Flash-Enhanced Atomic Layer Deposition: Basics, Opportunities, Review, and Principal Studies on the Flash-Enhanced Growth of Thin Films

Within this work, flash lamp annealing (FLA) is utilized to thermally enhance the film growth in atomic layer deposition (ALD). First, the basic principles of this flash-enhanced ALD (FEALD) are presented in detail, the technology is reviewed and classified. Thereafter, results of our studies on the FEALD of aluminum-based and ruthenium thin films are presented. These depositions were realized by periodically flashing on a substrate during the precursor exposure. In both cases, the film growth is induced by the flash heating and the processes exhibit typical ALD characteristics such as layer-by-layer growth and growth rates smaller than one A/cycle. The obtained relations between process parameters and film growth parameters are discussed with the main focus on the impact of the FLA-caused temperature profile on the film growth. Similar, substrate-dependent growth rates are attributed to the different optical characteristics of the applied substrates. Regarding the ruthenium deposition, a single-source process was realized. It was also successfully applied to significantly enhance the nucleation behavior in order to overcome substrate-inhibited film growth. Besides, this work addresses technical challenges for the practical realization of this film deposition method and demonstrates the potential of this technology to extend the capabilities of thermal ALD. © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0301507jss] All rights reserved.

Atomic layer deposition (ALD) is a specific thin film deposition method wherein film growth is based on a sequence of alternating saturated surface reactions. In a typical ALD process, this is realized by the alternating exposure of the substrate surface to two precursors, which are separated from each other by inert gas purging or chamber evacuation in order to prevent gas phase reactions. As a result, the film growth proceeds in a self-limiting manner and monolayer-by-monolayer, and thus, ALD enables the deposition of thin films with excellent uniformity and conformality as well as with sub-nanometer thickness control. Thanks to these unique characteristics, ALD has emerged as an important thin film deposition technique in semiconductor technology, e.g. in the fabrication of metal-insulator-metal (MIM) film stacks in high aspect ratio dynamic random access memory (DRAM) structures and gate dielectrics in transistors. In recent years, the use of ALD has also expanded into other fields such as optoelectronics, energy conversion and storage devices, micro-electro-mechanical-systems (MEMS) and nanotechnology. [1][2][3][4][5][6] Although a large number of materials have been deposited by ALD so far 1-6 for various applications, there are still a number of challenges in ALD. The deposition temperatures in ALD are mostly lower compared to chemical vapor deposition (CVD) processes due to the different mechanisms of film growth and the necessity to prevent precursor decomposition in order to keep the self-limiting growth behavior of ALD. As a consequence of the lower energy available for film formation, the films may not meet the properties required for the specific application. In these cases, a post-deposition annealing is necessary to improve the film quality, 3,6-13 e.g. to obtain a desired film structure, 6-10 density, 10-12 purity 12,13 or other related properties. [8][9][10][11][12][13] However, this high temperature processing is often impracticable due to a restricted thermal budget of the substrate, in particular when coating temperature sensitive substrates. Secondly, the reactants of an ALD process, e.g. oxygen, may react with the substrate itself leading to the formation of a parasitic interfacial layer. 14,15 In order to prevent this issue a priori, the proper choice of reactants or the use of an alternative deposition technique is essential. Furthermore, many ALD processes suffer from substrate-inhibited film growth accompanied by inefficient precursor consumption and the formation of films with unfavorable properties 9,14,16,17 like for example a high roughness. 16 Moreover, there are materials of interest, e.g. titanium, tantalum and aluminum, which so far cannot be deposited by thermal ALD at all. 6 These limitations may be overcome by providing additional energy, e.g. in form of a plasma, 18 light [19][20][21] or heat, [22][23][24][25][26][27][28][29][30][31][32][33] in an appropriate way to enhance the film formation. In the scope of this work, flash lamp annealing (FLA) is used to thermally enhance the film growth in ALD.
In FLA, a substrate is exposed to a light flash with durations typically in the millisecond range. The light is absorbed within the top layers of the sample, causing a rapid heating of the surface near region. In contrast, the bulk material experiences no or only a moderate heating. Thereby, the bulk acts as a heat sink and provokes quick cooling of the heated layers. As a result, the elevated temperatures of the top region are present on the millisecond timescale only, and thus, FLA is not only a thermal treatment method but a low thermal budget process as well. [34][35][36][37][38] Consequently, FLA is a suitable technology to power high temperature processes even on temperature sensitive substrates. 34 Examples for the application of FLA are the activation of dopants and the defect annealing after ion implantation to form ultra-shallow junctions, 37,38 the crystallization of amorphous silicon in order to produce crystalline silicon on various substrates, 36,[39][40][41] the thermal treatment of high-k dielectrics for memory and transistor applications 42,43 and of transparent conductive oxide films for solar cell applications, 34 the sintering of particles 44 as well as ink-jet printed films, 45 and the deposition of thin films. [22][23][24][25][26][27][28][29] The latter use of FLA is the object of flash-enhanced ALD, which we describe in the following also abbreviated as FEALD.
Within this work, the basic principles of flash-enhanced ALD are presented in detail. Thereafter, both previously reported work about this matter and comparable approaches are reviewed and classified. In the main part, results of our studies on the FEALD of aluminumbased and ruthenium thin films are presented and discussed with the main focus on the impact of the FLA-caused temperature profile on the film growth. Besides, this work addresses technical challenges for the practical realization of this method and demonstrates the potential of this technology to extend the capabilities of thermal ALD.

Principle of Flash-Enhanced ALD
There are several variants of FEALD, but all of them are based on the basic principle schematically illustrated in Fig. 1. First, a substrate is exposed to the precursor and the precursor molecules adsorb on the surface via either chemisorption or physisorption. Next, the substrate is irradiated with a high-intensity light flash, which is usually generated by an array of xenon flash lamps. The spectral distribution of this light mainly covers the visible and near infrared range, and only a minor part extends to the near ultraviolet range. 36 Consequently, direct interactions between the light radiation and gas molecules, i.e. photolytic effects, are negligible, and the FLA treatment only leads to a short term heating of the illuminated surface. Due to this heating, the surface temperature exceeds the threshold temperature, which is required to achieve the thermal decomposition of adsorbed precursor molecules or to activate chemical reactions between the adsorbed precursor molecules and a second reactant. After this step, a layer of atoms originating from the precursor or a layer of the compound formed during the chemical reactions remains on the surface and volatile reaction by-products are purged out of the reaction chamber. As a result, one layer of the film material is formed in this basic sequence representing one process cycle in FEALD. Like in conventional ALD, films can be grown in a layer-by-layer fashion by repeating this basic sequence until the desired film thickness is achieved. In addition, flash heating in each cycle results in the periodical annealing of the already grown film, and hence, may lead to an improved film quality.
On the basis of the described basic principle (Fig. 1), various process variants are possible in FEALD. They mainly differ in the way, how FLA is applied in the deposition process. Figure 2 shows the process sequences of five different variants of FEALD. For comparison, the scheme of conventional thermal ALD is shown as well (Fig. 2d).
A flash-enhanced deposition process can be realized as a singlesource process by flashing periodically on a substrate during a continuous precursor gas flow (Fig. 2a) or during every single precursor pulse (Fig. 2b). In both cases, FLA induces the thermal decomposition of the precursor molecules on the substrates surface without the need of a second reactant. Although these variants are in fact a kind of a pulsed CVD from the process sequences point of view, the film growth may proceed in an ALD-like manner. In another slightly different approach, the adsorption step and the precursor decomposition are more clearly separated from each other by performing FLA during the purge subsequent to the precursor pulse (Fig. 2c). In this way, the film growth in one process cycle is definitely limited to the layer of adsorbed precursor molecules like in conventional ALD, and thus, is self-limiting as well. However, both the adsorption of precursor molecules on the substrate as well as on the film material and a negligible desorption during the purge step are essential for the realization of this approach.
Other techniques are based on the four-step gas pulse sequence of conventional ALD, comprising a precursor pulse, an inert gas purge, the pulse of the second reactant and another inert gas purge (Fig. 2d) Figure 2. Process sequences for different variants of FEALD and conventional thermal ALD: (a) -(c) single-source FEALD processes with FLA during (a) a continuous precursor exposure, (b) every single precursor pulse or (c) the purge pulses, (d) conventional thermal ALD, (e) FEALD using a precursor and a second reactant, (f) combination of thermal ALD and in-situ FLA for the thermal treatment during the film growth process. The grey-shaded part of each sequence represents one single process cycle.
) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 207.241.231.81 Downloaded on 2018-07-21 to IP P279 second reactant to initiate a chemical reaction between adsorbed species and the second reactant (Fig. 2e). Similar to the variants described above, this approach is aimed at the direct deposition of a thin film induced by FLA. On the contrary, FLA is also a suitable method to thermally treat thin films not only subsequent to an ALD process but already during the film growth process as well. Thereby, this insitu annealing can basically be applied after a certain number of ALD cycles or even after every single ALD cycle (Fig. 2f). However, since this work mainly addresses the direct deposition of thin films, the latter subject will not be discussed here in detail. Related studies have been reported elsewhere. 32,33 Furthermore, flash-enhanced deposition processes can be realized by any combination of the described basic approaches as well.
Altogether, the application of FLA in a deposition process offers a high potential to remarkably extend the capabilities of thermal ALD. Promising potentials are the possibility to simplify an ALD process by the realization of single-source processes, and thus, to prevent the formation of parasitic interfacial layers, the reduction of growth delay in the initial phase of film growth by the enhancement of the nucleation behavior, the deposition of films with improved properties (e.g. highpurity thin films), the deposition of new materials with this special kind of thermal ALD, and the deposition of thin films on temperature sensitive substrates.

Literature Review
Flash-enhanced deposition processes.-The atomic layer growth of silicon thin films induced by FLA has been reported by Murota and Sakuraba et al. 22,23 Their process was realized by flashing periodically on a substrate in a continuous monosilane (SiH 4 ) gas flow according to the process sequence in Fig. 2a. The film growth is based on the saturated adsorption of SiH 4 molecules in the time intervals between flashing and the subsequent thermal decomposition during the FLA treatment. Observed influences of the SiH 4 partial pressure and the substrate orientation on the saturation behavior were quantitatively explained by Langmuir-type adsorption. Apart from the saturated adsorption, the process exhibits a sub-monolayer growth rate and a temperature window like in ALD. Furthermore, it allows the growth of epitaxial silicon films when providing an appropriate substrate. 22,23 The atomic layer epitaxy of germanium on silicon was investigated using the same technological approach and monogermane (GeH 4 ) as the source material. Again, the film growth is induced by FLA and proceeds in a self-limited manner due to a saturated precursor adsorption. Within the temperature window, the growth per flash cycle equates one germanium monolayer and is independent of the applied GeH 4 partial pressure. However, the adsorption behavior in terms of surface coverage velocity is determined by the GeH 4 partial pressure. Depending on the orientation of the silicon substrate, the obtained germanium films are single crystalline or amorphous. 24,25 The thermal nitridation of silicon by FLA was studied by Watanabe et al. 26,27 by flashing periodically on a silicon wafer exposed to an ammonia (NH 3 ) ambient. It was found, that FLA significantly enhances the reaction efficiency of physically adsorbed NH 3 molecules, while the Langmuir-type adsorption/desorption characteristics remains unaffected. As a result, this approach enables the step-by-step atomicorder nitridation of silicon. 26,27 Furthermore, the authors suggest the combination of the FLA-induced atomic layer growth of silicon as described above and the subsequent thermal nitridation of this layer by FLA in order to grow silicon nitride films in a layer-by-layer manner. 28 The use of FLA for the enhanced ALD of tantalum carbonitride and tantalum carbide films is claimed in a patent of Ishizaka et al. 29 It is stated, that FLA might be applied according to the process sequence in Fig. 2e to activate a chemical reaction between the adsorbed tantalumbased metal-organic precursor and a second reactant. In addition, they claimed that FLA during the purge steps of each cycle (see sequence in Fig. 2f) might be used to complete reactions of residual unreacted precursor, and that another FLA treatment after a certain number of cycles might be a suitable method to further improve film properties like adhesion, density or crystallinity. 29 Temperature-modulated deposition process by using RTA.-The co-reactant free deposition of ruthenium thin films was reported by Fafard et al. 30 using an approach similar to flash-enhanced ALD. This so-called thermo-cyclic ALD is based on the process sequence illustrated in Fig. 2c, but instead of FLA rapid thermal annealing (RTA) was applied to achieve the thermal decomposition of the adsorbed precursor. The process utilizes the ruthenium-based precursor CHORUS, a substrate temperature of 200 • C during the precursor exposure and an RTA temperature of 500 • C. Thereby, ALD-like saturation was observed and a linear growth behavior without any noticeable incubation period was demonstrated on both tantalum nitride and ruthenium substrates. Finally, the authors suggested the use of FLA instead of RTA in order to significantly shorten the process time. 30 Photo-enhanced deposition processes.-Other methods use radiation in the ultraviolet range to enhance the film growth. In this context, photons and gas molecules directly interact with each other, resulting in the excitation and/or the dissociation of the molecules. The related increase in reactivity and/or the enhancement of the adsorption behavior largely affect the film formation. With respect to ALD, the photo-assisted ALD of tantalum oxide 19,20 and boron nitride 21 thin films using an UV lamp and an ArF laser, respectively, has been reported in literature. Compared to thermal ALD, the application of UV light has led to a higher growth rate, a broadening of the ALD window towards lower temperatures, and partially different film properties. Furthermore, the application of UV light in photo-ALD of ZnO 46,47 and TiO 2 48 films, and the laser-based atomic layer epitaxy of GaAs 49 films has been reported as well. Regarding CVD processes, the deposition of various materials by laser-driven CVD has been studied extensively. [50][51][52][53]

Process Tool
The process tool employed for our studies was custom-made by FHR Anlagenbau GmbH (Ottendorf-Okrilla, Germany). The cluster tool consists of a vacuum chamber with one process station for ALD, PECVD and FLA, respectively. At the FLA station, the flash lamp unit with its 8 xenon flash lamps is located outside the vacuum chamber. Light irradiation can reach the substrate via a quartz glass window in the top of the chamber. On the opposite side of the window, a self-made cross-flow deposition reactor was installed inside the vacuum chamber in order to enable FEALD processes. This setup is schematically illustrated in Fig. 3.
The deposition chamber is a horizontal warm-wall reactor, which exhibits a circular opening in the top. This one is required to allow the illumination of a substrate by FLA. However, to keep the reaction gases inside the reactor, and thus, to separate the volumes of the reactor and the vacuum chamber from each other, the opening is sealed using a glass wafer. One key issue is to maintain the transmittance of this window in order to enable repeatable conditions during the light exposures. For that reasons, an inert gas purging unit is integrated just below the glass wafer aiming to avoid any film deposition on the window. However, despite intense inert gas purging, the coating of this window cannot be prevented completely. For this reason, the glass wafer is utilized as a sacrificial glass window and can be replaced easily.
The deposition reactor is capable to coat substrates with sizes up to 100 mm in diameter. The reactor walls can be heated up to 150 • C, while the actual substrate temperature is achieved by heating the substrate carrier from the backside. Regarding source materials, the tool has two bubblers for liquid precursors available. Applicable reactive gases are hydrogen (H 2 ), ammonia (NH 3 ), and monosilane (SiH 4 ). Argon (Ar, purity > 99.9999%) is applied as carrier gas as well as for inert gas purging. The gas flows and gas pulse sequences are controlled by mass flow controllers and ALD valves, respectively.  The base vacuum with a pressure in the range of 10 −6 mbar is provided by a turbo molecular pump.
The tool allows FLA processes with flash pulse times between 1.8 and 20 ms and with energies up to 25 kJ. The energy for a flash is provided by charging a capacitor system with an overall capacitance up to 8 mF. In this connection, the particular electrical energy is determined by the used capacitance and the selected charging voltage. The latter one can be varied between 1.0 and 2.5 kV in order to control the supplied electrical energy. Once the xenon flash lamps are ignited, an intense light flash is generated by discharging the capacitor system. The flash duration (FWHM of the light pulse) is defined by the current combination of the capacitance and inductance of the system, and thus, the use of different flash durations requires changes in the capacitor/inductor unit of the tool. The spectral distribution of the emitted light covers mainly the visible range with maximum intensities at wavelengths of about 470 and 570 nm. 36 The relationship between the provided electrical energy and the light energy at the substrate surface has not been determined so far. For this reason, all values concerning the flash energy density refer to the applied electrical energy, and thus, specify the upper limit in case of an energy conversion factor of 1 when illuminating a 100 mm wafer.

Studies on the Flash-Enhanced ALD of Aluminum-Based Thin Films Using the Precursor Trimethylaluminum
Basic studies on the FEALD of thin films were performed by using the precursor trimethylaluminum (TMA). TMA is widely used in ALD as well as CVD for the deposition of aluminum-based films, it exhibits a high vapor pressure and a high reactivity towards oxidizing reactants. However, so far no thermal ALD process for the deposition of metallic aluminum films has been realized. 6 Thus, the intention of this study is not only to investigate fundamental characteristics in FEALD but also to demonstrate the capability of FEALD to deposit new materials using aluminum as an example.
Experimental procedure.-The processes were realized by flashing periodically on a sample during continuous precursor exposure (see process sequence in Fig. 2a). In addition, hydrogen (H 2 ) was continuously applied as well aiming to serve as reducing agent during TMA decomposition. The precursor dose was defined by the TMA partial pressure. FLA was performed with a flash pulse time of 1.8 ms and time intervals of 30 s between flashing. In all processes, the initial substrate temperature was room temperature (∼25 • C) and the total pressure was fixed to 200 Pa. Silicon was used as substrate material. In these studies, the distance between the flash lamps centerline and the substrates surface was 41 mm and energy densities up to 80 J/cm 2 were applied.
The thicknesses of the grown films were determined by spectroscopic ellipsometry right after the deposition processes using a J. A. Woollam V.A.S.E ellipsometer and a spectral measurement range from 300 -1000 nm. However, the films were aluminum oxide (Al 2 O 3 ) like films instead metallic aluminum, which we attributed to the subsequent oxidation of the material. For this reason, the measured ellipsometric data were evaluated by utilizing optical parameters for Al 2 O 3 , and thus, the obtained values reflect the thicknesses of the oxidized films only. The growth rate or rather the growth per cycle (GPC) was calculated from the determined film thickness and the number of applied process cycles without taking into consideration any delay or changes during the film growth process. The film composition was analyzed by x-ray photoelectron spectroscopy (XPS) using a Multiprobe unit from Omicron NanoTechnology GmbH and non-monochromatic Al-Kα radiation. Prior to the XPS measurements, the samples were treated with a soft sputter cleaning in order to remove organic surface contaminations. The conformality of films grown in 3D structures was studied with a Hitachi S-4700 field emission scanning electron microscope (FE-SEM).
All processes were accompanied by the successive coating of the sacrificial glass window and the related gradual decrease in effective light intensity. This again directly affected the flash-enhanced film depositions in terms of a reduced growth rate. In order to monitor this drift, we applied a reference process regularly. This reference process comprised 300 process cycles with a TMA partial pressure of 10 Pa and a light energy density of 40 J/cm 2 . While having an uncoated glass window, this process yields a growth rate of about 0.2 Å per cycle. After applying this reference process several times, we observed a decrease in growth rate of about 10% after 3000 cycles and of about 20% after 5100 cycles. Anyway, since one experimental series mostly comprised less than 2000 cycles this drift has only a little effect on the results of the corresponding series. However, there are quantitative differences between experimental series because the glass window was not replaced after each single series. As a result, the obtained characteristics are not quantitatively comparable among each other and are interpreted only qualitatively.
Results and discussion.-The influence of the light energy density on the film growth was studied in a series of processes applying energy densities up to 80 J/cm 2 , a TMA partial pressure of 10 Pa and 300 deposition cycles each. As shown in Fig. 4, the growth per cycle depends strongly on the applied flash energy and increases with increasing energy density. No film growth was observed for an energy density of 0 J/cm 2 representing a process without any flashing. Consequently, the film growth in this FEALD process is only induced by the FLA. As already mentioned in the previous section, photolytic effects can be excluded in this context, because there is no overlap between the absorption spectra of TMA molecules 51,54 and the light emission spectra of the xenon flash lamps. 36 Instead, each single FLA pulse results in rising the surface temperature above the decomposition temperature of TMA for a short period of time, leading to the thermal decomposition of molecules that are in contact with the surface during that time. The temperature for the onset of a slight decomposition of TMA is given in literature in the range between 300 55 and 332 • C, 56 decomposition has been reported for temperatures above 380 • C. 56,58 This heavily temperature-dependent decomposition behavior, which extends further up to around 500 • C, 55,57,58 largely contributes to the observed characteristic shown in Fig. 4. The higher the flash energy the higher is the surface temperature, and thus, the more pronounced is the decomposition of TMA resulting finally in a larger growth rate. Furthermore, the growth rate might be not only determined by the maximum temperature but also by the period of time which the surface temperature exceeds the decomposition temperature. In this regard, a higher maximum temperature is accompanied by a longer time period enabling the thermal decomposition of TMA. Consequently, the film growth per cycle is determined by these two superimposing effects, and thus, depends strongly on the applied flash energy as visible in Fig. 4. All these processes exhibited a deposition non-uniformity of less than 12%, and the largest GPC was always obtained in the center of the reaction chamber. This feature might be attributed to a slightly non-uniform light distribution with maximum intensities in the center or to a circular heat dissipation in the substrate or the substrate carrier. The obvious dependencies between the FLA-caused temperature and the growth per cycle indicate, that this type of film growth can be assigned to a CVD-like deposition process. On the other hand, the growth rate is in the range of less than one monolayer per cycle, which is a typical characteristic of ALD processes.
According to XPS analyses, the films consist of mainly aluminum, a considerable amount of oxygen and some carbon (Table I). In addition, depth profiles of the films showed that the oxygen content is quite constant over film depth. However, in comparison to an Al 2 O 3 film grown by thermal ALD the fraction of oxygen is significantly lower. Consequently, the aluminum to oxygen ratio is much larger than for Al 2 O 3 . The XPS spectra of the corresponding Al2p signals are shown in Fig. 5. The signal of the FEALD film exhibits a peak representing Al-Al bonds. In contrast, this peak is not present in the signal of the Al 2 O 3 film. These results, namely the considerable higher aluminum to oxygen ratio and the existence of Al-Al bonds indicate, that metallic  aluminum was deposited by the flash-enhanced process and that the material oxidized subsequently. In this regard, oxide formation most probably occurred already during the film growth because of residual oxygen in the reaction chamber and the high reactivity of aluminum towards oxygen, meaning that each freshly grown aluminum layer is directly oxidized during the deposition process. Furthermore, the films had a carbon content in the range of 4 to 6 at-% as summarized in Table I. This is most probably related to incomplete reactions and the incorporation of methyl ligands into the film. However, this feature is not necessarily a characteristic of the flash-based process, but instead might be attributed to the specific decomposition mechanism of TMA. 57 For comparison, carbon levels up to 20 at-% have been reported for a CVD process utilizing TMA and hydrogen. 59 The influence of the precursor dose on the growth rate was studied by varying the TMA partial pressure in processes with flash energy densities of 40 J/cm 2 (see Fig. 6). When increasing the TMA partial pressure, the growth per cycle increases initially and reaches a saturation level of about 0.2 Å for TMA partial pressures of 10 Pa or above. Thus, the film growth is independent on the precursor dose indicating a self-limiting growth behavior like in conventional ALD. For the following experiments, we set the working point to an energy density of 40 J/cm 2 and a TMA partial pressure of 10 Pa. Figure 7 shows the dependencies of both the film thickness and the growth per cycle on the number of cycles for various types of substrates. Here, crystalline silicon (cSi) having a native oxide served as the reference material. Besides, also silicon wafers treated with diluted hydrofluoric (HF) acid in order to remove the native oxide and to create a hydrogen-terminated surface were used. Furthermore, silicon covered with a 1000 nm thermally grown silicon oxide (SiO 2 ) and with films of 10 nm ruthenium (Ru), 10 nm tantalum nitride (TaN) and 20 nm titanium nitride (TiN) deposited by sputtering and CVD, respectively, were applied in this study. Due to the storage at ambient air, the nitride films were already partially oxidized to tantalum oxide (Ta 2 O 5 ) and titanium oxide (TiO 2 ), respectively.
There is an almost linear relationship between the film thicknesses and the number of cycles for all tested substrates, however, the obtained thicknesses were different for different substrates and a given cycle number (Fig. 7a). This behavior is even more apparent when considering the growth per cycle as illustrated in Fig. 7b. In this regard, the largest growth per cycle was observed on thermally grown SiO 2 and the lowest one on the ruthenium film. In addition, a considerable higher growth rate in the initial stage of film growth and a subsequent decline to a constant value is noticeable for all oxide surfaces. This characteristic feature indicates a substrate-enhanced growth behavior, which might be attributed to the presence of hydroxyl (-OH-) groups on the oxides surfaces, to a layer of adsorbed water molecules on the initial surface or to a gradual reduction of the underlying oxide films by the growing aluminum at elevated temperatures. However, such surface-related effects should influence the growth during the first few cycles only, but in this case the change in growth rate is more pronounced and extends over several hundreds of cycles. A reasonable explanation for this behavior is the successive change of surface reflectivity due to the growing film and the related different surface temperatures that were achieved by the FLA treatment.
The different optical characteristics of the various substrates can be identified as the reason for the differences in the growth per cycle as well. This issue was investigated more detailed by performing simulations of the temperature profile. These simulations are based on the wave transfer matrix method described by Smith et al., 60 where the given distributions as well as the resulting reflectivities are averaged over the flash lamp spectrum. Figure 8 illustrates the results of these simulations with respect to the time-dependent profiles of the surface temperatures for a blank silicon wafer, a silicon wafer covered with 1000 nm SiO 2 and a silicon wafer coated with a 10 nm ruthenium film. It can be clearly seen, that the highest temperatures are achieved for the SiO 2 covered substrate, while the lowest temperatures are obtained in case of the substrate with the thin ruthenium film. These results are directly related to the reflectivity of the corresponding materials. While around 37% of the incident light is reflected at the blank silicon surface, the SiO 2 film lowers the surface reflectivity of the substrate to about 27%. Thus, the SiO 2 enhances the coupling of light energy into the substrate which leads to a higher temperature. The higher surface temperature again resulted in the observed higher growth rate for the same reasons already discussed above. In contrast, the ruthenium layer exhibits a reflectivity of around 58%, which is remarkably higher than that of silicon. Hence, a lower amount of light energy can be absorbed and effectively contributes to the heating in this case. Consequently, the achievable surface temperatures as well as the resulting growth rates are the lowest on the ruthenium coated substrates. However, it should be noted that the temperatures shown in Fig. 8 represent the theoretical maximal achievable temperatures only, which is related to the fact, that the energy densities given in this work are the upper limit in case of an energy conversion factor only H 2 only Ar Figure 9. Growth per cycle as a function of the hydrogen fraction. The flash energy density was 40 J/cm 2 , 300 process cycles were applied and silicon was used as substrate. The TMA partial pressure and the overall pressure were 10 Pa and 200 Pa, respectively. of 1. Consequently, the real temperatures are in fact lower, but the qualitative trends obtained by these simulations remain unaffected thereof.
While the previously described FEALD processes were realized with both TMA and hydrogen, we next studied the influence of hydrogen on both the film growth and the film properties. The hydrogen was intentionally applied as reducing agent to saturate released methyl radicals resulting from the TMA decomposition in order to form methane, and thus, to prevent reactions between methyl groups and the film leading to carbon incorporation. In order to investigate the impacts of the hydrogen dose, we varied the hydrogen fraction by gradually substituting hydrogen with the corresponding amount of inert argon gas. At the same time, the TMA partial pressure and the overall pressure were kept constant at 10 Pa and 200 Pa, respectively.
As shown in Fig. 9, the growth per cycle is the largest when applying only argon and no hydrogen at all. Then the GPC decreases gradually with increasing hydrogen fraction and reaches its lowest value in the case of undiluted hydrogen. According to XPS measurements (see Table II), this effect is accompanied by an approximately 20% increase in the aluminum to oxygen ratio. These characteristics might be attributed to differences in the film formation process, in particular with respect to the incorporation of incompletely decomposed TMA adsorbate or volatile reaction products, which should lead to changes in the carbon content of the film. However, no remarkable difference in the carbon content was observed for films grown with TMA and either only argon or only hydrogen. The carbon level was around 6 at-% in both cases. Apparently, the hydrogen has no beneficial effect during TMA decomposition in terms of carbon incorporation. The observed dependency between growth rate and hydrogen fraction might be explained as the result of a thermal effect instead. Since the thermal conductivity of argon is about one order of magnitude lower than that of hydrogen, cooling via heat convection proceeds remarkably slower in argon. As a result, the surface temperature after each single FLA treatment exceeds the threshold for the thermal decomposition of TMA for a longer period of time. This again enables more TMA  molecules to be decomposed in one process cycle, and thus, results in the higher growth rate when utilizing argon instead of hydrogen. The observed differences in the oxygen content might be explained by different film densities leading to a changed oxidation behavior after exposure to ambient air. Finally, we studied the deposition conformality of this FEALD process by growing a film on a silicon sample with deep trench structures. The original structures exhibited a depth of about 6 μm and a diameter of 150 nm, and thus, an aspect ratio of about 40:1. The film was grown with 2500 process cycles in totally, yielding a film thickness of about 32 nm on the surface of this sample. As visible in the SEM cross-sectional image in Fig. 10, the film covers the structures top edge and even the slight undercut at the holes openings very conformally. Furthermore, the side walls in the structures upper part are coated quite uniform as well. This film with a thickness of about 30 nm was clearly verified by SEM down to a depth of about 1 μm. However, in larger depths, in particular at the bottom regions, the film was hardly visible by SEM. Quite obviously the film exhibits a smaller thickness in these regions. One reasonable explanation for this is the development of different temperatures during FLA in the top and the bottom region of these trenches. In the top regions, the incident light experiences multiple reflections at the trenches side walls, and hence, a larger fraction of the light is absorbed there. Because of the narrow trenches, the light is basically trapped in the top region, resulting in an enhanced heating of this region. The bottom regions of the trenches, on the contrary, experience a remarkable lower light intensity, leading to a lower heating. The resulting temperature gradient then provokes the different film growth, since the growth rate in this process largely depends on the temperature achieved during FLA as discussed before. Another reason, that might contribute to the smaller growth in the bottom regions as well, is the change of the gas composition in the trenches right during the millisecond heating. Due to reaction, the volume at the lower part of the trenches experiences a reduction in the number of TMA molecules and an enrichment with reaction by-products. The by-products might be increasingly incorporated into the growing film, leading to an enhanced blocking of adsorption sites. Although the gas volume in the trenches can rebalance in the time period between two subsequent flashes, the blocked adsorption sites would cause a lower film growth in the next cycles. Most probably, both effects contribute to the reduced growth in the bottom region of the structures. Nevertheless, the top regions have been coated quite conformally. As a result, this study has demonstrated that the film growth in flash-enhanced processes proceeds in structured substrates as well, but the possibilities in terms of overall conformality need further investigations.

Studies on the Flash-Enhanced Deposition of Ruthenium Thin Films Using the Precursor CHORUS
Ruthenium thin films grown by ALD are of huge interest for the application as metal electrodes in MIM capacitors, as gate electrode of field effect transistors and as seed layer for the electrochemical deposition of copper. However, the thermal ALD of ruthenium faces two major challenges. On the one hand, oxygen is utilized as the second reactant, 6,14,16,17 which can lead to the oxidation of the substrate or the underlying film, and thus, to the formation of an undesired parasitic interlayer. 14 Another key issue is a poor nucleation and the related delay in the initial stage of film growth on various, in particular nonmetallic, substrates. 14,16,17 To overcome these challenges, it might be attractive to employ FLA for both the realization of a single-source process and the enhancement of the nucleation behavior. Concerning this matter, our studies are based on the work of Fafard et al., who reported the co-reactant free deposition of ruthenium thin films using the so-called thermo-cyclic ALD and the ruthenium-based precursor CHORUS. 30 This ruthenium tricarbonyl derivative was designed and synthesized by Air Liquide. It is a liquid precursor with a high reactivity and a vapor pressure of more than 0.1 Torr at 23 • C. In addition, it was already successfully applied for the deposition of highly pure ruthenium films by thermal CVD. 61 Here, we report about the flash-enhanced deposition of ruthenium thin films using the precursor CHORUS.
Experimental procedure.-The precursor CHORUS was provided by Air Liquide. The precursor containing bubbler was maintained at a temperature of 35 • C and an argon gas flow of 100 sccm was utilized as carrier gas in order to deliver the precursor into the reaction chamber. The precursor pulse duration and the subsequent purge time were 1 s and 60 s, respectively. No second reactant was applied at all. Instead, FLA should enforce the thermal decomposition of the adsorbed precursor molecules. For that purpose, FLA with a flash pulse time of 1.8 ms was performed during either the precursor pulse or the purge step. In these studies, the distance between the flash lamps centerline and the substrates surface was 84 mm and energy densities up to 160 J/cm 2 were applied. The purging of the reaction volume was realized with an argon flow of 700 sccm and the overall process pressure was kept at 120 Pa. Moreover, the sacrificial glass window was replaced after 1000 cycles at the latest, in order to achieve satisfying process reproducibility.
The film growth was studied on three different types of substrates. Besides crystalline silicon (cSi) with a native oxide, silicon covered with thin films of 5 nm ruthenium (Ru) or 5 nm tantalum nitride (TaN) was applied. These films were deposited by sputtering. Due to the storage at ambient air, the TaN was already partially oxidized to tantalum oxide (Ta 2 O 5 ). The film thicknesses were determined by spectroscopic ellipsometry using a J. A. Woollam V.A.S.E ellipsometer. For verification purposes, selected samples were characterized by x-ray reflectivity (XRR) as well utilizing a Bruker D8 Discover XRR tool and Cu-Kα radiation. The film composition and film roughness were analyzed by x-ray photoelectron spectroscopy (XPS) and atomic force microscopy (AFM), respectively, using a Multiprobe unit from Omicron NanoTechnology GmbH. The resistivities were calculated from sheet resistances obtained by four point probe measurements and the respective film thicknesses.
Results and discussion.-For a first approach, we directly adopted the process sequence and main process parameters from the work of Farfard et al. 30 Thus, a process cycle consists of the precursor pulse, a subsequent purge step and the FLA treatment during the purge (see the process sequence in Fig. 2c), enabling a clear separation between precursor adsorption and thermal decomposition. The base substrate temperature was 200 • C. In our studies, FLA was performed 30 s after the precursor pulse, i.e. in the middle of the purge step, in order to rise the surface temperature above the decomposition temperature of 250 • C, 30,61 and hence, to achieve film growth by the enforced decomposition of the layer of adsorbed precursor molecules. In a first experimental series, 100 process cycles were realized with different flash energy densities up to 160 J/cm 2 at a substrate temperature of 200 • C. While no films were detected on the silicon or tantalum nitride substrates, we observed a remarkable film growth on the ruthenium substrates. However, almost the same growth was noticed in processes without any FLA treatment as well. We assume that this feature is attributed to a catalytic self-decomposition of the precursor on the ruthenium surface at 200 • C, which indeed, would destroy any self-limitation of film growth. In order to avoid this issue, we lowered the substrate temperature to 150 • C and realized the series again. But under these conditions, no film growth was observed at all. One reasonable explanation for the obtained results might be a poor adsorption behavior of the precursor at the applied substrate temperatures. Since these processes did show neither FLA-induced film growth nor ALD characteristics, they were not investigated any further.
In order to achieve flash-induced film growth, the process sequence was changed and the following studies were realized by performing FLA directly during the precursor pulses according to the sequence shown in Fig. 2b. Here, the application of FLA at different points of time during the precursor pulses was examined at first. In this regard, deposition non-uniformities of around 24% were noticed if FLA was applied in the middle of the precursor pulses. On the contrary, the application of FLA right at the end of the precursor pulses resulted in an improved deposition uniformity with non-uniformities of 13% at the maximum. This correlation might be attributed to the precursor adsorption behavior again. As a result, in the following processes FLA was always initiated at the end of the precursor pulses.
In that way, we first studied the impact of FLA on the film growth at different substrate temperatures. Thereby, each process comprised 300 cycles and for FLA the maximal available flash energy density of 160 J/cm 2 was applied. As shown in Fig. 11, there is a strong dependency between film growth and substrate temperature, the higher the substrate temperature the higher is the growth rate. However,  this characteristic is not directly attributed to the substrate temperature itself, but instead is actually related to the effective temperature achieved during the FLA treatments. This one is composed of the substrate temperature and the FLA-caused temperature raise. Since the same flash energy was applied in all processes, the different substrate temperatures resulted in different maximal temperatures. And these, as already mentioned before, are directly associated to the period of time which the surface temperature exceeds the decomposition temperature. Consequently, the higher substrate temperatures enable the thermal decomposition of precursor molecules for a longer period of time, and thus, results in a higher growth rate. Furthermore, the growth depends on the particular type of substrate and is the highest in case of silicon. This feature can be explained with the different reflectivities of the substrates, leading to different surface temperatures as already discussed in the previous section.
In case of substrate temperatures up to 150 • C, films were grown on all used substrates with the flash-based process, but no film growth was observed in processes without any flashing. Thus, the film growth is induced by the FLA only. The corresponding growth rates are between 0.11 and 0.64 Å/cycle, and therefore, are in the range reported for the thermal ALD of ruthenium films. 14,17,30 On the contrary, for substrate temperatures above 150 • C a considerable film growth is noticeable on the ruthenium substrates without any FLA as well, but no films were observed on the silicon or tantalum nitride substrates in processes without any flashing. Again, we assume that this feature is most probably related to a catalytic self-decomposition of the precursor on the ruthenium surface at these substrate temperatures. For this reason, the film growth in the flash-enhanced process at a substrate temperature of 175 • C is based on two superimposing effects, namely the flash-induced growth as well as the growth via catalytic precursor self-decomposition on the ruthenium surface. However, it is evident that FLA is required to create initial nucleation in order to achieve any growth on silicon and tantalum substrates.
XPS measurements showed that the grown films are metallic ruthenium. Impurities are oxygen with a content below 2 at-% and also carbon. The carbon concentration, however, was not reliably determinable by XPS, due to overlapping of the low intensity carbon C1s peak by the high intensity ruthenium Ru3d 3/2 peak. 62 Films with thicknesses larger than 7 nm exhibit resistivities between 100 and 400 μ cm, which is remarkable high compared to literature. Knaut et al. for example reported a resistivity of about 24 μ cm for a ruthenium film with 11 nm thickness and grown by thermal ALD. 14 The film densities are in the range between 7.4 and 8.0 g/cm 3 as determined by XRR. Figure 12 shows the measured data and simulated curves from XRR analyses of films grown with 300 cycles at substrate temperatures of 125, 150 and 175 • C. Comparing the curves, the different periodicities of the oscillations clearly represent the different film thicknesses that were already shown in Fig. 11. The thicknesses determined from these XRR measurements are also in good agreement  with the thicknesses obtained by ellipsometry. In addition, all XRR graphs exhibit a remarkable attenuation indicating high film roughnesses. This result was confirmed by AFM measurements as well. Figure 13 shows exemplary the AFM image of a film grown with 300 cycles at a substrate temperature of 150 • C. This film has a root-meansquare roughness of about 0.9 nm. However, the maximum height difference of the profile was determined to 10.5 nm, which is about half the film thickness. Comparable characteristics were obtained for all films grown by flash-enhanced processes. These large roughnesses might be related to distinctive island growth in consequence of poor nucleation in the initial stage of film growth, or to the high temperature processing due to FLA, and they might be one reasonable explanation for the relatively high resistivies.
The impact of precursor dosing on the film growth in terms of self-limiting growth behavior was investigated by varying both the precursor pulse time and the precursor temperature. When increasing the pulse time while keeping the precursor temperature constant, a saturation-like characteristic was observed (Fig. 14), but complete saturation was not achieved within the limited number of experiments. In the particular case, the non-saturating behavior might be attributed at least partially to the contribution of precursor self-decomposition at 175 • C substrate temperature. However, increasing the precursor temperature from 35 to 50 • C while keeping the pulse time constant, led to an even more remarkable increase in growth rate of around 80%. Here, the larger growth rate can be explained as consequence of the increased precursor partial pressure enabling a larger number of precursor molecules to be decomposed at the substrate surface during FLA. These results clearly show that the film growth proceeds in the transport-limited state. Taking into consideration the results above as well, this flash-enhanced process can be described more appropriate as a pulsed-CVD, whereat the film growth is reaction time limited as well as transport limited. However, further studies are needed to fully understand the process and the film growth mechanisms.
Despite the non-self-limiting growth behavior and the slight contribution of precursor self-decomposition, ruthenium films were successfully grown in a layer-by-layer manner at a substrate temperature of 175 • C. As shown in Fig. 15, the film thickness depends linearly on the number of cycles. The growth rates are 0.84 and 0.68 Å/cycle on the silicon and the ruthenium substrates, respectively. In both cases, an incubation period of about 10 cycles is determined from the present thickness data. However, this is rather small taking into account that absolutely no film growth has been observed on silicon without any FLA at all (Fig. 11).
Finally, we studied the ability of flash-enhanced processes to improve the nucleation behavior, and thus, to significantly lower substrate-inhibited film growth. Here, the main idea is to apply the flash-enhanced deposition process in order to create a seed layer, that enables further film growth by other methods like conventional thermal ALD without an incubation period. However, while the applied process tool enables flash-induced depositions, no thermal ALD process for ruthenium was realizable due to the non-availability of oxygen as the second reactant. Instead, we utilized the observed catalytic precursor self-decomposition as the largely substrate-inhibited deposition process, which did not enable the deposition of ruthenium films on bare silicon substrates at 200 • C substrate temperature (see Fig. 11). The overall process of this study was performed in two stages and the results are schematically illustrated in Fig. 16. First, a thin ruthenium layer was grown on the silicon substrates by a flashinduced process comprising 50 cycles at a substrate temperature of 150 • C. This layer exhibited a thickness of about 1 nm as measured in a preceding reference process. The lower growth rate in comparison to the results shown in Fig. 11 is attributed in this case to an already partially coated glass window and the related lower effective light energy. After growing the seed layer, we increased the substrate temperature to 200 • C. Subsequently, 300 cycles without FLA were applied to grow the film via catalytic precursor self-decomposition at 200 • C substrate temperature. In doing so, the growth of about further 4 nm ruthenium was observed. According to Fig. 11, this film growth is related to the pre-deposited seed layer, since no growth would occur on the silicon substrates otherwise (Fig. 16). The successful realization of this approach clearly demonstrates that flash-enhanced deposition processes offer the potential to significantly enhance the nucleation behavior, and thus, are a promising method to overcome the substrate-inhibited growth in ALD. Future studies, however, should also focus on the possibilities of flash-enhanced processes to increase the nucleation density, and thereby, to obtain smoother films.

Conclusions
In this work, flash-enhanced ALD was presented as a promising approach for thin film deposition. Here, flash lamp annealing is applied to thermally enhance the film growth in ALD. Thereby, the FLA treatment in each process cycle raises the surface temperature for a short time, enabling the thermal decomposition of adsorbed precursor molecules or the activation of chemical reactions between adsorbed precursor molecules and another reactant. Various possible process sequences were presented and previously reported work concerning this matter was summarized. The processes described in literature were mostly realized by performing FLA during the precursor exposure.
Our studies on the FEALD using the precursor TMA were realized by flashing periodically during the precursor exposure as well. The studies revealed that the film growth was only induced by the FLA and proceeded in a self-limiting manner. Although the films were Al 2 O 3 -like films, XPS results indicated that metallic aluminum was deposited, but the material oxidized most probably due to residual oxygen already during the deposition process. The growth per cycle depended strongly on the applied substrates, which we attributed mainly to their different optical characteristics leading to different temperature profiles during the FLA. Furthermore, we demonstrated that the film growth proceeds in structured substrates as well.
The flash-enhanced deposition of ruthenium films using the precursor CHORUS was successfully realized as a single-source process by flashing during the precursor pulses. The film growth was induced by FLA again. However, for substrate temperatures above 150 • C we observed that catalytic precursor self-decomposition contributed to the film growth as well. The studies revealed that film growth proceeded in the transport-limited state, and hence, these processes were described more appropriate as pulsed-CVD-like processes. Despite this characteristic, ruthenium films were grown layer-by-layer with a linear relation between film thickness and number of cycles, and with a short incubation period on both silicon and ruthenium substrates. The grown films were metallic ruthenium and exhibited a minor impurity content, but had high resistivities and high roughnesses. Finally, FLAinduced deposition was successfully applied to significantly enhance the nucleation behavior.
While these studies directly presented the capabilities of flashenhanced ALD to grow new materials, to realize single-source processes and to overcome substrate-inhibited film growth, further promising potentials are the deposition of films on temperature sensitive substrates and the deposition of films with improved properties. Altogether, the application of FLA in a deposition process offers a high potential to considerably extend the capabilities of thermal ALD.
On the other hand, there are some key challenges for the practical realization of flash-enhanced deposition processes. Most importantly, the coating of the glass window must be prevented in order to achieve process reproducibility. The substrate-dependent growth rates arising from different optical characteristics and the related different temperature profiles might be another issue in particular when coating patterned substrates. On the other hand, this effect might be also a promising opportunity for the realization of area-selective ALD.