RU2175761C2 - Method for measuring surface relief by means of scanning probe type microscope - Google Patents

Abstract

FIELD: electronic measuring technique, namely probe type scanning microscope. SUBSTANCE: method is realized due to use of surface features as reference points at making motions. Motions are performed from one reference point to another adjacent reference point in order to form chain of features arranged one relative to another. Identifying program performs searching, detection and calculation of position coordinates of feature (reference point). Scanning of comparatively small area around each feature and placing of corresponding fragments of surface according to respective positions allow to construct actual relief of surface. Information related to position coordinates of features and their arrangement system allow to perform precision positioning of probe. EFFECT: enhanced accuracy and linearity at measuring surface relief, improved precision positioning of probe and resolution of instrument. 3 cl, 15 dwg

Description

 The invention relates to electronic measuring equipment and is intended for use in a probe scanning device both for the purpose of high-precision measurement of the surface topography and the precision positioning of the working body — the probe — when performing technological actions in the processes of nanolithography. The basis of the microscope are a scanning tunneling microscope (STM); atomic force microscope (AFM); near-field scanning optical microscope; scanning electron microscope, etc.

 Known methods for the precision positioning of the probe used in scanning tunneling and atomic force microscopes [1, 2, 3].

 Despite the high resolution achieved by scanning probe microscopes (SPM), the latter cannot be used as a metrological tool or as a nanolithograph without serious modernization, since the results of the measurements are negatively affected by thermal drifts of the mechanical and electrical parts of the device, residual non-orthogonality, and spurious relationships between manipulators (coupling), as well as hysteresis and creep characteristic of piezoceramics.

 To date, many methods have been developed to get rid of the above distorting factors, but methods [1, 2, 3] in which dynamic dynamic correction of the operation of the piezoelectric manipulators of the microscope are performed are suitable for solving the problems of interest to us. Correction is carried out by introducing into the control circuit of the device three closed servo systems - one for each manipulator. During operation, the tracking system seeks to compensate for the difference that occurs when the manipulator moves between the specified amount of movement and the actual measured using a linear sensor.

 As analysis shows, the methods used in the analogues of the invention [1, 2, 3], fundamentally no different from each other. The use of different types of sensors and the associated framing electronics determines the sensitivity of the measuring system, its linearity and the noise-free minimum step size of the manipulator. Thus, a miniature capacitive sensor is used in system [1], an optical one in system [2], and a combination of a tunneling current sensor and software recognition of elements of the reference surface in system [3]. The optical sensor is a pair: a light-emitting diode-two-section photodetector mounted on the base of the device and separated by a screen with a narrow slit that is mounted on the manipulator. The tunnel sensor consists of a conductive surface, a sharp needle, a piezomanipulator and a current stabilization system. The rms noise value measured in the frequency band of several kHz is 10 nm for the system [1], 0.6 nm for the system [2] and 0.01 nm for [3].

 As a prototype of the invention, the scanning and positioning method used in a probe microscope [3]. In this method, each device manipulator is mechanically connected to a tunnel current sensor manipulator. The surface of highly oriented pyrolytic graphite (HOPG) is used as the conductive surface of the sensor. HOPG is convenient in that, due to its physicochemical properties, it is able to keep the surface clean for a long time, easily chops off, has extended atomically smooth sections, and most importantly a stable and highly ordered crystal lattice.

угол между кристаллографическими направлениями: 60^o). Выполняя программное распознавание атомов графита полученного изображения, можно сформировать управляющий сигнал коррекции для манипулятора микроскопа.In the method [3], any movement of any of the microscope manipulators leads to the movement of the needle of the corresponding tunnel sensor along the graphite surface, while the sensor feedback system, using its own manipulator, maintains the tunnel gap in it at a constant level. Simultaneously with scanning an unknown surface of a sample, an analog of the invention produces an STM scan of a known surface of graphite (lattice constants: a = b = 2.464 ± 0.002

angle between crystallographic directions: 60 ^o ). Performing software recognition of graphite atoms of the resulting image, it is possible to generate a control correction signal for the microscope manipulator.

The disadvantages of the prototype used for measuring the surface topography and positioning of the probe include:
1. The need to use three additional tracking systems, each of which is equipped with a highly sensitive tunnel sensor and manipulator. Such an approach greatly complicates and increases the cost of the SPM nanolithograph;
2. The inability of the system to determine the displacements of the probe tip in situ, as a result - the occurrence of a cosine error and an Abbe shift error [4]. These errors appear due to the fact that the measuring system of the sensor detects the manipulator’s displacement relative to the device’s body and, therefore, the results are affected: firstly, the imperfection of the body itself (mainly due to its thermal deformation), causing a cosine error (additional unaccounted displacement the manipulator due to the mismatch of the actual direction of movement with the measuring axis of the sensor); secondly, the bending of the manipulator tube during the sweep (in the AFM also the additional influence of the cantilever bending), leading to a violation of the Abbe principle (during bending due to the small intrinsic length of the probe, its tip does not lie on the measuring axis of the sensor, as a result an additional unaccounted bias);
3. The inability to perform accurate measurements of the relief at the limit of resolution of the microscope;
4. The probability of measurement error due to the fact that due to the presence of defects it is almost impossible to ensure the invariance of the structure of the reference surface at each point of the scanner field;
5. Lack of critical ability for nanolithography to arbitrarily hold the probe for a long time at a selected surface location;
6. Lack of an effective mechanism for suppressing microscopic noise arising both in the tunnel current stabilization system (forces in the case of AFM) and in the lateral positioning system, and therefore, the inability to improve the vertical and lateral resolution of the microscope;
7. The inconvenience in the work, due to the fact that the characteristic scale of the relief of the investigated surface does not always correspond to the atomic scale of the reference surface of graphite. As a result, for example, the scan time due to the need to use a small step may be unreasonably long. Replacing graphite with standards with larger structural elements and larger lengths is not always suitable, because in this case we are forced to use, as a rule, not a natural standard, but an artificial non-ideal structure, which itself turns out to be made with some errors.

 From the foregoing, it follows that the currently existing systems are not able to provide high-precision measurements of the surface topography and perform precise positioning of the working body at the site of exposure during nanolithography.

 To eliminate the above disadvantages, a method is proposed, the main idea of which is to use the features of the investigated surface as reference points when performing displacements. Movements are carried out from one feature to another, located in the neighborhood. As a result, a connected sequence (chain) is formed in which features are placed relative to each other. The features may include both features of the "hill" type and features of the "pit" type. The search, detection and calculation of the coordinates of the position of the feature is performed by the recognition program. Scanning a small area (segment) around each feature, and then, laying out the obtained fragments of the surface at the corresponding positions determined during recognition, you can reconstruct the real surface relief.

 In the method, a transition is actually made from an absolute coordinate system associated with a manipulator to a relative coordinate system associated with surface features. The class of admissible surfaces here can be determined through the characteristic size of the feature and through the characteristic distance between adjacent features. It is necessary that the thermal drift and creep in the microscope used would slightly affect the results of single measurements of the relief in the vicinity of the feature and the relative distance between adjacent features and, therefore, would not lead to errors when performing averaging.

 Further, for definiteness, we assume that the surface is atomically smooth, and by features we mean those convexities on it (changes in electron density) that are usually taken as atoms. Note that, as shown by real measurements, all atomically smooth surfaces fall within the description of this class and can potentially be scanned by the proposed method.

 In this invention, obtaining a surface scan consists of two stages: the first is the construction of a seed — a quasilinear chain of atoms of a given length, oriented in a given direction (the direction is counted counterclockwise from the axis of the X-manipulator); and on the second, the rest of the scan “grows” onto the formed seed, using one of the following two methods of local binding.

 In the first method (see Fig. 1), the circuit (seed) is circumvented with the addition of new circuit elements (the circuit remains all the time, for example, on the left). Moreover, the movement of the probe during scanning takes place exclusively along the nodes of the growing chain (the process is similar to laying a rail with a track stacker: it moves along a path that it builds itself).

 In the second method (see Fig. 2), the probe moves along quasistrings, which play an auxiliary role here. The first quasi-string is the seed, the rest are formed during the scanning process. The body of the chain is formed from the elements of the first quasistring and neighboring atoms for each position of the probe in the quasistring when it moves from the last atom of the quasistring to the first (a possible analogy here may be as follows: the tracker moves along the finished track (quasistring), creating in parallel a new one). After the probe reaches the first atom of the current quasistring, the system switches to the last atom of the next quasistring.

 If we take an arbitrary atom as elementary seed, then going around it will give a spiral image of the surface, the shape of which will be determined by the shape of the cell formed by neighboring atoms (see Fig. 1a). This method is the simplest scanning method used in the described invention.

 If a one-dimensional rectilinear (in some cases conditionally rectilinear — see, for example, Fig. 1d, e) chain of atoms is used as a seed, then its bypass will lead to the construction of an image in the form of an elongated spiral (see Fig. 1b, d). Moreover, if the direction of the crawl at the end of each “line” (the atom chain oriented during the crawl by seed) is reversed, we obtain a surface image “expanded” along quasistrings (see Fig. 1c, e) and located in the general case arbitrarily with respect to the axes of the manipulator. In all methods, after finding the "next" atom of the chain, the "current" atom is added to the chain and becomes part of the bypass loop.

 Thus, regardless of the bonding method, the seed plays the role of the original driver. Its length actually determines the size of the scan, and the orientation - its position relative to the coordinate system of the manipulator (see Fig. 1b, c, d, e).

 In the claimed invention, depending on the experimental or technological task to be solved, the positioning of the probe is carried out in two ways, which differ in that the first of them requires a preliminary scan of the surface, as was briefly described above, and in the second, no. Accordingly, the positioning process in the second method requires significantly less time, while ensuring accurate probe movement within the scanner range. In fact, the second positioning method is the process of seed formation, the description of which will be given below.

 In the first method, positioning is carried out only within the scanned surface area. To perform the movement, a subchain is selected from the entire chain (a route is laid), which connects the current position of the probe with the final desired path defined by the experimenter. Further, at each step of the path scanning the aperture and recognizing the nearest neighbors, the probe moves to the position that corresponds to the data of the selected route. The process is performed until the probe reaches the final trajectory feature. If further it is required to perform more accurate positioning within the vicinity of the final feature, then it is carried out by moving the probe relative to the position of this feature, but for the time during which the combined action of creep and drift will not lead to the loss of this feature in the next probe binding cycle. We note that the real path traveled by the probe is an approximation of a given trajectory by a broken line running through the "centers of gravity" of the features.

So, the method of precision scanning and positioning is to perform the following actions:
1. Periodic binding of the microscope probe to the "current" atom of the surface.

 The binding function allows you to hold the probe over the selected atom of the surface for an almost unlimited period of time. It eliminates the influence of the lateral thermal drift of the microscope and the creep of the X, Y piezoelectric manipulators and is actually a digital tracking system in the lateral plane, implemented in software. Local “surface features” act as “sensors” of the position, and the relative distance between them can be measured accurately as a result of multiple averaging (see point 4 of the method).

 Binding actions are activated automatically after a certain period of time T, measured by the readings of the system timer of the control computer and associated with the magnitude of the total instantaneous drift of the microscope. By the full drift of the microscope or simply by the drift we will further understand the result of the combined action of thermal drift and creep.

 The interval T is determined by the results of measurements of the drift, which varies in time and depends both on the specific design of the microscope and on the conditions of the experiment. Comparing the absolute coordinates of the current atom of the chain with the bindings obtained in the previous cycle and using the given thresholds, the interval T either increases by a discrete if the drift is small, or decreases if it is large, or remains unchanged (deadband control characteristics) if the drift value is within tolerance. Therefore, the lower the drift speed, the less often the inclusion of the snap function will be performed, the faster the scan will generally occur. Thus, the method automatically adjusts to the changing magnitude of the drift and takes into account the temporal fluctuations of the latter.

 To perform the binding, first a line scan of the square of the neighborhood, hereinafter referred to as a segment, of the current atom is performed, then the resulting image is subjected to a recognition procedure [5], after which among the recognized atoms there is “current” (located closest to the center of the raster), at the end of the microscope probe moves to the center of the "current" atom.

 2. Scan aperture and recognize nearest neighbors.

 In this operation, the initial horizontal scanning of the surface is performed within the specified square window (2Rx2R), covering with some margin the neighbors closest to the current atom. Although scanning is actually performed inside a square window, subsequent work with the image of the surface is carried out inside the circle of radius R inscribed in it, hereinafter referred to as the aperture. After that, the neighbors are recognized and their coordinates relative to the center of the aperture are determined.

 3. Definitions of the next atom of the chain (local binding of features).

 When creating a seed, first the "previous" atom is removed from the list of neighbors, and then a search is made for such an atom that is located at the smallest angle to a straight line that sets the direction of motion. The atom found gets the label "next." Moreover, when moving along the crystallographic direction, the position of the next atom of the chain is used to clarify the direction of motion if the change in the local curvature of the chain lies within specified limits. Thus, this technique allows you to keep the selected direction, easily bypassing those places on the surface of the crystal where point or linear defects of the lattice are located.

 In the case when a loop is traversed with attachment, first the previous atom is removed from the list of neighbor atoms recognized in the aperture, then sorting is performed: among the atoms, those that do not belong to the already passed chain are selected. After that, among the selected ones, one is searched for that is located at the largest (smallest) angle (the angle is counted around the “current” atom in a counterclockwise direction in both cases) to the segment connecting the current and previous atoms when the loop is circled counterclockwise (by clockwise). The atom thus found is taken as the next atom in the chain.

 The fundamental and most significant drawback of the method of forming a surface scan by bypassing a contour with attachment is the impossibility, starting from a certain length of the chain, of sorting neighboring atoms correctly. The reason is the error accumulated in the chain. Moreover, the larger the noise and drifts of the microscope, the more reasonable the number of averages, the specified restriction comes, and therefore the limit on the maximum scan size that can be obtained with this microscope. In order to suppress the accumulation of errors, when adding new links in the chain, it is necessary to correct by “dissolving” the residual in the chain from the previous atom to one of the detected neighboring atoms included in the chain at the previous turn.

 As an alternative method to avoid scanning errors associated with the accumulation of errors in the chain, the previously mentioned binding using auxiliary quasistrings can be proposed (see Fig. 2). The main idea underlying it is that the smallest possible number of chain elements would potentially take part in the sorting operation of neighbors, that is, that the chain would be viewed to the minimum possible depth. For this purpose, a group is formed from the neighbors of the current atom of the current quasistring, which includes atoms lying on one side defined before the start of the scan from the quasistring and not belonging to the quasistring. To bind adjacent groups of neighboring atoms belonging to adjacent atoms in a quasistring, a special acknowledgment atom is used.

 Note that if no suitable atoms were found among the neighbors in the binding operation, the aperture radius is increased and its scanning and analysis are performed anew.

 4. Measurement of differences and averaging of segments (skipping).

 This operation is designed to perform high-precision measurements of x, y, z coordinates of surface atoms. Since, due to the presence of destabilizing factors, it is impossible to accurately measure the absolute coordinates of atoms by averaging, regardless of whether the binding mechanism works or not, the technique in question measures the coordinates of the "next" atom of the chain relative to the coordinates of the "current" one. This approach is justified, since the drift consists mainly of the STM head drift, and it does not lead to a relative displacement of neighboring features, the same part that is introduced by the sample by reducing its size can be made negligible.

 At first, the probe binding is blocked, since now the bulk of the functions of the latter will be performed by the difference measurement procedure itself. Then the microscope probe is moved to the position (absolute) of the next atom, after which a regular line scan of the square (mxm) of its neighborhood - segment is performed. After that, the next atom is recognized. Having determined its coordinates, the procedure calculates and averages the differences Δx, Δy, Δz, and then moves the STM needle back to the position of the current atom, after which it performs actions similar to those that it just did with the next atom. Having found the coordinates (absolute) of the current atom, the procedure again calculates and averages the difference of coordinates between the atoms.

 Further, the sequence of actions described above, called skipping, is repeated as many times as averaging (the number of averaging is essentially unlimited) was specified. At the end, the binding is unlocked, the “current” tag is assigned to the “next” atom, and the counter of the number of chain atoms is increased by one. Recall that the relative coordinates for the first movement of the probe to the position of the next atom are determined by scanning and recognition within the circular aperture.

 If during the skipping cycle the change in the drift velocity is small, then almost complete compensation occurs. Suppose, for example, that the component of the surface drift along x is aligned with the component of movement along x of the probe when it moves from the current feature to the next, then the measured difference will be more true, when moving in the opposite direction it will be the same amount less. Therefore, the average value of these differences will be equal to the true distance between these features. As for the absolute value of the drift, it can be quite large - if only during the time the probe moved from one feature to another, the latter would each time lie within the boundaries of the segment.

 Note that when the probe moves between the current and next atoms, the feedback loop is always closed, and when scanning a neighborhood, depending on the set mode, it can be either closed ("direct current" mode) or open ("constant height" mode).

 In addition to accurately determining the relative coordinates of the next atom of the chain, averaging of surface segments is also performed during skipping, which allows you to effectively deal with noise and achieve high resolution in the vertical plane.

 It should be noted that the size of the segment should be selected so that the segments of adjacent atoms overlap slightly. This is required so that after completing the "assembly" the resulting surface would not have gaps.

 If the drift during the scan was not very strong and / or the requirements for measurement accuracy were not too high, then for the subsequent surface assembly the averaged pieces of the surface — segments and averaged differences — are saved, otherwise, in each skipping cycle, one-time measurements of the absolute coordinates of the “current” and “next” atoms together with the corresponding once measured segments.

 The fact is that the scanner is “attached” to the surface and, therefore, drifts along with it, moving along the scanning field, at each point of which different calibration coefficients and oblique angles act. Therefore, in order to avoid errors during averaging, first, using the calibration grid, it is necessary to correct the absolute coordinates and segments. It is impossible to do this during scanning due to the fact that the task of searching in the calibration grid, approximating coefficients and, finally, correcting segments require significant computational resources.

 Since the relative distance between the features determined during skipping is a real number that varies according to a random law, it is possible to reconstruct the relief by performing a large number of measurements of differences and segments and introducing more pixels into the image at the stage of assembling the real surface (it is assumed that the needle is quite sharp ), the extreme lateral details in which will be finer than those that the used microscope is able to detect during conventional scanning.

 A few words should be said about the calculation of the absolute positions of atoms ("current" and "next") in the vertical plane, which are defined as the average in a given circular neighborhood of the atom. It is clear that the values obtained in this way for the z coordinates of the atoms most likely weakly reflect their actual positions, but are quite suitable for obtaining the relative coordinates used both in constructing a stylized surface and a real one made up of segments.

 Note that in the described scanning method at each moment of time there is no more than one atom having absolute coordinates. The coordinates of the remaining atoms of the chain are relative. Although in case of strong drift, the relative coordinates contain errors and are not suitable for particularly accurate measurements, meanwhile, it is possible to carry out the linking and also to plot a route when positioning with their help.

 Correct calibration of the microscope allows eliminating distortions caused by the nonlinearity of the manipulators, their nonorthogonality and parasitic influence on each other at the stage of surface assembly. The calibration procedure consists in scanning the reference surface by the proposed method, which can be, for example, HOPG, and determining for each absolute position the features in the scan of the local calibration coefficients and the local oblique angle (necessary to take into account coupling of the type X <------ > Y). As a result, we obtain the distribution (grid) of calibration coefficients and oblique angles in the scanning space. Since the influence of thermal drift and creep is suppressed during the measurement process, the calibration grid is independent of the scanning speed and its direction.

и углов косости α в плоском поле. Кроме нелинейности и неортогональности сканера, в найденном распределении также происходит учет паразитных связей типа: X<------>Y, Z--->X, Z--->Y.Consider the calibration of the scanner in the lateral plane, it is carried out according to the triangles [5] formed in the aperture of the feature under consideration and its nearest neighbors (6 equilateral triangles for the graphite lattice). If the surface has a small trend, then for some fixed position of the Z-manipulator we obtain the distribution of calibration coefficients

and oblique angles α in a flat field. In addition to the non-linearity and non-orthogonality of the scanner, the distribution also takes into account spurious links of the type: X <------> Y, Z ---> X, Z ---> Y.

и углов косости α в пространстве. Помимо нелинейности в вертикальной плоскости, в найденной сетке будет учтен кауплинг типа X--->Z, Y--->Z, обычно проявляющийся на изображении наиболее сильно.To perform the calibration of the microscope Z-manipulator, a calibration structure is required that has on the surface not only a system of features with a known period, but also a height (graphite is not quite suitable for these purposes, since the value of the surface corrugation in it can differ greatly from experiment to experiment). By scanning the reference surface for different positions of the Z-manipulator, set by the coarse step of the scanner, we obtain the distribution of calibration coefficients

and skew angles α in space. In addition to non-linearity in the vertical plane, the found grid will take into account coupling of the type X ---> Z, Y ---> Z, which usually manifests itself most strongly in the image.

 To visualize the data obtained during scanning, it is necessary to “assemble” the real surface. In those places where segments of neighboring atoms are stacked with a slight overlap, the relief is averaged. During the assembly process, using the approximated calibration coefficients and the oblique angle calculated from the values in the grid closest to the current position of the segment, this position is corrected, as well as the imbalance and obliquity of the image are corrected [5] in the segment itself. In the case when the surface features are so great that the segments are noticeably distorted by the non-linearity and the coupling of the manipulators, the image is corrected at each point of the segment using approximated calibration coefficients and oblique angles calculated from the values in the grid closest to this point.

 Note that the scan results do not depend on the type of unit cell, nor on its size, nor on the presence of defects or any disordering of atoms on the surface. These properties are obtained solely due to the local binding of surface features.

 The proposed scanning method can be classified as adaptive, because it contains links that provide self-adjustment of the process to the current environmental conditions, surface features and specific equipment.

 It should be noted that this description of the method of precision scanning and positioning involves the use of a fixed segment size, specified before starting work. In the case when the distances between the features are very wide, the segment size becomes variable and must be determined dynamically during the scanning process.

 The methods for precise positioning of the STM probe incorporated in the method are the basis for the construction of nanolithographs. The ability to autonomously follow along a chain of atoms can be used to automatically search for various defects on the surface: vacancies, inclusions, monoatom steps, disordered regions, regions with different phase states, as well as their statistics. Moreover, the trajectory here can be chosen random, for this it is necessary to sequentially randomly set the length of the chain and the direction of movement along it.

 Holding the direction of motion along the chain of atoms so that it constantly coincides with the crystallographic direction on the surface, the presence of microdefects and stresses in the crystalline body can be detected by the curvature of the chain.

 Another useful application of the method is the automatic return of the probe to the operating area. This function is necessary in experiments on nanolithography, where after any exposure to the surface by a probe, the sample is removed from the microscope, processed (film coating, etching, annealing, etc.), and then set back to see what changes occurred with the surface at the site of exposure. In such cases, in order to ensure automatic search-displacement to the place of impact, it is necessary to make a branched system of chains on the initial surface of the sample from features (see Fig. 3) converging to this place. Then it is enough to “capture” some feature at the edge of the field, set the direction of movement to the zone and wait for the moment when the probe reaches the last element of the circuit. The moment of arrival in the zone corresponds to the absence of the following “suitable” feature among the neighbors (in this mode, the aperture size is fixed).

In FIG. 1 illustrates the simplest methods of local binding by traversing a loop with attachment. Seed: (a) single atom; (b), (c), (d), (e) one-dimensional rectilinear chain of 10 atoms. The seed is oriented along: (b), (c) crystallographic direction; (d), (e) a direction constituting an angle of 0 ^o with the X manipulator. Circuit bypass direction: (a), (b), (d) fixed (counterclockwise); (c), (e) variable (with switching at the end of each quasistring).

 In FIG. 2 schematically illustrates a method for locally binding surface features using auxiliary quasistrings.

 In FIG. Figure 3 gives an example of a system of converging chains of features converging to the place of impact on the surface of a sample. The letter Z marks the operating area.

Скорость сканирования в апертуре: 10000

Скорость сканирования в сегменте: 500

Скорость перемещения при скиппинге: 100000

Средняя скорость латерального дрейфа около 0,5

Интервал привязки зонда не более 500 мс. Время измерения: 23 мин.In FIG. 4 shows a surface scan of HOPG obtained by the method of the invention. (a) Stylized image. Carbon atoms are conventionally depicted as hemispheres. (b) Real surface reconstructed from segments (constant height mode: I _tun = 998 pA; U _tun = -10 mV). The chain length is 400 atoms. The number of averagings at a point: 2. The number of averagings of a segment: 5. The averaged lattice constant a = b = 2.299

Aperture Scan Speed: 10000

Segment Scan Speed: 500

Skipping speed: 100000

The average lateral drift velocity is about 0.5

The probe binding interval is not more than 500 ms. Measurement time: 23 min.

Скорость сканирования в апертуре: 7500

Скорость сканирования в сегменте: 750

Скорость перемещения между атомами цепи: 75000

Средняя скорость латерального дрейфа около 0,2

Интервал привязки зонда не более 500 мсек. Время сканирования: 39 мин.In FIG. 5 shows the surface of HOPG obtained in high resolution mode. (a) Stylized image. (b) Segment image (constant height mode, U _tun = 15 mV, I _tun = 401 pA). The number of averagings at a point: 3. The number of averagings of a segment: 1000. The average lattice constant a = b = 2.299

Aperture Scan Speed: 7500

Segment Scan Speed: 750

The speed of movement between the atoms of the chain: 75000

The average lateral drift velocity is about 0.2

The probe binding interval is not more than 500 ms. Scan time: 39 min.

 In FIG. Figure 6 shows the STM probe moving along a chain of atoms in a given direction. The direction retention submode is on. Length traveled: 1 micron. The number of atoms in the chain: 4060. Movement speed: 3 atoms / sec.

Коэффициент асимметрии k_s = -0,09. Эксцесс k_k = 1,27. Время измерения: 7 часов 30 мин.In FIG. 7 shows a histogram of the distribution of the error of measurement of the HOPG lattice constant. The number of measurements I = 2.5 • 10 ⁵ . Single measurement error (3 σ): ± 0.255

The asymmetry coefficient k _s = -0.09. Kurtosis k _k = 1.27. Measurement time: 7 hours 30 min.

и 0,334

соответственно. (b) Эволюция кончика зонда в пространстве (направление перемещения обозначено стрелкой). Вероятное направление дрейфа в латеральной плоскости: 269^o. Вектора и положения зонда обозначены значком "+". Количество привязок зонда микроскопа: 500. Интервал времени между привязками T = 700 мсек.In FIG. Figure 8 shows the results of measurements of the STM drift. (a) Distribution of drift vectors in the velocity space. Lateral average and maximum drift velocities: 0.120

and 0.334

respectively. (b) Evolution of the tip of the probe in space (the direction of movement is indicated by an arrow). Likely direction of drift in the lateral plane: 269 ^o . The vectors and positions of the probe are indicated by a “+”. The number of bindings of the microscope probe: 500. The time interval between the bindings is T = 700 ms.

 The described method of precision scanning and positioning was implemented and tested when working with the structural elements of the surface — atoms, which are extremely distinguishable by a scanning tunneling microscope.

Число точек в апертуре: 27х27. Число точек в сегменте: 13х13. Иглой служила кусаная проволока NiCr, а размеры образца ВОПГ (0001) составляли (2х2х0,5) мм³.The measurements were performed on a NT-MDT Solver P4 scanning probe microscope in air. An IBM compatible 486DX4 100 MHz computer was used as the manager. Resolution of the microscope in the lateral plane:

The number of points in the aperture: 27x27. The number of points in the segment: 13x13. The bite wire NiCr served as a needle, and the dimensions of the HOPG sample (0001) were (2x2x0.5) mm ³ .

 Performing local binding of features using auxiliary quasistrings (the length of the quasistring is 20 atoms, orientation along the axis of the X-piezomanipulator), the HOPG surface was scanned (see Fig. 4). The "torn" edges in the image (b) give out on its segment structure. Due to the insufficiently high speed of the control computer, strong drift, and a small number of averaging of chain links, the positions of some atoms in the scan are noticeably shifted. During surface assembly, non-linearity, non-orthogonality, and cowling were not corrected. Artifacts associated with the segmented image structure (b) are practically absent. Attention should be paid to the low noise level in the segmented image, although no mathematical data processing has been performed.

 The next experiment was to obtain a scan of the HOPG surface in high resolution mode (see Fig. 5). For each of the 7 carbon atoms, 1000 measurements of the segment and the differences were performed. After assembly (correction of non-linearity, non-orthogonality, and cowling was not performed), the lateral resolution of the microscope is “improved” by 20 times. Linking is done by traversing the contour with an attachment. The type of seed is a single atom. In the upper left and lower right corners, the initial sizes of image elements corresponding to the limit of resolution of the microscope are clearly visible.

 As an experimental confirmation of the possibility of precision positioning on an atomically smooth surface (a method of positioning without preliminary scanning of the surface), as well as to demonstrate the ability to accurately measure distances of several microns by counting atomic periods, the STM needle was moved along the crystallographic direction on the HOPG surface (see. Fig. 6), the distance traveled was 1 μm (4060 carbon atoms).

На фиг. 7 показана функция распределения погрешности измерения постоянной решетки графита, там же для сравнения дана функция нормального распределения, построенная с использованием эмпирических значений среднего и среднеквадратического.To prove the possibility of using the proposed method for high-precision measurements of distances and directions on the surface, 250,000 averagings of the HOPG lattice constant were performed. It took 7.5 hours of continuous operation of the STM to complete the work, while the measurement error was ± 0.0005

In FIG. Figure 7 shows the distribution function of the measurement error of the constant lattice of graphite; the normal distribution function constructed using empirical mean and rms values is also given there for comparison.

Вероятное латеральное направление дрейфа: 269^o. Близость вероятного направления дрейфа к направлению "медленного" сканирования указывает на присутствие значительной составляющей крипа Y-манипулятора.To measure the drift of the microscope, a sequence of bindings of the probe to the selected atom on the surface of graphite was performed. In the process of tracking an atom, its drift vector was determined from its movement, i.e. drift direction and its velocity (displacement value related to the anchor interval T). In FIG. 8a shows the spatial distribution of the obtained drift vectors. In FIG. 8b shows the trajectory of the probe tip (binding atom) in space during the measurement. The average drift velocity of the microscope in the lateral plane was about 0.12

Likely lateral drift direction: 269 ^o . The proximity of the probable drift direction to the direction of the "slow" scan indicates the presence of a significant component of the creep of the Y-manipulator.

Sources of information
[1] JE Griffith, GL Miller, CA Green, DA Grigg, PE Russell, A scanning tunneling microscope with a capacitance-based position monitor, J. Vac. Sci. Technol. B, vol. 8, N 6, p. 2023, 1990.

 [2] R.C. Barrett, C.F. Quate, Optical scan-correction system applied to atomic force microscopy, Rev. Sci. Instrum., Vol. 62, N 6, p. 1393, 1991.

 [3] H. Kawakatsu, Y. Hoshi, T. Higuchi, H. Kitano, Crystalline lattice for metrological applications and positioning control by a dual tunneling-unit scanning tunneling microscope, J. Vac. Sci. Technol. B, vol. 9, N 2, p. 651, 1991. H. Kawakatsu, H. Kougami, Automated calibration of the sample image using crystalline lattice for scale reference in scanning tunneling microscopy, J. Vac. Sci. Technol. B, vol. 14, N 1, p. 11, 1996. H. Zhang, F. Huang, T. Higuchi, Dual unit scanning tunneling microscope-atomic force microscope for length measurement based on reference scales, J. Vac. Sci. Technol. B, vol. 15, N 4, p. 780, 1997.

 [4] J.E. Griffith, D.A. Grigg, Dimensional metrology with scanning probe microscopes, J. Appl. Phys., Vol. 74, N 9, p. 83, 1993.

 [5] R.V. Lapshin, Automatic lateral calibration of tunneling microscope scanners, Rev. Sci. Instrum. , vol. 69, N 9, p. 3268, 1998. R.V. Lapshin, Procedure for the recognition of atoms in STM images, Proceedings of the third international scientific and technical conference "Microelectronics and Informatics", Issue 3, Moscow, Zelenograd, 1998.

Claims (3)

 1. The method of measuring the surface topography with a scanning probe microscope, which consists in scanning the test surface in a horizontal plane while maintaining a constant gap between the probe and the sample in the vertical plane, as well as linearizing the measurements by compensating for the difference that occurs when the manipulator moves between a given amount of displacement and the actual measured using linear sensors, characterized in that instead of position sensors to compensate for the thermal drift of the head and creep of zomanipulators, local features of the surface topography are used, which serve as reference points when performing probe movements, for which the aperture is first scanned and software recognizes the nearest neighbors, among which the current feature is located, and in accordance with the selected local binding method, the following: the skipping operation is performed, then moving the probe to the position of the next feature relative to the position of the current one, after which the process described above is repeated I until a scan of a given size is received; Now using a grid distributed in the scanning space from local calibration coefficients and oblique angles found during the measurement of the reference surface using this method, image distortions caused by non-linearity, non-orthogonality, and piezo manipulator coupling are corrected.  2. The method according to claim 1, characterized in that in order to move the probe from the current feature to any other predetermined one lying within the scanned area, on each section of the route, the aperture of the current feature is scanned, recognition of neighboring features, identification of the current and the determination of the next, coinciding with the given route, after moving the probe to the found position of the next feature, the process is performed cyclically until the probe reaches the final feature of the route.  3. The method according to claim 1, characterized in that, in order to automatically return the probe to the operating area on the sample, where a branched system of chains converging to the area is preliminarily made from the features, the initial feature and the direction of movement to the area are indicated, the size of the aperture is fixed, and the moment the probe reaches the zone is determined by the absence of the following features among the recognized neighbors.

Images