Atomic layer deposition (ALD) is a thin film coating technique in which the experimentalist relies on well-defined surface reactions of molecular gaseous precursors. The self-limiting nature of the reactions controls the amount of material deposited at each step in a digital manner. Repeating a set of complementary reactions in a cyclic manner therefore allows one to set the thickness of a coating with outstanding accuracy down to 1 nm and below even on non-planar surfaces. This unique capability renders ALD particularly attractive for applications in which a device functionality relies on interface engineering, even more so if the interfaces feature complex geometries. Among them, depositing the gate dielectrics of state-of-the-art transistors represents a high-volume commercial application for which ALD is indispensable. Emerging applications exist in the realm of energy conversion, including batteries and photovoltaics (solar cells). Here, however, the slow nature of ALD limits its use for prototyping new solar cell concepts and optimizing the thicknesses of individual semiconductor layers systematically.
To overcome this limitation, the ALAMS project has demonstrated the applicability of atomic-layer additive manufacturing (ALAM) in the photovoltaics field. ALAM represents an extension of the ALD concept in which the self-limiting surface chemistry of ALD is performed in a spatially constrained manner. A microfluidic nozzle delivers the precursor gas flows to the surface on the micron scale so that motions of the nozzle with respect to the substrate result in the deposition of a monolayer of material with sub-nanometer thickness, just as in ALD but along a line defined by the motions. Multiple passes of the nozzle add material as in classical 3D printing, maintaining the sub-nanometer thickness control. At the project start, ALAM had only been demonstrated with two materials, a wide-bandgap metal oxide and a metal, and semiconductor applications of the technology had never been tested.
The ALAMS project developed the technology to make it adequate for use in photovoltaics applications, specifically for the prototyping of ‘extremely thin absorber’ (ETA) solar cells in an additive manufacturing approach. To achieve this, an ALAMS prototype was rebuilt with several significant novel enabling features. Firstly, a motorized stage enabling 10-cm motions and the gas delivery nozzle were enclosed in a box under inert atmosphere, which prevents deleterious aerobic oxidation of the precursors and semiconductors. Secondly, the gas delivery system was designed for handling up to six different precursors without vacuum interruption. Subsequently, deposition processes were developed on the prototype for the ALAM processing of the necessary semiconductors. For titanium dioxide (electron transporter), zinc oxide (electron transporter), zinc sulfide (interfacial layer), antimony(III) sulfide (light absorber), vanadium(V) oxide (hole transporter) and platinum (electrical contact metal), a number of process parameters had to be explored and optimized, in particular the temperatures of the precursors, vapor delivery lines and control valves, gas delivery nozzle, and substrate, as well as the flow rates of the carrier gas and inert gas. For each material, we found the set of parameter settings with which the thickness deposited increases linearly with number of passes and the growth rate is independent of small variations in the parameters. We demonstrated that for known ALD processes, the chemical reaction can be transferred from ALD to ALAM while maintaining characteristics such as the growth rate (in units of thickness per pass or per cycle) and the major materials analytical fingerprints. A significant lesson learnt concerns the consumption of precursors, which is extremely low in ALAM when compared with ALD. The precursor containers must be maintained at temperatures 40 to 50 °C lower in ALAM than in ALD, which translates into consumptions reduced by several orders of magnitude. Simultaneously, the growth rate in units of thickness per time can be much higher in ALAM than in ALD for simple line patterns.
To conclude the demonstration, we deposited functional miniaturized solar cell stacks by ALAM consisting of the semiconductors mentioned above. Electrical characterization under simulated solar light demonstrated the appearance of a photovoltage (in open-circuit conditions) and a photocurrent (in short-circuit conditions). The performance is below the state of the art of classically manufactured cells consisting of the same materials, likely due to the imperfect lateral definition of the lines and Ohmic losses.
