TWI286196B - Methods and systems for determining overlay error based on target image symmetry - Google Patents

Abstract

In systems and methods measure overlay error in semiconductor device manufacturing based on target image symmetry. As a result, the advantages of using very small in-chip targets can be achieved, while their disadvantages are reduced or eliminated. Methods for determining overlay error based on measured asymmetry can be used with existing measurement tools and systems. These methods allow for improved manufacturing of semiconductor devices and similar devices formed from layers.

Description

1286196 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明 发明[Prior Art] 'Modern semiconductor devices' such as integrated circuits 'are generally made of wafers of semiconductor material. The wafers are fabricated by a series of patterned semiconductor material layers. The circuit pattern is made using a variety of long established techniques, such as lithography. • The overlay measurement in the fabrication of semiconductor devices is used to determine the appropriateness of the alignment of a printed layer with the previous printed layer. Accurate alignment of each layer at all locations in each layer of the device is important to achieve design goals in order to increase process efficiency and quickly and accurately measure any alignment between two patterned layers on a wafer. error. This is also very important for measuring any alignment error between successive exposures in the same layer. The misalignment between the layers is called the overlap error. The overlay pair metric tool or machine is used to measure the overlay error. This information can be fed into a closed loop system to correct the overlay error. • Current overlay metrics use optically interpretable overlay marks or patterns that are printed onto layers of a substrate (usually a semiconductor wafer) during fabrication. The patterns are imaged at high magnification, the images are digitized, and the image data is processed using various known image analysis algorithms to quantify the overlay error to measure the relative displacement of the two successive layers. Therefore, the overlay measurement includes direct measurement and misalignment between the patterns directly associated with each layer. As semiconductor devices become smaller, accurate stack-to-measurement is becoming more and more difficult. And the equally important system, the new measurement method and system must be able to match the extremely high speed of the existing stack-to-measurement to a large amount of time, otherwise it is not desirable to repeat the measurement. Therefore, ^, non-needed metrics and systems will encounter major technical challenges. Sweat I, two improvements The microelectronic device with a smaller and smaller volume is used in a small and can be placed in the active area of the device. A pair of indicia of the heart is disclosed in U.S. Patent Application Serial No. 11/035,652, the disclosure of which is incorporated herein by reference. The overlay marks, which are small enough to be placed in the active area of the device, can be measured at the actual desired location. With these stacked pairs of marks, the small dimensions of the stacked pairs must be in line with the practice. Ideally, the existing stacked pairs of tools must be used to measure the overlapping pairs of marks. These overlapping pairs of marks do not have the only correct size. However, the smaller the volume of the stacked pairs of marks, the more useful they are. The overlap of the area is quite practical. The stack "size" must contain any space that is needed to achieve the correct measurement. The current optical overlay measurement system uses visible light, and it is compatible with the design of the pattern. Although changing these classes can improve optical resolution, system changes are time consuming, silky, and potentially uncertain. Therefore, it is advantageous to change the existing one. Operate to accurately measure smaller overlays

[Summary of the Invention] The measurement system is super-operational and stylized, and the existing system can be accurately and memorized. The case surprisingly found that there is a relationship between asymmetry and the overlap of pairs 1 1286196 = so that the mismatch error can be determined or calculated exactly based on the mismatch. Therefore, the advantages of the notes can be reduced or even eliminated. = Ming k for the method of determining the overlap error by the measured asymmetry. This method can be combined with the current (four) amount of heart (4) silk. The method of the method can improve the fabrication of a semiconductor device composed of a plurality of layers and a similar device. [Embodiment]

Prior art techniques have been designed to place wafers in the wafer, ϊ ί. It will produce a film characteristic in the structure of the lower layer and the surface of Fig. 2, which proves to be detrimental to the modernity (4) to the m: Ιί pattern in the large thickness of the Japanese nuclear money (four) below the pattern of the second degree = 1. The test results show that 'use these small markers for accurate measurement. However, there is a need for new image processing methods to perform these speculations.

In the experimentally obtained optical intensity curves, the asymmetry of the overlapping marks in a wafer was observed. This asymmetry is clearly due to the proximity effect. If the proximity effect is not specifically considered, then the overlapping deviations from the asymmetric image will contain the overlap of the measurements.凊 See R. Attota's 4th issue of the SPIE meeting in May 2004, Volume 5375, pages 395 to 4〇2

New In-Chip and Arrayed Line Overlay Target [)esigns Λ. However, as explained below, the overlay error can actually be determined by the measured asymmetry. 8 1286196 Execution = One of the methods is as follows: L defines a Region of Interest (ROI) that essentially surrounds the marker or pattern to be measured. The ROI should also allow for uncertainty in the location of the marker after the pattern is identified. 〆, and δ \ this will make the R 〇 I wider than the design size of the mark by about i, 2, 3, 4, / / micron. The figure-picture shows the cross-m ame mark in a 2 micron frame. The cross design within the frame maximizes the > within the tag. Other tag shapes and sizes can also be used. It is also possible to use an outer φ~, 1 micron mark, which shows the ROI around the mark in the seventh figure. ΓΥ ;; axis or first-axis measurement. Simply change the "X" to ι"ι. Measure the vertical axis or the second axis in the vertical measurement. ## on the detection device, change in the measurement system or tool in the second, each line: ====-the number of consecutive scan lines is usually equal to 2^1 measurement direction' as the seventh The figure shows. The number of camera lines. Although Bo 5 Hai / Zhuang 01 is perpendicular to the direction of the measurement direction, CCD camera), however, the second clothing is usually a two-dimensional camera (for example, #说, with raster scanning laser two with other Detective The measuring device, for example, is a single photodetector for each scanning line. The center point is selected by the one shown in (s) (y), and the initial candidate degree is selected. In the middle of the measurement, the scanning line of the image (4) on the other side of the intensity of the image - corresponding to the measured light intensity and the center point. *, compared with the first intensity of the above =, then the implementation The degree of symmetry of +f(S,y) is taken as the reference, and the image point of the image is repeatedly executed in the R〇1, and the other candidates in the scanning line are used in the initial case. 9 1286196 in the center point is added and there is: U value can be selected as the next candidate center point. Let the two beauty =: Γ candidate, the heart point will be regarded as the actual center; the point called i degree, that is Port du: The highest intensity curve t on the line is close or symmetrical to the other side along the other side of the point (8) = the intensity curve formula has several kinds of ways to calculate Difficulty. Among them, the method is based on the difference between the intensity of all points at the same distance from the point of the point. The difference formula of degrees is ___ f(s, y)= Icy(s+X^y)-I^~^^y))2dx where y) is the image intensity, (s) is the center point, and the symmetry value at point s is f(s,y) ° The 1st person works with the image at the scan line (y) in the tenth figure. # In the eighth and eleventh figures, each scan line (>0's f (s brighter line). The maximum value of these curves (labeled as Wy) is related to how much information they have. As shown in the seventh figure, 'the ROI extends beyond the pivot of the stacked pair of markers. As shown in Figure 8 and Figure IX As shown, the end of some or all of the scan lines may not pass through the mark and will not contribute to any calculation or measurement. IV. The result of the stack will be proportional to the thread along the line of the mark. The minimum value of f(s'yc), where the center represents the vertical measurement direction 2. The minimum value of f(s, y) can be called fmin(y), and the value of the fmi handle is called two kin. The fourth figure is a theoretical relationship diagram, which shows that in the available range , 1286196 The difference between the difference and the simulated symmetry is a straight line or a straight line 'and has a meaningful relationship. The tenth—the symmetry of the mark drawn in the figure and the experimental pair of the measured pair show the relationship between them. The eleventh hull '^车由' material is used - the box in the box (b〇X raise b〇X) is marked by the known stack of fi, and the ❿ γ axis data is used in the figure - π η has been produced. This data supports the theoretical relationship diagram in the fourth figure. The line shown on the t ' m« line has no analog symmetry of noise. Heart: no 'no miscellaneous tTL, the model rewards Correspondence or correlation between the pertinence and the overlap II. The upper trajectory shows a two-dimensional analog symmetry with 1% noise (a random variation in this signal is approximately equal to 1% of the maximum signal level). After substantially determining the symmetry value of each scan line, it can be combined in one of several different ways to derive a combined single value that can be used to calculate the label overlay error. The symmetry values of the scan lines can also be averaged or combined by other means: As described below, one of the ways of combining the values is a weighted average method. Instead of combining the symmetry values, another minimum symmetry value is selected from among all of the symmetry values for determining the overlay error. - Regardless of the method used by β, the resulting single-symmetric value can be used to determine the overlay error. The resulting single-symmetric value can be used to find the overlay error using the fourth graph, type. If the built-in initial branch of the mark has a block difference of 20 to 80 nm, 30 to 7 〇 nm, 4 〇 to 6 〇 nm, m "" γ γ 1 ^ 骑 会 — - ship money _ curve this, even if there is a miscellaneous Another equivalent of the existence of the signal, which can also achieve a more precise search for the error, is to use the experimental model of 11 1286196 shown in Figure 11. The extra steps described in the paper help the necessary steps of the 3 series. The square n degree 'but these steps can also be carried out in the following steps. 糸 The basic form, even if there is no J, 'can be \\ line two Bu Jing 3 to form the center of the knot., hidden (y) g line is located v. of the mark in the direction, for each scan line (), calculate the signal, the weighting factor w(y); there is = two = place (4) method, the weighting factor is VI. Use the above weighting factor Calculate a more accurate estimate for fmin · , J / max (y) f (s, y) dy f (4): gas - 瘫 J / max iy) dy

-ROI where 'f(s) is the weighted average of f(s, y), ROI is the region of interest, fmaX(y) is the weighting function, and the minimum value of f(S) is proportional to the overlap error. The reason for implementing these steps is because the center of the mark (xc, yc) is not known. If you know, you can use the following formula to calculate the asymmetry π, f(s, y) = (I(xc+s, yH(xc_s, y)) where I(x, y) is the image intensity. The definition of f(s,y) in the previous section is mirrored, 12 1286196. However, the more favorable one is that the new definition here is unmarked. The square of the intensity difference is used to count #f( s, y) does not need (xc, yc) 〇νιι· use fmin to measure the overlap, the fmin is proportional to the signal strength〗. Therefore, if the Xiang (four) degree normalizes the result to remove its dependence, The measurement accuracy can be improved, from which the overlap can be calculated (for example, the alignment error of the overlay on different layers), where: f'min=fmin/I(X,y) is at X,y General on

This aggregate calculation is performed for all points within the R〇I. • :1· If ^ makes the symmetry function of the $ * mark, then the H Middle Ages noise will never be prepared. This is the case in the right hand. The line is in the heart of the $ trajectory. For example, if the difference is normal, it is expected to overcome this result.

With the symmetry vs. shown in the four figures, all measurements can be made. · And on the straight part of the deviation curve, refer to the third figure of the fish, and the landscape image changes with the pair. The third partial map ^^ measured the minimum hemiplegia: 3⁄4 Mingzhong will be the second count = fruit. Using the averaging method "heart: the signal is -2 〇 mark, shouting no = 13 1286196 A system for performing the above measurement may include a digital camera for capturing the mark image. The digital camera is usually equipped with an optical magnifying lens (for example, a microscope), produced and connected to a computer that is programmed to perform the above steps. The system may further include: a light source for illuminating the mark; a microscope stage for supporting the substrate; And an automated robotic arm fixture or actuator for moving the microscope stage, usually by moving the mark to align the lens. If a digital camera is used, the mark image is acquired in pixels. In the example shown, one pixel is about two. The interpolation can be used to resolve the (equal) position, and the point of the pixel is used as a reference to calculate the symmetry value: Knife said, ί广上^ Substituting the overlap error or the overlap bias, for example, the difference or the alignment error. == The position of the other-lower mark on the other layer is performed on the scan line of each step or the actual example of each line is slightly 1; = This: The small value mentioned also covers the maximum value, $. In this paper, the effective calculation can be exchanged with the maximum and minimum values = ^ for various equal-numbered layers; the square:: in the system ί thin = The invention includes the micro-mechanical device, the micro-motor; the spirit and the scope of the present invention; the following, the sequel, the singularity, etc., are not changed, replaced and modified. Therefore, the present invention makes various examples and descriptions. , the present invention:: this: (b): 14

The full range of equivalent components and equivalent steps of 1286196.

15 1286196 [Simple description of the diagram] The first picture, the first picture B, the first. A plan view of the figure and a pair of marks on the first board. Alpha base map. The second figure is the cross-section of the cross-marks shown in the first to the right-graphs. The third figure is the relative light intensity measurement vs. one-fold pair of mark lines (--the most correct: 己) - The model on the side is used to show the asymmetry in the measurements: the model of the model. '

= Type symmetry vs. overlap versus deviation and image noise effects. The ϊΐ: positional relationship diagram of the section through the overlay mark shown in the illustration in the 丄 丄 diagram. The slice in the picture shown in the fifth figure, which is not shown in the figure c, has the smallest overlap error.卞 园 其

Figure 6 is a cross-sectional view of the image orientation of the image shown in the illustration in the figure, in which the symmetry is shown by the change in the deviation of the overlay. The illustration in the sixth figure is the overlap of the first c-picture. In this figure, there is a large overlap error in the X-axis. Formula..., Month f Seven is a plan view of the stacked pairs of marks in the fifth figure in the region of interest. The eighth figure is a graph of f(s, y) based on the x-axis for each scanning line 7 in the seventh figure (shown by the bright gray line). The eighth graph also shows the weighted average of f(s, y) for all, y (dark lines). In the eighth figure, fmin(s) is at or near zero, that is, at the center of the pair of marks, which indicates that the deviation in the X-axis is small or even without any deviation. The graph in the ninth graph is identical to the eighth graph, but is based on the γ-axis, and the graph shows a curve having a large overlap bias. In the ninth figure, the 1286196 fmin(s) system appears at about -5 microns, and the minimum symmetry value is about 25, which indicates that the stacking error is very large. The tenth figure is a schematic diagram of the selection of a candidate center point, and the calculation is also performed to indicate the degree of symmetry in the image in the region of interest based on the selected center point. In the eighth to tenth drawings, the X-axis corresponds to the position in the region of interest, where the X-axis is distributed at -25 mm to +25 mm. The degree of symmetry calculated by the system corresponding to the Y-axis (based on the comparison of the intensity values), wherein the lower the value, the greater the symmetry, and the higher the value, the smaller the symmetry. ❿ The eleventh figure is a graph of the measured symmetry vs. overlap versus deviation. [Main component symbol description] None

17

Claims (1)

1286196 Pickup, Patent Application Range: 1. A method for measuring overlay error, comprising: A] placing a first marking element on a first layer of a substrate; B] placing a second marking element Placed on a second layer of a substrate, the first marking element and the second marking element form a mark; C] acquire an image of the mark; D] measure light along a first scan line of the image Intensity; E] calculating a first symmetric value of the first scan line according to the measured light intensity along the first scan line; F] measuring the light intensity along a successive scan line of the image; Calculating the contiguous symmetry value of the successive scan lines of the image according to the measured light intensity along the contiguous scan line; H] repeatedly performing steps F and G to obtain a symmetric value for substantially all scan lines of the image And I] calculate the deviation between the first and second ► mark elements using the minimum symmetry value or the maximum symmetry value among the symmetry values determined in steps D to Η. 2. The method of claim 1, wherein the image of the mark is captured by a camera, and further comprising dividing the mark into a pixel array, and a scan line substantially passes through the bit Each pixel on a single line. 3. The method of claim 1, further comprising determining the minimum symmetry value along the center of the mark in a direction perpendicular to the direction of the measurement, and then determining the deviation based on the value. 18 1286196 4. If the first item of the patent application scope is for the first of each of the sweeping lines, it further includes: a weighting coefficient W; W for substantially each scanning line is used to generate a substantial value of ^^ multiplied by the The weighted coefficient weight average; the sum of the symmetry values of the sum and the e scan lines is used to find the minimum value of the weighted average of the scan line symmetry values; the heart is determined according to the minimum value of the weighted average value 5. If the method of the third paragraph of the patent application is applied, the strength is normalized to the symmetry values. ^ 乂6. : The method of the first item of the scope, wherein the symmetry value of the 5 hai is calculated by the sum of the squares of the intensities of all the points at the center point I:: distance. The method of the third item, wherein the step further comprises providing a known number of built-in deviations in the target f and then calculating the overlapping bias using a linear function 攸 the measured symmetry. The method of the item i of the range includes the following steps for substantially each scan line: [i] remotely selecting a first waiting center point and measuring light on both sides of the first candidate center point Intensity, then calculating a first symmetric value of the _th candidate center point; and [ii] selecting a contiguous candidate center point, measuring light intensity on both sides of the contiguous candidate center point, and then calculating a splicing of the splicing candidate center point Symmetric value; [iii] repeatedly performing step [ii], finding the center of symmetry by confirming that the symmetry value represents the maximum degree of symmetry in all of the waiting points 19 1286196 Q points; and '[iv] according to the symmetry Center to calculate the deviation. 9·—planted in two Method for measuring stacking error in dimensions, comprising: A] placing a first marking element on a first layer of a substrate B] placing a second marking element on a second layer of a substrate The first marking element and the second marking element form a second note; ^ C] select an area of interest near the mark; D] obtain an image of the area of interest; E] by the following way The first (X) axis determines the deviation of the first reed from the second layer: [i] the intensity of the first series of sweep lines (iO) extending across the region of interest, the scans The line (y) substantially hangs the first axis; The center point [ii] calculates a degree of symmetry of substantially each of the scan lines (y) according to the measured light intensity; '[iii] is calculated according to the symmetry value calculated in step E[ii] The deviation value of the first (X) axis; F] determines the deviation along the second (y) axis by [i] extending substantially across the region of interest along the alignment and substantially perpendicular to the alignment a second series of scan lines (x) of the first axis to measure the light intensity; [ii] calculating a degree of symmetry substantially for each of the scan lines (x); 20 1286196 [iii] according to step F[ii] The symmetry value calculated in the calculation calculates a y-axis deviation value. 10. A method for measuring a stacking error, comprising: A] selecting an initial candidate center point of a first scan line, B] measuring a light intensity of the scan line; C] comparing the initial candidate center Pointing the measured light intensity on one side and the measured light intensity on the other side of the initial candidate center point; D] calculating φ the first scan line based on the measured light intensity along the first scan line An initial first symmetry value; E] selecting another candidate center point on the first scan line; F] calculating a symmetry value of the candidate center point selected in step [E]; G] by repeatedly performing the step [E And [F] and determine the candidate center point with the smallest symmetry value to find a center of symmetry point; H] repeatedly perform steps [A] to [G] for additional scan lines on the mark; A symmetric callout is calculated for each of the centers of the symmetry points; and J] determines a minimum symmetry value from step [I], and then calculates a mark deviation based on the minimum symmetry value. 11. A measurement system comprising: an imaging member for acquiring a marked image; a light measuring member for measuring light reflected from the mark; and a symmetry calculating member for measuring the measured Light calculates a symmetry value; and 21 1286196 deviation calculation means for calculating the deviation based on at least one symmetrical value. 12. The system of claim 11, wherein the imaging member comprises a camera. 13. The system of claim 11, wherein the symmetry calculation component and the deviation calculation component comprise a computer, the computer is programmed to perform the following steps: A] selecting a mark on a first scan line The initial candidate center point; • B] measures the light intensity on the scan line; C] compares the measured light intensity on one side of the initial candidate center point with the measured light intensity on the other side of the initial candidate center point D] calculating an initial first symmetry value of the first scan line according to the measured light intensity along the first scan line; E] selecting another candidate center point on the first scan line; The symmetry value of the candidate center point selected in step [E]; G] finds a center of a symmetry point by repeatedly performing steps [E] and [F] and determining a candidate center point having a minimum symmetry value; Steps [A] through [G] are repeatedly performed for the additional scan lines on the mark; I] calculating a symmetric value for substantially each of the centers of the symmetric points; and J] wherein the deviation calculation means includes the computer, The computer will be For the minimum of the asymmetry value is determined from the step [I], and then by symmetry according to the minimum value to calculate a deviation mark. twenty two