The quantification of thin film atomic layer deposition conformality in high aspect ratio nanostructures
The integrated microcircuit technology  is the driving force behind the development of modern precision technological processes. As these processes dynamically develop and the technologies become more complicated, the technique and equipment for thin film materials fabrication of different nature (dielectrics, semiconductors, metals, etc.) have been intensively developed . In integrated circuit technologies (ICT) thin films (TF) are the basic construction material to prepare insulated transistor structures, capacitors, resistors, multilevel metal coating system and final etching. Thin films have been prepared on substrates of different types heated to the required temperature, mainly by chemical vapor deposition method as a result of irreversible reactions of initial vaporous substances. Reactions at CVD process at depressed temperature levels could be activated by plasma discharge. Main trends of CVD methods development applied to ICT from the 1970s till present time have been considered . Particularly, the time stages of CVD methods development and tasks were identified and directions of the development, research results and TF properties for each one were generalized. At first stage the main task of CVD methods was to provide uniform thickness coatings on an almost flat surface of Si-substrates, more and more increasing in size. Starting from the second stage the IST development began with the formation of ICT elements on semiconductor substrates (see Fig.1). It is possible to distinguish steps (see Fig.2a) and grooves (holes) (see Fig.2b) as typical simplified structures in reliefs of 3D integrated micro circuits (IMC). Such simplified structures can be characterized using the value known as "aspect ratio" (AR). In general, aspect ratio is the ratio of height or depth of the groove (H, µm ) to its width or diameter (G, µm).
At the third stage of the IMC technologies development (in the middle of 1990s) the problem of "voids-free" quality filling of gapes on grooves of IMC relief (relevant for sub-micron IMC) has emerged. The problem known as "voiding" is that in case of irregular TF growth the unacceptable for IMC technologies voids appear (see Fig.2b). Moreover, problems described above are compounded due to development and complication of modern IMC technologies, which lead to appearance of reliefs with high values of aspect ratio (AR). Obviously, filling of rectangular cross section reliefs without voids is possible only in 100% of conformity value of TF growth on all structure surfaces.
Different ways of solving problems of conformity and filling of relief’s voids at CVD at this stage were discussed by author with co-authors in [3–6]. The main idea of the author’s empirical method of summary of results of TF investigations is to find out quantitative interrelations between TF conformity and structure characteristics and conditions of CVD processes. Analyzed were about 200 original publications from different sources in order to select a quantitative data (values of conformity, AR and G) or photos of TF structures with scale which allows such estimates to be made. The author’s own research at the time was made on test structures of sub-micron IMC with height of steps of the order of a micrometer. Studies have been carried for different CVD processes for TF mostly based on silicon dioxide. We characterized deposition processes quantitatively using "effective constant of growth rate" for CVD proposed by author, called keff,. The expression for keff (dimension is cm/s) is determined from keff = χ×W/[Si], where W – rate of TF growth, [Si] – concentration of Si-component in the gas phase, χ – proportionality factor.
See detailed explanations concerning these expressions in . Values of keff for known CVD processes in case of silicon-based TF have shown the difference of more than three orders of value . It was proved by the author that quantitative interrelation between conformity of deposition and keff may be presented as: d2/d1(%) ~ 45,4×keff-0,51. Slow processes with rate-controlling heterogeneous stages (small keff) allow of distinguishing the maximum, close to 100% value of TF deposition conformity. On the contrary, the activated high-rate CVD processes (high keff) present the unacceptable conformity of TF.
We have treated the experimental data for voids in the vertical rectangular grooves of IMC testing structures with H~0,25…1,1 µm and G~0,1…1,1 µm using SEM. Critical points (ARcrit, Gcrit) on curve in coordinates AR – G [4–6] have been found (see an example on Fig.3). When analyzing the structures with different values of H plotting of similar graphs can help to determine the critical points set and gives the possibility to draw straight lines through these points. So, the fields under straight lines correspond to conformity deposition of TF, above – nonconformity deposition on the studied relief. In this case, the tangent of the angle of inclination of the straight lines reflects the ability of the investigated CVD process for TF to fill the relief of IMC with known characteristics under specific experimental conditions.
We proposed to call tangent of the angle of inclination as a "structure complexity" parameter (Structure Complexity, SCcrit = ARcrit/Gcrit = Hcrit/Gcrit2, dimension is µm-1). As a result of the analysis and generalizations, it was possible to relate IMC structure complexity with keff of CVD process according to the following expression: SCcrit ~ 3,03×keff-1,47. It is important that SC parameter can be applied to characterize structure complexity of initial structures (before TF deposition). Indeed, if H, G and AR values are known, the structure complexity can be expressed as SCstruct. Correlation of SCstruct value and the experimentally obtained SCcrit allows of comparing the tasks and real possibilities of the studied CVD process. Unfortunately, plotting graphs similar to those shown in Fig.3 is quite laborious and expensive, both in terms of producing structures for research and in terms of the need to use SEM of high resolution. In author’s opinion, in order to make an estimation of CVD methods possibilities it is permissible to operate with SCcrit value obtained experimentally for one point. Further development of IMC technologies led to the emergence of reliefs with deep grooves, reduced linear dimensions G (down to 0.02 μm) and, thus, with much larger (up to 100) aspect ratios (see the analysis in ). The unconformity of TF in such complicated structures during CVD process have led to appearance of voids inside of these structures (see Fig.2b, i.e. SCcrit < SCsrtuct). To solve these problems in the early 2000s, the researched began to actively explore and use the method of so-called "thermally activated atomic layer deposition" (ALD) . The advantage of the method, strictly speaking, which is a variant of CVD, makes it possible to localize the chemical reaction of TF formation strictly on the surface of heated objects. Localization takes place in a narrow temperature range, called "ALD-window". The use of ALD has allowed for some types of TF and for some period of time to solve problems when working with structures of increased complexity, but with the further development of IMC technologies and complication of structures, the relevance of the problem is still present.
METHODS OF THIN FILMS CONFORMITY CONTROL AND FORMULATION OF THE PROBLEM
Most authors use testing structures like "groove" (or "hole") prepared by conventional microelectronic technologies (see Fig.2b) to study the TF conformity (description and methodology of the analysis see in [2,4–6]). Verticality of the walls of the structures is very important for analysis, because it is impossible to obtain conformal coating for grooves with a backward slope, and in grooves with inclined walls, the conformity of TF deposition is very high. However, due to a difference in the wall slopes of the original structure relief as reported by different authors, it is difficult to classify the data for structures with inclined walls in practice.
Developing this approach [4–7], the author and co-authors in  proposed a method for analyzing conformity of deposition for the CVD and ALD processes, taking the SCcrit parameter as the basis. It was proposed at the time to plot the experimental data presented in a few publications as a chart with coordinates d2/d1 – SC. We have calculated examples of different TF prepared by ALD and have shown the result in Fig.4a. The obtained curves prove that structures with SCcrit ~ 20…100 have the conformity of 100%. However, it was discovered, that in case of metallic ruthenium obtained with CVD process at some values of structure complexity the systematic conformity declined (see Fig.4a). The point of deviation of the conformity value from 100% corresponds, in a sense, to the critical values of SC in Fig.3, and was also designated as SCcrit. However, systematic data on the reduction of conformity for TF ALD on complex structures at that time was not found in the literature. By present time there were many studies of ALD processes for various materials and their properties. ALD methods are improved, for example, direct or remote plasma is used to activate one of the reagents, sequences of oxidation and reduction components of the reaction are used, surface treatments are proposed to improve nucleation of the deposited material on the surface, etc. The total number of publications exceeds several thousand; there is a number of generalizations on ALD (see, for example, ).
Along with that, publications dedicated to investigation of TF conformity obtained by ALD method are few. Usually, authors supply data about TF conformity as a part of other information and present it as demo slides of the photos obtained using scanning electron microscopy (SEM), schematically shown in Fig.2b. As a rule, a thickness of TF in upper and lower parts of grooves or holes for such structures have been indicated. In some cases a data for a few groove depths conformity had been given for upper, central or lower parts of the structures only. Authors of these publications conclude that AR value for the structures of certain size (G) where deposition conformity is equal to 100% (or not) has been achieved. Only in  the profile of metal platinum TF thickness distribution in a vertical hole of 3.8 μm and dia. of 0.023 μm has been drawn as a point-to-point graph. It was shown that deposition close to conformity with the use of ALD method may be obtained up to the groove depth of ~0.5 μm, after that a TF thickness rapidly decreases.
Vertical nanoscale high aspect structures are difficult to produce and analyze and, due to this fact, very expensive, although, it corresponds well for real device structures. Recently proposed and applied practically were so called flat "horizontal" testing structures (see Fig.2d). These "gap" testing structures (hereinafter called microstructures) may have linear dimensions H up to some hundred micrometers [12,13] and G about 0.5 μm. In  testing structures with dimensions of up to tens millimeters and gaps about some millimeters (hereinafter called macrostructures) were described. Flat horizontal structures present a base with supports and a cover. After TF deposition a cover is removed and distribution of TF thickness in a bottom of the structure (which is, practically, the analog of a vertical groove wall) can be controlled both visually and by conventional control methods. As a result, it is possible to register the profiles of thickness along the wall (see an example in a Fig.5b). The final aim of horizontal gap structures is the attempt to predict TF growth regularities in nanostructures.
Presently the experimental data for analysis and generalization of the results pertaining to conformity problems connected with coatings of high aspect structures are being collected. While analyzing TF conformity at ALD process, the one of the most important goals is the adequate interpretation of the experimental data for different types of thin films and under various experimental conditions on diverse types of testing structures.
We set the problem of finding simple quantitative approaches applicable to analysis of the experimental results for quite different experimental conditions. The supplied material presents a development of the earlier time-tested approach to analysis of TF conformity on the IMC reliefs ALD .
DESCRIPTION OF THE PROPOSED APPROACH AND INTERPRETATION OF THE RESULTSIn high aspect structures prepared by ALD method the grooves and holes are filled as it is shown in Fig.2b, where voids are formed inside these structures at non-conformity deposition. For some practical goals it is possible, but in the majority of cases it is inadmissible. To develop the approach to analyzing SCcrit values (see Fig.4a), the data presented in publications for such high aspect grooves (holes) can be re-calculated based on the known data of AR and G, and SC values. Examples of such re-calculation in d2/d1(%) – SC coordinates are shown in Fig.4b. The data for metal ruthenium are shown as separate dots (it should be remarked that different precursors of ruthenium and ALD conditions  were used). It was shown that conformity in some cases is definitely lower than the desirable one (100%). In that case we cannot to determine the SCcrit value using experimental dots with d2/d1 <100% in order to achieve the 100% real deposition of TF conformity using the ALD method.
When authors present more interesting cases concerning the data on conformity values for groove depths, the re-calculation makes it possible to see real trends and to determine the SCcrit values. Indeed, the profiles of depth in grooves/holes can be re-calculated and presented as the dependence of d2/d1(%) – SC, (see Fig.5b). When we normalize the current value of TF thickness to the zero coordinate (the beginning of a groove, starting from reagents input), we obtain the conformity parameter in percentage. Afterwards the current coordinate (i) of the vertical groove depth or point in a thickness scans for horizontal microstructures (in fact, Hi) may be easily transformed into ARi and SCi values for known G value. The results of this re-calculation are shown in a Fig.4b for four different TF materials as curves which indicate values of SCcrit. The curves of Al2O3, TiO2 and metal iridium were obtained using flat structures. It can be seen that the maximum values for them are SCcrit~200 μm-1. It is interesting to note that, according to Fig.4b, the maximum possible structures of SCcrit~1100 μm-1 were obtained when ozone was used as the second reagent , which in recent years has been considered a promising reagent for the production of metal TF in ALD process. Generally, trends in Fig.4b correspond to the data shown in Fig.4a for TF of metal ruthenium obtained by CVD method and published in . Based on these data, the earlier proposed approach for analysis of conformity TF growth on the reliefs using dependencies like d2/d1(%) – SC seems to be consistent for ALD process in case of thin films. This approach makes it possible to move from specific testing structure sizes to the unified value of their complexity (SCstruct) and estimate their possibility to be filled with TF materials for any conformity value (SCcrit) and for any size of microstructures. Currently, there are few experimental data for nanoscale structures, and an unambiguous assessment of the applicability of the proposed approach to their description is still to be done in the future. However, as the grooves of the TF are filled, their size G decreases down to zero, that is, until the closing fronts of the TF are fully closed, these effects are also present at the level of microstructures. In this regard, the author suggests a possibility and adequacy of the application of the proposed methodology for nanostructures.
As concerns macrostructures, their data processing , according to the proposed method, gives similar trends, but the absolute values of the complexity of such structures turn out to be 4–6 orders of magnitude less than for micro and nanostructures. In this connection, the question of applicability of the results for such structures in predicting the conformity of nanostructures remains open. At the same time, such easily manufactured and cheap structures turn out to be very convenient for tracking the growth trends of TF in gap structures, including for express control of processes.
The data presented in the paper and the published data make it possible to delineate the areas of conformity deposition of TF using different methods of deposition from the gas phase, see Fig.6. Inclined sections of fields on the right are conditional and indicate the falling trend of conformity in general when SCstruct > SCcrit. The data presented unambiguously prove the advantages of ALD methods from the viewpoint of the obtained TF conformity in high aspect nanoscale structures with maximum SCcrit values. The proposed approach, though based on singular published data, allows of a comparative assessment of the parameters for filling structures with one or another method of obtaining TF coatings for initial structures with known parameters. The approach proposed by the author enables to adequately assess and numerically compare the growth conformity of thin films for high-aspect nanoscale structures of varying complexity, for various options and modes of thin film production using the ALD method. ■