Spatial Atmospheric Atomic Layer Deposition : A new laboratory and industrial tool for low-cost photovoltaics

Recently, a new approach to atomic layer deposition (ALD) has been developed that doesn't require vacuum and is much faster than conventional ALD. This is achieved by separating the precursors in space rather than in time. This approach is most commonly called Spatial ALD (SALD). In our lab we have been using/developing a novel atmospheric SALD system to fabricate active components for new generation solar cells, showing the potential of this novel technique for the fabrication of high quality materials that can be integrated into devices. In this minireview we will introduce the basics of SALD and illustrate its great potential by highlighting recent results in the field of photovoltaics. Atomic Layer Deposition: unique pros, significant cons Atomic Layer Deposition (ALD) is a powerful deposition technique offering unique assets: deposition of high quality films at low temperatures, very precise control over film thickness and highly conformal coatings, i.e. complete and uniform coating of high aspect ratio features. Page 1 of 17 Materials Horizons M at er ia ls H or iz on s A cc ep te d M an us cr ip t These properties have rendered ALD the technique of choice in the microelectronics industry (e.g. deposition of gap dielectrics for magnetic heads, high-k oxide dielectrics and metal electrodes for Dynamic random-access memories, DRAM). But the burst of nanoscience and nanotechnology in the last decades has also brought an associated boom of ALD popularity, since it allows the nanoengineering of surfaces with precise nanoscale control. As an example, the first report on the homogenous coating of nanoparticles by ALD dates from 2000, In the case of nanoparticles, a fluidized bed reactor is more convenient. This increase in the use of ALD is testified by the publication of 7 general reviews dedicated to ALD in the last decade. 12 Apart from those, reviews have been written covering the application of ALD in specific areas such as nanotechnology, the environment and energy (lithium batteries, fuel cells and photovoltaics), and medical and biologic applications. Despite this increasing success, ALD still faces two main drawbacks which compromise both its generalised use in the laboratory and its implementation in large scale, high throughput production lines. First, ALD is a very slow technique (up to 2 orders of magnitude slower than chemical vapour deposition (CVD)) and, second, it usually involves processing in vacuum, thus making it complicated and expensive to scale up. Both the unique advantages and drawbacks of ALD are the consequence of how it works: ALD is a particular case of CVD in which the reaction is restricted to the sample surface, thus being self-limited. This is achieved by exposing the sample to the reactants at different time, i.e. in a sequence of pulses (figure 1a). In this way, the metal precursors (e.g. Al(CH3)3 in the case of fabrication of Al2O3) are supplied and react with the surface, ideally forming a monolayer. Excess precursor is then purged, usually by evacuation. The second precursor (e.g. H2O) is then injected and reacts with the chemisorbed layer forming a monolayer of the desired material (i.e. Al2O3) plus by-products (CH4) that have to be purged along with the excess precursor. The cycle is then repeated the necessary number of times to obtain a very precise film thickness. Thus, the unique advantages and drawbacks of conventional ALD are a consequence of Page 2 of 17 Materials Horizons M at er ia ls H or iz on s A cc ep te d M an us cr ip t separating the different precursors in time (figure 1a, conventional ALD is thus also called temporal ALD). Figure 1a and b: Temporal vs. Spatial ALD. Copyright 2011, AIP Publishing LLC; Reprinted with permission from ref. 22 The Key to atmospheric, high throughput ALD: from temporal to spatial Although there have been several approaches to realizing ALD without the need of a high vacuum, Spatial ALD (SALD) has been the most popular in the last years due to its conceptual simplicity and the high deposition rates that can be achieved, and thus has also been called Atmospheric ALD (AALD). Conversely, atmospheric temporal ALD approaches based on fluidized reactors show comparable deposition rates with conventional vacuum ALD since those still rely on injection-purge cycles. The key idea allowing high throughput atmospheric spatial ALD is not new and was indeed patented in 1977 and 1983 by Suntola et al. It simply consists in separating the precursors in space rather than in time (figure 1b), as Page 3 of 17 Materials Horizons M at er ia ls H or iz on s A cc ep te d M an us cr ip t opposed to temporal ALD. Thus, in SALD the different precursors are supplied constantly in between inert gas regions. Films are then grown by alternatively exposing the substrate from one precursor region to the other going across the inert gas regions (figure 1b). In this way, the oscillation of the substrate (or of the gases injector) from one precursor zone to the second one, going across the inert gas regions, reproduces the classical temporal ALD scheme: the first metal precursor reacts with the surface forming a monolayer while any unreacted precursor is swept away and purged in the inert region; Then the second precursor reacts with the previous monolayer forming a layer of material; Finally the sample returns to the first precursor region again going across the inert region where any by-products and excess precursor are purged. This approach allows much faster deposition rates, which are mainly limited by the precursor reaction kinetics, so that a full reaction can take place while the sample is within the precursor zones. As recently reviewed by Poodt et al., from 2008 there has been a variety of SALD designs exploiting this idea, some of them already being commercialized. The versatility of Suntola's approach can be appreciated by the multiple designs and configurations presented so far. One of the approaches used in several of the recent SALD designs is the use of a short distance between the precursor delivery zone and the substrate (typically below 100 m, Figure 2a). In this configuration, referred to as "close proximity SALD", the different gas regions are located along alternative channels where the gases flow parallel to each other, as illustrated in Figure 2a. The close distance of the substrate, i.e. the small gap between injector and substrate, prevents the different precursors from mixing since the inert gas channel acts as an effective barrier. This approach has several advantages including: i) The possibility of having smaller designs since the precursor regions are very small; ii) Simpler designs which are compatible with roll-to-roll processing; iii) The possibility of using gas-bearings. Although several patents on close-proximity SALD where filed from 2008 to 2011, Levy et al. from Kodak were the first ones to present and publish their results on the deposition of ZnO thin film transistors, thus Page 4 of 17 Materials Horizons M at er ia ls H or iz on s A cc ep te d M an us cr ip t showing the suitability of SALD for fabricating device components. Later Poodt et al. also demonstrated high quality Al2O3 films obtained by SALD. These films were used to passivate the surface of conventional silicon solar cells. The design used by Levy et al. consists of a deposition head in which the different gas flows are distributed along the parallel channels (see scheme in Figure 2b). This design is very simple and can be made very small, being indeed reminiscent of a printer head. Furthermore it is very easy to scale up, modify and customize. Finally, it is readily compatible with roll-to-roll processing. Such an approach is therefore very readily applicable both in laboratory and industrial settings. A similar approach has already been commercialized for the deposition of Al2O3 passivation layers on silicon and Cu(In,Ga)Se2 thin-film solar cells. Figure 2. a) close-proximity SALD approach. The sample oscillates under the different flows reproducing the cycles in conventional ALD (numbered arrows); b) Scheme of the deposition head designed by Kodak. Copyright 2013 John Wiley & Sons, Ltd. Adapted with permission from ref. Page 5 of 17 Materials Horizons M at er ia ls H or iz on s A cc ep te d M an us cr ip t The potential of SALD: Fabrication of active components for low-cost solar cells. During the last years, in our group we have been using and developing a Kodak close-proximity SALD system, pioneering the deposition, study and implementation of high quality SALD films in low-cost/new generation photovoltaic devices. Our results include: i) The demonstration of the feasibility of using SALD for the ultra-fast deposition of high-quality films that can be integrated in devices; ii) The deposition of materials with superior properties compared to other low temperature and scalable methods; iii) The fundamental studies of new-generation solar cells by using doped oxide films; and iv) The nano-engineering of surfaces with SALD thin films to improve solar cells efficiencies. These results are succinctly reviewed next. In the first example we used SALD to grow high quality TiO2 blocking layers for organic (poly(3hexylthiophene-2,5-diyl):[6,6]-phenyl-C61butyric acid methyl ester (P3HT:PCBM)) solar cells. Dense, uniform thin TiO2 films were grown at temperatures as low as 100 °C in only 37 s (~20 nm/min growth rate). Incorporation of these films in P3HT-PCBM-based solar cells showed performances comparable with cells made using TiO2 films deposited with much longer processing times and/or higher temperatures (SALD at 350 °C and spray pyrolysis at 450 °C). The high quality of the amorphous films obtained at 100 °C allowed the use of only 12 nm thick blocking layers, thus compensating for the lower conductivity of amorphous TiO2 with respect to crystalline TiO2, and resulting in two orders of magnitude faster deposition than other low-temperature scalable methods


Materials Horizons Accepted Manuscript
These properties have rendered ALD the technique of choice in the microelectronics industry (e.g.2][3][4] But the burst of nanoscience and nanotechnology in the last decades has also brought an associated boom of ALD popularity, since it allows the nanoengineering of surfaces with precise nanoscale control.As an example, the first report on the homogenous coating of nanoparticles by ALD dates from 2000, 5 In the case of nanoparticles, a fluidized bed reactor is more convenient. 6This increase in the use of ALD is testified by the publication of 7 general reviews dedicated to ALD in the last decade. 3,7- 12 part from those, reviews have been written covering the application of ALD in specific areas such as nanotechnology, 13,14 the environment 15 and energy (lithium batteries, 16 fuel cells 17 and photovoltaics 18,19 ), and medical and biologic applications. 20Despite this increasing success, ALD still faces two main drawbacks which compromise both its generalised use in the laboratory and its implementation in large scale, high throughput production lines.First, ALD is a very slow technique (up to 2 orders of magnitude slower than chemical vapour deposition (CVD)) and, second, it usually involves processing in vacuum, thus making it complicated and expensive to scale up.
Both the unique advantages and drawbacks of ALD are the consequence of how it works: ALD is a particular case of CVD in which the reaction is restricted to the sample surface, thus being self-limited.This is achieved by exposing the sample to the reactants at different time, 21 i.e. in a sequence of pulses (figure 1a).In this way, the metal precursors (e.g.Al(CH 3 ) 3 in the case of fabrication of Al 2 O 3 ) are supplied and react with the surface, ideally forming a monolayer.
Excess precursor is then purged, usually by evacuation.The second precursor (e.g.H 2 O) is then injected and reacts with the chemisorbed layer forming a monolayer of the desired material (i.e.Al 2 O 3 ) plus by-products (CH 4 ) that have to be purged along with the excess precursor.The cycle is then repeated the necessary number of times to obtain a very precise film thickness.Thus, the unique advantages and drawbacks of conventional ALD are a consequence of  The Key to atmospheric, high throughput ALD: from temporal to spatial Although there have been several approaches to realizing ALD without the need of a high vacuum, Spatial ALD (SALD) has been the most popular in the last years due to its conceptual simplicity and the high deposition rates that can be achieved, and thus has also been called Atmospheric ALD (AALD). 235][26][27] The key idea allowing high throughput atmospheric spatial ALD is not new and was indeed patented in 1977 and 1983 by Suntola et al. 28,29 It simply consists in separating the precursors in space rather than in time (figure 1b), as

Materials Horizons Accepted Manuscript
opposed to temporal ALD.Thus, in SALD the different precursors are supplied constantly in between inert gas regions.Films are then grown by alternatively exposing the substrate from one precursor region to the other going across the inert gas regions (figure 1b).In this way, the oscillation of the substrate (or of the gases injector) from one precursor zone to the second one, going across the inert gas regions, reproduces the classical temporal ALD scheme: the first metal precursor reacts with the surface forming a monolayer while any unreacted precursor is swept away and purged in the inert region; Then the second precursor reacts with the previous monolayer forming a layer of material; Finally the sample returns to the first precursor region again going across the inert region where any by-products and excess precursor are purged.This approach allows much faster deposition rates, which are mainly limited by the precursor reaction kinetics, so that a full reaction can take place while the sample is within the precursor zones.As recently reviewed by Poodt et al. 22 , from 2008 there has been a variety of SALD designs exploiting this idea, some of them already being commercialized.The versatility of Suntola's approach can be appreciated by the multiple designs and configurations presented so far. 22,30e of the approaches used in several of the recent SALD designs is the use of a short distance between the precursor delivery zone and the substrate (typically below 100 m, Figure 2a).In this configuration, referred to as "close proximity SALD", the different gas regions are located along alternative channels where the gases flow parallel to each other, as illustrated in Figure 2a.The close distance of the substrate, i.e. the small gap between injector and substrate, prevents the different precursors from mixing since the inert gas channel acts as an effective barrier.This approach has several advantages including: i) The possibility of having smaller designs since the precursor regions are very small; ii) Simpler designs which are compatible with roll-to-roll processing; iii) The possibility of using gas-bearings. 22Although several patents on close-proximity SALD where filed from 2008 to 2011, 22 Levy et al. from Kodak were the first ones to present and publish their results on the deposition of ZnO thin film transistors, thus

Materials Horizons Accepted Manuscript
showing the suitability of SALD for fabricating device components. 31Later Poodt et al. also demonstrated high quality Al 2 O 3 films obtained by SALD.These films were used to passivate the surface of conventional silicon solar cells. 32The design used by Levy et al. consists of a deposition head in which the different gas flows are distributed along the parallel channels (see scheme in Figure 2b).This design is very simple and can be made very small, being indeed reminiscent of a printer head.Furthermore it is very easy to scale up, modify and customize.
Finally, it is readily compatible with roll-to-roll processing.Such an approach is therefore very readily applicable both in laboratory and industrial settings.A similar approach has already been commercialized for the deposition of Al 2 O 3 passivation layers on silicon and Cu(In,Ga)Se 2 thin-film solar cells. 22,33

The potential of SALD: Fabrication of active components for low-cost solar cells.
5][36][37][38][39][40] Our results include: i) The demonstration of the feasibility of using SALD for the ultra-fast deposition of high-quality films that can be integrated in devices; ii) The deposition of materials with superior properties compared to other low temperature and scalable methods; iii) The fundamental studies of new-generation solar cells by using doped oxide films; and iv) The nano-engineering of surfaces with SALD thin films to improve solar cells efficiencies.These results are succinctly reviewed next.
In the first example we used SALD to grow high quality TiO 2 blocking layers for organic (poly(3hexylthiophene-2,5-diyl): [6,6]-phenyl-C61-butyric acid methyl ester (P3HT:PCBM)) solar cells. 23Dense, uniform thin TiO 2 films were grown at temperatures as low as 100 °C in only 37 s (~20 nm/min growth rate).Incorporation of these films in P3HT-PCBM-based solar cells showed performances comparable with cells made using TiO 2 films deposited with much longer processing times and/or higher temperatures (SALD at 350 °C and spray pyrolysis at 450 °C).The high quality of the amorphous films obtained at 100 °C allowed the use of only 12 nm thick blocking layers, thus compensating for the lower conductivity of amorphous TiO 2 with respect to crystalline TiO 2 , and resulting in two orders of magnitude faster deposition than other low-temperature scalable methods (figure 3a).We also studied thin ZnO (<200 nm) films grown by SALD in a matter of minutes as a hole-blocking layer in P3HT:PCBM based inverted solar cells.Despite the very rapid, open atmosphere growth method, the performances of the cells were comparable with the best cells of the same type reported in the literature.The performances were also maintained after 200 days of storage in air in the dark (figure 3b), 34 showing the high stability of the ZnO films made by SALD.

Materials Horizons Accepted Manuscript
Secondly, we successfully deposited phase pure Cu 2 O films of thickness 50 -120 nm using our SALD system at temperatures between 125 and 225 °C (figure 3c, inset). 35Growth rates were ~1nm• min −1 , which is two orders of magnitude faster than conventional ALD.The films grown at 225 °C had carrier mobility values above 5 cm 2 V −1 s −1 .Good carrier concentrations of around 10 16 cm −3 were obtained for all deposition temperatures.The mobility values achieved were much higher than those obtained for electrodeposited films and approached the obtained for thermally oxidized films grown at 400 °C (figure 3c).Subsequently, we demonstrated SALD grown p+ Cu 2 O films on top of electrodeposited ZnO/Cu 2 O solar cells (figure 3d) to build in the electric field, thereby permitting the use of a thinner (~1mm compared to ~3mm) electrodeposited Cu 2 O layer.This thinner cell design gave reduced recombination losses and increased charge collection from both incident light and light reflected off the back electrode, resulting in a 29% increase in efficiency. 36other of the advantages of SALD is that doping can be achieved very easily by adding precursors to the different precursor lines.As an example, nitrogen doped ZnO (ZnO:N) can be deposited by using an ammonia solution instead of water. 41N doping decreases the carrier concentration of ZnO and this has allowed us to perform fundamental studies of PbX (X = S,Se) quantum dot (QD)/ZnO solar cells.For QDs where there is a significant density of intragap states, recombination at the ZnO/QD interface can be limited by reducing the ZnO carrier concentration through N doping since it prevents the back-transfer of electrons and/or enables the collection of trapped electrons via sub-bandgap states in the ZnO.As a result, both the cell open circuit voltage (V OC ) and efficiency increased with increasing N dopant level (figure 3e). 37 have also used SALD to engineer the surface of ZnO by depositing a 20 nm coating of N:ZnO on plain ZnO.The reduction in electron concentration in the surface results in enhanced exciton disociation in ZnO/P3HT cells due to a more efficient light-induced de-trapping of electrons for the ZnO:N surface.This enhances exciton dissocation and electron transfer from

Materials Horizons Accepted Manuscript
the P3HT to the oxide, resulting in 4 times higher short circuit density(figure 3f). 39Finally, Mg doped ZnO has been successfully deposited using SALD, allowing a continuous tuning of the ZnO band gap from 3.3 eV to 5.5 eV by adjusting the Mg percentage in Zn 1-x Mg x O.We then used Zn 1-x Mg x O in PbSe/ZnO solar cells, and we could increase the V OC from 408 mV up to 608 mV at a Zn 1-x Mg x O bandgap of 3.55 eV, the highest V OC reported to date for this type of cells (figures 3g-h). 40

Future challenges for SALD: towards a materials printer
SALD is a technique that has only just recently begun to be used for the deposition of high quality materials in functional devices.Indeed, the first theoretical [42][43][44] and kinetic 45,46 studies on SALD have been published in just the last months and reports evaluating the effect of deposition conditions on the properties of SALD ZnO and Al 2 O 3 were also published recently. 47- 49 o date, Al 2 O 3 , Cu 2 O, TiO 2 and ZnO (including doping such as Al:ZnO, Mg:ZnO and ZnO:N) have already been demonstrated (table 1 presents a summary of these materials and the application to solar cells detailed above).Hence, while it is early days, it is clear that there is tremendous potential for SALD, as briefly discussed in this minireview.The emergence and importance of SALD has also been recognized by the inclusion of a dedicated symposium on high throughput ALD at the AVS60 meeting (Fall 2013).

Materials Horizons Accepted Manuscript
One of the main limitations of SALD compared to ALD is that only relatively high volatility precursors can be used in order to have a sufficient amount of precursor molecules in the carrier gas.As the technique becomes more popular, we expect novel precursors to be designed specifically for SALD systems with higher volatility and reactivity.Of course, the price of novel SALD precursors will need to be reasonable in order not to compromise the low cost associated to SALD.But in this sense, SLAD is intrinsically more efficient in terms of precursor utilisation, thus allowing the use of precursors with slightly higher prices than conventional ALD/CVD precursors.Other limitations that will have to be overcome include the ability to maintain the high deposition rates of SALD when high aspect ratio features have to be coated, since precursor diffusion plays a key role in these cases (particularly in the case of high aspect ratio cavities, such as supported alumina templates 51,52 ).The coating of freeform samples will require novel SALD designs, especially in the case of the close proximity approach.
Additionally, while plasmas have already been implemented in some SALD systems, 22 their use in atmospheric systems will have to be developed by implementing atmospheric plasmas.Finally, the deposition of complex materials such as multicomponent oxides remains to be explored using SALD.
One of the main advantages of SALD is that it is based on a very simple idea that allows for freedom of design modification, and that it is a very simple and cost-effective research tool, implementable in many different laboratory settings.Thus, customized systems for specific materials or substrates will surely be developed in the coming years.We expect that over the next years we will surely see a proliferation in the utilization of SALD, both in terms of the number of system designs and applications targeted -the areas of flextronics and barrier layer coatings being primary examples.

Materials Horizons Accepted Manuscript
Fast air printing of functional oxide films: spatial atomic layer deposition, a new technique with a high potential the field of low cost photovoltaics.

Materials Horizons Accepted Manuscript
separating the different precursors in time (figure1a, conventional ALD is thus also called temporal ALD).

Figure
Figure 1a and b: Temporal vs. Spatial ALD.Copyright 2011, AIP Publishing LLC; Reprinted with permission from ref.22

Figure 2 .
Figure 2. a) close-proximity SALD approach.The sample oscillates under the different flows

Figure 3 :
Figure 3:a) Best SALD cell efficiency compared with equivalent cells from the literature in terms . Musselman, G. J. Mulholland, A. P. Robinson, L. Schmidt-Mende, and J. L. MacManus-Driscoll, Adv.Mater., 2008, 20, 4470-4475.Fast air printing of functional oxide films: spatial atomic layer deposition, a new technique with a high potential in the field of low photovoltaics.148x71mm (150 x 150 DPI)

Table 1 :
Summary of the different materials deposited by SALD to date and their application to Despite this success, the ultimate challenge is to create a versatile materials printer based on the SALD approach where all the characteristics, materials and applications that are currently achievable with ALD can also be obtained with SALD, while maintaining the vacuum-free environment, use of simple equipment, and the orders of magnitude faster deposition rates.