Quick link to open access paper: Applied Physics Reviews 9, 041313 (2022)
Earlier this month I published my review paper entitled Atomic layer deposition of conductive and semiconductive oxides. It was an invited review for Applied Physics Reviews and I was very happy to see it picked as Featured article!
In case you are not familiar with the terms (semi)conductive oxide, amorphous oxide semiconductor (AOS) or transparent conductive oxide (TCO), you can find a brief introduction in the text box below.
What are (semi)conductive oxides?
(Semi)conductive oxides are a class of metal oxide materials that can interestingly combine optical transparency and electrical conductivity, two properties that typically don’t go hand in hand. Think of glass, which is transparent but electrically insulating, or metals, which are conductive but not transparent. These metal oxides are mostly transparent in the visible region of the spectrum because of their relatively high band gap (> 3 eV). At the same time, they can be made conductive by (typically n-type) doping.
These materials are often labelled as amorphous oxide semiconductors (AOS) or transparent conductive oxides (TCOs). This depends on the doping level (or: carrier density), they either behave as a semiconductor (AOS) or as a conductor (TCO). As schematically shown in the figure below, at relatively low carrier densities these materials act as a semiconductor: the Fermi level EF resides in the band gap and is relatively far away from the conduction band (more than the thermal energy kT). In this case, the carrier distribution can be described by Boltzmann statistics and importantly, the position of the Fermi level can be relatively easily changed by external influences: for example, by gating in a transistor or by gas exposure in a gas sensor. These AOS materials find their application in thin-film transistors, sensors, and more. Interestingly, these materials can have high carrier mobilities in their amorphous state. This is related to the orbital overlap in the material; I won’t go into such detail here, more info in the paper!
Conversely, for high doping levels (typically >1018-1019 cm-3), the Fermi level comes closer to the conduction band or even enters the conduction band. In this case we speak of a degenerate semiconductor and the carrier distribution is dictated by Fermi-Dirac statistics. The Fermi level becomes much more resilient to external influences, since now a small change in EF implies a large change in carrier density. In other words, the material behaves more as a (metallic) conductor rather than a semiconductor. In this case we speak of TCOs, which find their application as transparent electrode in for example solar cells and displays.
These AOS and TCO materials are often based on (a combination of) indium, zinc, tin, titanium and gallium. Common AOS materials are InGaZnO, ZnSnO and InZnO, while popular TCO materials are for example ITO (In2O3:Sn) and AZO (ZnO:Al). Note the difference in notation: AOS materials are typically more equally mixed and hence noted as for example InGaZnO. TCO materials typically consist of a host metal oxide that is doped with a few at. % of a dopant atom, and hence written with the dopant notation using the colon (:).
What I think is an interesting aspect of the review paper is that it discusses ALD of both the AOS and TCO material systems on an equal footing. This is different from other works that discuss either systems (e.g. , ), while to me it makes a lot of sense to address both: there is a very strong overlap in the elements used, in their optoelectronic properties, and even more so in the important ALD aspects and approaches. I personally realized this at the moment I started exploring the field of ALD AOS and noticed how much I benefited from my background in the field of ALD TCOs (For solar cells, i.e. high-mobility hydrogen-doped In2O3– and ZnO TCOs, as well as TCO-based passivating contacts –) As such, it would be great if this review paper could also help in bringing these communities closer together.
I think the second unique feature of the review paper is its scope: As you would expect of a review paper, an important aspect of the paper is to review the many ALD processes and approaches for these conductive oxides available in literature, including graphs and tables on precursors, growth-per-cycle values, and so on… In the review paper I tried to go a step further by drawing a connection: of how ALD aspects (surface chemistry, supercycles, nucleation, …) relate to the resulting film structure (crystallinity, doping density and dispersion, stoichiometry…) and thereby the optical and electrical properties (carrier density, mobility, free-carrier absorption, …) . Moreover, examples are given of how such insights can be leveraged to make better films.
Since the review paper is open access for you to read, I don’t intend to reiterate the contents in this blog post. Instead, I would like to take a different angle here by highlighting four of the interesting (recent) applications of ALD of (semi)conductive oxides that I came across. I’ll focus on those applications that are uniquely enabled by ALD. Interestingly, in these four applications, four distinct merits of ALD play an enabling role. These merits are (1) composition control, (2) the ability to make ultrathin films, (3) conformality, and (4) soft deposition. Therefore, I decided to structure the blog based on these merits! Finally, I reckon there are other cool examples that I might have missed, so feel free to add in the comments section below!
ALD merit #1: Composition control
Nanolaminate In-Ga-Zn-O channels for TFTs
InGaZnO is an AOS channel material used thin-film transistors (TFTs), for example in display technology. What is interesting about this material is that the film structure and electronic properties can be accurately tailored by changing the stoichiometry of the metal centers. Briefly put, In-rich films exhibit a higher mobility due to a favorable overlap of the In 5s orbitals that build up the conduction band. However, In-rich films also suffer from a high carrier density (and hence from a negative threshold voltage Vth). This can be understood from the intrinsic tendency of oxygen vacancies to form in In2O3 owing to a relatively weak In-O bond. These oxygen vacancies act as electron donors. Ga can be added to the film in order to reduce the carrier density, yet at the expense of carrier mobility. Similarly, the main role of Zn in InGaZnO is to suppress crystallinity. Therefore, accurately balancing the composition of these InGaZnO films is vital for making the best films possible.
With conventional deposition methods (mostly PVD), the metal atoms are in principle incorporated randomly within the matrix, where the composition of the deposited film is determined by e.g. the (sputter) target composition. See also the figure below. Interestingly, with ALD you easily obtain an additional degree of freedom to design the material! Because of the cyclic nature of ALD, where films are grown atomic layer by atomic layer, you can design nanolaminate structures and graded compositions in a rather straightforward manner (note that there are many exceptions and subtleties, which I review in the paper). For example, Cho et al. showed improved TFT characteristics by making a bilayer InGaZnO structure of varying composition. Here, I would like to focus on another interesting approach I encountered that is really enabled by ALD. Sheng et al. showed very good transistor properties (e.g. high field-effect mobility μFE of ~74 cm2/Vs) by making a clever nanolaminate that I will explain below. Interestingly, these results were also recently reproduced by spatial ALD by TNO at Holst Center and SparkNano (previously known as SALDtech, see also the interview as part of the Company talk series).
Sheng et al. obtained superior transistor characteristics by making a nanolaminate that is made up by so-called ALD “supercycles”. One ALD supercycle consist of an integer n ALD In2O3 subcycles, followed by what they call “spacers” consisting of one ALD Ga2O3 subcycle and one ALD ZnO subcycle. The “In2O3-rich” layers contribute to a high lateral carrier mobility, while the Ga2O3 and ZnO subcycles still sufficiently suppress the carrier density and crystallinity, respectively. Ostensibly, a better trade-off between properties can be made using a nanolaminate instead of an isotropically-dispersed InGaZnO layer, and the best mobility of ~74 cm2/Vs was found for an In2O3 interlayer thickness of 1.8 nm.
What I also found insightful is that they mention a clear benefit of using plasma ALD in their work. I was aware that plasma ALD is commonly used in these AOS materials, mainly since the use of a reactive oxygen plasma results in less oxygen vacancies and thereby helps suppress the carrier density (as oxygen vacancies form electron donors). What Sheng et al. highlight, is that by using a plasma-based process, the growth-per-cycle of the various materials in the In-Ga-Zn-O nanolaminate is equal to that in the pure compound. In other words, for example the growth-per-cycle of the ZnO step within the nanolaminate is the same as when growing pure ZnO. This is certainly not the case for most ALD supercycle processes! Actually a large portion of the review paper is devoted to such growth effects: basically almost always the growth-per-cycle is different (enhanced or inhibited) when combining ALD materials in a supercycle. This renders it difficult to target for example a certain stoichiometry without prior knowledge of such effects. These changes in growth-per-cycle arise due to a myriad of mechanisms that can occur when changing ALD process within the supercycle and are mostly related to surface chemistry. Again, more details in the review paper. Most likely, in the case of Sheng et al. the O2 plasma helps to combust any remaining ligands and to fully oxidize the surface, thereby “resetting” the surface chemistry for the next subcycle.
ALD merit #2: Ultrathin continuous films
Ultrathin In2O3 transistor channels
Another cool application of ALD that I came across is the use of ultrathin (down to 0.7 nm!) In2O3 channels in TFTs that are made by plasma ALD. The key perk of ALD here is the ability to provide such ultrathin layers with wafer-scale uniformity, as also shown in the cited paper. Although not reported in the paper, I would expect the use of plasma ALD is quite vital here too, as its reactivity helps expedite a fast nucleation. This is critical for obtaining continuous, closed films at these very low thicknesses.
The authors also mention two strong benefits from keeping the In2O3 so extremely thin:
Firstly, by keeping the In2O3 thickness down to the nanometer regime, the formation of the polycrystalline phase is suppressed. Having an amorphous channel has its advantages. Even though polycrystalline In2O3 can have a high mobility, the grain size is typically on a similar order of magnitude as the device dimensions. This leads to variability in TFT characteristics, as some TFTs might have few or no grain boundaries, while others have many.
Secondly, In2O3 has the downside that donor defects are easily formed that raise the carrier density to the order of 1020 cm-3, sufficient to reach degeneracy. By making the In2O3 ultrathin (quasi-2D) films, quantum confinement effect kicks in, basically increasing the band gap of the material (as also used in many quantum dot applications). As the conduction band shifts upwards, the trap neutrality level moves within the band gap, as opposed to within the conduction band for bulk In2O3. Because of this, unintentional defect doping no longer leads to degeneracy. For more details and visualization, be sure to check out the original paper. 
ALD merit #3: conformality
Ferroelectric RAM with oxide semiconductor channel
One area where it seems ALD oxide semiconductors will play a defining role is in monolithic 3D integration of back-end-of-line (BEOL) compatible AOS transistors. Basically, this entails 3D memory that can be placed vertically above the Si transistors on a chip. This is an alternative scaling method to use chip area more efficiently, and was one of the “hot topics” that Intel mentioned in their keynote during last month’s IEDM conference (see Intel website). Specifically ferroelectric RAM (FeRAM) is mentioned. In FeRAM, the gate dielectric is ferroelectric, which means that a remnant polarization can be induced in the dielectric by biasing the gate. The electric field from this remnant polarization in turn controls the carrier population in the oxide semiconductor channel. By switching polarity of the gate, “program” and “erase” operations can be performed.
An example of explorative work in this direction is shown in the figure below. While quite a lot can be said about the workings of such a device (see the paper), I would like to focus here on some of the ALD aspects of this structure. As can be seen in the figure, the FeFET is based on a trench structure. The rationale behind this is that the use of vertical structures helps increase the areal density of the transistors. This is also where ALD shows its merits, as it can conformally coat the trench with the (anti)ferroelectric gate dielectric as well as the oxide semiconductor (In2O3 in this case). While the aspect ratio of the trench shown in the figure might not appear extremely challenging, this will likely change in the future: firstly, further downscaling beyond such a demonstrator device is expected and second, as the paper also mentions the envisioned application is to stack multiple of such transistors in a single trench.
Interesting to note is that ALD is used both for the FE-HfO2 (HfO2 often paired with ZrO2) and the In2O3 channel. Prior to ALD In2O3, the ferroelectric layer is crystallized by annealing at 600 oC. This yields crystalline domains that play a key role in the ferroelectric effect. The In2O3 layer is deposited at 200 oC and then annealed in O3 at 200 oC, presumably to lower the carrier concentration by the suppression of oxygen vacancies. As a detail (not noted in the paper), a close look at the TEM images therein seems to suggest the In2O3 layer is also crystalline, potentially already as-deposited or after annealing. Either way, in this specific case it is thus best to speak of an OS (oxide semiconductor) rather than an AOS.
ALD merit #4: soft deposition
ALD SnO2 buffer layers in silicon-perovskite tandem solar cells
|Actually, I already wrote about this application of ALD on AtomicLimits back in 2019! That blog gave a summary of a tutorial I gave at the ALD conference on “Atomic Layer Deposition for solar cells”. Back then I wrote “…, I think ALD SnO2 will keep on playing a key role in these tandem structures.” This is certainly still true, which I why I think it deserves mentioning in this blog again. Also, in three years a lot has happened, so a brief updated overview here.|
One application where I really see a role for ALD is for SnO2 in silicon-perovskite tandem solar cells (see the figure below). Such cells are a hot topic, where a solar cell based on a perovskite semiconductor is placed on top of a solar cell that is based on a conventional silicon solar cell. This can make the total solar cell more efficient: briefly put, the top perovskite solar cell is more efficient for absorbing high-energy photons, while it transmits low-energy photons to the silicon bottom cell that is more efficient in this low-energy part of the spectrum. For more info, see the previous blog.
What is the role of the ALD SnO2 here? It serves mainly two functions:
- The typically 10-20 nm thick ALD SnO2 act as a buffer layer that prevents (plasma) damaging of the perovskite during sputter deposition of the top TCO contact.
- As a dense, amorphous material, it improves environmental stability of the perovskite cell which is susceptible to ingress of ambient oxygen and water.
What seemingly makes ALD an enabling technique here is that is a soft deposition technique. The SnO2 is typically deposited by thermal ALD from tetrakis(dimethylamino)tin (TDMASn) as precursor and H2O as co-reactant. Importantly, the fact that this ALD process is thermal (so only gases supplied) obviously ensures no plasma damage, while also the ALD process can be performed at very low temperatures (<100 oC) which makes it compatible with the temperature-sensitive perovskite materials.
Two main things have changed since my previous blog. The record efficiency is now at 32.5%, which is much higher than the 28% in 2019. Because of this, these tandem cells now have a very clear efficiency premium over silicon solar cells, which are levelling off at around 26.7%. Relatedly, there seems to be a lot of momentum to work towards mass production of these cells. Most notably, various EU consortia comprising manufacturers and knowledge institutes are teaming up in this respect, also in an effort to bring solar cell production back to Europe. For example, earlier this month Meyer Burger announced partnerships with various institutes in activities that “… center around industrialization of the new technologies, moving from the laboratory to mass production …”. Similarly, last month the PEPPERONI EU project was kicked off, where PV manufacturer Hanwha Qcells and partners strive to set up a pilot line for the production of tandem cells in Thalheim, Germany. Interestingly, whereas Meyer Burger (and most record cells) opt for so-called silicon heterojunction solar cells as bottom cells, the PEPPERONI project targets the use of PERC (passivated emitter and rear contact) silicon bottom cells (see image below). These PERC cells are by far more standard in industry than heterojunction cells, which make them an appealing option from an industry point of view. And while it appears that the record efficiency of PERC-perovskite solar cells is slightly lower thus far at 28.7%, it should be noted that this approach is relatively newer and it would be interesting to see how far this technology will go.
Either way, in both cases ALD SnO2 plays a vital role! Who knows, this application of ALD might very well be the next big contribution for ALD in photovoltaics, after the success of ALD Al2O3. (See also two of our earlier blogs on ALD Al2O3 for silicon solar cells here and here)
 B. Macco and W. M. M. Kessels, “Atomic layer deposition of conductive and semiconductive oxides,” Appl. Phys. Rev., vol. 9, no. 4, p. 041313, Dec. 2022, doi: 10.1063/5.0116732.
 T. Tynell and M. Karppinen, “Atomic layer deposition of ZnO: a review,” Semicond. Sci. Technol., vol. 29, p. 043001, 2014, doi: 10.1088/0268-1242/29/4/043001.
 J. Sheng, J.-H. Lee, W.-H. Choi, T. Hong, M. Kim, and J.-S. Park, “Review Article: Atomic layer deposition for oxide semiconductor thin film transistors: Advances in research and development,” J. Vac. Sci. Technol. A, vol. 36, no. 6, p. 060801, 2018, doi: 10.1116/1.5047237.
 B. Macco, Y. Wu, D. Vanhemel, and W. M. M. Kessels, “High mobility In2O3:H transparent conductive oxides prepared by atomic layer deposition and solid phase crystallization,” Phys. status solidi – Rapid Res. Lett., vol. 8, no. 12, pp. 987–990, Oct. 2014, doi: 10.1002/pssr.201409426.
 B. Macco, M. A. Verheijen, L. E. Black, B. Barcones, J. Melskens, and W. M. M. Kessels, “On the solid phase crystallization of In2O3:H transparent conductive oxide films prepared by atomic layer deposition,” J. Appl. Phys., vol. 120, no. 8, p. 085314, 2016, doi: 10.1063/1.4962008.
 B. Macco, H. C. M. Knoops, and W. M. M. Kessels, “Electron Scattering and Doping Mechanisms in Solid-Phase-Crystallized In2O3:H Prepared by Atomic Layer Deposition,” ACS Appl. Mater. Interfaces, vol. 7, no. 30, 2015, doi: 10.1021/acsami.5b04420.
 B. Macco, H. C. M. Knoops, M. A. Verheijen, W. Beyer, M. Creatore, and W. M. M. Kessels, “Atomic layer deposition of high-mobility hydrogen-doped zinc oxide,” Sol. Energy Mater. Sol. Cells, vol. 173, pp. 111–119, Dec. 2017, doi: 10.1016/j.solmat.2017.05.040.
 B. W. H. Van De Loo, B. Macco, J. Melskens, W. Beyer, and W. M. M.. Kessels, “Silicon surface passivation by transparent conductive zinc oxide,” J. Appl. Phys., vol. 125, no. 10, p. 105305, 2019, doi: 10.1063/1.5054166.
 B. Macco et al., “Atomic-layer-deposited Al-doped zinc oxide as a passivating conductive contacting layer for n+-doped surfaces in silicon solar cells,” Sol. Energy Mater. Sol. Cells, vol. 233, no. September, p. 111386, Dec. 2021, doi: 10.1016/j.solmat.2021.111386.
 B. Macco et al., “Temporal and spatial atomic layer deposition of Al-doped zinc oxide as a passivating conductive contact for silicon solar cells,” Sol. Energy Mater. Sol. Cells, vol. 245, no. June, p. 111869, Sep. 2022, doi: 10.1016/j.solmat.2022.111869.
 J. Melskens, B. W. H. van de Loo, B. Macco, L. E. Black, S. Smit, and W. M. M. Kessels, “Passivating Contacts for Crystalline Silicon Solar Cells: From Concepts and Materials to Prospects,” IEEE J. Photovoltaics, vol. 8, no. 2, pp. 373–388, Mar. 2018, doi: 10.1109/JPHOTOV.2018.2797106.
 J. Ibarra Michel, J. Dréon, M. Boccard, J. Bullock, and B. Macco, “Carrier‐selective contacts using metal compounds for crystalline silicon solar cells,” Prog. Photovoltaics Res. Appl., no. January, pp. 1–34, Mar. 2022, doi: 10.1002/pip.3552.
 M. H. Cho, C. H. Choi, H. J. Seul, H. C. Cho, and J. K. Jeong, “Achieving a Low-Voltage, High-Mobility IGZO Transistor through an ALD-Derived Bilayer Channel and a Hafnia-Based Gate Dielectric Stack,” ACS Appl. Mater. Interfaces, vol. 13, no. 14, pp. 16628–16640, 2021, doi: 10.1021/acsami.0c22677.
 J. Sheng et al., “Amorphous IGZO TFT with High Mobility of ∼70 cm 2 /(V s) via Vertical Dimension Control Using PEALD,” ACS Appl. Mater. Interfaces, vol. 11, no. 43, pp. 40300–40309, Oct. 2019, doi: 10.1021/acsami.9b14310.
 M. Si et al., “Why In2O3 Can Make 0.7 nm Atomic Layer Thin Transistors,” Nano Lett., vol. 21, no. 1, pp. 500–506, 2021, doi: 10.1021/acs.nanolett.0c03967.
 Z. Li et al., “A Vertical Channel Ferroelectric/Anti-Ferroelectric FET with ALD InOx and Field-Induced Polar-Axis Alignment for 3D High-Density Memory,” in 2022 IEEE Silicon Nanoelectronics Workshop (SNW), Jun. 2022, pp. 1–2, doi: 10.1109/SNW56633.2022.9889037.
 K. Sveinbjörnsson et al., “Monolithic Perovskite/Silicon Tandem Solar Cell with 28.7% Efficiency Using Industrial Silicon Bottom Cells,” ACS Energy Lett., vol. 7, no. 8, pp. 2654–2656, 2022, doi: 10.1021/acsenergylett.2c01358.