TWI326476B - Differential critical dimension and overlay metrology apparatus and measurement method - Google Patents

Description

1326476 IX. Description of the Invention: [Technical Field of the Invention] The general semiconductor process of the present invention, in particular, monitoring and control, is used in lithography and etching processes for microelectronics fabrication. [Prior Art] During the manufacture of microelectronics, the semiconductor wafer is processed by a series of machines, and a lithography process and a subsequent etching process are performed to form features and devices in the substrate on the wafer. This process is used in a wide range of applications, including semiconductor fabrication, flat panel displays, micromachines, and disk heads. The lithography process causes the mask or reticle pattem to be spatially modulated (aerial image) into a film of photoresist (also referred to herein as impedance) on the substrate. These absorbed spatial image segments, whose energy (called actinic energy exceeds the critical energy of the chemical bond in the photoactive component (PAC) of the photoresist material, produces - latent image _ plus mage in the impedance) In the impedance system, the latent image is directly formed by ρΑ〇; in other (called acid-catalyzed photoresist), the photochemical reaction first generates an acid, and then reacts with other photoresist compounds during exposure baking to form a latent image. In these examples t, the latent image marks the volume of the impedance material, whether in the development process

4IBM/04134TW 6 1326476 (during the example of positive photoresist) is removed or left after development (in the case of negative photoresist) to create a three-dimensional pattern in the resistive film. The resistive film pattern produced by the subsequent etching process is used to transfer the patterned openings in the impedance to form an etched pattern in the underlying substrate. It is important here to be able to monitor the realism of the patterns formed by the lithography process and the etch process, and then control or adjust these processes to correct any defects. The manufacturing process therefore involves the use of multiple metrology stations and the monitoring of pattern features formed on the wafer. The data collected by these measuring machines can be used to adjust lithography and etching process conditions to ensure compliance with product specifications. Referring to Figure 1, a general lithography and I-insulation manufacturing line 10 is disclosed for fabricating a semiconductor. - or a plurality of wafers are processed in the direction of the manufacturing line 1Q by 1 。. - Photocluster u〇 contains a lithography machine, with a built-in machine for deposition and fire, baking b阻抗 round impedance (10), so t〇〇i) 1 η, the pattern is imaged on the wafer, such as exposure Machine m), and after baking and developing the exposure pattern to the impedance thief, the wire is exposed _纟113. After the use, several sets are used to measure the pattern features formed on the impedance. For example, H(〇LM) 120' is used to confirm that the ugly formed on the impedance layer can be aligned with the _ on the crystal. Scanning electron microscopy MSEM;) 130 ′ - the key dimension (cd) of the measurement characteristics

4IBM/04134TW 7 1326476 Width. The measured values from the measuring machines 120, 130 may be communicated with the light cluster 110 r· \ and the surname cluster 140 (which generally includes the surname cavity 141), as represented by the data flow path ·, the path 135, according to these The measured value adjusts the process conditions. These measurements are evaluated in a deposition step 125 where it is determined whether the wafer 5 must be reworked. The photoresist in the repair process 101 is stripped from the wafer 5 and sent back to the cluster 11 Oh, in the modified lithography condition, the impedance pattern is covered again. If the impedance pattern meets the product specifications, the wafer 5 can continue to be processed by the money cluster 140. This decision is typically based on a limited number of measurements per wafer, such as about 2-3 stacks of measurements in about 2 locations per wafer, and only 1 CD measurement at 5-10. A limited number of measurements is needed to maintain a reasonable throughput of approximately 30 seconds per wafer or 1 wafer per hour. If wafer 5 meets the requirements for overlay and CD measurements, wafer 5 can continue to be processed by etch cluster 140, at which point the impedance pattern is transferred to the wafer substrate again, again, resulting in a pattern on the material. Measurements are made on a line 130 or an atomic force microscope (AFM) 150 measuring machine. The surname measurement obtained from the subsequent measurement machine may be sent back to other online machines by the data flow path 135 to adjust the process conditions. Regularly, a wider range of offline measurements 25 may be used similar to online

4IBM/04134TW 8 machines such as OLM GO, SEM ISO and AFM 150 and others such as thin thickness measurement (FTM) 160 and electronic probe measuring machine (EPM) 170 are expected on all wafers More point measurements are taken to obtain more measured values, so, referring to Figure 2, a more satisfactory assumption is that in the wafer processing system 20, 'a light cluster 110 may contain other metrics such as FTM 160 and OLM 120'. Machines and methods, such as Scatter Measurement Measure (SCM) 180 and Microscope (MCR) 185, may be more helpful, and are not provided in general. While it is assumed that the wafer processing system 20 will increase the ability to measure compared to a typical system, it adds complexity and expense. In recent years, 'called' scatterometry, the technology has developed 'optical measurements for periodic structures without the need for complex hardware such as SEM or AFM. The principle of scatterometry is about the details of small patterns. Information can be extracted from the reflection of the grid pattern or the zero-order diffraction energy. The general SCM 'uses the reflection energy from the pattern on the wafer and compares the signal of the reflection energy to determine the pattern characteristics. The SCM has speed and The advantages of simplification, but the need to build a large library of signals for the matching of the reflected signals. This signal library is expensive, time-consuming, and requires a computer server 190 and related databases to perform the alignment. Measurements may also be added to the offline measurement system 25 'to improve the quality and quantity of information' and subsequent lithography and

4IBM/04134TW 9 Control of the etching process. For example, Littaau et al (U.S. Patent No. 6,429,930) describes the use of scatterometry to determine the center of focus, by measuring-diffraction signatures, and at different angles of incidence, wavelengths, and/or phases, A summary of the library of the miscellaneous notes. _, scattered (four) amount is dense. Ten counts require a server farm and a database containing the signal library, thus increasing complexity and cost. For the independent parameters of the film stack and the target pattern, the amount of the side of the gap needs to be determined, and the degree of understanding of the prior knowledge of the fine layer and the target pattern features is generally unknown. For general-purpose bulk material, the difference is _ifferential. When it is applied to CD measurement, it is susceptible to noise interference: such as illumination, wavelength, controller response, and target alignment measurement; such as film thickness and light characteristics. Process changes. The general scatterometry is also limited to the zero-order diffraction. This zero-order diffraction detection is used to interpolate the thickness of the green film, but it is used to distinguish the signal characteristics of the target CD and film stack. And there is a poor signal bottle (signal tG nQlse ra (10). The target of the scattering measurement must be large enough to make the illumination energy within the target (ie the illumination must be all within the target area) compared to the typical CD Or overlay the target, the target of the scatterometry has more wafer areas. Furthermore, when these target features become more isolated (the ratio of the target CD to the target period is reduced), the scatter measurement is gradually degraded. Because of the degree of isolation, cd is out of focus

4IBM/04134TW 1326476 (de-focus) sensitivity increases, the ability to measure out-of-focus, critical lithography process parameters require measurement of isolation characteristics. It is expected to control lithography process conditions (such as exposure dose and out of focus) to ensure image quality. The primary determinant of lithographic images is a surface on which the exposure energy is equal to the critical energy of the photoresist of the impedance film. Exposure and focus are variables that control the shape of the surface. The exposure is set by the illumination time and intensity to determine the average energy of the spatial image per unit area. 1. The reflectivity and appearance of the substrate (7) can result in The local variation of the exposure, and the focus set by the position of the photoresist film corresponding to the focus plane of the imaging system determines the amount of modulation reduction relative to the image of in-focus (in-focus). , may cause local changes in focus. The microscope (MCR) 185 can be used in conjunction with specially designed metrology targets to monitor dose and focus, as detailed below. lithography of semiconductor fabricated wafers, Depending on the control of the lithography process, it ensures that various pattern features fall within the general process range 0r〇cess whdow). The process scope is parameter space, and the tolerance of all patterns is consistent between the applicable range of the process. Therefore, proper measurement and control requires two basic parameters in the lithography process, especially dose and focus (out of focus). Chi amount, while the lowest loss of raw image deterioration level domain

4 Old M/04134TW 11 Aberration. The control of lithography must be based on the measurable _ attribute response to dose and out of focus. Therefore, Lin's is the dose and focus on the process control line. - A description of the method of __quantity and loss of f, shouting, through the use of the focus exposure matrix (fQCUS expQSU "em FEM". Form the test T case of the light field (four) 'in the towel grating component verification fine focus and agent The processing of 1 is based on the characteristics of the lithography process to describe the characteristics of the lithography process. Scanning electron microscopy (SEM) or optical machine is generally used to make patterned wafers (such as FEM wafers). Imaging performs measurement of pattern features. However, performing SEM measurements is expensive, relatively slow in operation, and difficult to automate. Methods for obtaining doses and focusing using a microscope have been described in

Ausschnitt et al. (eg, C. P. Ausschnitt, SPIE, Vol. 3677 pp 14 (M47 (1999)'' for the determination of dose and out-of-focus for on-line lithography control,';

U.S. Patent No. 5,965, 309 to Ausschnitt et al.; U.S. Patent No. 5,976, 740 to Ausschnitt et al. Ausschnitt et al. have revealed "dual-tone", the measurement target (referred to as "schnitzls") for depicting dose and focusing characteristics. The "hue" of the lithographic pattern is made of the presence or absence of an impedance material. Decided that it is typically deposited on the surface of the substrate of the wafer or thin 12

4 Old M/04134TW 1326476 in the film 'and will be side later. The pattern is an impedance shape on a clean background, or an impedance shape that is absent in the background of a resistive material. Complementary color production is formed by the face-to-face interchange in the lithography process. Preparing a mask having an opaque and light-transmissive region according to a shape or space to be built in the resistive material, and then using a source from one side of the mask to illuminate or project the shape of the mask or Space to the impedance layer on the other side of the mask, these tonal patterns may be built into the impedance material. The two-tone metric target disclosed by Lu Ausschmtt et al., the dose and focus have a difference in response to the age difference and the advantage of the hybrid, which can be measured by a microscope. A further advantage is that the phase system can also be used to measure overlays, doses and focusing. However, the sensitivity of schnitzl_ to the gamma focus bias is roughly symmetrical, resulting in a positive or negative value for the focus bias. Furthermore, this method requires a high-quality microscope and focusing ability. It is necessary to obtain schnitzl〇metry and overlay the exact image of the sin target to obtain the required measurement. Obtaining accurate, accurate, and focused images will increase the time required for measurements, making measurements susceptible to process and environmental changes that can occur in the clusters and clusters of light. The general stack-to-measurement is also dependent on the microscope and is susceptible to similar mirror-month quality, focus, and process variations. In particular, the use of a microscope causes a source of error (to〇l_induced shift, TIS) error.

4IBM/04134TW 13 1326476 Error caused by machine calibration and optical alignment change, wafer-induced shift (WIS), under-layer or stack-to-target error. Therefore, for the measurement and control of patterned lithography and etching processes, there is still a need for an inexpensive, fast, on-line method and system; one is substantially sensitive to the pattern size in a single layer and relative to the pattern layer, and The method of patterning and underlying film laminate and substrate film insensitivity or film thickness is insensitive to the problems and deficiencies of the prior art. Therefore, it is an object of the present invention to provide an integrated measurement system comprising - On-line measurement and control machine, test pattern and evaluation method for determining lithography and etching process conditions and overlay error, whereby the pattern group can distinguish exposure, focus and side problems' and borrow - second scale scale (10) The measurement of the two-dimensional stacking error of the semiconductor project process and the measurement of the two groups are carried out. SUMMARY OF THE INVENTION The object of the present invention is to provide a method for evaluating lithographic parameters such as focusing and exposure, and paste parameters such as rate and isotropic, which are inexpensive and easy to use. The purpose of this is to provide a fresh-device that determines key dimensions, rim characteristics (such as sidewall angle, thickness loss), exposure and focus conditions, stacking error, and film thickness characteristics.

4 IBM/04134TW 14 Another object of the present invention is to provide means for determining lithographic parameters and currency parameters to maintain desired efficiency. There is still another object and advantage in the present invention, and a part of it is obvious to those skilled in the art, and a part is explained in the specification. SUMMARY OF THE INVENTION For the prior art, the foregoing and other objects and advantages, by the present invention, provide a method for measuring the size of a substrate 7 on a substrate. Nominal _, this nominal pattern contains a series of features with a major pitch in a main direction, wherein the 仏_ case is characterized by a nominal feature size, the nominal feature size is in the main direction (eg The x direction) is repeated with the main period p and has a predetermined variation perpendicular to the main direction (such as the y direction). Corresponding to the nominal pattern 'this nominal' is formed on the substrate - the target _, such that the target pattern has a substrate feature size corresponding to the nominal feature size. The characteristic size of this pattern is the size of the 瞧 ( ( ( ( ( ( ( ( ( ( ( ( 。 。 。 。 。 。 。 。 。 。 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽 纽Provides the size of the focus (ie, the measured size) and the _straight direction—the difference between the detected changes in the number of diffraction stages, and the value of the nominal feature size.

4 Old M/04134TW 15 Deviation of the characteristic dimensions of the substrate. Measured by one or more non-zero singularity series based on - or more than zero singularity series 'measures one or more non-zero-order diffractions, and the number of related changes' and then according to the provided (four) lining Focus on size. In the case of multi-wavelength or broadband illumination, the change in the parallel direction (such as the X direction) - or more non-zero, the level of diffraction, the number of stages is determined to determine the variation of the diffraction intensity for the target size and contour characteristics. Response. Preferably, although the present invention contemplates the possibility of using any or more, detectable non-zero-order diffraction orders, the intensity variation of the non-zero-order diffraction series can still be accommodated. The method of the present invention allows for the measurement and control of pattern contour features and dose and out of focus by enabling wavelengths or broadband counties. The invention is suitable for the measurement of critical dimensions (CD). While the two-tone target pattern is used in accordance with the design of the present invention, the present invention provides measurement and control of lithographic parameters such as dose or out of focus. The invention is also suitable for the measurement of overlapping pairs. In the example of the overlay measurement, the phase and intensity variations of one or more non-zero order diffraction orders are used. In another aspect, the present invention provides a device for performing metric measurement, comprising a Koda field source for illuminating a target pattern; a detector for detecting a change in one or more non-zero-order diffraction orders; and a means for fixing Substrate; a means for locating the source, the substrate and the detector, causing the source to illuminate the target pattern, and causing the detection of the light to detect one or more non-zero-order diffraction orders, the radiation pattern

4IBM/04134TW 16 Target diffraction. J's device can be money-step and "second-system configuration, for the zero-order number of radiation of the pattern target diffraction, and includes - means for positioning the second _, the second detector is relatively Two detectors _ zero-level number light # 衧吏仔. To, and - means, based on the detection of zero-level, . and number: the second size of the final note. For example, the film thickness 1 is determined by (iv) zero-order diffraction from the substrate of the substrate, or an unpatterned area of the target. Once another aspect of the invention provides a measure of the difference measure, the difference is configured to operate on a wire during semiconductor manufacturing, such as the manufacture of a circuit semiconductor. The apparatus of the present invention includes means for determining a deviation between process conditions (e.g., dose, out-of-focus or money engraving speed and isotropic) and nominal process conditions based on one or more non-" and VO-number changes" And the difficulty of providing the process conditions after the P process to scream the deviation of the decision of the miscellaneous parts. [Embodiment] Difference key size (Χ: ττ > See Fig. 3 'Describe a single, integrated' according to the present invention The optical degree (ΙΜ) machine port is 2〇〇, which may be used to perform sequential or simultaneous measurement, stacking and film thickness. The 1M machine of this machine may be pl〇y)

41BM/04134TW 17 1326476 In-line processing system 30, and IM machine 200 eliminates the need for SEM machine station 130, OLM machine 120 and FTM machine 160 or SCM machine 180 (see Figure 1). , or 2). The IM machine 200 may be configured to be integrated with a lithography process or an etch process machine, so it can be measured during on-line processing. This novel difference target and measurement method is provided by the IM machine 2 to provide in-situ CD and overlay correction at each measurement point. The IM machine 200 is used in conjunction with the measurable measurement target to quickly and reliably measure more measurement points on all wafers in Lu without increasing cost and complexity, and compared to conventional techniques. The method can maintain or increase the throughput of the wafer. For example, it can be expected to perform at least 50 measurements on each wafer while still maintaining 1 wafer throughput per hour, if converted to Move, Align and Measure (MAM) time, then It takes about 0.5 seconds to measure 50 times on each wafer. At present, the maM time is about 3-5 seconds at each wafer measurement point. The IM machine can be deployed in the off-line system 35 instead of the FTM machine, the SEM machine, and the 〇l machine. Taiwan, or SCM machines, thus reducing overall cost and complexity. In accordance with the present invention, an integrated metric (IM) system is described that includes its method of clothing and target structure for performing optical measurements of CD, dose, out of focus, and overlay. The phase device may be used with all appropriate measurements in conjunction with appropriate design target structures and methods. The IM device and system of the present invention may be used

41BM/04134TW 18 1326476 Hunting by reflection or scattering in the common method of the month b (such as the two kinds of reflectometer, (reflectometry), ellipsometer (eiiipsometer) or scatterometry (scatterometry) is suitable for obtaining other Measured values such as film thickness and pattern profile. The system of the present invention is suitable for use in on-line processing, in photodusters and etch clusters, or during off-line wafer processing. Since the system of the present invention is configured to detect different levels of diffraction diffracted from the design target metric structure on the wafer, the metrics of the present invention and related target structure systems are referred to as "diffraction measurements" for convenience. (diffractometry)". The reference figures are used to illustrate that the method according to the invention 'is not necessarily drawn to scale. An embodiment of the diffraction measuring system 4 according to the present invention is illustrated in FIGS. 4A-4D. An illumination source 410 is preferably a multi-wavelength source such as a group of light emitting diodes (LEDs) or lasers. Or a fimte-band source, such as a xenon lamp, which is projected onto the wafer 45〇- by an illumination member 413 (which may include a reduced object and a collimated object that is not disclosed). A target of 455. In Fig. 4A, the wafer 450 is in a horizontal plane having an x&y direction (in the direction of the y4 plane), and /, the axis of the 曰S plane is the vertical z direction. The wafer may comprise a substrate 451 and at least a film stack 452, typically a plurality of layers of film 452' comprising, for example, a layer of photoresist (impedance) material. Selectively

4IBM/04134TW 19 provides a polarizer 414 which, in the absence of the target 455, can be set to minimize the diffraction efficiency of the diffraction order and/or the reflectivity of the wafer 45〇. In particular, the transverse magnetic field (10) viscous north magneticfield, TMfidd) polarization, which enhances the first-order diffraction efficiency from the grating target 455. The shirt color filter 4i2 is selected to be cut, and the width of the illumination near the main irradiation wavelength is reduced so that the range of the irradiation energy is at least one wavelength of the wavelength Δ. The energy T needs to be wide enough to ensure sufficient signal reflection from the patterned target 455. In contrast, the signal reflection is distinguished from the unpatterned region 475 from the wafer 45 flavor. For the target formed at the impedance, the illumination energy bandwidth; 10 ± Δ λ must be outside the range of the actinic energy so that the impedance does not require additional modification. The Lai ship turns into a completely monochromatic color because the monochromatic illumination may change due to internal reflection in a layered thin layer of pine with a certain thickness. Therefore, the filter 412 provides at least one wavelength of the wavelength; one of the gentleman's △ energy and one energy band, wherein the selected one is, so that the target milk has a non-zero level (track coffee) diffraction series has - mainly The pitch will be described in detail below) and will be collected by a collection light or an objective. The M-planes collected in Fig. 4A collect light at a wavelength that is dispersed into a non-zero order diffraction order and are projected at a normal incidence angle to the side array 46A. In the y-z plane illustrated in the figure, the collecting light member 430 images the target 455 々 in the side array. This _ plane is described by coordinates ("’), with coordinates (χ

4IBM/04134TW 20 y) Description of the substrate plane distinction. An image processor 49〇 may be provided to detect the signal detected by the detection array 460 and determine the size of interest. The analytical metrics used for image processor 490 are dependent on the measured dimensions and will be described in more detail below. An example of a novel diffraction measuring device 40, illustrated in Figures 4-8 through 4D, is used to image a novel diffraction measurement target 455. The novel diffraction measurement target 455 has a main period P (or a comparable pitch P) of the repeating element 6〇1 in the X direction, and the main period p is selected such that ρ > λ. For example, in accordance with the present invention, a target 455 is considered which consists of one or more sub-regions 600 where each sub-region 6〇〇 consists of elements 6〇1, repeated in cycles in the X direction (eg Figure 4B illustrates the size Η in the y-direction (as illustrated in Figure 4B). This goal is suitable for measuring CDs. In accordance with the present invention, the diffractive measuring device 40 is configured and mounted in a similar manner and can also be used to image a stack of targets. In a preferred embodiment, the apparatus 4 is configured to illuminate in the X direction (ie, the direction of the main period P), and the angle of incidence at the target 455 is: ^ = arcsin(»A0/P) (1) relative to the Z direction , where the detector 460 is located. It should be noted here that the condition Ρ>ηλ〇' must be made to make the illumination angle 0 have a true value between 〇 and 9〇. This grading ray 440 will reflect with the 夹 axis clamp _0, while the ηth stage firing 4 ΙΒΜ / 04134 ΉΛ / 21 ray 441 will be substantially parallel to the z axis. Equise (1) is established, and the diffraction of the fourth-order diffraction order 441 is approximately symmetrical with respect to the x-axis direction and angular distribution (angular dustxibution). (2) For the imaging object 430 The nth-order diffracted ray 441 is captured without interfering with the incident and anti-hysteresis 440, and the nth shunt lung line is projected onto the detection array 460. The detection array 46〇 may be one of the conventional techniques of a CCD or other similar artifact. Selecting the irradiation energy bandwidth and the main period p of the earth Δ λ and the target 455, so that the n-th order diffracted rays are in the main periodic direction (ie, the direction, direction) of the detecting array 46G, and can be distinguished from other levels of diffraction; f will weigh 4. It is suspected that if ρ=ι〇〇〇奈米_), the broadband exposure is in the range λ〇±Λ λ=5〇〇 side nano (four) and the incident angle is 30 degrees, with a range of ±36±3 degrees. The first level of diffraction angle. The collection lens 430 is designed to project the y dimension of the target 455 at a magnification 投射 to the detection array. The diffraction seam in the detection column spans a small size 'it is always the target sub-region _ y size Η 。. In the X direction, the target 455 is composed of a solid component coffee with a period ρ interval. For plane wave monochromatic illumination, project to the side array

4IBM/04134TW 22 X326476 Zero-order diffraction energy span / size «λ. ), fine lines are generated by the diffracted light at the surface of the collecting lens 43. The angular width of the principal fringe of the diffracted beam of age 4c is:

A iVPcos(8)(3) constitutes the period P of the target 455 and the number N of the elements 6〇1, and provides an angular dispersion of (4) at the irradiation wavelength, so that the first-order diffraction can be combined with the other-stage diffraction area. For example, a pitch of about 1 micron in the X direction provides a sufficient angular dispersion of about ± 2 degrees. Therefore, the preferred embodiment of the target must be a level of 1 〇 micron in the X direction, for Ν = 1 〇, ρ = ι μm, u'o nano, 0 = 3 〇, then The ω Ξ 33 degrees obtained by the equation (3). The beam on the surface of the lens extends into a smock (9) where ζ〇 is the distance from the substrate to the collection lens. The intensity of the non-zero-level diffraction will vary with the size of the detection array, such as the first-order intensity/; (shown in Figure 491 of the scoop. The span of the diffracted energy in the detection of the brake train 46〇 ( Span) is the length h, which varies depending on the energy of the illumination, as shown in Fig. 4C. Therefore, for the wavelengths A〇, Z (f=10 monochromatic illumination of house meters, the diffraction energy η crosses in __ Distance 丄, (λ600 μm. For multi-wavelength or broadband illumination, in the range of △ λ, the wavelength of the derivation of the divergent system is dispersed to further extend the projection energy in the χ direction. In the broadband riding _ 巾, this causes The projection energy is continuously distributed continuously over the angle ±Δ 0 in the X direction, for the above Δ 0 =

4IBM/04134TW The example of 23 degrees of 23 degrees, the total extension of the amount in the / direction is the heart... soil △ 5 microns' for the case of discontinuous multi-wavelength illumination. The wavelength of the celestial wave can be 16 or overlapped, and the angle of the ship's touch is dispersed. In order to achieve the above characteristics, the numerical aperture of the collecting lens 430 is saturated, and the y-sense wide meets the criteria of Λ^51ηΔθ and called (10). The standard ensures that the collecting light member 43G can capture the χ direction at the divergence angle ±&; 绕 绕 绕 441 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 第二 新颖 新颖 新颖 新颖 新颖 新颖 新颖 新颖 新颖 新颖 新颖Contains a few pieces of 601 - or more finite gratings, sub-patterns, or sub-regions _, yuan

The piece 601 is in the main cycle (the interval is interval, and has at least a design nominal width "°. - the shape of the finite grating _ domain" - generally rectangular, can be borrowed - total minimum times _ height limbs _ long gift The feature of the side member 430 must be included in the full-angle dispersion ΔΘ of the broadband illumination in the x-direction, and does not interfere with incident and reflected rays. In the above example of ^413.3 degrees, ουsay 5 is needed. There is a direct trade-off between Μ and target size, which is needed at λ=7〇〇_ and off 5' Λ#αΐ. In order to maximize the degree of plucking (Ling 匕 focus), it is better to allow paper, Operating at a lower limit in y, accompanied by a diffracted ray that is nearly perpendicular to the substrate, as shown in Figures 4A-4C.

The 4IBM/04134TW 24 array has a physical image (ie side component) size of approximately 1Q. For the image, the magnification m of the imaged object 45 must be at least rib so that the detection array 46G spans 4Q images in the } direction. Thus, for the previously described example of wideband, the projection of the first-order diffraction spans the U,) region. If there are more than one and sub-regions _ separated by - pitch %, the span of the projected image 490 will be increased by the example. For the example of the two sub-regions, the span along the 乂 direction The distance will be Mx((?0+//)' as shown in Fig. 4D. It is disclosed in Fig. 4A 量 _ _ 4Q also scales zero or reflected 440 respectively. If the zero-order ray 44 〇 through a wavelength dispersion light The piece 435' and the angled scattered radiation, line 445 (as disclosed as the non-zero order diffraction order of the light-transmitting light in FIG. 4A) is detected by the detector 480 (such as a CCD array) by collecting light Collected by piece 430, the diffraction measurement system can be used for: a light reflectometer or an ellipsometer and a general spectroscopic scattering technique, and a general spectroscopic reflector or an ellipsometer for measuring the thickness of the film, which is generally used for spectral scattering techniques. A CD of Ρ Ρ Λ 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 The device is 48 〇, and the target size in the y direction is imaged 'it has one or more sub-regions 600 in the y direction. 25 4ΙΒΜ/041 34-ΠΛ/ 1326476 In accordance with the present invention, the design of the target 455 will depend on the determined characteristics, such as 〆CD (including profile characteristics) or overlays. The specific target design will be enhanced for process parameters such as dose and focus, The measured CD and overlay characteristics are responsive to help with feedback and feedforward correction of those process parameters during various patterning steps, ie, immediate tracking during the process. Performing measurement of film thickness can be done without a target pattern' The diffraction measurement system 4G used on the film stack 452 is achieved by the 敝 zero-order diffraction order (reflected ray) 44q · Further, 'multiple targets can be grouped in the irradiation area of the substrate, To simultaneously measure CD, stack and film thickness, as described below. In order to measure CD, target 455 is designed according to the present invention to be formed on wafer _, for process conditions such as lithography and out of focus and money engraving Isotropy, with different responses, which will be described in more detail below. The design principle of the present invention can be more clearly understood by referring to Figure SA and the guilty] / Figure 5A shows an idealized optical thumb 5 Oh, this light thumb has a week Line 51 and line-shaped space 53, which line 51 has a width y, the space having a space width of 53 · secret listening, characterized by the edges of the grating lines described in the cycle of the horizontal axis of the user. The vertical axis 57 in Figure 5A illustrates the relative composite reflectance amplitude. If the reflectance of the line 51 (e.g., impedance line) & reflectance of the exposed substrate in the space 53, the relative reflectance H is performed. This relative reflectance is a function of the illumination wavelength and the angle of incidence. For the sake of simplicity, let's assume

4IBM/04134TW 26 The conversion of the two reflectance states is not _, and the assumption of the vertical sidewalls on the thin components is _. For example, the presence of a non-vertical interest in the conversion region of the variable reflectivity of the strips 4 and _ will complicate the analysis, but does not fundamentally change the result determined by the average π, that is, the average on the side wall of the optical element. Question CD. Non-vertical sidewall effect enabling profile characteristics and the determination of the average π. A finite grating has a periodic corpse, and the cycle household is made up of N long

A component (such as a line) whose degree (or height) is Η (where η is parallel to the y-axis) and the finite grating is illuminated by the amplitude A〇, and the amplitude a0 may be the reflection amplitude of the wavelength function on the finite grating surface a The spatial variation of (x, y) is illustrated by the following equation:

^,y)-A,\Rs+RLS rect yΜ, rect\ W) ® combi recti NP. (4) For λ, this number of diffraction theoretical close values is valid and can be calculated by Fourier transform formula (4) Obtain the far-field amplitude at the ηth order at each wavelength: A (u, v) = A, h §{Uj v) + RLSHWNPsin c{Hv) sin c(Wu)^ Sin c L n where (u,v) is the far-field coordinate from the grating z, defined by (ιΐΞχ/λζ, ν_λ. For η#0, the entangled series strength Λ=|Αη|2 is used by Provided by the following formula:

Respect (5)

4IBM/04134TW 27 1326476 n{u,v) S][nc2(^v)sinc2(.^s.nc2 ^ A^HNP ' 1 IW f·4 . f —, [1 acos(2ttJ^w)] k xsI sine2(Hv)sinc^ ( \-\ NP n tl —— _ 1 pJi NP 1 pj\ (6) Square private (6) detachable for wavelength λ correlation and target component width "related term. , 'v = 〇, J, the plane of the detector 460 in FIGS. 4A-4D, according to the equation (1), the intensity is dispersed in the V direction according to the wavelength, and is imaged in the / direction at a magnification of Μ. W): Α^ΗΝΡ2 7ΓΠ

13⁄4 W|2|l-C〇S 2mWs ~P~. (7) In equation (7), the wavelength and linewidth dependent components are separated. In the 0,) plane of the detector, the intensity distribution 4 (Λ') of the / direction can be obtained by integrating the ash: 八0')

J_Y ^HNP^ PI and (2)丨Α>ηνρ2 (8) 7ΠΊ The intensity distribution in the / direction provides a relative reflectivity | ί^{χ 〇} 丨 direct measurement 'on the other hand, the intensity distribution in the y direction / 〇 /) can be obtained by entering the points:

/ M f HNP2'' (1 1 /ϊ〇+ΔΑ _ \\^(λ}2άλ _^〇-ώΛ 1 —cos 2mW y [m) ^ 70Ί y {2Αλ) P > \ J

4IBM/04134TW 28 (9) 1326476 The intensity distribution in the 乂 direction is a function of r. The radiance of the old and the Canco wire is 1) „, which illuminates the energy to a fraction of a certain number of stages. DEn{w): ^(Λ) = -_ΖιΜ____ί Ρ \A\\hnpJ7,(y) (p'

Kf (do MM2 (10) λ2ί 1] VAA Ί r J JK^WIW L \ p J Figure 5B shows that the diffraction order is n=〇, i and 2 are the diffraction efficiency sinks (7) 56, 57, 58. As shown in equation (1), the diffraction efficiency concept (11)

w (7) 56, 57, 58 are functions of the standard line width ^, in this case, we assume RZ#1. The diffraction efficiency wind (7) 56, 57, % reveals the standard line width ^ 〇 .5 (when the patterned area is 5Q% of the total grating area, this light thumb is called 50% load cycle grating (duty cyde grating n-level diffraction ^ system is located The peak intensity is while the second stage diffraction 58 is at zero.

Figure 6 illustrates an embodiment of a diffractive target design 6G for use with a diffraction measurement system 4 (see Figure 4a_milk) for measuring critical dimensions (CD) in accordance with the present invention. CD target 6 〇 contains __) element 6 〇 1 - sub-pattern area ' N elements are spaced by period p (here, period P is measured in the X direction 'which is the direction of the main period of the target pattern) ' The central axis 65 of the grating element 601 is measured between turns, wherein the grating element 6〇1 has a length of 卩 in the y direction, and the real direction is in the direction of the target period (i.e., in the X direction). Preferably the width of the design of each element 6〇1

41BM/04134TW 29 1326476 degrees ~ linear change in the direction of the 卩 and the dimension H:

V ξΑ ± Α and 2 , is the size where ζ is the midpoint of the thin angle between the series and the y-axis, where % (eight) = my. Within the range of print linearity adjacent to y0, the print width % is determined by the following formula:

JV(y) = (y~ym)tanC + - 2 (13) where ym is the maximum or minimum measurable position of the diffraction at the non-zero level.

The image of the first stage diffraction measured by the detection array 460 and used for the tapered raster CD target 6 of Figure 6 is illustrated in Figures 7a-7f. The imaging lens 43A of the diffraction measuring system of the present invention assumes a magnification M, typically 50-100, within the range required to solve the X, y intensity. Figure 7A illustrates an example of monochromatic illumination, plotted on a plan view of the debt measurement array 460, detecting the intensity of the image intensity (;c,,y) within the region 710. Figure 7B illustrates the intensity (less ') of Figure 7A, which is the integral or sum of the detectors in the 兀 direction. Figure 7C plots the intensity of the intensity of the image in the y direction of the detector y in the case of monochromatic illumination (〇 790 ', which is determined by the divergence ω of the light member 430 (see the front program ( 3) shown) - wide production 2 / 〇). For the example of broadband illumination in the range λ 〇 ± Δ 》, VL and white

4IBM/04134TW 30 integral or total strength 妒) 795, spanned by the angular scatter - - length Z / (; U soil △ again) (as shown in the front program (2)). FIG. 7A illustrates image intensity/ten, /) 78 782, tear, respectively, in the case of multi-color, discontinuous, wavelength ^ is an illumination example, which is plotted on a plan view of the detection array 460 of the side region 71〇, And its first level image is dispersed on the array 460. The intensity center (y) of Figure 7D is the integral or sum of the directions along the X of the tester. The intensity of the specific implant order as the wavelength function, the intensity distribution 781, 782, the tip of the tear, or the zero, may be different. In the example of multicolor, discontinuous illumination in Fig. 7, the intensity distribution W, 78, 782, 783 intensity w) 791, u is the integral or total intensity along the side ^ direction. For broadband illumination of fine λ ϋ ± Δ λ , the integral or sum of the intensity centers (1) 795 along the y direction is also plotted in the figure. Since the diffraction occurs only in the patterned area of the target, the intensity of the detection is zero except for the image title, the height field, and the length of the postal 710. The condition w)=p/2 depends on the detected non-zero-level diffraction, which is consistent with the peak of the debt measurement intensity (odd-level diffraction) or zero (even-order diffraction); after that, compared to the design A change in the value of %,) will result in a positional offset relative to the peak or zero around the fixed size. In the example of the first-order diffraction, each -

4IBM/04134TW 31 1326476 is a measurement of the measured size #(/〇) relative to the center of the target,

.= Zl±>V image center. The peak position of 2 is provided. 'W'oXCD) of equation (10) can be obtained by the following equation / + Ρ 2* (14) The measured peak position can be measured by the known equation r. Column (9), and then Equation (13) is obtained for qy) expansion:

, (y: a〇, _y'J = a0 f Γ 2m ao' 1-cos -k MP (15) In equation (15), except for a〇, /m, all parameters are known, therefore, strength The position of the maximum or minimum A may be determined by a suitable curve fitting method 'if only %, meaning as a free parameter, equation (15) is measured as a least square approximation of 7 々.) ). When there is a fixed and midpoint of the image center, the target component is deleted by the print width 〒. ), determined by equation (M). Another embodiment of the target design 80 is disclosed in FIG. 8, which uses a two-centered-under region 'region i and region 2, each of which contains a meta#, 802' repeated in the X direction with a period p, available For the display of the hourglass ^• grating element 8〇〇' or a cylindrical grating element not shown. i 801 of area i and component _ of area 2, with a distance ^ ^

4IBM/04134TW 32 1326476 The position of the design center of the knives and the knives of the knives and knives is measured. The position of the two peaks of each other is measured, thereby doubling the measured Weidu. The advantage of the step-by-step is that the position of the image center is ^, 2 is not _read due to the known pitch size % (G〇(4)J does not change with the process conditions, and the target design order is 2.) The measurement of the maximum lateral rotation of the stage, ... is determined by the curve preparation method of Equation 05), so that in the image center %) = %) = "(6) The target component width is measured by: (16), CD, y. ) is determined by the known imaging size of the target size of the magnification of the magnification]^ and the ruler. The target design of Figures 6 and 8 is used in the 7-year-old that is close to the focus of ρ/2. For the conditions required to enable optical measurements, _ (3) is set - lower limit. Fine, in the other side of the gamma - the boundary of the target element can be described by the array of secondary components. (4) The pitch 々 (cut) is comparable to or smaller than the fine structure of the circuit pattern. : The distance between the pupils is determined to have the target of the circuit pattern to be printed - the wavelength at which the measurement is provided, and the ray can be diffracted, ^

4 Old M/04134TW 33 1326476 Rough, the main pitch P is limited, and the _-like pitch of the circuit pattern may be significantly smaller. Determining that there is sufficient target sensitivity (equal or greater sensitivity to circuit pattern) for process variations. This example of target 900 is disclosed in Figure 9A, where the target component frequency, 902, 903, 904 is similar to the component of Figure 8, X. 〇2, which is described by a tight nest line parallel to the main period p, has a fine period 々 and a width of about 々/2. In the example of the target 900 of the fine pitch, the target 9〇〇 is composed of the sub-region labeled as the region 丨 and the region 2, and the region 丄 and the region 2 are respectively arranged by the component 9 which is spaced by the period p in the x direction. 2 composition, wherein the element 901 along the y direction (ie, the region P} between the region 1} and the 9〇2 (ie, region 2) is high in a similar manner to the target 8〇〇 of FIG. The complementary color w-period pattern is labeled as region 3 and region 4, respectively composed of elements 903, 904 at periodic intervals in the y-direction, wherein the points in region 3 and region 4 are separated in the y-direction to follow the y-direction. The predetermined interval 仏 is separated. The separation distance 91 between the complementary color sub-regions is not important for the design of the target _ but sufficient to properly separate the side signals. By observing the diffraction measurement, according to FIG. The print target of design target 9(8) now contains three different reflectivities: 1} the reflectivity of the thinner elements 9〇1, 902, and the width, for example, by impedance

4IBM/04134TW 34 line exists to represent; 2) thinned complementary tone element 9 (B, 904 reflection rate · Rr, and has a width, for example, by the opening / or groove in the impedance; and 3) The reflectivity of the region 9〇5 is Rs′ and is filled with tight nested parallel lines 9〇9 and spaces 908. Referring to FIG. 9B' is an enlarged view of the circled area 9〇6 of the area 1, respectively showing the first and second areas (areas of the remnant material) and the second=fine shapes 901 and 902 , which is described by the end of the nominal period 々<&lt;p's tight nested #2 908 (i.e., the region of the linearly patterned material having a line shape having a width s of preferably about 々/2). Its length is in the direction of the main cycle household (ie, the X direction of Figure 9A). As shown in the detailed diagram of region 906, in FIG. 9B, the target shape regions 9〇1, 902 are tapered, by shifting the position around the end of the spatial line 908 at a continuous pitch. ~ Fixed increase w. The circled region 907 of the region 3 is illustrated in the enlarged view of Fig. 9C, showing the first and second tapered spatial elements 903, 9〇4 (areas of the residual pattern r·material), in a nominal period々 &lt;&lt; ρ tight nested space - 909 (i.e., the region of the line-form residual patterned material having a width of two: preferably about 々/2), the length of which is along the main period ρ Direction (ie the X direction of Figure 9). As shown in the detailed diagram of the looped region 9〇7, as shown in Fig. 9C, the target space regions 903, 904 are tapered, by shifting the position around the end of the space line 909 in the continuous section. From ^

4IBM/04134TW 35 fixed increase ~. The effective thinning angle of the line between different reflectivities is provided by ο. The target design of Fig. 9 is to make the width of the tapered shape 90 902', and simultaneously measure the Ζ with the width of the tapered spatial shape 9〇3, 9〇4. The extreme position of the non-zero-order diffraction pair (/y complex, w, Lm2) and (y rmh y rmd can be borrowed by f_~iG〇~G1T) tanCs 2 J (17) The target element in the image k is determined by the following equation Width ^, r(y〇) Sun丄r = - L'Tml+y, z-Tm^ where 〃 Μ According to the region (10) of the spatial region 3 in which the mask target is tapered in Fig. 9A, the printed substrate pattern is printed. The flat view ugly is shown in Figure 10A. Revealed in the plane view _ on the plane of the figure _, indicating that the thinned space 9G3 ' (according to the shape of the mask disclosed in Fig. 9A) has a width '' by the shape of the impedance _ (root _ % shape line 9 (9) It is formed by staggering the end of the line. On the x_y plane along the line a_a, the profile of the profile is revealed in the picture brain, and the 憎 聊 具有 has the structure of the main chat P. On the y_z plane along the line M, the profile touches the 3 lines reveals In Figure loc, the structure of 10()9 has a main period P/. In the cross-sectional view, the structure is in the impedance film on the substrate 450, and the substrate is packaged with a layer 452. 1〇B and Figure 1〇c in the Shixi wafer as above and above

4IBM/04134TW oxide layer 452 is indicated. In general, the thickness of the impedance 1〇〇9 and the film stack 452 is much smaller than the thickness of the germanium wafer, i.e. &amp;ί()χ&lt;&lt;4ί. Figures 11 through 16 provide operational simulation examples of the target pattern 9 of the present invention and the associated printed structures 1〇〇1, 1〇〇2, 1〇〇3 disclosed in Figures 9A-9C, 10A-10C. The response of the diffractive measurement system of the present invention, as well as the measurement of the CD to the general process variation - the dose and focus changes in the formation of the impedance image, and the thickness of the oxide layer and the impedance film - are described in the co-authorization and filing date. The simulation is carried out for U.S. Patent No. 10/353,900, issued on Jan. 28, the entire disclosure of which is incorporated herein by reference. For the _ consisting of _ _ _ _ _ _, the simulation depends on the micro-age system and the diffractometer to measure the diffracted energy, resulting in the imaging of the impedance towel. The dimensions, pitch and transition of the mask elements, the characteristics of the impedance, the optical characteristics of the lithography system, the lamination of the substrate, and the wavelength of the illumination in the diffraction measurement are all chosen by the user. In the ® U to 16 _ sub, the positive impedance of the prototype has a thickness of _ 25Gnm to 35Gnm, a reflectivity of 1 &gt; 73 and a critical mode (the hearing material is assumed to be corrected by the seam exposure - the partial impedance is equal to or greater than The use of the critical mode is consistent with the assumption of the vertical impedance of the specific impedance material. Assume that the wafer contains a -Shi Xi substrate with a reflectivity of 3 passes of 5, where the image component corresponds to the amount of Wei, and the oxide layer below it has 600 nm.

4IBM/04134TW 37 thickness and 1.46 reflectivity. The target is based on a two-tone target pattern, similar to: the target of A: 9 〇〇, but with a smaller number of diffraction arrays _ , 兀 9 〇 9 〇 2 or complementary tonal elements 903, 904), in the X direction Repeat, the slinging piece (9 W and the syllabus are touched by the end of the ', nest-like τ ο οf (similar to the line 9 (10) of Figure 9A and the 9G8 of the work room). &amp; some times along the y direction In the cycle &amp; repeat, the main miscellaneous of the Congxian-group domain, that is, the region|region contains the wireless array of the elements 901 in FIG. 9, and the region 2 contains the wireless array of the elements 902 in FIG. Region 3 contains the wireless array of components in Figure 9, and is separately simulated. In the direction of the main period of the tapered main diffraction array (in the X direction), the target_ has - pitch 〇 nm. Each of the main diffracting elements (such as 9 (n, 9Q2, 9ω, and 9 () 4), in the direction of ^ (the direction of the width of the elements, the direction of the thinning) is described by the sub-cultivation _. Each-array: The underlying element 909 has a -th pitch of ~=25 〇nm and a sub-width of s=125 nm. The width of the thinned main array elements can vary from 35 〇 nm to (4) coffee, and the y direction The direction at which is increased by δ = 1.25 ηιη to the printed array (such as region 1, 2, 3 or 4) from the tapered portion (such as the printed shape corresponding to the design target region 90 902, 903 or 904, respectively) A simulated diffraction, which is similar to that described in Figure ι.

4IBM/04134TW 38 1326476 A space tree raster system - mask background collision and - pattern conversion 〇 simulation, - shape element grating is converted by - mask background 〇 and - graph wood conversion 1. The designed 5〇% load cycle width y is /2=500nm on the mask, but the 50% load cycle width r(y〇) of the print is due to the line shortening effect, while her large mask size is used for the button. The critical mode of the impedance pattern, the size of the line shortening is a feature of the spatial image and the nominal dose. In order to ensure that the 5〇% load cycle width of the printing is in the center of the array element in the latent image, the thickness of the mask must increase with a predetermined amount of line shortening, wherein the latent image is formed in the impedance under the nominal (four) . The simulated lithography exposure system assumes that I has a numerical hole (four) of 0.7, a partial adhesion of 〇 6 and an illumination wavelength. The nominal dose is normalized to G.32, which is the dose that is fully exposed to the large sputum region. The simulated developed image was calculated as a condition for the exposure dose of the normalized dose varying between m + 10%. These three dose conditions were simulated according to the defocus of 〇, 100 nm and m. The calculation of the final zero-order and first-stage diffracted signals of the simulated developed image assumes that a plane wave of 300 nm-700 mn φ is irradiated with θ=3 人, and consists of ΊΈ and ΤΜ polarized light of the equal stem. The calculation of the second-stage diffracted signal is based on the assumption that the plane wave illumination of the 300nm-400nm bandwidth is incident at θ=444 degrees.

4IBM/04134TW 39 1326476 and consists of equal amounts of TE and TM polarized light. The differential line shortening response of the fine-grain features 908, 909 to the dose level out of focus is illustrated in Figure 11, where the features 9 〇 δ, 9 〇 9 are shaped (tear, 902) and space ( 903, 904) The area is green. In the clean field (9), the simulated latent image at the end of the tail end shape 9b is shown in Fig. 11A. In the case of zero out of focus, the tip of the line 915 is at the length ^ and the nominal exposure dose _' Marked as private outline. Note that when the exposure is from (the outline of '') to 0% to +1% of the outline of the goods, the width of the space opening 903 is increased, and the impedance line is shortened (the figure is compared with FIG. 9 and 9C). Wherein the spatial shape is 9〇3, which corresponds to the relative shape of the nominal shape 903 'relatively' for the opposite tone pattern (such as the tear-off, moving), when the exposure is changed from the _ _ _ _ _ _ _ _ The space length 4 increases 'the line end space should be 6 formed in the impedance field. Note that at the nominal dose _, the impedance shape 915 has a length ^ which is different from the spatial length center (i.e., longer). On the other hand, in the case of out-of-focus, the length of the end of the line is 9% in the case of Jiesong Cong) and the length of the end of the wire Z/) (in the impedance field (10)) is shortened at 2 〇〇 nm out of focus. This is compared with the length 4, &amp; ‘Now see Fig. 12’ reveals that 〇% dose error, zero out-of-focus, and assuming a resistive thickness of 300 nm in your _ _ sub, 苐 one level (n=l) and second-level (n=2) winding

4IBM/04134TW 40 Various eyes 2 (7) (determined by equation (11) ^), the edge is in the direction of the direction Ld (...' direction) and in the average wave ·, direction 'for the zero level and the first - spinning ^, is 3〇〇-7〇〇nm, and is 3〇〇4〇0_ for the second level. The true diffraction angle of the towel * in the equation is required to be between the degrees to limit the allowable wavelength band. The diffractive performance of the tapered spatial targets 903 and 9〇4 are respectively determined by the number of stages (4), 2 (2), and (4). The number η = 1, 2 of the transfer of coffee] said that the move, marked as above. According to the present invention, the target sizes "also", %(6) corresponding to the peak positions of the first-order curves ΐ22, (4), and the size %(6), "win" corresponding to the zero position of the second-order curve 1222, respectively, are Independent of the reflectivity of the substrate and target (as shown in Equation 9), and is particularly useful for CD determination and various analysis of dose and defocus, by using the slow-release performance of the measurement (analog data in this example) These dimensions are determined by fitting to the parametric curve of the form in equation (4): DEn(W :aQ>Wm) = a〇

(18) where (less w), the target design size required to make the material ash 2 on the substrate. In Figure 12, the discontinuous data points 1231, 1241, 1232, 1242 describe the simulated values of the diffraction performance, and the continuous line curves 12〇1, 1221 are

4IBM/04134TW 1326476 is the fit of equation (18), and the dotted line 12〇2 ' 1222 is the fit of n=2. The suitable parameters are: Μ η 1 2 LTLT a〇[%] 0.55% 0.52 % 0.14% 0.14% ^m[nun] 0.5692 0.616 0.5785 0.6206

As demonstrated in Figure 12, a wide range spanning from π, from 〇35 to 〇85 μm, is excellently matched, even at the ideal dose (the dose required to print tiny features s = 〇 125 nm) and best focus, The % of the number of shots rm and the signal of zero are significantly offset by the design value % = 〇 · 5 μm. As determined by the maximum value of n = 1, the space end 908 depicting the target shapes 9〇1, 9〇2 is shortened to a width % of 69 nm, and the space end 9 9 of the target shapes 903, 904 is shortened to the width % It is 116 nm. Although in terms of quality, the critical contour of the spatial image is recognized, the spatial image shortening is only about one-half of the measured shortening. Since the boundaries of the regions of different reflectivity are spatially adjusted by the fine features, this measurement is too high for shortening the estimate | so the average of the adjustments is measured. Note that from the table j can be seen = by

4IBM/04134TW 42 丄(10)476 n~2 The minimum mosquito test is described as the shortening determined by the maximum value of n=4. This is because the first-stage diffraction, the second-order diffraction increases the average edge correction, and for the edge adjustment, the different diffraction orders have different sensitivities.匕 Revealing the measurement of the horizontal diffraction order provides useful information based on the coarseness of the edge of the line. Figure 13A reveals the % response to the dose, while Figure UB reveals the response to the focus (B) of the exposure station. Referring to Figure 13A, for shape 90 902, curve 1301 is the response to the dose at n = 1, and curve 13 〇 2 is the response to the dose at that time. For complementary shapes 9〇3, 9〇4, curve 1311 is the response to the dose at n=l, and curve 1312 is the response to the agent at n=2. Referring to Figure 13B, for shape 9 (U, 904, curve 1321 is in response to out of focus at n-1, and curve 1322 is in response to out of focus at n = 2. For complementary shapes 9〇3, 9 〇4, curve 1331 is the response to the out-of-focus on the day, and curve 1332 is the response to the out-of-focus at n=2. For the diffraction order n=1, the response of the 2' dose is roughly linear (Fig. 13A) The response of the water coke is roughly parabolic and symmetric to the best focus (Fig. 13B). The slope of the response to the two target tonal doses is reversed, and the curvature of the response to the focus is the same. Due to the shape and spatial structure, it is possible to distinguish between micro-quantity and focusing energy. For example, 4' can be based on Ausschnitt (US Patent No. 5,965, 3〇9).

4IBM/04134TW 43 1326476 the contents of which are hereby incorporated by reference. Figures 14A and 14B show that the modulus of the diffracted energy is zero (4) and the first (fourth) in the X direction.

The oxide layer 452, and spans the range of the m ^ thickness is W in the detector and the so-called measurement, for the inclusion of the map = from Figure 4A Λ ^ 0 Β rp pj I - G domain 3 and ^ ~ ~ = 0.5 /^72 pl has a design ruler - close to

I _ size value. In the example of (4), the diffraction grating element 435 of Fig. 4 is hunted, and in the example of (4), the axe is linearly dispersed by the printing grating 455. Therefore, the wavelength of the X in the direction of the side of the __, the wavelength of the direction is equal to the wavelength of the x-direction of the gamma of the zero-order of the object 。. For example, the equation (5) is for nG ' The optical error response is a function of the substrate reflectivity &amp; and the corresponding reflectance ~. As shown in equation (10), for (d), the spectral response is a direct measure of the reflectivity of the substrate associated with the substrate. The sensitivity of the diffraction efficiency to the thickness change of the underlying oxide layer is revealed. Figure 14Α 'For the first-order system, it is revealed in Figure ,, which is measured by the width and the edge of the curve shown in the thickness of the oxide layer from 45〇. Between nm and 55 〇 之 之间 , , , , , , , = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = The approximate measure of relative latitude. The zero-order extension is just significantly larger than

4IBM/04134TW 44 1326476 The extension of the first level of 14G1 'zero level is significantly more sensitive than the first level. This quantitative measure of relative sensitivity can be obtained by varying the range of diffraction efficiencies over the midpoint of the range. For the extension of the zero level disclosed in Fig. 14, for the extension of the first stage of Fig. 14Β, the value of the ratio ADEn 1400, wind &quot; ^«0.4 ]T)J\. Therefore, in our example, for the thickness of the lower oxide layer, the diffraction efficiency of the zero-order is three times the sensitivity of the first-order diffraction efficiency. This indicates that for the measurement graph _ sex, non-zero The grade is the preferred choice' and for the lower film thickness measurement, the zero order is the preferred choice, and in fact, in the absence of the pattern, the zero level is better than the first level. Figures 15A and 15B illustrate similar to Figures 14A-14B! In addition to the court, we will reveal the shattering of the number of diffraction orders guided by the dose variation for pattern exposure within the range of the optical dose defined above. Kawasaki is set at the best focus. Then we can H. The extension of the zero rhyme line (10) and the extension of the j-th curve 15Q1 'evaluate the relative greenness of the suspension', and can be derived for the measurement pattern characteristics, The zero order is the preferred choice. As in the example of Fig. 14, the quantitative measurement of the relative sensitivity can be obtained by the ratio of the diffraction efficiency to the ratio of the midpoint values of the norm.

4IBM/04134TW 45 ADEn's zero-order extension 1500, ~^E~n 6DE. ;〇.l ' is shown in Figure 15B first and L exhibition 1501'. Therefore, in the example disclosed in FIG. 15, for the exposure dose, the sensitivity of the zero-order firing efficiency is about three times that of the first-stage diffraction efficiency, and the energy of the sister irradiation is decomposed to be higher. Pointing the zero level is mainly sensitive to pattern changes. Therefore, when the relative reflectance is ~~&gt;. The zero-order sensitivity is toward 〇. Figure 16A and Hall reveal the sensitivity of the change in the hard focus of the target pattern to the 'first-stage firing efficiency' at the nominal value of the two different mask sizes, the maximum focus and the range of 200 nm. The dose is fixed at the ideal. In phlegm 6/Zm, the impression is negative and the sensitivity to the dose is low, as described in Figure 16a. The % m 'printed grating is approximately 20% loaded with a ring, and the sensitivity to the dose is relatively high, as described in Figure 16B. Therefore, a high focus sensitivity is achieved, along with the more isolated grating elements, and the focus control is to measure these relatively isolated structures. On the other hand, dose sensitivity is a powerful function of non-target negative loops. Therefore, we have derived the ideal spectral sensitivity for simultaneous dose and out-of-focus for relatively isolated grating elements, which leads us to a simpler target embodiment, as described below.

4 IBM/04134TW 46 1326476 A preferred embodiment 1701 for measuring a discontinuous grating target of a CD is disclosed in Figure π. This illustrated target 1701 provides one of the preferred means of CD measurement, with no limitation on the nominal CD size. The target grating 1701 is cut into two or more pattern regions, such as a region 元件 (element symbol 1731) and a region 2 (element symbol 1732) ′ respectively containing lines such as 1711, 1712, and having a length ugly in the y direction. The nominal widths in each of the pattern regions 1731, 1732, respectively, are the same as the lines 1711, 1712, but are different in the sub-pattern regions 1731, 1732, respectively (eg, the first pattern region 1731 includes the width) Line 1711, second pattern area 1732 includes line 1712 having a width &amp; This feature mi, 1712 is preferably joined in a continuous region 1750 which can help avoid shorting of the wires of the components 1711, 1712 and also provides elements 1711, 1712 that are structured to support the printed structure. Note that the width of the elements nil, 1712 will generally be much smaller than the width of the continuous region 1750. The separate sub-pattern areas (e.g., 173, 1732) are preferably, but not necessarily, aligned with each other in the y-direction. The features of the following sub-pattern regions are defined and normalized by the pitch: u = (^+w2) 1. 1P is the unknown average width, and the nominal size of the design is ΠΤ. 5A^-W2) 2·1P is a predetermined (design) offset value between the nominal widths of the design.

4IBM/04134TW 47 by as shown. The positive solution % of equations 20a, 20b, with cr = cTexp being the average line width. The square of the contrast C2 is not shown in Figure 21; 9 is dominated. When the light grid is more isolated, that is, hG, the contrast increases. Note that when y — ρ, the raster becomes an isolated space. In semiconductor applications, the accuracy of measuring CDs is on the order of 1 nm. The accuracy of the CD measurement technique of the present invention depends on the sensitivity of c to alpha variations. For the small variation we are concerned with, the rate of change f is provided as: AC Am — — _ C^(21) Obtained from equations (20a, 20b): c&gt;o: dm__F3 ~dC~ 2π C&lt;〇: dw ~dC~ 2π (22a) (22b) where 2C equations 22a, 22b indicate the speed of change with contrast, (23), therefore,

4IBM/04134TW 50 provides a measure of the sensitivity. For a good recording _ quantity is not a companion. With the contrast 'L speed, which is exactly the opposite in the equation,) a small change of preferably 1 can make the contrast change large. The equation is similar, orphaned - the important feature is that the w is wide, and the width is wide (ie, the width w is close to the corpse), and the sensitivity is increased, where 妾 is near zero, as shown in Figure 。. At one extreme, Each grating is characterized by a reflectance ~ one, and the core is defined by the existence of the film at the other extremes, and each of the gratings is a feature of &amp; the hunting is defined by the absence of the film. (4) The wire 3 job has the expected characteristics 'that is, as the feature width decreases, the sensitivity increases. At %_5, the detailed observation of the sensitivity effect is revealed in Fig. 22B. Instead of the equation 21_23, at P=l〇〇Qnm and (4)·15, Corresponding to △&quot;Xie,|Taste' C &gt; 〇.25, the change of CD lnm at 5〇nm in Figure 17 (such as when 5〇TM, then ^ is greater than 15%' or about 57.5 nm, and ~ is less than scratch, or about 42 5 nm), making f &gt; 〇〇 2. The change in nominal contrast 2 〇 /: (change in absolute contrast is 5%) is measurable Therefore, it is achievable to see the financial property of the metrics of the metrics. As mentioned in the play, one of the differences in CD measurement is to form a pattern on the wafer. Good light, can make a difference to process conditions, such as dose and defocus change

4IBM/04134TW 51 1326476 Value response. Other embodiments of the diffractive measurement target of dedication and focus separation are disclosed in Figures 23-29. In the target 23 of FIG. 23, as shown, the difference shape of the four regions 2301, 2302, 2303, and 2304 is defined (ie, the layer of patterned material 'is still present in the figure) and the space (pattern area) 'Where the pattern layer of the photoresist has been removed). The width n%, % of the grating elements 231, 2312, 2313, 2314 is much smaller than the grating period, i.e., m%, separating each element from its neighboring elements. In the first and second regions 23〇1 and 23〇2, the reflectance of the substrate (open space 2320) is 4, respectively, and the reflectances of the grating impurity elements (such as the impedance shapes 2311 and 2312) are respectively, and the third And the fourth regions 23〇3 and 2304, the reflectances of the surrounding regions (such as the large impedance region 233〇) are respectively off, and the reflectances of the grating space (substrate) elements 2314 are respectively. Note that these reflectances are "effective, reflectivity, which is affected by the contour characteristics and edge effects of the narrow feature regions. There are two pairs of oppositely toned regions, that is, the substantial opening of the sub-pattern regions 2301, 2302 (removal) The pattern area) is 232 〇, and the area 2330 is substantially filled (ie, filled with the pattern layer material such as impedance) composed of the sub-pattern areas 2303, 2304. The two pairs of two-tone areas (232 〇, 2330) can be processed separately. To measure the shape and space of the isolation according to the invention described above. When forming a latent image or a developing pattern in the impedance, the dose is increased from -10% (24〇1) to 0% (2402), the isolated shape width

4IBM/04134TW 52

WL iT

Degree 2 decreases in the direction indicated by the dashed arrow 2405, as shown by the curve of Fig. 24A, in the opposite direction (shown by the dashed arrow 2406 of Fig. 24B).

The isolated shape width of w - w^wA is r 2 , which increases as the dose increases from -1〇% (2411) to 0% (2412) to 10% (2403). On the other hand, the spatial dimensions (curves 2411, 2412, and 2413 of Fig. 24B) and the shapes (curves 24〇1, 24〇2, and 24〇3 of Fig. 24A) are in the same φ depending on the change in out-of-focus (or focus). The direction changes. Figures 24A and 24B illustrate the application of the measurement method of the present invention to the simulation of the focus-exposure matrix (F〇cus_EXp0sure Matrix) of the target 2300 of Figure 23 applied to the spatial elements 2313, 2314 to determine the same nominality. At the time of the dose, the size of the shape elements 2311, 2312 is printed. Methods of dosage and defocusing are described in the publication of Aussdmitt (U.S. Patent No. 5,965,309) or C.P. Ansschnitt. The resolved dose and out-of-focus of on-line lithography are disclosed in Pr 〇c spffi, v() 367 ent.: 140-147 (1999), which is incorporated herein by reference in its entirety. In the case of a two-tone color, the response of the 230G side of the wire is inferior to the process for the process, and in the ugly size produced by the process, the difference between the nominal change measurement can be changed by the above-mentioned class. Binding to and out of focus. - The hue is characterized by an opaque (or dark) line, or a characteristic of a clear (or bright) field (presenting a patterned material that still exists), and the opposite color

4IBM/04134TW 53 1326476 The pattern is characterized by a opaque depiction. The clear (or bright) characteristic of the month (or taste) field and the difference CD / dose / poly ^ ^ ^, another example of the 払 2500 reveals that the square = V: this 'one in the first and the first Secondary pattern _. Bu 2502, residual patterning. The effective isolated shape area of the material 2511, f has a nominal width ^ (5) 'by reflectivity _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Good and have the shape or line of the enamel, the area where the patterned material is removed) the end of the thief, the area of the lion, the length of the lion is 'the length along the main Zhou Shu (ie, the X direction of Figure 25), and In a manner similar to the shape area smoke and gauge, perpendicular to the effective shape area 2511, plus the edge, the shape area view, 902 is drawn by the thin periodic line of Fig. 9B. In the secondary_region 2503 ' 2504, - the effective riding space area (the area where the patterned material is removed) 25U, 25M (the nominal width %, (6), the reflectivity ^ characterizes it by the nominal period) The net parallel narrow square of the household (i.e., the line with the width of the patterned material removed, preferably having a width) is formed rigidly, the length of which is parallel to the direction of the main period P, and is similar to the spatial regions 903, 904. Description, perpendicular to the edges of the effective spatial regions 2513, 2514, the spatial regions 903, 904 are defined by the thin periodic line 909 of Figure 9C. Viewed by the diffraction measurement 40, by a fine period parallel to the rectangular line and space

4IBM/04134TW 54 1326476 The area covered by the line 2523' is characterized by the effective reflectance %. In a similar manner to the target 23 of FIG. 23, the two pairs of regions (25 (U, 2503) and (2503, 2504) in FIG. 25 can be processed separately, as described above, according to the present invention, U.S. Patent No. 5,965,309 to Ausschnitt, which measures the effective isolation shape %, %, and the space width, the effective width of %, the effective isolation shape %, % for the dose call, and the effective separation space %% response to the dose, and they The response to the out-of-focus is the same, as shown in Figures 24A, 24B and Figures 27A, 27B. According to Ausschnitt, US Patent No. 5,965, 3,9, the apparent response of the shape and spatial structure makes the age control and the shadow Dosage and Focusing. The benefits of Target 2500 in Figure 25 are related to Figure 23 of Figure 23, which is disclosed as follows: 1) the shape and the end of the space have sensitivity to dose and accumulation; and 2) compared to The wafer pattern needs to be determined by other process steps, such as CMP (Chemical Mechanical Polishing), to maintain a more uniform pattern density in the figure. Others are measured by the difference (:1) of the pattern formed in the impedance, so that the target of the Moonlight &amp; 1 cover and χκ focus difference is based on the starik〇v (integrated circuit measurement, double detection and Process Control IV (1990) vol. 1261 Exposure Monitoring Structure (SPIE) and Inoue, et. al (Dose Sensitivity Design of US Patent No. 6,251,544, published June 26, 2001) and Suwa (US Patent No.)

4 Old M/04134TW 55 1326476 4,908,656) and Ausschnitt (US Patent No. 5,953,128) focus sensitivity design. Starikov and lnoue promote sub-resolution assist features (SRAFs) by using sub-resolution on the mask to greatly enhance sensitivity and suppress focus sensitivity. Ausschnitt and Suwa introduce thinning The end of the wire is used to enhance the sensitivity of the out-of-focus. These designs have been adapted to the differential CD measurements of the present invention as previously described in Figure 26. In Fig. 26A, the 2600 system is designed to have a dose sensitive area % 1 〇 and a defocus sensitive area 2620 '. The dose sensitive area 2610 is composed of two pattern areas, a region 1 and a region 2, respectively, having repeating elements. And the out-of-focus sensitive area 2620 is composed of two pattern areas, a region 3 and a region 4, respectively, having repeating elements. Target 2600 has a major period p in the X direction and has N repeating sub-pattern regions 2630 (for clarity, only two repeating sub-pattern regions 2630 are shown). The open patterned regions or spaces (between the remaining patterned materials) have widths %, %% corresponding to the secondary pattern regions 26u, 2612, 2623, and 2624, respectively. A mask design 2650 for forming an element region 2630 is illustrated in FIG. The upper mask region 2670 includes a primary feature 2651, 2653 that abuts the secondary resolution promoting features (SRAFs) of the primary features 2611, 2612 for imaging on the wafer, respectively. As illustrated in Figure 27A, the width % of the space in the dose sensitive area 2610 is very sensitive to the dose, as shown by curve 2701 (example)

4 Old M/04134TW 56 1326476, as described by Starikov and Inoue), and the %% change in width does not change rapidly with the out-of-focus, as shown by curve 2703 in Figure 27B. By contrast, the focus sensitive area 2620' formed by a mask design such as 2680 includes, for example, tapered lines 2661, 2662 (as described by Ausschnitt and Suwa) to form patterned regions 2623, 2624, respectively. The % change in the space of the focus sensitive area 262 is a function of dose, which is similar to the curve 2702 of Figure 27A and, therefore, relatively insensitive to dose changes but rather sensitive to out of focus, as depicted by curve 2704 of Figure 27B. . In accordance with the design of other diffraction targets for achieving focus sensitivity in accordance with the present invention, phase shifting elements on the mask as evidenced by Brunner (U.S. Patent No. 5,300,786) are used. The same is true of these embodiments in that they rely on CD or overlay measurements. A variety of designs have been adapted for use in the differential measurements of the present invention as previously described, as well as the measurement of the difference stacks described below. In addition to dose and focus control, the objectives for overcoming specific applications can be gathered in Figures 68, 9, 23, 25 and 26, and other described embodiments. For example, the target includes a series of width distances ' to approximate the wedge element 6 disclosed in Figure 6 which is provided in a wide range of CDs, a means of linearity of the quantitative lithography process. A comparison of wafer measurements with mask measurements on this target, called quantitative MEFF (mask error enhancement factor, mask _r

4IBM/04134TW 57 enhancement effect), the application of FF effect is critical to the cognitive source of CD changes in sub-wavelength imaging. The target embodiment can be designed according to the characteristics of the specific patterned layer, and the figure reveals that the seed has a fine 々 in the X and y directions, and the target 诵 including the complete contact hole surface is formed to form a larger width with a nominal width 仏Element 2811, and 2812 having a nominal width gray, has a major period P in the &amp; direction. In addition, Fig. 29 discloses a target 29GG having regions 1 and 2 of two pattern regions, wherein the background reflection region is composed of thin spatial lines, and the thin spatial lines have a fine period dan, parallel to measurable The versatile period p of the grating elements 2901, 2902, the measurable grating elements 2901, 2902 respectively have a nominal width center; and another object embodiment 3 of the invention is disclosed in Figure 3, which makes the difference of the present invention The diffraction measurement can be applied to a similar pass-through (thr〇ugh_pitch) CD test1. The dependence of printed CDs on the measurement of pitch or period is critical to the decision of the Optical Proximity c〇rrecti〇n (〇pc) rule. The 〇PC rule determines the correction of the product mask design to determine that features of different lay lengths are printed to the same size. The limitation of the 〇pc rule is that the general SEM CD measurement method is slow and laborious, especially in the case of SEM CD measurement, which prevents the CD from being sufficiently consistent with the applicable range of the process. The target of Fig. 30 is composed of a plurality of difference gratings 3

4IBM/04134TW 58 3003, 3004, each of which is similar to the design disclosed in Fig. 17, in which the periods Pa, Pb, Pc, and corpus d' are changed from one to the next difference grating, respectively. In the monochromatic illumination λ〇, the target 3〇〇〇 from Fig. 30 will occur as shown in Fig. 31 'zero-order intensity/this, / book, / this, / like the path along the path, and the first level of strength // b, heart, / ^ is collected by the light member 430. However, the first-order intensity 7ya of the sub-region 3001 is not collected 'because the pitch of the sub-region &&lt; The relationship between the angles can be expressed by the following grating equation: • ^2_ = sin0 + sina p (24a) The period of the grating period of the non-zero-order diffraction measured by the device must conform to the case of A^x&lt;Sina&lt;7y^x, where :

_ηλ〇 η\ sin0+ NAX sin0-NAX For shallow angle illumination Θ =70. And satiety = 〇 · 5, corresponding to the maximum collection angle a max = ± 30. The range of the first-level detection is 〇 7 and the household < 2 3 is 〇, so the DUV source with a wavelength of 2 〇〇 nm can be used, which can be measured in the range (10) Na. A range with a longer period can be achieved with long wavelength illumination. - Multi-wavelength or wide-area can provide the full range of the button, the period is between 15G and chat. The period defined by the equation is smaller than the diffraction at the zero level. Figure 32A reveals that the first level of intensity 7 from the target 3_ is not detected by +

4IBM/04134TW 59 In the three periods ρδ&lt;ρ, the outer band listens to the plane view of __. Since the diffraction angle varies with the period, the intensity of the turn-over is along the χ, and the direction is staggered' corresponding to the direction of the various wavelengths of the detector. The measurement can be made by measuring the intensity/integral or summation in χ, (m is achieved in each-period CD, as illustrated in Figure 32B. Therefore, the present invention can achieve measurement over a wide range of pitches) . The zero-order detection path 44 of the device 40 in Figure 4A enables simultaneous measurement of CD and film thickness. As disclosed in FIG. 33A, for a target similar to the design of the target 1701 in FIG. 17, the zero-order image of the plane view of the ccm side is divided into regions and passes in the direction of y, corresponding to the target 1701 in FIG. The patterned area Π 3 卜 1732, and the area 3305 corresponds to the unpatterned area 175 目标 of the target (4) in the figure. The intensity spectrum of the unpatterned region 3305 can be used to measure film thickness. As shown in FIG. 33B, 'the unpatterned image area 3305 corresponding to the unpatterned target area 1750' between the two raster image areas 3301, 33〇2 (corresponding to the target areas 1711, 1712, respectively) along line AA Zero-order intensity spectrum / 乂x') 3307 has a unique mark, depending on the nature of the film __ the refractive index of the Z-th film 屯 (again), kz. ( λ ) - the true or imaginary composition - and each Depending on the thickness of the film. In the case of the known η, (又作), kzU), the thickness is regarded as a free parameter, and the multi-layer response is generally used to measure the spectrum.

4IBM/04134TW 1326476 The thickness is determined by the fit. In the case where the value of ❹(1), ki.U) is unknown, the thickness can be determined by the general formula of the dispersion behavior in the fitting formula, such as the Cauchy formula. Of course, in the case where the non-zero-order diffraction does not exist, the same method can be used to determine the film properties and thickness by using the detector gamma of the shock in Fig. 4A. The difference stacking method of the present invention is disclosed in Fig. 34. Explain the ability to illuminate from two opposite directions, all along the direction of the main cycle of the target, ie, the X direction, where 'Fig. 34A illustrates the illumination from the target 455 in the negative χ direction (the direction of the main period), and Figure 33 Β illustrates Illumination of the target 455 in the direction of the direction, whereby the number of diffraction orders from positive and negative can be detected by a single device. The detection is accomplished by configuring device 40 to reposition illumination 41 or reposition target wafer 450 to achieve an appropriate illumination direction with respect to the target. When the angle is θ = _@) and the illumination is from left to right, the +1 diffraction order 441 can be detected. When the angle e = arcsin(-f), the illumination k is right to left, then the -1 diffraction order number 44 Γ can be detected. The ability to detect both positive and negative diffraction orders is necessary to measure the overlay error and will be described below.

Old M/04134TW 1326476 explained. : 曰 曰 private grating 35 〇〇 (similar to the ideal CD of Figure 5A: the photographic repeat is disclosed in Figure 35a, where, for the sake of clarity. Only (four) - a repeating element. The stacked pair of target grating 3500 is printed on the crystal The circle ' contains two features 3501, 3502 in the cycle household, one of which is formed by a patterning process A having a width K, and the feature 3502 is formed by a process B having a width. The patterning process B may represent the first layer of the _ characteristic gamma, and the figure is

Process A may represent the second, overlapping layer of features. For the aforementioned CD catalog, feature 350 3502 may be comprised of an array of lines, trenches, or smaller patterns, as long as these features are applicable to the printed process layer. The decision about the width % of the period P is measurable stacking error. In the picture, the ideal object is surveyed. 3. There are two lines in the water weikezhi. 35 〇 〇 35 〇 2 in its period: the line measurement has a width % and the line pass has a width %, and its cap is separated and edited. The value sand is designed to have a nominal value of i/2. Therefore, the X-direction overlap error &amp; between the two features 3501, 3502 normalized to the period p can be expressed as . =^-1 p 2(25) The vertical axis 35〇7 of Fig. 35 is illustrated as having one True amplitude] and zero phase line

4IBM/04134TW 62 35〇ι ′ and a normalized synthetic reflectance with a true amplitude γ and a line of phase 0〇2, where 0<(5), κ.... The normalization of the reflectance of the line and the arrest is related to the example of the composite reflection of the underlying film stack and the structure relative to the CD (see Figure 5). The target features are formed in a single layer, the corresponding features 3501, 35〇2 It is possible to form different layers in the film stack. Therefore, the "reflection rate" of the two features 3501, 3502 related to the structure is not _, and the amplitude and phase are different as described below. The optical variation of the substrate reflectance A and the associated line reflectivity &amp; reflectance amplitude as a) on a fine optical surface is determined by an amplitude Λ and the following heterogeneous (10), fine grating It consists of a line Ν period with a length Η (length Η is in the y direction, the direction of the main period of the straight line):

Rs+KtS + R

, BS rec\^\rect \rect\

X ~WA

x D ® combi ~ rec(-L·) ® comb rect\ NPt (26) where we define a ^ and ~, the number of diffraction theory approximations: in the ~ effective 'nth level far field secret system by Fourier transform Represented·· 4·(»,ν) ~Χ~ =JisS(u, V)+HNP^inciHv)^

Sine

41BM/04134TW 63 where ♦, }) is the far-field coordinate of the distance grating z, where the grating system is defined as (usX/^ZiVsy/^ in the direction

= -, v = 〇) P ′ The intensity of the plane of the detector 460 in Fig. 34 is dispersed according to the wavelength of the equation (1) in the χ direction, and is imaged at the magnification μ in the y direction. For _, the first order amplitude is provided by the following equation: and the corresponding intensity is: (28)

(29) hniKDx) = \Κ{λ)\2 \ + γ\λ) + 2γ(λ)c〇s(±^i + φ{λ))' where, at a single wavelength, the following definition is used: p (30a) m WB sinc(-

P =cos' WA sin K again) ^BS (^) and 〇&lt;m) illusion and -π<咐

Rea/( and Ay (again)

RasW 3bs (^) ^AS (^) (30b)

(30c) In multi-wave illumination, equations 30a-30c are branched as a function of their associated wavelengths at each wavelength, whereas the amplitudes and intensities represented in equations 28-29 are still valid. In other words, it is determined from the diffraction intensity measurement (fourth) that the four additional unknown parameters, that is, the amplitude of the normalized composite reflectance and W,

4 IBM/04134TW 64 1326476 The present invention provides a means for detailing the amplitude, phase, and landing error based on (4) fine 妓, which will be described in detail below. In Π-. In the example of ', the intensity in equation (29) is similar to the equation (7) m. 4 pairs of errors ^ approximations can be used, this near 撼 money is set on the red CD reading (see equations 20a, 20b). However, in the general case (r&lt;1 and 妗〇), it may be encountered that the product 4 is worthy. Illustrated in Fig. 35A is the correlation refracting power provided by equation (29), and the correlation phase difference between two values (r = 1 and 〃 = 〇 5 respectively), the ε of the pairwise error is &amp; (P 2), diffraction order number == 1 normalized intensity plot 3513, 3515, under these conditions, the normalized intensity of „ = ±1 diffraction series is consistent (ie each curve 3513, 3515 is a pair of positive and negative series) 'In &amp;=0.5 Both positive and negative series have a minimum fixed, but the master is 'max — 'min / π ΒΧ + / π ώ, and decreases as 7 decreases. Figure 35C shows the plots 3517, 3519 of the normalized intensity of the diffraction order „ = at r = ] l and π / S, respectively. The intensity curves 3517, 3519 of n = + l and -1, respectively, in space Separated 'when it is 0 = 0 and r = l, the average value is equal, as shown in Figure 35B, curve 3513, and the minimum value in the opposite direction relative to the nominal value of a P 1/2 is based on =〇 offset. In „ = ±1 series, non-crossing

4IBM/04134TW 65 1326476 Multiple intensity curves 3517, 3519 may also be different, and this distinct function will have a zero crossing at the location of the zero redundancy error. However, note that equation (29) shows that this method will fail when 彡 approaches a multiple of zero *π/2. Therefore, for the change value of 彳, the best decision of the stack error is to be described in the lower deck by the number of positive or negative diffraction orders, intensity change or average phase shift. Figure 36 illustrates an embodiment of a diffraction measurement target 36 suitable for use with a diffractive measurement system milk (see Figure 34) in accordance with the present invention, which can be used to measure the overlay. The measurement overlay pair target 3600 includes a pair of elements 361 〇 and 362 〇, opposite to each other at an angle (tilt ' and repeated with a period p (ie, in the χ direction, the direction of the main period of the target pattern). Between elements 3 (10) and 362 The relative distance (10) of the 〇 varies linearly with the y direction. Each element is designed to be tilted relative to each other so that A(y) is 5 when it is along the y direction, and then referred to as y〇. The designed distance will be: 2 (31) and when there is a stack of errors B, the print can be expressed as: DM=(y-^tm^ + ~ = (y~y0)tanC + l+Ps 2 X (32) where κ is the offset relative to '= 7. In the case of a stack of target targets consisting of several CD targets with a period of P, it will be close to ι〇 or greater.

4IBM/04134TW 66 1326476 Assuming that the monochromatic illumination is at wavelength seven, the image of the grating-to-target diffraction of the grating stack of Figure 36 is shown in Figures 37(A)-(D), which uses a y-direction amplification. The diffraction measurement system 40 is made so that 圮 = tree. Figures 37A and 37B illustrate the intensity of the +1 and "diffraction series, 37〇1, 37〇2, respectively, plotted on the plane view of the detector 46G. This is because the illumination is from the directions illustrated in Figures 34, 34, where β The relative phase of the process pattern is zero. The correlation strengths that are average or total in the X' direction are illustrated in plots 3711, 3712, respectively. For the intensity of 4 = 〇 ' +1 and ] diffraction series 37 〇 3, 37 〇 4 is a plan view of the detector 460 shown in Figures 37C, 37D, and related χ, the average or total strength of the direction They are illustrated in plots 3713, 3714, respectively, because the diffraction occurs only in the patterned region of the target, and the intensity detected outside the region (4) is zero. In the area of the plane of the detector 46, &lt;4(4), extending from 3V to 3V, as shown by the plots 3711, 3712, 3713, and 3714, the intensity in the / direction is the same, but according to the equation (29) And as shown in Figures 35A and 35C, change along y. In, the diffraction order „ = ±1 is consistent, and

/ Λ P The minimum intensity occurs at ^ 2, and the result of a non-zero-order stacking error S will simultaneously produce an offset at the minimum intensity position of the positive and negative series " = ±1, with 2 (see Figure 37A and Figure 37B). The plots 3371, 3712) the minimum value ν' indicates that the design position has the same direction, where. 2 is a fixed parameter

4 Old M/04134TW 67 1326476 defined. In general, the stacking error is expressed by the following equation: ^ =^y-(yv-y'〇) (33) where in this case, however, 'when ~〇' = the smallest of ±1 series (y~〇 Take

Dx[y 2 position and symmetric offset (see Figure 37C, 370% Figure 3713, 37 steps so 'stacked error b is only directly related to the average of the minimum and maximum of the positive and negative diffraction orders. pvp~ ~y〇j (34) ,,._y-v~y+u . ν' _-v' +. , ^v=—r— y where 2 and P 2 are generally achieved using non-zero diffraction Determining the stacking error is consistent with the change in the number of diffraction orders in the function shown in equation (29). At a fixed wavelength of 4, it is possible to match the known dependence of the positive and negative diffraction series. Suitable methods, such as least-squares fit, determine the unknown &amp;, 匕, phase and overlap error &amp; the direction is perpendicular to the target period (y, direction) of equation (29): / +n(y)+/_n(y)-2|i:j2^r^+n[C0S〇+n(y)+C0S〇n(y)jj=^

4IBM/04134TW 68 φ+η(/)=^^ζ2+^ D (Vl=(less 'one less'TM)+, household (ν'-ν,, 1 meaning (8) Μ—tan( + == —~~ ^-tanC+P(- + sx) M 2 μ h K2 xJ (35b) and the remainder, which must be minimized by considering the ruler, x0 and 匕 as free parameters. Figure 38A shows the difference Wave illumination, wavelength 4 + ΔΑ, Λ Λ — Α 正 positive first-order intensity 3801, 3802, 3803 plane view of the detector 460 plane. y in the diffraction intensity of the positive series, dependent measurement The least squares fit is performed at each location of the detection array 460, as shown in Figure 3818. The associated negative first order intensities 3806, 3805, 3804 are illustrated in a plan view of the detection array 460, and Figure 38D illustrates One of the negative series is the least squares fit. A similar analysis can be performed on broadband illumination. When the ® error is wavelength independent, the use of multiple wavelengths or broadband can only be used when the wavelength is greater than a predetermined threshold close to the agreement. Determine the measurement accuracy. «The values that can be used depend on the relevant width of the target component and the reflectance in the illumination wavelength range as shown in equation (30B). The ® pair error is the difference detection position Χ '(4) the specific wavelength overlap error ~ 匕. The standard deviation of the specific wavelength overlap error can be estimated, which provides in-situ monitoring of measurement accuracy. According to the present invention, In the example of CD measurement, another embodiment of the stack of targets 3900 printed on the wafer is designed with, hourglass, or "barrel"

4IBM/04134TW 69 1326476 The heads 39CU, 3902 ' consist of two relatively inclined areas, for the A lithography process, for the B process conditions of 3912, the sub-area side with the nominal pitch The side, as shown in Figure %, is overlaid on the brush target side. The use of the two regions _ and the side will double the measurement sensitivity, and the vertical pitch _ is divided by the center bit g 4 (the plane to the position which is not fine with respect to the target of FIG. 39) demand. Equations) The least squares of p ~2 are based on 2, that is, Μ - 1 ' determines the distance between offset positions. Therefore, relative to the known week (four), the overlap error: G. ε _. (^ 〇 ~ G) tanC * - one (36) revealed in Figure 40 of the discontinuous difference grating target 4〇0〇, provided based on Another embodiment of the overlay measurement principle of the same differential diffraction measurement principle of the present invention. The grating 4_ is divided into three or more sub-regions, such as = 0 (4010), region '(1), region 2 (4_, where each sub-region, the component mask of the A process is %% and positive P-correction, having the same position in the X direction (position of the main period), and the element 2 of the b process has a width % and a tangent pitch P, which is located at a nominal position different from the TMi.

4IBM/04134TW 70 1326476 Direction has a - nominal length Η 'in the - sub-region, region 0 (4010) of material B element 4〇〇2 has a nominal fixed offset DX = 0.5P with respect to the A component. In adjacent sub-regions (such as region chat n) and region 2 (4〇12), the B elements have relative offsets &, &amp; respectively, where 2 &quot; 1 ' is preferably Δι=Δ2 ' Each zone is imaged separately in the detection array 46 of the diffraction measurement system 40, with a magnification Μ in the positive direction, as shown in Figure 34A, taking the first level, and then subtracting as shown in Figure 34A. First level. For monochrome illumination, the image detected by the detection array 46 is illustrated in Figures 41A and 41, where the intensity of each region is approximately the same. The η=+1 (positive) image is illustrated in Fig. 41Α, regarding the region r4111+), the region 〇(411〇, and the region entanglement, and the η=](4) image, regarding the region 1 (4111), the region 〇(411〇-) And area 2 (4112-) is illustrated in Figure 41. As discussed in Figure ,, each area is of size AxZ, where ugly and 7. The positive and negative images are transmitted by making all regions of the difference target 4000 After a certain level of imaging, the illumination direction is sequentially switched and imaged. Each image may be stored and analyzed, and the image sequence may have various changes, including but not limited to, imaging multiple targets printed in different directions, and from Illuminate multiple directions of parallel target periods, or reposition the wafers sequentially to obtain another imaging direction. For each diffraction order, separate target areas (such as 4011, 4010, 4012) are imaged. In the detection array 46〇

4IBM/04134TW 71 1326476 relative area (for +1 series, image intensity / 〇 +, / / 1 + 1, respectively, in the area 411 〇 +, 4111 +, 411 + 2 for +1 series, image intensity / 〇1, 、, A are in areas 4110., 411Γ, 4112-). For the example of overlapping target 4〇〇〇 in Fig. 40, the three values in the relative phase 0 are not (i.e., respectively, 0, π/s, and ΓΜ) and r = 1, by equation (29), six. The strong mm /2 ι pairs the error ~ in the direction of the main period, such as the x direction, the response is tied to Figure 42A_42C. The strength of 曰' is limited to the dynamic range (dynamic coffee (4). This range is the range of the overlap error &amp; when the split A and B process patterns are overlapped on the target side in Figure 40. In the example of the target column in Figure 4, the dynamic range is close, and the response of the relative intensity to the error of the stacking error will be linear within this range. Outside the dynamic range, the intensity-to-money error may not be presented. Miscellaneous. The width % of the target component is preferably chosen to be wide enough to make the proximity effect - eff = not simple, so that the impression can be substantially as designed - preferably 'nominally equal' and selected as the period - The brewing score, % and % is preferably G.2P. For the six kinds of strong Qiuqi three kinds of overlapping sub-regions 4〇11 4012 +1 and -1 diffraction series) stacking error &amp; Normalization varies with a single function of phase 0. It may be possible to classify the singularity of the six degrees of view according to the relative state of the brush I.

4IBM/04134TW 72 1326476 For example, Figure 42A illustrates the intensity response of #=0. Here, for each of the target sub-regions 4010, 4011, and 4012, the positive and negative diffraction orders are superimposed to each other as shown by the intensity curves 421 421, 4211, and 4212, respectively. However, when π/8, six kinds of intensity/〇+i, /1+i, claw^, and y, respectively, the response curves 4220+, 422 (Γ, 4221+, 422 Γ, 4222+, 4222_ are different and This is shown in Figure 42. On the other hand, for # = π/4, /1+, (4232) and /2" (4233) the response is discernible, but /〇ι, π_^/ο+1, / 2_丨 is overlaid on curves 4230 and 4231, respectively, as shown in Fig. 42C. Furthermore, the present invention uses the relative response of the intensities to obtain relative relative and relative amplitudes 7*', which will be described in detail below. + ri + rx Ik: IK _ l + pcos(^+ +^Ί l + c〇s(/+ (37a) 川 +1 _ l + pcos(^+ ~S) 1 + cos〆(37b) ik. ' ' _ l + yOCOS〇- +&lt;J) 1 + cosy- (37c) _ l + pcos(y- + l + cos^~ (37d) where we will define the parameters:

4IBM/04134TW 73 1326476 δ = 2πΑ Ρ P = _ 2r Ι + γ2 ψ + Ίπϋχ + Φ Ρ ψ~ _ 2tzDx Ρ ~Φ (37e) Rearrange equations 37a-37e to get unknowns &lt;, f and p : ψ * = arccos(-) P (38a) _ τ~ ψ = arccos(——) p (38b) ±, 2 C^l + aT^ (1-α2) (38c) where: a = cos δ κ ± soil η Γ± c1 ± _ri —ri 2 Λ: soil-Ci (38d)

K For the particular difference in the overlap pair metric, the equations (38a)-(38c) can be simplified to: Δ

P 4, therefore 〇: =0 and square T+ ψ+ = arccos(——) p (39a) 74

4 old M/04134TW

Ψ = arcc〇s(^~) P where: (39b) (39c) household = 丄 - 1 K ± (39d) = for simplification purposes, we will be limited to 简化 a similar simplified analysis. Λ Now we can solve equations (25) and (37)-(view), and (/0±1, /W2± A means that the error is unknown, the susceptibility and the relative phase: (40) ( 41) The functions defined in equations (38a) and (38b)t W lead to the simulation of the solution, but these solutions can be solved by stacking errors and relative phases: rationality and stagnation. 39a) The ratio of the intensity ratio of _(39d)^ is 'can' determine the stacking error &amp; and phase 0 of positive _ (and theoretically intentional A), and equations (39a)-(39d) need to transform (transformi〇) n:)

+ -P is used in (V strict), and this transformation is in time, with different possible phase systems

4IBM/04134TW 75 (regime) at (-πια) 3 :::, the equation (41) is a fixed value to complete A #私(4()) in Qianxian County, which has a slope of The result is a fixed and appropriate change in the shirt. For the flow chart of (7), the transformation of Lai Mi® 43·45, we define the parameters: ACsC+ ~C~ ά, ψ ~ψ^ +ψ- _π ΑΤξΤ+ ~τ~ C++C- ^ ( 43) &quot; The flow chart of Figures 43-45 describes a phylogenetic tree (1〇gicaltree) in which the various parameters of equation (43) are tested and compared with zero to determine whether 〆 and v 需要 need to be corrected ( Or change) 'The parameters are derived from the measured intensity. Referring to Fig. 43, the parameter Δ(: (block 4301) is first tested, if △ [= 〇, then the first, degenerate, #=0 or π, and the two diffraction orders η = ±1 It is similar. Next, the test parameter _, if block 4304)', is set to be equal to 〇 (block 43〇6), then test parameter c (block 4306). If 5&gt;0 (block 4307), then No conversion by 〆 or factory is required. If ? &lt; 〇 (block 4309), 〆 is converted to ^ + &lt; and f is converted to 4IBM/04134TW 76 1326476

Π-疒' is shown as block 4310. If y &lt; 0 (square line 4311), then 0 is set equal to π (block 4312) and the parameter ^ is tested (block 4313). If ^0 (block 4314), then a conversion to and without Κ is performed (block 4315). If "&lt;〇 (block 4416)' then f is reversed and 〆 is converted to τ-〆 (block 4317). For the example of △ & 〇 (block 4318), a branch is executed (block 4400) to determine 〆, such as Figure 44, again, 'execute a branch (block 4500) to determine &lt;, as shown in Figure 45. Referring to Figure 44, for the △ &amp; 〇, 〆 branch (block 4400), first test c + (square 4401). If 〇〇 (block 44〇2), then test δγ (block 4403) 'if ^&lt;〇 (block 4404), no conversion of 〆 is needed (block 4405), if (block 4406), then test (block 4407), if ΔC &gt; 0 (block 4408), no conversion of K is required (block 4405). If Δc &lt; 0 (block 4409), 〆 is converted to π (block 4410). In short,

If c+ &lt; 〇 (block 4W1), ΔΓ (block 4412) and (block 4416) are tested successively to determine the appropriate transition of 〆 (block 4414 or block 4419). Similarly, for the example of 0, the determined appropriate conversion (at blocks 4505, 4510, 4514, and 4519) is illustrated along the system of Figure 45 and begins at block 4500. By the logic of the conversion of the f and the factory, the equations (39a)-(39d) can be detected by the innovative overlapping stack target 4000 as shown in Fig. 40, and the strength of the six kinds of "two ranks and one series" is detected: team, chemical, The measured values of A, /0-丨, 71-ι, &amp; (as shown in Figure 41 4 IBM/04134TW 77) are used to determine the exact value that can be accommodated. For △, not to move the pure post and, after the conversion, square = 〇) (10) Shaona thin - rounded value ~ has a single - slope and zero intercept, as shown in Figure 46A. To produce this: graph, the selection is increased by αι between au(5) and increased by 1 correction. In the example of continuously changing the target as shown in 11 36, the use of multi-wavelength or wide-band illumination can simplify the intensity at the position from the maximum to the maximum χ. The value of the difference between the error &amp; pairs of relative intensity ratios defined in equation (38d) and the sum &amp; is: 9ε^κ δκ (44) From equation (39), we can get: dV 4πΚβη+ 9ψ\ 'δη-} (45a) 1 , ΰψ + 'dKJ (45b) Replaced in equation (38): soil p3(^)2 ll-(— V \ρ j 2 (46a)

4 old M/04134TW 78 1326476

The ^ and the width calculated by equation (44) are governed by the same transformation on the side of equation (4). As shown in Fig. 46B, for (10) and _, the overlap of the error metrics of the current day and the month (4) is '^ at the time of the intersection, at the time of zero, so that the purpose of this will be ensured; ^ set &amp; ten and measurement techniques The overlap error _ d^x has a sensitivity to rain near &amp; The dependence on ~ is the relative vibration of the two pattern levels a, b and the phase k function, where the two pattern levels a, b define the overlapping targets. Since it does not cross zero, the contribution to sensitivity is not as good as stone K. The measurement of the error requires the detection of the n=±i series intensity in at least two directions to determine the vector component U of the overlap error. Therefore, the diffraction device of the present invention preferably includes material illumination and _ at least two The ability of the direction of the W to diffract the order, the two directions correspond to the two directions of the target as illustrated in Figures 39 and 40. Figure 47A and Figure are diagrams of an embodiment of a diffractive device of the present invention. The diffractive device utilizes a general illumination source 410 (having a bandwidth of ι ± ΔΑ as previously described), an optical color

4IBM/04134TW 79 The 409, the illuminating optics, the optical polarizer 3!4 and the general diffraction series detection ϋ 460 (such as the CCD array), so that the required intensity data can be quickly obtained. A rotatable mirror 398 directs illumination, the initial illumination is along direction 310, and then directed to the mirror (e.g., 301, 302) to direct illumination along path 321 to the diffraction target 455 on substrate 450, which is characterized by There are periodic corpses along the X direction. The rotatable mirror 398 is rotatable to provide illumination from different directions to the diffraction target 455. Thus, for example, in Fig. 47A, the rotatable mirror 398 is guided in the positive X direction by the mirrors 3〇1, 302 as again guided to collect the +1 diffraction order number 441. By repositioning the rotatable mirror 398 by 180 degrees, the initial illumination in the 32 〇 direction is directed toward the redirected mirrors 3〇1, 3〇2 to follow the path 3 ι, so that the target 455 can be illuminated from the negative X direction. Therefore, the diffraction order is 44 Γ as shown in Fig. 30B. The detecting light element 43 and the unnecessary detector CCD1' themselves rotate in synchronization with the rotatable mirror 398 to maintain the non-isotropic (in x_y different) image imaging capability and the fixed plane of the illumination plane of the light member 43 0 . It is possible to provide an additional re-directed mirror (not shown) in the y-direction to also obtain a stack-to-measure measurement in the y-direction. Substrate 45A may alternatively be supported or otherwise secured by a rotatable mirror 380 for positioning the substrate to provide illumination from different directions. A polarizer 314 is selectively provided that rotates along the mirror 398 to provide a corresponding

4IBM/04134TW 80 1326476 Ideal for the polarization of the first-order diffraction performance. The preferred embodiment is to direct illumination to four different directions on the vertical axis. For a target grating having elements perpendicular to the angle of incidence, the first order is diffracted and imaged on the gamma CCD1 in the direction of the vertical grating period, while the wavelength is linearly dispersed in the horizontal direction of the grating period. A second detector array (e.g., CCD 2) may be arbitrarily provided with a 'zero order energy' that can be collected through the dispersing element 435 and the second optical member 436. Furthermore, a third detector array 485', such as CCD3, may be provided arbitrarily, which collects the dispersion element 435 for reflection and passes through the zero-order energy 440 in the direction of the third optical member 486 to obtain additional measurements, as follows Said. A plurality of targets are configured to be applied to the metric device of FIG. 47, and the CDs and overlay targets of the aforementioned main (fourth) state can be grouped together in the single- ❹ process layer to form a synchronized illumination target, cluster, and, corresponding to a specific Manufacturing or process feature applications. One of the aforementioned clusters is disclosed in Fig. 3 by the pitch (thr〇UgMtch) target. In order to establish the target cluster 'the plane of the imaging shot in the vertical direction of the main target period/period, the neighboring gratings in the rich direction do not interfere with each other'. Therefore, the cluster can be massive in a single direction (only limited to the side seen) The main target is touched and stacked in the vertical direction of the illuminating surface. _ 'The main target has a period parallel to the illumination plane, and the wavelength of the main target's diffraction intensity is dispersed and extended, requiring illumination

41BM/04134TW 81 The same orientation of the plane, the main target will not be placed adjacent to the other main target. "Isolation of the stacking target in the illumination plane is achieved by changing its orientation. * The main targets of the simultaneous illumination are parallel Or perpendicular to the illumination plane 'is the best isolation between the main targets in the illumination plane. An effective cluster configuration that enables simultaneous measurement of stack, CD and fine thickness is disclosed in Figure 48. The invention is shown in Figure 48 An embodiment of the difference to the target 48〇〇, possibly having two stacked pairs of gratings 481〇, 483〇 in layers B and A (immediately using the grating element of process B and using the aforementioned printed grating elements of process a) Each of them is similar to the grating of Fig. 39, with a double grating sub-region, wherein the X-azimuth grating 4810 has an X-direction pitch, and the y-direction grating 4830 has a y-direction pitch. In most of the examples. 'Ρ0Ζχ=Ρ0办=p〇i. The same target 4800 is also designed to incorporate B-level gratings 4820, 4840, each of which is similar to Figure 6, 8, 9, 17, 23, 25, 26 or 28-30, divided Don't have a medium-distance CDx, Ρα&gt; γ, which is in most of the examples.

The sequential images 4901, 4902, 4903, 4904 on the detector 460 (CCD1) may appear in Figure 49B, depending on the orientation of the rotatable mirror 398, for example, assuming that the mirror is rotated clockwise Direction 322, which is sequentially fixed to positions 1, 2, 3, and 4, as shown in FIG. 49A, is formed for each of the 4IBM/04134TW 82 images 4901, 4902, 4903, and 4904, respectively. The overlay image process includes storing the intensity in each of the grating regions, each orientation and the surrounding, and solving the stack error of each pair of raster regions 481〇, 4830 using equations (35a)-(36). . Embodiments of the overlay device 340 and the difference CD illustrated in Figure 49 may be assembled to perform general spectral scattering techniques and film thickness metrics. By making both sides of the rotatable mirror 398 reflective, the zero stage 440 of reflection in each orientation can pass directly through one wavelength dispersion element 435, second light element 436 to second CCD detector 480 (i.e., CCD 2). In the case of conventional or differential grating targets, the zero-order spectrum can be analyzed by general scattering techniques to determine the characteristics of the pattern (CD, sidewall angle, etc.) and the underlying film. As previously noted, the zero-order intensity spectrum of the unpatterned target area, as shown in Figure 49, can be used to measure film thickness, wherein the unpatterned target area is image 4905 that is revealed on CCD2 array 480. The CCD2 image 4905 is disclosed to have an oval shape because it is a substantially circular illumination of a zero-order intensity of 41 Å, and this illumination 410 is imaged in a direction perpendicular to the period of the transmission grating 435, and in a direction parallel to the transmission grating period. The wavelength is dispersed and elongated. All of the target images may be protected by a third array of debt detectors (ie, CQ33), and the third detector array and optical member 486 are configured to collect zero-order energy 4906 for target pattern recognition and wafer alignment. , or other measurements, borrow

4IBM/04134TW 83 ^26476 This does not require the target light element to be solved. The apparatus Mo of the present invention, configured in the manner shown in Figure 47, provides the ability to simultaneously direct CD, overlay and film thickness measurements at a single target location, yielding approximately twice the measurement speed in product applications, wherein these three measurements It is required. Although the CD and film thickness do not need to have both positive and negative diffraction series measurements, it is possible to average the positive and negative secrets and to simulate the accuracy of the measurement. Therefore, in some cases, Wei Feng can guide the measurement of CD and film thickness from the stacking. In either case, the difference diffractive measuring device of the present invention and the green cluster target layout can have a wide range of applications' ideal for measuring the throughput and capacity. In the absence of any target pattern on the wafer 450, as shown in FIG. 5A, the embodiment of the present invention 40 may be used for the general spectral film thickness measurement 'where the second side device 480 (FIG. 5〇c) A zero order spectrum 5001 of collected and predetermined film thickness properties (true or imaginary components of the 'refractive film rate at the measured wavelength), as known in the art for the paste film and the thickness of layer 451. In the absence of the target raster, no non-zero order diffraction orders are collected at the first debt measurement location 460, as shown in Figure 50B. The optional device 340 may be configured with a light member Yang and an observation detector c(10) for (4) observation and adjustment (via adjusting the mask 5〇〇2) to position the illumination corresponding to the target position on the substrate. As revealed, light is in the detect

4IBM/04134TW 84 The image is formed on the detector 485, which is reflected from the surface of the conversion grating. An optical diffuser (not disclosed) placed in the zero-order beam path 371 may also direct light to the imaging system. In a preferred embodiment of the diffractive measuring device 340 (Fig. 47) of the present invention, 'as shown in Fig. 51' provides adjustment mirrors 302, 303 to determine first order diffracted rays 441, 44 (depending on the X direction, respectively) The positive 321 and negative 331 illuminating rays will be substantially perpendicular to the substrate surface 475 such that the diffraction order is directed to the first detection array 460. When the center wavelength in device 340 and/or the main pitch of target 455 changes, as shown in FIG. 47, the angle between incident ray 321, 331 and first order diffracted rays 441, 441 according to equation (1) One will change. In order to maintain a fixed first-order diffraction direction Θ such that the first-stage detector array 46 〇 may be positioned in the direction of the vertical substrate surface 475, it is preferred that the adjustment device 34 has a low mirror 3 〇 2 The height h of 3〇3 and the inclination angle ξ (relative to the vertical or z-axis), as detailed in the geometric relationship of the device 340 of Fig. 51, in the middle: Η ', h = roc〇\0 (47) where K) The horizontal distance from the good center horizontal position to the low mirror 302 303 is read from the first detector array. As shown in Fig. 4C, the angle ^ shows the wavelength dispersion of the target of the illumination. Note the many implementations of the device 34 of the present invention.

4 Old M/04134TW 85 Μ中's available function. It does include, but is not limited to, replacing a fixed beam with a single rotatable mirror, properly directing the beam by different means, such as a dithered mirror, and adjusting the fiber channel to properly direct the beam, roughly to the target. The substrate is rotated for the center to replace the detecting optical instrument. Figure 52 illustrates a flow chart incorporating one embodiment of the object design process of the present invention. First, a target design pattern is provided. The target design pattern has a key pattern to be printed (as in the case of semiconductor fabrication, the minimum value of the circuit features ash and pitch), and the process begins (block 52〇〇). Collecting and imaging object 413 for an anisotropic diffraction measurement (see FIG. 4A) provides a selected illumination center wavelength λ 〇 and a band width Δλ and a correlation low, iV^y 'the grating target 455 of the present invention will It may be decided to ensure that the first stage diffraction is isolated from other diffraction orders and can be detected (block 521A). For example, the height 11 of a grating region (or sub-region) (the direction of the vertical main period 0.7 乂 P) is preferably greater than about 'where the numerical aperture in the imaging direction is better between 〇〇5 and 0.5, and when It is preferably about 0.2 when compromising between focus, target size and correctness of the figure. The main period of the grating λ μ is necessary for the condition A &gt;|«|zU to collect the total light |"|△, λ &lt;ρ&lt; is preferably Ν. Spectral, numerical aperture factory is necessary. For „=±1 And; ^2 to,

4IBM/04134TW = α5 is preferred. The number N of repeated elements is preferably about 10 or more. Regardless of the application, the target design is further dependent on the CD or overlay metric (block 5225). For a CD target, the component size is preferably based primarily on the target feature key size % (block 5240). For example, as shown in Fig. 17, 'on the design of the CD target 1701, there are two grating sub-regions 1731, 1732, in which the elements 1711, 1712 respectively have a width. Here, ash &gt; % + Δ, and) ^ = % Δ, wherein preferably, the difference between the sub-grating target width ^, the heart △, 0.01 Γ (^ Δ &lt; 0 · 25 ash, And preferably about 〇1%. In order to enhance the sensitivity of the dose and the out-of-focus, the grating element (ie, the target 2500 shown in Figure 25) may be surrounded by the secondary element, which is perpendicular to the main pitch. Direction, with period Ρ /, where /7 (tick 7y«p. For secondary resolution assistance

Subresolution assist feature (SRAF), as shown in Figure 26A

And the target grating 2600 of 26B, the SRAF gap is preferably limited to less than (1 + σ&gt;, wherein λβ, σ6, and muscle are respectively the wavelength of the exposure light source, the coherence and the numerical value. Aperture. For the agent of Figure 26, the target resolution of the target 2600 must be less than the resolution of the exposure source. 0 If the target type is used for the measurement of the overlay (block 523〇), then design the number of critical dimensions for the design. Low dependence, but limited by

4 Old M/04134TW 87 system (ie λ' NAy) resolution factor or acceptable substrate surface on the wafer: (real estate) 'Therefore the size of the stack will depend on the main pitch, and for the stack The target 4000 is illustrated in FIG. 4A. For example, the nominal gap between the A process component and the B process component (ie, the horizontal bundle or the vertical direction A) is preferably 0=〇5P and The corpse is 〇+△ and is the corpse τΔ' where △ is preferably between 〇〇1 household and 〇25p. The width %, % of the individual elements of the stack to the target is preferably j^^#〇 25jP. The apparatus of the invention of Luben (for example, as illustrated in Figures 34, 47, 5, 51) can be used to perform multiple measurements at multiple metric locations on the wafer, as illustrated in the flow charts of Figures 53-55. induction. An example of such an application is a metrology module 2 on a displacement or etching machine, as illustrated in Figure 3, but the integrator 200 of the present invention can also be used offline 35. Referring to the figure brother, a substrate having a plurality of positions is loaded and aligned with the machine (block 5300). When inputting the machine, it can provide the number and type of measurement positions of the wafer, and there is a difference between the targets (the secret amount of the target cycle), which can be used to measure the machine-port. The type of goal. A confirmation is first performed to determine if the wafer has a pattern (block 5310). Typically, if the wafer is patterned, all of the patterns on the wafer will be aligned in a similar, comprehensive manner, so the rounds are generally aligned by appropriate movement and rotation (block 5320), for example Ting can use the observation detector 485 (such as CCD3), as shown in the figure

4IBM/04134TW 88 1326476 47. If the wafer is not patterned, the alignment step 53i2 may be skipped. The wafer is then positioned such that it is to be measured [positionally illuminable (7) block 5330). After that, the mosquito in the target position will make (4) the analysis path (square side). If there is no position pattern (5346), only the zero-order measurement path 5346 is performed. If the position is patterned, a non-zero level measurement can be performed by path 53 方块 to square illusion % or 5360, or zero path can be obtained by performing path 5347. Execute zero and (4) and analyze the smoked measurement target or the general scattering technique target (square side) of the present invention, as shown in detail in Fig. 54. If the location includes a CD diffraction measurement target in accordance with the present invention, then the CD analysis path is performed (block 535〇), or if the location includes the stacked diffracted measurement target of the present invention, then the overlay analysis path is performed (block 5360) ) 'Details See Figure μα, 55B for details. The same machine can then be used to continuously analyze all of the selected measurement locations (block 5390), and after all locations have been measured, the wafer is unloaded (block 5399), or continue along the displacement or etching machine program. See Figure 54 'If a zero-order measurement is appropriate for the selected metric position (block 5345) ' then a general film thickness measurement or diffraction test 1 ' may be performed depending on whether there is a target (block 5400) . In either case, the properties of the film 5403 will be provided as an analysis of the obtained zero-order reflection measurements (block 5405). If no target exists, then

4IBM/04134TW 89 1326476 is the etch rate and isotropic, sensitive sub-area, from the average cd measurement (block 5516), to obtain the process conditions (block 5517), where the average CD measurement system uses process conditions and CD changes ( The block (5) is obtained by a module which is exemplified by the relationship between the Aussdhnitt method described in U.S. Patent No. 5,965,309. , ° Fine by the intensity analysis as a function of wavelength (ie / (χ,)) to obtain process conditions 'such as dose or defocus measurement, where the wavelength function is the sum or average in the y direction ... signal library (square MU) can be provided by experimental green, for example, by using a -3⁄4 exposure matrix (FEM). or by analog means. This signal library can be compared and adapted to the intensity spectrum (block 5523) to obtain process conditions such as metering and line, or _speed and isotropic 1* (blocking). Because of the intensity of the lion's slogan (10) and the special _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ To measure other positions on the wafer (the non-zero-level intensity measurable by the device of the invention by the device of the invention = all the frequency signs of the target surface, such as _ wall angle, only missing, base chat, etc. , Simulation or experience process... The underlying film stack is not sensitive. The ring library can be used to determine these contour details.

4IBM/04134TW 91 and the precise point of the process conditions corresponding to the test. The differential diffraction measurement of the present invention has the advantage of three directions compared to the general-type adiabatic operation: the sensitivity to the contour change is increased, and the dose and focus sensitivity confirmed above are not for the underlying film laminate. Sensitive, improving the signal-to-noise ratio for determining contour features. With respect to the average CD', the difference can be determined in isolation by the difference technique of the present invention, thereby increasing the signal-to-sound ratio (block 5524 of Fig. 55a). This is disclosed in the arrow section of Figure 55A, where the arrow is simply connected to the CD to the contour decision path. Insensitivity to the underlying film stack, enabling more efficient simulation or two-column generation, which is achieved by passing the focus-exposure matrix in the lithography and the spectral change in the etch rate-isotropic matrix The decision. Established in the spectrum and process slaves - the direct system for the use of the remaining control. The path 5519 of the path 5519' shown in Fig. 55A is from the spectrum and signal library 5522 to the process setting 5527. Referring to Figure 55B, for an example of a stack-to-target (block 536〇), the overlay analysis (block 5380) includes the following steps. In order to measure the overlap, intensity measurements are required from 4 different directions (block 5538). Comparing the measured intensity (block 5540) with the stack-to-target module response (block 5539), for example, as described in equations (35a)-(35b) for the y-direction (as shown in Figure 5%)

4I8M/04134TW 92 1326476 3_) A target having a continuously varying feature size, or a stack-to-target with a-feature size for a non-continuous change in the y-direction (target 23 如图 as shown in Figure 23). In order to ensure the best signal-to-noise ratio for analysis, it is possible to select a long-term plan to make the view as large as possible (eg, Equation 35, for the goal of continuous variable dip size, or as Equation 42, for discontinuous variation feature size) Goal). If the target is of a non-continuous change type (block 5543), then a phase shift analysis of the ^ (Fig. 43-45) is required to enable the selection of substantially meaningful overlay errors. Thereafter, according to Equation 35, for the target of the slanting sizing size, _ according to Equation 42 for the target of the discontinuously varying feature size, the overlay error 'is calculated as the average of the 4 pairs of errors determined at the selected wavelength 5545) . If there are still more metric locations, the rider can continue (block 5390) until all selected locations have been processed. The method is suitable for implementation in a computer readable access medium for implementation in a computer system such as an image processor 49_4), the computer system having a central processing unit, an input/transport/report device and a storage device The present invention can be implemented to perform the method and to receive the data, and to control the apparatus of the present invention, as illustrated in FIG. The present invention relates to the current microscopy technology for CD and overlay control and dose and focus control, including a simplified and more robust and

4IBM/04134TW 93 1326476 The advantages of superior precision and speed measurement devices. The difference stacking and CD stacking technique of the present invention only provides the correlation of the measured intensity, providing an in-situ correction of the measured target period, thereby reducing the dominant source of TIS and ws and in the typical stacking machine. Matching error between stations. Further, the present invention provides a dose m, a balance, and a shop without a library of procedures required for scatter measurement of the (3) or wheel metric. The present invention has been described above with reference to various embodiments, and it is not intended to limit the invention to the specific forms and examples disclosed herein. Therefore, the present invention is intended to cover modifications and variations of the present invention. Industrial Applicability The present invention is useful for lithographic processes of integrated circuits, particularly by ensuring proper exposure and cost conditions for each wafer fabrication to achieve wafer design dimensions and control critical dimensions. BRIEF DESCRIPTION OF THE DRAWINGS The present invention provides a method for controlling critical dimension (CD), stacking, and film thickness and process conditions, and the method of the present invention will be described with reference to the drawings.

4IBM/04134TW 94 1326476, which is described in more detail with the application of the present invention. It should be noted that the accompanying drawings are similar to the component symbols used to illustrate the corresponding elements, and the drawings and the alternatives are drawn in equal proportions. Figure 1 illustrates a typical component of a semiconductor patterning system consisting of lithography, money engraving, and various metrology machines. Figure 2 § Xi Mingming Semiconductor Patterning System, which includes the increased recombination produced by the use of scattering technology (scm) machines. Figure 3 illustrates a semiconductor patterning system produced using the integrated metric (IM) device of the present invention. 4A and 4B respectively illustrate vertical side views of one embodiment of a device in accordance with the present invention. Figure 4C illustrates a detailed view of the device diagram of Figure 4A. Figure 4D illustrates a detailed view of the device diagram of Figure 4B. Figure 5A is a schematic illustration of the two-dimensional reflectivity of a raster target. Figure 5B illustrates the results of the first-order intensity as a function of the width of the grating element of Figure 5At. Figure 6 is a general view of a grating target consisting of tapered grating elements. Figures 7A-7F illustrate an image of the first-order diffraction intensity of a tapered grating target on the CCD array of the apparatus of Figure 4, the towel 11 7A illustrating the shadow

41BM/04134TW 95 1326476 Image, Fig. 7B illustrates the intensity of the sum in the direction parallel to the target period, and Fig. 7C illustrates the intensity of the sum in the direction perpendicular to the target period in the monochromatic illumination. Figures 7D-7F are related diagrams illustrating multi-wavelength illumination. Figure 8 illustrates that the -grating target is split into two regions, the grating elements having opposite tapers, but having a common period. Figure 9A illustrates that the -grating target is divided into four regions, with the relatively thinned grating elements having a relative hue in each pair of regions, but having a common _ period. 9B and 9C respectively illustrate the tapering and tone inversion. Xin Figures 10A-10C illustrate the physical targets corresponding to the simulation results. Figure (1) is a top view, and Figure 10B-C is a cross-sectional view of the substrate and the target. Figures 11A-11D illustrate a simulated wheel gallery at the end of the space and shape, which is depicted by the objectives of Figures 10A-10C.

Figure 12 reveals the distribution of the simulated, wavelength average intensity in the direction of the vertical target period. I Figures 13A-13B reveal the simulated dose and focus response at the target of Figure 9. Figures 14A-14B illustrate the simulated response of zero order and first order diffraction efficiencies as a function of wavelength and oxidized thickness. Figures 15A-15B illustrate the simulated response of zero-order and first-order diffraction efficiencies as a function of wavelength and exposure dose. Figures 16A-16B illustrate the width of the grating elements in two different designs, as a wave

4 Old M/04134TW 96 1326476 Response. The dose target uses a secondary resolution-enhancing feature on the mask to enhance the dose sensitivity on the substrate and to suppress focus sensitivity. The focus target uses a thinner line to enhance the focus sensitivity of the shortened end of the line. Figures 27A-27B are graphical representations of the response of the vessel and focus dose and focus of Figure 16. Figure 28 reveals a difference raster target layout that includes tight contact holes. Figure 29 discloses a differential light thumb target layout comprising closely parallel lines perpendicular to the grating period. Figure 30 reveals a different goal, with multiple types of targets of Figure 17 at different cycles. Figure 31 discloses an optical ray of the object of Figure 30 in the present invention. Figures 32A-32B disclose the detected intensity distributions according to the apparatus of Figures 3A and 31, respectively. Figures 33A-33B reveal the zero-order predicate strength of the target layout of Figure 17. Figures 34A-34B are skirts disposed in the first and second stages of the positive and negative diffraction imaging. 35A-35C reveal that the second embodiment of the continuous grating target suitable for the _ metric is the reflectivity of the reflection 'and under the different vibrating ❺ and phase (four) different cakes, the final first, and the intensity of the graph 'the final The primary intensity depends on the relative position of the optical thumb element. Figure 36 is a two-order photo-glycan target comprising two opposing, inclined grating elements.

4 IBM/04134TW 98 1326476 Figures 37A-37D illustrate a monochromatic image of the diffracted first and second order diffraction intensities of a two-step oblique grating target on a detection array of the apparatus of the present invention. 38A-38D are multi-wavelength images of the positive and negative first-order diffraction intensities of a two-stage oblique grating target on a side array of the apparatus of the present invention. Figure 39 illustrates a two-order hourglass target consisting of a pair of relatively tilted grating targets. Figure 40 e brother - divided into three regions of the two-order grating target, the grating element b process steps have different positions relative to the elements in the A process. 41A-41B illustrate images of the normal and negative first order diffraction intensities of the three grating regions in accordance with the objective of Fig. 4(). Fig. 42A-42C(d), (d) with the same reflectance phase as: a: an image of the positive and negative first-order diffraction intensities of the three grating regions as a function of error. In the case of consistency, the transformation γ is determined. Figure 43 is in the case of Lu = (〇, 〇 and the flow chart. Determine the transformation of the number of series of revolutions &lt; Determine the transformation of the number of series of conversions. Figure 44 is in the Flowchart in the example of AC* 。 Figure 45 is a flow chart in the example of AC*〇.

Figure 46A is the edge of all allowed W values, the pair of errors

4IBM/04134TW 99 1326476 The value of the value of ~ will be displayed in all allowed dynamic ranges -F = ^ in 〇 Figure 46B to draw the measurable component v, &quot; change, overlap error Sensitivity fe*. Figures 47A-47B(d)-devices use the _common source and detector to enable rapid sequential measurement of the positive and negative first order diffraction intensities. Figure 48 illustrates a combination of CD, overlay and film thickness targets for measurement, X and y orientation CD, χ and y stack error between two pairs of steps, and film thickness in unpatterned areas. Figures 49A-49D illustrate the positive and negative first-order diffraction intensities, the fixed detected zero-order intensities, and the target images for pattern recognition and alignment for sequential detection of the targets in Figure 48. Figures 50A-50D illustrate schematic views of a device of the present invention lacking any target pattern on wafer 450. Figure 51 is a detailed view of the geometric relationship that must be maintained between the incident and reflected beams of the device to ensure that the zero- and non-zero-level diffracted energy can be simulated as the center wavelength changes. Figure 52 is a flow chart of the target design process of the present invention. Figure 53 is a flow diagram of a measurement mode that may be accompanied by the measurements of the present invention. Figure 54 is a flow diagram of zero-order measurement analysis when using the apparatus of the present invention.

4 Old M/04134TW 100 1326476 Figures 55A-55B are flow diagrams of data analysis of the present invention for use in the measurement methods and apparatus of the present invention to determine CD, dose, focus, and overlay. [Major component symbol description] 10 lithography and etching production line 110 light cluster 111 trajectory machine 112 exposure machine 113 post exposure trajectory machine 120 stack of measuring machine 130 scanning electron microscope 140 etch cluster 141 etching chamber 5 Wafer 150 Atomic Force Microscope 160 Film Thickness Measurement Machine 170 Electronic Probe Measuring Machine 180 Scattering Measurement Metrics 185 Microscope 190 Computer Server 15, 25, 35 Offline Measurement System 2801 Contact Hole 3 (Π, 302, 303 Mirror 314 Optical Polarization Sheet 340 diffraction device 380, 398 rotatable mirror 40 diffraction measurement system 410 illumination source 412 color filter 413 illumination light member 414 polarizer 430 collection light member 435 wavelength dispersion light member 436 collection light member 450 crystal 5] 451 base 4IBM/04134TW 101 1326476 452 film stack 460 detection array 490 image processor 908 space 455 patterned target 480, 485 detector 486 third light piece 2650, 2680 occult design 2612, 2623, 2624 times pattern area 475 , 1750, 3305 unpatterned areas 60, 80, 900, 1701, 2300, 2500, 2600, 2800, 2900, 3000, 3600, 3900, 4000, 4800 Target 6 (Π, 8 (M , 802, 901, 902, 903, 904, 231, 2312, 2313, 2314, 2901, 2902, 3001, 3002, 3003, 3004, 3901, 3902, 4810, 4820, 4830, 4840, grating elements 909, 1711, 1712, 3501, 3502 line 173 1173, 1801, 1802, 2501, 2502, 2503, 2504, 2611, 2611, 2612, 265 2 2653 Main features 4IBM/04134TW 102

Claims (1)

1326476 X. Patent Application Range: L A method of measuring the size of a substrate, the method comprising the steps of: providing a nominal pattern comprising a series of features 'in a primary direction Having a primary pitch of one of the periodicities, wherein the nominal pattern is characterized by a nominal feature size, the nominal feature size being repeated in the main direction at the period p, and the nominal One of the main directions of the feature size has a predetermined variation in the vertical direction; corresponding to the marriage case, a target pattern (four) pattern is formed on the substrate, wherein the target_ has a substrate feature size, The substrate feature size corresponds to the nominal feature size; The 乂 characteristic is that the target _ is irradiated to a fine wavelength to generate a diffraction order from the target pattern; and between the supply-focus size and the vertical-to-side variation of the edge-to-edge level _, should be the substrate feature size deviation 'the new ship size is scaled to the nominal feature size; the side along the direction of the one of the more than one zero non-zero diffraction series And according to the relationship 'determining the size of interest based on the one or more non-changes. The method of claim 1, wherein the illumination comprises more than one wavelength, and the method further comprises detecting the primary direction. 4 IBM/04134 TW 103 1326476. The change in the number of non-zero order diffraction orders. 3. The method of claim 1 or claim 2, wherein the target size comprises a critical dimension. The method of claim 3, wherein the detected change is included in the one or A change in the strength of many non-zero-order diffraction orders. 5. The method of claim 4, wherein the nominal pattern comprises a first sub-pattern and a second sub-pattern feature, the first pattern comprising a first sub-pattern feature The pattern feature is a nominal length and width. The second pattern includes a complementary color tone having the nominal length and width, and wherein the method further comprises selecting a feature length and width based on the target secondary pattern. The deviation determines a deviation between the process condition and the nominal process condition, wherein the target sub-pattern feature length and width are relative to the nominal length and width. 6. The method of claim 5, wherein the target pattern comprises a first target portion and a second target portion 'the first target portion is formed on a first layer of the substrate, and the first The target portion corresponds to one of the nominal target patterns, and the first target portion is formed on the second layer of the substrate, and the second target portion corresponds to one of the nominal target patterns Second 4IBM/04134TW 104