Features of control of silicon nanostructures technology
• accuracy and reproducibility of the result;
• efficiency of obtaining the result;
• high performance;
• high degree of automation;
• safety of application of the methodology and equipment on which it is implemented;
• high degree of reliability of the result obtained.
These requirements impose limitations on both the techniques used in the operational process control and the equipment on which they are implemented. In the production of nanoscale products to meet these requirements, it is necessary to take into account the following circumstances:
• changes in the mechanical properties of objects and structures during control operations, as well as the effect of surfaces and interphase boundary, changes in the structural properties that were considered in the paper ;
• influence of diffusion processes, blurring of material interfaces;
• impact of boundary regions of a nanoscale object;
• reduction of the degree of impact on the sample during the control measures;
• increasing the complexity of the result obtained, in the sense of increasing the number of simultaneously determined parameters.
The need to take into account such a wide range of factors in the construction of technological control operations forces us to develop new control methods, since the conventional methods, even those that are well developed, cease to meet the requirements, which change as technology develops.
CONTROL OF THIN FILM PROCESSES
As a first example, let us consider the ellipsometry technique , classically used for technological control of thin film thicknesses. This technique has a number of important qualities, including: high accuracy and reproducibility of the result, high performance, low impact on the sample. However, with the development of technology, for some processes even ellipsometry ceases to meet the requirements for control operations.
Let's consider the process of the gate dielectric control of the MOSFET and the changes that occur with the development of VLSI production technology. Historically, silicon oxide was used as the gate dielectric in CMOS technology. As the technology developed, to reduce the length of the shutter channel, it was also necessary to reduce the thickness of the dielectric layer. When its required thickness approached 1 nm, the growth of parasitic effects made it necessary to switch to new materials. Among these effects, one can distinguish, firstly, the depletion of the gate, the tunneling of carriers into the shutter, which leads to an increase in the leakage current, and, secondly, the diffusion of the dopant (such as boron ) from the gate to the channel, which also leads to degradation of properties of the IC. Therefore, starting from the 90 nm technological level, nitrogen was introduced into the gate insulator in small amounts (about 10% by weight). This allows both to prevent diffusion of boron  and to increase the dielectric constant of the film. At subsequent levels of CMOS technology, the transition to so-called high-k materials began, such as hafnium silicate and hafnium oxide. Thus, the control of the processes of obtaining a gate dielectric must include not only the measurement of the thicknesses of these films, but also the determination of their composition, which can be achieved with the use of an integrated approach involving simultaneous control of the thickness and refractive index of the films.
With the help of classical laser ellipsometers with a rotating polarizer, analyzer or compensator, traditionally used for control of thin films with sub-angstrom reproducibility , it is only possible to measure the optical thickness of films, that is, the product of the thickness by the refractive index. This problem has traditionally been solved with the help of a transition to spectral or angular ellipsometry [5, 6], which allows to register ellipsometric values for the spectral range or angles of incidence, respectively, and thus allows simultaneously to determine more than one independent parameter, that is, the film thickness, and its optical properties.
However, for such processes as the formation of a gate insulator in which the allowable technological spread of the layer thickness should not exceed 1–2 angstroms, the method of spectral ellipsometry (SE)  with simultaneous control of thickness and optical properties is insufficiently accurate. This is due to the fact that recording the values of ellipsometric quantities even for a wide spectral region does not completely eliminate mutual correlations of the parameters when finding the result by numerical methods, and, therefore, does not allow to eliminate ambiguity completely, since the values of ellipsometric angles for different parts of the spectrum are not completely independent quantities.
The use of complex measurement methods, where several separate techniques complement each other in order to find a reliable result, allows such ambiguities to be eliminated. Such a control process for high-k gate dielectrics based on a combination of several optical techniques in a single data processing cycle is considered in . The problem of the exact experimental determination of the optical properties of thin films over a wide range of the spectrum (190–800 nm) using such techniques as ellipsometry, spectrophotometry and scatterometry to objects of modern technology is discussed in [9, 10]. Traditionally, the SE method [11, 12] is used for this purpose, however, when the absolute value of the thickness of the controlled layers decreases less than 10 nm, the use of cross-checking is necessary, since the accuracy of the SE is insufficient . The value found is then used in ellipsometry to determine the optical properties of the film. X-ray reflectometry (XRR) [13, 14], which has excellent sensitivity to thin films, is non-destructive and does not require calibration to external standards, is proposed as such a cross-checking method by independent teams of the authors of [8, 9].
Such a complementary principle for the construction of measuring techniques can also be used in the control of multicomponent films and films of complex composition. For example, one can consider the control of the obtaining films of fluorine-containing silicate glass. In this case, simultaneous measurement of the refractive index is necessary to control the thickness with satisfactory accuracy, since the optical properties of this film, determined by its composition (namely, by the fluorine content), can change with successive implementation of the deposition. For simultaneous determination of the thickness and optical properties of these films, the SE technique was used . The addition of X-ray fluorescence spectroscopy (XRFS) to this control scheme allows to reveal the dependence presented in Fig.1. Thus, it was possible to use the SE technique for the rapid evaluation of the composition of these films by changing their optical characteristics.
The presented dependence also makes it possible to increase the accuracy of determining the thickness of such films. In this case, the XRFS determines the concentration of fluorine in the film, which is then recalculated into the refractive index used in the SE as a fixed value. This allows to exclude the influence of the correlation of "optical properties – thickness", and to increase the accuracy of determining the thickness by 2–3 times.
CONTROL OF LATERAL SIZES
OF FORMED ELEMENTS
The next example may be the control of the lateral dimensions of the elements obtained by photolithography and etching. Traditionally, for this purpose, scanning electron microscopy (SEM) was used. With the development of photolithography, the transition to exposure at 193 nm using excimer lasers (ArF) was performed to increase the resolution of the structure formation processes. This entailed the development of new photoresists with the corresponding properties (the so-called "chemically-enhanced photoresists"). However, when measuring the dimensions of photoresist lines or gaps between such lines by SEM methods, the effect of "burning" is observed, which consists in decreasing the observed line size or increasing the gap as a result of probing the sample. According to modelling, the reduction in the size of photoresist line is proportional to the dose of irradiation of the resist by electrons. For 193 nm resistors, with a single scan of a photoresist line with a nominal size of about 100 nm, this decrease reaches 3–10 nm  in one measurement. The effect is affected by the magnification and accelerating voltage, but it is observed even at such low values of accelerating voltages as 300 V . Further reduction of the accelerating voltage in the SEM leads to a sharp degradation of the image contrast . Thus, the measuring method itself makes an error in the measurement result.
For 120 nm CMOS technology, the aspect ratio of the thickness and length of the gate of the transistor exceeded the threshold value, which was equal to one. For this technological level, the required reproducibility of measurement of critical dimensions is 1.8 nm . In the structure of the gate having an aspect ratio of 1:1, the deviation of the angle of the wall by 1° will lead to a change in the observed size, measured from the top and bottom of the shutter, by more than 10% of the permissible deviation of the process (Fig.2). Thus, it is necessary not only to measure the critical size of the obtained shutter with the required reproducibility, but also to know precisely at what height the measurement takes place. To this end, a technique is required that makes it possible to measure the overall profile of the resulting three-dimensional structure of elements, including the determination of their lateral dimensions, thicknesses, and the angle of inclination of the walls. This can be achieved with the transition to diffraction optical techniques for measuring critical dimensions, such as scatterometry  (Fig.3 and 4).
In the practice of modern microelectronics, optical scatterometry is used as a standard method for controlling the formation of three-dimensional structures. In particular, scatterometry is actively used to control critical dimensions in photolithographic  and etching processes [17, 19]. This technique allows one to analyze diffraction patterns from periodic structures in order to restore their spatial profiles. In this case, it is possible not only to measure the lateral dimensions of periodic structures, but also the depth of the grooves of etching, the slope of their walls. In the world practice, the scatterometry has become widely used in the transition to small design rules [19, 21], since it allows the replacement of several traditional means of control, and also has high performance and accuracy . The technique is successfully applied at technological levels up to 20 nm [20, 21], and the effect of the roughness of the element boundaries on the result has been studied [18, 22] and can be taken into account.
However, the successful introduction of scatterometry in each technological case requires a comprehensive study of the structure by cross methods. Such a study, which made it possible to introduce this technique to control the etching of gap insulation, was described in detail in . The scatterometer used in this study can be used to measure critical dimensions up to a design rule of 45 nm. This is due to the fact that its spectral range of 250–800 nm is limited by the source used (xenon lamp), as well as by the transmission characteristics of optical elements. The technique of optical scatterometry for smaller design rules can be realized by switching to a part of the spectrum with a shorter wavelength using a deuterium source having a typical spectral range of 150–400 nm. Such an installation will allow monitoring of critical dimensions down to 22 nm .
Further shift of the spectral range of optical scatterometers is problematic, since it will require vacuumation of the optical path, the use of ultrapure nitrogen for its constant purging, or other technical solutions that will lead to a sharp increase in the cost of the equipment and the cost of its operation. Therefore, leading world manufacturers of metrological equipment are developing X-ray devices for measuring critical dimensions. An example would be a system for measuring critical dimensions by analyzing critical-dimension small-angle X-ray scattering (CD-SAXS), which is currently being developed by the Japanese company Rigaku . This unit is designed for use at design rules of 10 nm or lower. Another group of researchers is currently developing a scatterometer operating on focused ion beams (FIB) . This approach allows us to determine not only the spatial geometry of the periodic structure under study, but also the distribution profile of the chemical elements in it. This is necessary for the operational control of the distribution of doping impurities on the vertical walls of the structure, which is important in modern CMOS technology in the formation of so-called FinFET transistor structures.
The first experiments described in [24, 25] indicate that the control of both the smallest and the other IC elements requires a combination of optical and X-ray or ionic techniques, the possibilities and limitations of which must be investigated additionally. In this connection, it should be recalled that it is necessary to reduce the degree of influence on the sample, with which we began our examination of the optical scatterometry. When passing in scatherometry to the X-ray range, it is necessary to take into account many factors that may make application of the technique in the production difficult or impossible. It is necessary to take into account the radiation defects, the negative effect on the photoresist and the underlying layers. At the transition to the FIB, the nomenclature of parasitic effects is further increased. The effect of FIB on the material structure was discussed at a recent ConFab-2017 conference  and in one of the latest issues of SolidState Technology journal , and was also considered in recent works by both foreign authors  and us .
CONTROL OF BLURRING
An important feature that must be taken into account when controlling nanoscale structures is the blurring of the interfaces of the material that occurs when they are interdiffused. When the absolute values of the element sizes decrease, the relative contribution of these effects increases. And in some cases, they can lead to a complete loss of functional properties of the structure and, as a consequence, to the complete failure of the product. An example of such a case may be the diffusion-barrier structure used in the technology of copper metallization – the Damascene process. It uses a double structure based on tantalum and its nitride. The properties of these DBC structures, as well as the reasons for choosing them for copper metallization, are described in detail in [30, 31] and other works. For our purposes, we only need to mention the functional purpose of each of the layers in this structure: TaN provides adhesion to the interlayer dielectric (both to oxide and to low-k dielectrics), has satisfactory diffusion-barrier properties, and also determines the growth of the alpha phase (cubic) of Ta, which has a lower contact resistance compared to the beta (tetragonal) phase; tantalum acts as the basic DBC layer, provides good adhesion of copper and, in addition, the necessary crystallographic orientation of the subsequently deposited seed layer of copper. The layers of TaN and Ta can be deposited in one chamber, which makes the use of this DBC structure economically justified. A typical structure of copper metallization, in which DBC on the basis of tantalum and its nitride is used, is shown in Fig.5.
At present, interferometric (ellipsometry), X-ray (XRR, XRFS) and photoacoustic techniques  are used for the operational control of the processes of obtaining this structure, which do not allow us to determine the separate thickness values of the Ta and TaN layers for the following reasons:
• similarity of optical properties – for ellipsometry;
• blurred interface – for XRR;
• low sensitivity to nitrogen – for XRFS;
• similarity of acoustic impendence values – for photoacoustics.
However, the thickness of the tantalum layer is a critically important parameter, since it determines the diffusion-barrier properties of the structure.
These circumstances led to the fact that the common practice is to measure in the process cycle only the overall thickness of the double structure. The thicknesses of the individual layers are obtained indirectly – from measurements on the satellite wafers. For this, two additional wafers are used: one contains a film of tantalum nitride, the other – of tantalum deposited in the same modes as for obtaining each of the layers in the integrated structure. Such a method does not allow taking into account the blurring of the interface, which inevitably occurs when an integrated structure is obtained. As the technology develops, the thickness of the layers in the DBC structures decreases. For example, for design rules below 65 nm, the total thickness of the diffusion barrier structure should be less than 5 nm . This makes it necessary to take into account the interface between materials.
The team, with the participation of the authors of this paper, proposed a technique that makes it possible to obtain separate values for the thicknesses of Ta and TaN films. It involves the combined use of XRR and small angle X-ray scattering (SAXS), which, in order to find solutions, are combined into a single data processing cycle. This technique is described in detail in . It made it possible to obtain a distribution profile of the electron density along the depth of the sample, and also to establish how it changes in this sample. The latter was done thanks to SAXS, the addition of which allowed to resolve the ambiguity "density-roughness" [35, 36] and, ultimately, to reveal the fact of diffusion of nitrogen from the tantalum nitride film into the tantalum film. This phenomenon should be taken into account in the design of devices, as it leads to the degradation of the barrier properties of the structure. Inadequate barrier properties of such structures can lead to a decrease in the reliability of IC up to the total loss of consumer qualities.
The paper considers the process of control of the production of nanoscale products of solid-state electronics and it is shown that in many cases a revision of existing control methods is required to satisfy the requirements for this process. In order to increase the reliability of the results obtained, in some cases a transition to complex control techniques is required, implying a complementary use of several different methods. Sometimes this approach is the only one possible to obtain the desired result. If when controlling processes, for example, of submicron technology, you could choose from the existing methods one that is optimal for the given case, in modern technology this can often not be done, since such a technique simply does not exist.
The paper shows the need to take into account the specific phenomena inherent in the transition to the production of nanoscale elements and structures. Among them, firstly, we should mention the change in mechanical and structural properties of nano-objects, as well as diffusion phenomena and surface effects. With the development of technology in the direction of reducing the size of the elements produced, these effects begin to play a decisive role in the properties of the structures obtained and, consequently, they must be taken into account when organizing the process control.
The development of technology and the movement of design standards in the direction of reduction, as has been shown repeatedly, are associated with a fundamental change in the properties of materials and structures, which was the subject of discussion in . Effective process control is based on taking these changes into account. Some of them should be specially noted as having a special impact on the process control and the need to develop new tools.
First of all, it is necessary to take into account the change in the structure of the material and the transition to a new state, in particular quasi-liquid. Recently, IBM introduced a picture of a transistor structure manufactured using 7 nm process  (Fig.7). The recent information shows that at such sizes the structure appears to be completely ordered amorphous, or, according to the existing terminology, quasi-liquid, and its mechanical properties differ substantially from the properties of the single-crystal silicon known to us.
At the same time, at the recent conferences  and in the published papers [27, 28], the problems connected with the influence of the measuring techniques themselves on the structure of nano-sized objects are considered in detail. These effects must be carefully taken into account when constructing a process control system to produce reliable measurement results and, as a consequence, have a high percentage of yield.
And, finally, the technological possibilities for the formation of small-sized (nano-sized) electronic or electronic-optical structures require the use of new processes and new measurement techniques. In modern literature, samples of such structures are tried to obtain using such well-controlled and in some cases unique methods, as molecular-beam epitaxy. However, in real mass production, this technology is of little use and may be inferior to well-known methods, in particular, radiation technology and, above all, ion implantation. As an example, let's look at the history of the development of technology for creating light-emitting structures based on SiGe quantum dots. Many published studies on the obtaining of such structures have been carried out using molecular beam epitaxy. At the same time, we are deeply convinced that the mass production of such devices can and must be based on the use of ion implantation (ionic synthesis), as has been repeatedly shown in our works [38–41]. ■
The authors thank D.Smirnov for taking part in measurements by X-ray methods and for discussing the results obtained.