“Plasma ions create electronic defects in sensitive substrates as well as generate a large electronic trap density in deposited films, especially detrimental for gate dielectric applications.” This is a sentence that I read about 10 years ago (in January 2007 to be more precise) in a white paper published by an ALD tool vendor. This white paper was on direct and remote plasma ALD, which the vendor at that time would not recommend to their customers as was stated in the paper.
The statement might as well be true under certain circumstances but it is not really nuanced. It would have been better to state the influence of ions on the material properties a bit more carefully. Plasmas have been ubiquitous in semiconductor materials processing for decades and certainly also in gate dielectric applications. Just think about the plasma etching steps to etch the gate stack. To avoid or minimize damage, one needs to understand the plasma-surface interaction very well and know how to control the ion impact by tuning the plasma conditions. I guess the latter was realized by the ALD tool vendor soon enough after publishing the white paper as they also started selling plasma ALD tools a year or so later. Nowadays, almost all ALD tool vendors have plasma ALD tools in their portfolio and these plasma ALD tools account for a good chunk of their sales. There are several ALD tool vendors who even told me that the plasma ALD tools make up at least 50% of ALD tools sold. Due to the importance of ALD in patterning this seems certainly to be the case for the 300 mm tools used in the logic and memory industry. By the way, if you are interested in database of plasma ALD publications, please see the website of the plasma ALD guy (Thanks Mark for providing such a helpful database!).
To demystify the role of ions during plasma ALD we began investigating the ions in plasma ALD reactors through a PhD project in 2008. Using Langmuir probes, we measured the ion density in various plasmas used for ALD (especially O2, H2 and N2 plasmas) and determined their flux to the substrate. We also measured the energy of the ions arriving at the substrate with a retarding field energy analyzer (RFEA) [1]. The PhD student who conducted these experiments was Harald Profijt and the experiments were carried out in our home-built ALD reactor and our Oxford Instruments FlexAL and OpAL reactors (note that at that stage we had only one of each, compared to the three home-built reactors and three FlexALs currently in our labs). All these reactors had a grounded substrate table at that time. At the substrate level, we found ion fluxes ranging from 1012 to 1014 cm-2s-1 and ion energies up to 35 eV depending on the plasma power and pressure conditions. For the oxide films investigated, we calculated that the ratio of ions impinging on the surface and the number of atoms deposited varied from 0.05 to 1 during a cycle whereas the ion energy provided per atom deposited could be as low as 1 eV and as high as almost 20 eV [2]. To illustrate the possible role of ions during the deposition process and to put it in perspective of other plasma processes, we added our ALD results in the famous plot by Takagi [3]. This graph has appeared in our review article on plasma ALD in 2011 [4].
Ion-surface interactions during plasma processes including plasma ALD. The plot is based on the one published by Profijt et al. [4], which by itself is based on the plot published by Takagi [3].The arrows show typical trends: (1) the energy of ions impinging on the substrate can be reduced by increasing the plasma pressure; (2) the ion density in the plasma and the flux towards the substrate can be increased by increasing the plasma power; and (3) the energy of the ions impinging on the substrate can be increased by substrate biasing in which RF bias power is coupled to the substrate table.
The bottom-line of the results obtained by Harald Profijt was that the role of ions during plasma ALD cannot simply be neglected. Under certain conditions, they can have a substantial flux towards the substrate and provide a fair amount of energy per atom deposited in plasma ALD. On the other hand, the ion energies are quite low and lie below the typical threshold energies for defect generation. Harald showed that the role of high energy photons (photons in the vacuum ultraviolet, VUV) can be much more detrimental under regular plasma conditions [2]. Moreover, the energy of the ions impinging on the substrate could be reduced by increasing the pressure of the plasma. This effect is well-known by plasma physicists (a higher pressure typically leads to a reduced sheath voltage and also to more collisions of the ions in the plasma sheath) and it was also confirmed experimentally in our ALD reactors. Hence, to rule out any damage by ions, it is advisable to operate the plasma at higher pressures (and to preferably use a remote plasma ALD reactor). I have schematically illustrated the effect of the pressure in the modified plot from Takagi shown above.
At that stage we started wondering what would happen to the plasma ALD process and the resulting material properties if we would increase the energy of the ions instead of decreasing it (see the modified plot of Takagi). We therefore decided to implement substrate biasing in our home-built ALD reactor. Substrate biasing is very common in the field of plasma processing, especially in plasma etching but also in plasma-enhanced chemical vapor deposition (PECVD). It means that the substrate table is not simply grounded but connected to an electric circuit such that the potential across the plasma sheath can be raised considerably. For plasma conditions with a collisionless plasma sheath (that means for sufficiently low pressures), the ions are accelerated over this larger potential leading to significantly enhanced ion energies. We implemented two configurations for substrate biasing: substrate-tuned biasing using only a matching network (with inductive and capacitive components) connected to the substrate stage and RF biasing. In the latter configuration a second RF power supply is connected to the system: it is connected to the substrate table through a matching network [5,6]. These methods have the advantage over dc biasing (which could be considered more straightforward) as they can also be applied when using non-conductive samples and/or films. With both configurations, we were able to considerably increase the energy of ions impinging on the substrate as can be seen in the figure below.
Remote plasma ALD reactor with an inductively coupled plasma (ICP) source. The plasma is generated by coupling RF power to the ICP coil through a matching network consisting of inductive and capacitive components. Substrate biasing is achieved by applying RF bias power to the substrate stage, also through a matching network. Additionally, substrate biasing can be achieved at zero RF bias power by simply tuning the matching network (substrate-tuned biasing). The plot on the right shows the ion energy distribution function (IEDF) obtained under certain biasing conditions. It clearly shows that the (average) energy of the ions arriving on the substrate can be significantly enhanced by biasing the substrate stage with a negative bias voltage (0 V means that the substrate stage is grounded) [5,6].
Next we evaluated the influence of the enhanced ion energy on the properties of the ALD films prepared. Very interestingly, we found that the enhanced ion energy could have profound effects on the properties but with big differences between materials: for TiO2 films we found that we could switch between anatase TiO2 and rutile TiO2 by increasing the ion energy (keeping all other conditions constant) [5,6] whereas Co3O4 films could be densified, and for Al2O3 the film stress could be controlled [6]. So it could be concluded that controlling the ion energy yields an additional “knob” during plasma ALD to tailor the ALD material properties.
Triggered by these results, our research partner Oxford Instruments decided to implement RF substrate biasing in their FlexAL systems for which their customers could opt-in. For Oxford Instruments, this was a relatively small step as they had ample experience with RF substrate biasing in their plasma etch and ICP-CVD tools. Only some of the hardware in the FlexAL system had to be adjusted. The second FlexAL system that was installed in our clean room was upgraded with this option of RF substrate biasing a few years ago (see the photograph below).
The second Oxford Instruments FlexAL system that was installed in the NanoLab@TU/e clean room in 2011. A few years later this system was upgraded with RF substrate biasing. In the meantime a third FlexAL system was installed which also has the capability of RF substrate biasing.
Since then we have been testing the influence of RF substrate biasing on a wide range of materials prepared with the FlexAL ALD reactor, see also the table below. The results have appeared in two journal papers that have just been published. One paper was written by the PhD student Tahsin Faraz [7]. He is the student that really carried out the RF substrate biasing work and demonstrated the influence of the ion energy on several materials, both on oxides and nitrides. The other paper was written by the PhD student Saurabh Karwal and he carried out in-depth investigations on (conductive) HfN films prepared by plasma ALD [8]. If you want to know the details of their work, I recommend reading their papers.
Overview of materials prepared by plasma ALD with substrate biasing, either in the home-built ALD reactor or in the Oxford Instruments FlexAL system. The influence of the enhanced ion energies varies greatly for the various materials: from densification to crystallization and from material deterioration to significant improvement of the conductivity.
Two articles on RF substrate biasing during plasma ALD that have recently been published [7,8].
Both students did not only test the influence of the RF substrate biasing on planar samples but they also investigated what happened to films deposited in 3D trenches. Due to the acceleration in the plasma sheath the flux of ions is directional when arriving on a substrate. If this substrate has submicron trenches, this means that the ions (almost) only hit the horizontal surfaces (the planar top of the wafer and the bottom of the trenches) and not the vertical sidewalls of the trenches. Interesting effects have therefore been seen for TiO2, HfO2, HfN and SiNx. For these materials, we even found that RF substrate biasing can be used for topographically-selective processes on 3D trenches. See some transmission electron microscopy images from Refs. 7 and 8 below. Please read the papers recently published to find out more details. For convenience, a schematic overview of the effects caused by enhanced ion energies is given in the figure below.
Films deposited on 3D trenches by plasma ALD with substrate biasing. The top left figure shows a TiO2 film which is conformally covering the trench. Due to the enhanced ion energies impacting the horizontal surface, the TiO2 is rutile on these surfaces whereas it is amorphous on the vertical surfaces. The bottom left figure shows also such a topographically-selective process but now for SiNx. The SiNx prepared with substrate biasing is of poorer quality than the films prepared without substrate biasing. This means that a diluted HF wet etch easily removes the SiNx from the horizontal surfaces leaving the SiNx only in place on the vertical surfaces. Both images come from Ref. [7]. The right images are from Ref. [6]: it shows a stack of HfNx films deposited in a trench with lower aspect ratio. The bottom HfNx film is deposited with substrate biasing and the top HfNx film was deposited without substrate biasing. It can be observed that the enhanced ion energies reduce the oxygen contamination of the HfNx on the horizontal surfaces but not on the vertical surfaces. The lower oxygen contamination increases the nitrogen content of the film.
Schematic illustration representative of the material properties and process control enabled by substrate biasing during plasma ALD on planar and 3D substrate topographies (after Ref. [7]).
Also from these recent results obtained on the FlexAL reactor, it is clear that additional control of the ion energy during plasma ALD is advantageous for tailoring the film properties and that this can be easily done on large substrate areas (the experiments on the FlexAL were done on 200 mm wafers). Yet it is also clear that the influence of enhanced ion energies typically varies from material to material. Also note that the experiments carried out are still reasonably limited in terms of the parameter space explored. As described in the supplementary information of Ref. 7, it is shown that many more configurations of the plasma ALD cycle with enhanced ion energies are possible. Furthermore, it is of interest to also study the impact of enhanced, but still reasonably low ion energies in much more detail. So far ion energies up to ~250 eV have been studied (achieved with average bias voltages up to -200 V) but ion energies between roughly 25 and 75 eV might be most interesting: this region is expected to cover energy thresholds for several ion-surface interactions. So this work is to be continued, not only by us but also by several other labs that currently have the substrate biasing option installed on their plasma ALD tools.
The influence of enhanced ion energies will be presented by Tahsin Faraz at the ALD conference in Korea (Incheon, July 29 – August 1, 2018).
References
[1] D. Gahan, B. Dolinaj, and M. B. Hopkins, Rev. Sci. Instrum. 79, 033502 (2008). DOI: 10.1063/1.2890100
[2] Ion and Photon Surface Interaction during Remote Plasma ALD of Metal Oxides, H.B. Profijt, P. Kudlacek, M.C.M. van de Sanden, and W.M.M. Kessels, J. Electrochem. Soc. 158, G88 (2011). DOI: 10.1149/1.3552663
[3] T. Takagi, J. Vac. Sci. Technol. A 2, 382 (1984). DOI: 10.1116/1.572748
[4] Plasma-assisted Atomic Layer Deposition: Basics, Opportunities and challenges, H. B. Profijt, S. E. Potts, M. C. M. van de Sanden, and W.M.M. Kessels, J. Vac. Sci. Technol. A. 29, 050801-1 (2011). DOI: 10.1116/1.3609974
[5] Substrate biasing during plasma-assisted ALD for crystalline phase-control of TiO2 thin films, H. B. Profijt, M. C. M. van de Sanden, and W.M.M. Kessels, Electrochem. Solid. State Lett. 15, G1 (2012). DOI:10.1149/2.024202esl
[6] Substrate-biasing during plasma-assisted atomic layer deposition to tailor metal-oxide thin film growth, H.B. Profijt, M.C.M. van de Sanden, and W.M.M. Kessels, J. Vac. Sci. Technol. A 31, 01A106 (2013). DOI: 10.1116/1.4756906
[7] Tuning material properties of oxides and nitrides by substrate biasing during plasma-enhanced atomic layer deposition on planar and 3D substrate topographies, T. Faraz, H.C.M. Knoops, M.A. Verheijen, C.A.A. van Helvoirt, S. Karwal, A. Sharma, V. Beladiya, A. Szeghalmi, D.M. Hausmann, J. Henri, M. Creatore, and W.M.M. Kessels, ACS Appl. Mater Interfaces 10, 13158 (2018). DOI:10.1021/acsami.8b00183
[8] Low resistivity HfNx grown by plasma-assisted ALD with external RF substrate biasing, S. Karwal, M.A. Verheijen, B.L. Williams, T. Faraz, W.M.M. Kessels, and M. Creatore, J. Mater. Chem. C 6, 3917 (2018). DOI:10.1039/C7TC05961B
This is great, the HfN paper is a nice addition!
Thanks Nick!