35, Figure S3 · p. 34, Figure S4 · p. 35
• Ada pseudomobility data were compared with manually measured hole mobility from hole-only devices, with both datasets’ maxima scaled to match for visual comparison.The supplied passages do not state the strength or direction of the relationship.Source: Figure S5 · p. 36, Robotic pseudomobility measurements… · p. 36
• Ada repeatability samples used a nominal doping ratio of 0.247 and a 59-second annealing time, with 10 replicates prepared and measured in one 3.75-hour run.The supplied table caption gives the test conditions but not the repeatability values.Source: Table S3 · p. 43