System and method for full-scope real-time detection of lithographic characteristics

Abstract

The present disclosure relates generally to systems and methods for full-scope real-time detection of lithography characteristics. An extreme ultraviolet (EUV) lithography system generates EUV light by irradiating a droplet with a laser. The system includes a collector and a plurality of vibration sensors coupled to the collector. The vibration sensors generate sensor signals indicative of shock waves from the laser pulse and impacts from debris. The system utilizes the sensor signals to improve the quality of the EUV light generation.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to systems and methods for full-scope real-time detection of lithography characteristics. BACKGROUND

[0002] There is a continuing demand to increase the computing power of electronic devices, including smart phones, tablets, desktop computers, laptop computers, and many other kinds of electronic devices. Integrated circuits provide the computing power for these electronic devices. One way to increase the computing power of integrated circuits is to increase the number of transistors and other integrated circuit features that can be included in a given area of a semiconductor substrate.

[0003] Features in integrated circuits are produced, in part, by lithography. Conventional lithography techniques include generating a mask that outlines a pattern of features to be formed on an integrated circuit die. A lithography light source illuminates the integrated circuit die through the mask. The size of features that can be produced by lithography of an integrated circuit die is limited, in part, at the lower end by the wavelength of light produced by the lithography light source. Smaller wavelengths of light can produce smaller feature sizes.

[0004] Extreme ultraviolet (EUV) light is used to produce particularly small features due to its relatively short wavelength. For example, EUV light is typically produced by illuminating a droplet of a selected material with a laser beam. The energy from the laser beam causes the droplet to enter a plasma state. In the plasma state, the droplet emits EUV light. The EUV light is directed toward a collector having an elliptical or parabolic surface. The collector reflects the EUV light to a scanner. The scanner illuminates a target with the EUV light through a mask. However, if the droplet is not properly formed and illuminated, there can not be enough EUV light to perform the EUV process. As a result, the lithography process can fail and the resulting integrated circuit will not function properly. SUMMARY

[0005] According to one embodiment of the present disclosure, there is provided a method of operating an extreme ultraviolet lithography system, comprising: performing a lithography process by generating extreme ultraviolet light in an extreme ultraviolet light generation chamber; generating sensor signals with a plurality of vibration sensors coupled to the extreme ultraviolet light generation chamber; analyzing the sensor signals; and adjusting a parameter of the extreme ultraviolet light generation based on the sensor signals.

[0006] According to another embodiment of the present disclosure, there is provided a method of operating an extreme ultraviolet lithography system, comprising: coupling a plurality of vibration sensors to a collector mirror of an extreme ultraviolet lithography system; generating extreme ultraviolet light by illuminating a droplet with a laser inside an extreme ultraviolet light generation chamber; generating, with the vibration sensors, sensor signals indicative of a distribution of debris from the droplet on the collector mirror; and adjusting one or more parameters of the extreme ultraviolet light generation based on the distribution of the debris.

[0007] According to yet another embodiment of the present disclosure, there is provided an extreme ultraviolet lithography system, comprising: a droplet generator configured to output a droplet into an extreme ultraviolet light generation chamber; a laser configured to generate a plasma from the droplet by irradiating the droplet with a laser light; a collector mirror configured to focus extreme ultraviolet light emitted from the plasma; and a plurality of first vibration sensors coupled to the collector mirror. BRIEF DESCRIPTION OF DRAWINGS

[0008] Aspects of the disclosure can be best understood from the following specific description together with the appended drawings. It is to be noted that the various features illustrated in the figures are not necessarily drawn to scale. Indeed, the dimensions of the various features can be arbitrarily expanded or reduced for the sake of discussion.

[0009] FIG. 1 is a block diagram of an EUV lithography system according to some embodiments.

[0010] FIGS. 2A-2B is an illustration of a lithography system according to some embodiments.

[0011] FIG. 2C is a block diagram of a lithography system according to some embodiments. FIG. 2A and FIG. 2B is a bottom view of a collector of

[0012] FIG. 2D is a block diagram of a lithography system according to some embodiments. FIG. 2A and FIG. 2B is a bottom view of a collector of

[0013] FIG. 3 is a flowchart of a process for operating an EUV lithography system according to some embodiments.

[0014] FIG. 4A is an illustration of a droplet debris impacting a collector of an EUV light generation system according to some embodiments.

[0015] FIG. 4B is a graph showing sensor signals provided by one or more vibration sensors according to some embodiments.

[0016] FIG. 5 is a block diagram of a vibration sensor coupled to a signal processor according to some embodiments.

[0017] FIG. 6 comprises a plurality of graphs of sensor signals from one or more vibration sensors according to some embodiments.

[0018] FIG. 7 is a graph of sensor signal strength in the frequency domain according to some embodiments.

[0019] FIG. 8 is a diagram of a collector according to some embodiments.

[0020] FIG. 9 is a diagram of a collector according to some embodiments.

[0021] FIG. 10A shows a process for constructing a two-dimensional representation of plasma evolution according to some embodiments.

[0022] FIG. 10B shows a process for constructing a two-dimensional representation of plasma evolution according to some embodiments.

[0023] FIG. 10C shows a process for constructing a two-dimensional representation of plasma evolution according to some embodiments.

[0024] FIG. 11 shows a process for generating a 3D holographic and time-resolved plasma model according to some embodiments.

[0025] FIGS. 12A-12H is a diagram of a 3D plasma model and corresponding flattening and plasma- tization pulses according to some embodiments.

[0026] FIG. 12I shows a 3D model of an EUV light generation chamber and a droplet impacted by a plasma-tization pulse according to some embodiments.

[0027] FIG. 12J shows a 3D model of an EUV light generation chamber and a droplet impacted by a plasma-tization pulse according to some embodiments.

[0028] FIG. 12K shows a 3D plasma model associated with the generation of EUV light from a droplet according to some embodiments.

[0029] FIG. 12L shows a 3D plasma model associated with the generation of EUV light from a droplet according to some embodiments.

[0030] FIG. 13 is a plot showing the separation distance of a flattening pulse and a plasma-tization pulse relative to the energy of the flattening pulse according to some embodiments.

[0031] FIG. 14 is a flowchart of a method for operating an EUV light generation system according to some embodiments.

[0032] FIG. 15 is a block diagram of a control system according to some embodiments.

[0033] FIG. 16This is a flowchart of a method for operating an EUV lithography system according to some embodiments.

[0034] FIG. 17 This is a flowchart of a method for operating an EUV lithography system according to some embodiments. Detailed Implementation

[0035] The following disclosure provides numerous different embodiments or examples for implementing various features of the provided subject matter. Specific examples of components and arrangements are described below to simplify this disclosure. Of course, these are merely examples and are not intended to be limiting. For example, in the following description, forming a first feature on or over a second feature can include embodiments in which the first and second features are formed in direct contact, and can also include embodiments in which an additional feature can be formed between the first and second features such that the first and second features do not need to be in direct contact. Furthermore, reference numerals and / or letters may be repeated in various examples throughout this disclosure. Such repetition is for simplicity and clarity and does not in itself indicate a relationship between the various embodiments and / or configurations discussed.

[0036] Furthermore, spatially related terms (e.g., "below," "under," "down," "above," "up," etc.) may be used herein to readily describe the relationship of one element or feature shown in the figure relative to another element(s) or feature(s). These spatially related terms are intended to cover different orientations of the device in use or operation other than those shown in the figure. The device may be oriented in other ways (rotated 90 degrees or in other orientations), and the spatially related descriptors used herein shall be interpreted accordingly.

[0037] In the following description, certain specific details are set forth to provide a thorough understanding of various embodiments of the present disclosure. However, those skilled in the art will understand that the present disclosure can be practiced without these specific details. In other instances, well-known structures associated with electronic components and manufacturing techniques have not been described in detail to avoid unnecessarily obscuring the description of embodiments of the present disclosure.

[0038] Unless the context otherwise requires, throughout the specification and the following claims, the word “comprising” and its variations, such as “including” and “comprising”, shall be interpreted in an open-ended, inclusive sense, i.e., as “including but not limited to”.

[0039] The use of ordinal numbers such as first, second, and third does not necessarily imply a sequential order, but can simply distinguish between multiple instances of an action or structure.

[0040] References in the specification to “some embodiments” or “an embodiment” mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments. The appearances of the phrase “in some embodiments” or “in an embodiments” in various places in the specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0041] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” include plural referents unless the content clearly dictates otherwise. It should also be noted that the term “or” is generally employed in its sense including “and / or” unless the content clearly dictates otherwise.

[0042] Embodiments of the present disclosure provide many benefits to EUV lithography systems. Embodiments of the present disclosure utilize vibration sensors coupled to surfaces of the EUV light generation chamber to determine characteristics of EUV light generation and debris distribution. The vibration sensors can detect characteristics of the shockwave generated by the laser irradiating the droplet. The vibration sensors can also detect the distribution of dispersed debris on surfaces within the EUV light generation chamber. Based on this information, a control system can construct a model of the plasma generated by the laser irradiating the droplet. The control system utilizes a machine learning process to determine adjustments that can be made to improve plasma generation based on the plasma model. The control system then adjusts various aspects of the EUV system to improve EUV light generation. The result is that the lithography process is properly performed, resulting in improved yield of wafers and better performance of integrated circuits. Furthermore, collector mirrors will not need to undergo extensive cleaning procedures or replacement.

[0043] FIG. 1 is a block diagram of an EUV lithography system 100 according to some embodiments. The components of the EUV lithography system 100 cooperate to generate EUV light and perform a lithography process. As will be set forth in greater detail below, the components of the lithography system utilize vibration sensors and machine learning techniques to improve the generation of EUV light. As used herein, the terms “EUV light” and “EUV radiation” are used interchangeably.

[0044] The EUV lithography system 100 includes a droplet generator 102, an EUV light generation chamber 104, a droplet receiver 106, a scanner 108, a laser 112, and a collector 114. The droplet generator 102 outputs droplets into the EUV light generation chamber 104. The laser 112 irradiates the droplets with laser pulses inside the EUV light generation chamber 104. The irradiated droplets emit EUV light. The EUV light is collected by the collector 114 and reflected toward the scanner 108. The scanner 108 conditions the EUV light and focuses the EUV light onto a target 110. The target 110 can include a semiconductor wafer. After being irradiated by the laser 112, the droplets exit the EUV light generation chamber 104 and are received by the droplet receiver 106. Further details about each of these components and processes are provided below.

[0045] The droplet generator 102 generates and outputs a stream of droplets. The droplets can include tin, but droplets of other materials can be used without departing from the scope of the present disclosure. The droplets move at high speed toward the droplet receiver 106. The droplets have an average velocity between 60 m / s and 200 m / s. The droplets have a diameter between 10 pm and 200 pm. The generator can output 1000 to 100,000 droplets per second. The droplet generator 102 can generate droplets having different initial velocities and diameters than described above without departing from the scope of the present disclosure.

[0046] In some embodiments, the EUV light generator 102 is a laser-produced plasma (LPP) EUV light generation system. As the droplets pass through the EUV light generation chamber 104 between the droplet generator 102 and the droplet receiver 106, the droplets are irradiated by the laser 112. When the droplets are irradiated by the laser 112, the energy from the laser 112 causes the droplets to form a plasma. The plasmaed droplets generate EUV light. The EUV light is collected by the collector 114 and passed to the scanner 108 and then to the target 110.

[0047] In some embodiments, the laser 112 is located outside of the EUV light generation chamber 104. During operation, the laser 112 outputs laser pulses into the EUV light generation chamber 104. The laser pulses are focused on a point through which the droplets pass on their way from the droplet generator 102 to the droplet receiver 106. Each laser pulse is received by a droplet. When the droplet receives the laser pulse, the energy of the laser pulse generates a high-energy plasma from the droplet. The high-energy plasma outputs EUV light.

[0048] In some embodiments, the laser 112 irradiates the droplet with two pulses. The first pulse flattens the droplet into a disc-like shape. The second pulse forms a high-temperature plasma from the droplet. The second pulse is significantly stronger than the first pulse. The laser 112 and the droplet generator 102 are calibrated so that the laser emits a pair of pulses to irradiate the droplet with the pair of pulses. The laser can irradiate the droplets in a manner different from that described above without departing from the scope of the present disclosure. For example, the laser 112 can irradiate each droplet with a single pulse or more than two pulses. In some embodiments, there are two separate lasers. The first laser delivers the flattening pulse. The second laser delivers the plasma-tizing pulse.

[0049] In some embodiments, the light output by the droplet is randomly scattered in many directions. The lithography system 100 collects the scattered EUV light from the plasma with a collector 114 and outputs the EUV light to a scanner 108.

[0050] The scanner 108 includes scanner optics 109. The scanner optics 109 include a series of optical conditioning devices to direct the EUV light to a reticle. The scanner optics 109 can include refractive optics, e.g., a lens or a system of lenses (zone plates). The scanner optics 109 can include reflective optics, e.g., a single mirror or a system of mirrors. The scanner optics 109 direct the ultraviolet light from the EUV light generator 102 to the reticle.

[0051] The ultraviolet light reflects from the reticle back to other optical features of the scanner optics 109. In some embodiments, the scanner optics 109 include a projection optics box. The projection optics box can have refractive optics, reflective optics, or a combination of refractive and reflective optics. The projection optics box can have a magnification less than 1, thereby reducing the patterned image included by the EUV light reflected from the reticle. The projection optics box directs the EUV light onto a target 110, e.g., a semiconductor wafer.

[0052] The EUV light includes a pattern from the reticle. Specifically, the reticle includes a pattern to be defined in the target 110. After the EUV light reflects from the reticle, the EUV light contains the pattern of the reticle. During extreme ultraviolet lithography exposure, a layer of photoresist typically covers the target. The photoresist helps to pattern the surface of the semiconductor wafer according to the pattern of the reticle.

[0053] The effectiveness of the lithography process performed using the reticle depends largely on the quality of the EUV light generation in the EUV light generation chamber 104. The quality of the EUV light generated within the EUV light generation chamber 104 is affected by various parameters or characteristics of the components of the EUV light generation system 100. Some of these characteristics relate to characteristics of the droplet, characteristics of the laser pulses, and characteristics of the interior surfaces of the EUV light generation chamber 104. As previously described, EUV light is generated from the droplet as follows: the droplet is first irradiated with a flattening laser pulse, and then a plasma is generated from the flattened droplet by irradiating the droplet with a plasmafying laser pulse. The droplet emits EUV light while in the plasma state. The light is scattered by the droplet and collected by the collector 114 and reflected by the collector 114 to the scanner 108.

[0054] It is beneficial for a very large amount of EUV light to be generated from each droplet. It is also beneficial for the EUV light to be scattered from the droplet in a radially symmetric manner about an axis corresponding to the plasmafying laser pulse. It is also beneficial for the surface of the collector 114 to be free of debris so that the reflectivity of the collector 114 is high. It is also beneficial for other interior surfaces of the plasma generation chamber 104 to be free of debris.

[0055] One characteristic that can severely impact EUV light generation is the effectiveness of the flattening pulse on the droplet. The flattening pulse from the laser 112 flattens the droplet into a disc or pancake shape. It is desirable for the bottom surface of the flattened droplet to be nearly horizontal in the X-Y plane, or nearly perpendicular to the direction of travel of the laser pulse. This is most likely to occur if the flattening pulse hits the center of the bottom hemispherical surface of the droplet. If the flattening pulse hits the droplet off-center, the bottom flattened surface of the droplet will be tilted with respect to the horizontal plane by an undesirable amount. As will be set forth in greater detail below, this results in undesirable EUV light generation quality. Whether the flattening pulse has the desired effect depends in part on the velocity of the droplet, the size of the droplet, the timing of the flattening pulse, and the energy of the flattening pulse.

[0056] Another characteristic that can severely impact EUV light generation is the effectiveness of the plasmafying pulse. It is desirable for the plasmafying pulse to hit the center of the bottom surface of the flattened droplet, and for the bottom surface to be nearly in the horizontal plane. If the plasmafying pulse hits the droplet off-center, the droplet can not fully plasmafy. In addition, debris can be ejected from the droplet in various directions, contaminating the surface of the collector 114 and other interior surfaces of the EUV light generation chamber 104. A droplet that does not fully plasmafy can generate a reduced amount of EUV light. The EUV light can also be scattered in an undesirable profile. Whether the plasmafying pulse fully and effectively plasmafies the droplet depends in part on the velocity of the droplet, the size of the droplet, the timing of the plasmafying pulse, and the energy of the plasmafying pulse.

[0057] The EUV lithography system 100 utilizes the vibration sensors 118 and the control system 116 to determine the current EUV light generation quality and adjust the EUV light generation parameters to improve the EUV light generation. As will be set forth in greater detail below, the vibration sensors 118 detect vibrations of the collector 114, and other surfaces of the EUV radiation generation chamber 104. The vibration sensors 118 generate sensor signals indicative of the vibrations and communicate the sensor signals to the control system 116. The control system 116 analyzes the sensor signals and generates EUV adjustment data indicative of adjustments to the EUV light generation parameters to improve the EUV light generation.

[0058] A large number of vibration sensors 118 can be distributed on the outer surfaces of the EUV light generation chamber 104. In many cases, the outer surfaces of the collector 114 are part of the outer surfaces of the EUV light generation chamber 104. Thus, multiple vibration sensors 118 can be distributed on the outer surfaces of the collector 114.

[0059] Each vibration sensor 118 senses vibrations of the surface of the EUV light generation chamber 104 in the vicinity of the vibration sensor 118. The vibrations can be generated by debris from the droplet impacting the inner surfaces of the EUV light generation chamber 104 as a result of the plasmafication pulse or the flattening pulse impacting the droplet. These impacts deliver a large amount of energy to the droplet and eject debris from the droplet. The time of flight of the debris can be calculated based on the sensor signals and the known timing of the laser pulses. The vibration sensors 118 can also detect shockwaves emitted from the droplet in response to the radiation of the laser pulses. The shockwaves will have a particular directionality based on the orientation of the droplet (e.g., the orientation of the surface of the flattening droplet). The vibration sensor signals are communicated from the vibration sensors 118 to the control system 116.

[0060] The control system 116 analyzes the sensor signals. Based on the sensor signals, the control system 116 can determine the time of flight of the droplet debris. The control system 116 can also determine the distribution of the debris on the inner surfaces of the EUV light generation chamber 104. The control system 116 can utilize this information to construct a 3D model of the plasmafied droplet.

[0061] The control system 116 can utilize a machine learning process to understand the laser plasma interaction. The machine learning process can utilize a large number of EUV light generation parameters (e.g., droplet velocity, droplet size, flattening pulse timing, flattening pulse energy, plasma formation pulse timing, plasma formation pulse energy, debris distribution, shockwave distribution, and other parameters) to understand how these parameters work together to generate a plasma and how they affect the overall EUV light generation. The machine learning process can enable the control system 116 to select good EUV light generation parameters to improve the quality and stability of the EUV system 100. In addition, the control system can predict a contamination distribution in the EUV system. This can enable active cleaning actions to clean highly contaminated areas of the collector 114 or other interior surfaces of the EUV light generation chamber 104.

[0062] The orientation of the surface of the flattening droplet is positively correlated with the droplet contamination distribution. Specifically, a relatively large amount of contamination occurs in the direction facing the flattening surface of the flattening droplet.

[0063] Accordingly, the control system 116 and the vibration sensor 118 provide real-time, full-scope, correct plasma conditions as well as droplet debris formation and distribution. With machine learning and big data analysis, the control system 116 can effectively improve EUV power, droplet contamination, and EUV energy stability. The control system 116 can also utilize deep learning for efficient EUV light generation, power decay, and dose error control. Further details regarding the machine learning process are set forth below in connection with FIG. 14 and FIG. 15 Further details regarding the machine learning process are set forth below in connection with

[0064] FIGS. 2A-2B is an illustration of a lithography system 200 according to an embodiment. The lithography system 200 is an extreme ultraviolet lithography system that generates extreme ultraviolet radiation through a laser plasma interaction. The plasma can be generated in a manner substantially similar to that described above in connection with FIG. 1 is generated in a substantially similar manner as described above. FIG. 2A The lithography system 200 is shown without extreme ultraviolet radiation. FIG. 2B The lithography system 200 is shown with extreme ultraviolet radiation.

[0065] Reference is made to FIG. 2A The lithography system 200 includes a plasma generation chamber 104, a laser 112, a collector 114, a droplet generator 102, and a droplet receiver 106. The EUV light generation chamber 104 is defined by the collector 114 and an enclosure 124 coupled to the collector 114. The components of the lithography system 200 work together to generate extreme ultraviolet radiation and perform a lithography process using the extreme ultraviolet radiation.

[0066] The collector 114 forms the bottom of the EUV light generation chamber 104. The cone housing 124 is coupled to the collector 114 and forms the top of the EUV light generation chamber 104. The collector 114 includes an inner surface 131 and an outer surface 133. The cone housing 124 includes an inner surface 135 and an outer surface 137.

[0067] A plurality of vibration sensors are coupled to the EUV light generation chamber 104. In particular, vibration sensors 118 are distributed on the outer surface 133 of the collector 114. Vibration sensors 118 are also distributed on the outer surface 137 of the housing 124. The vibration sensors 118 are sensitive enough to sense vibrations at the inner surfaces 131 and 135 of the collector 114 and the housing 124 due to debris and shock waves, as will be described in more detail below.

[0068] The droplet generator 102 generates and outputs droplets 128. The droplets can include tin, but droplets of other materials can be utilized without departing from the scope of the present disclosure. The droplets 128 move at high speed toward the droplet receiver 106.

[0069] The droplet generator 102 emits the droplets 128 periodically. FIG. 2B The view of FIG. 1 shows three droplets 128. One of the droplets is at the location of the laser shot point within the EUV light generation chamber 104.

[0070] After passing through the laser shot point, the coalesced droplet 128 is received by the droplet receiver 106. The droplet receiver 106 can include a droplet reservoir. The droplet 128 enters the droplet receiver 106, hits a back wall of the droplet receiver 106, and falls into the droplet reservoir. Other configurations of the droplet receiver 106 can be employed without departing from the scope of the present disclosure.

[0071] The laser 112 is located behind the collector 114. During operation, the laser 112 outputs laser pulses 130. The laser pulses 130 are focused on a point that the droplets pass on their way from the droplet generator 102 to the droplet receiver 106. Each laser pulse 130 is received by a droplet 128 at the laser shot point. When a droplet 128 receives a laser pulse 130, energy from the laser pulse generates a high-energy plasma from the droplet 128. The high-energy plasma outputs extreme ultraviolet radiation.

[0072] In some embodiments, the laser 112 is a carbon dioxide (CO2) laser. CO2 lasers emit radiation or laser 130 with a wavelength concentrated around 9.4 pm or 10.6 pm. The laser 112 can include a laser other than a carbon dioxide laser, and can output radiation with a wavelength different from the above-mentioned wavelengths without departing from the scope of the present disclosure.

[0073] In some embodiments, laser 112 irradiates each droplet 128 with two pulses. The first pulse flattens the droplet 128 into a disk-shaped form. The first pulse may be referred to as a "flattening pulse." The second pulse may be referred to as a "plasmaizing pulse." The second pulse causes the droplet 128 to form a high-temperature plasma. The second pulse is significantly stronger than the first pulse. Laser 112 and droplet generator 102 are calibrated such that laser 112 emits pulse pairs, thereby irradiating each droplet 128 with pulse pairs.

[0074] exist FIG. 2A In the example, the droplet 128 at the laser irradiation point has been irradiated by a planarization pulse. The irradiated droplet 128 is planarized into a roughly disk shape. FIG. 2A In the view, the flattened droplet is tilted relative to the horizontal direction.

[0075] Although FIG. 2A A single laser 112 is shown, but in reality, two lasers can exist. The first laser emits planarization pulses. The second laser emits plasma pulses.

[0076] Laser 112 can irradiate droplets 128 in a manner different from that described above without departing from the scope of this disclosure. For example, laser 112 can irradiate each droplet 128 with a single pulse or more than two pulses. Furthermore, the primary laser described herein can not only form droplets into a disk shape, but also into a mist or vapor state.

[0077] FIG. 2B EUV light 132 emitted from a droplet 128 receiving a laser pulse 130 is shown. When the droplet 128 transforms into plasma, it emits EUV light 132. In an example where the droplet 128 is tin, the output wavelength of EUV light 132 is concentrated between 10 nm and 15 nm. More specifically, in some embodiments, tin plasma emits EUV light with a center wavelength of 13.5 nm. Materials other than tin can be used for the droplet 128 without departing from the scope of this disclosure. Such other materials can generate extreme ultraviolet radiation with wavelengths different from those described above without departing from the scope of this disclosure.

[0078] In some embodiments, the EUV light 132 emitted by the droplet 128 is scattered in many directions. The lithography system 100 uses a collector 114 to collect the scattered EUV light 132 from the plasma and outputs the EUV light 132 to the lithography target.

[0079] In some embodiments, collector 114 is a parabolic or elliptical mirror. The scattered EUV light 132 is reflected by a parabolic or elliptical mirror having a trajectory toward the scanner. FIG. 2A and FIG. 2BThe EUV light 132 is reflected from the mask onto a lithography target. The EUV light 132 reflected from the mask patterns photoresist or other material on a semiconductor wafer. For the purposes of this disclosure, the details of the mask and various configurations of the optical equipment in the scanner are not shown.

[0080] In some embodiments, the collector 114 includes a central hole 129. The laser pulses 130 pass through the central hole 129 from the laser 112 toward the stream of droplets 128. This enables the collector 114 to be positioned between the laser 112 and the scanner.

[0081] When a plasmafied laser pulse illuminates a planarized droplet 128, the planarized droplet 128 will become a plasma. The energy of the plasmafied pulse also causes particles from the plasmafied droplet 128 to scatter from the planarized droplet. Some of the particles from the plasmafied droplet 128 will land on the inner surface 131 of the collector mirror 114. Some of the particles from the plasmafied droplet 128 will land on the inner surface 135 of the cone housing 124. These droplet particles are debris that contaminate the collector mirror 114 and the cone housing 124. The accumulation of these debris can significantly reduce the reflectivity of the inner surface 131 of the collector 114.

[0082] It can require a great deal of effort to clean the debris from the inner surface 131 of the collector 114 and the inner surface 135 of the cone housing 124. One solution is to clean all areas of the inner surfaces 131 and 135 thoroughly. However, this can be very time consuming and expensive.

[0083] As previously described, the scattering pattern of the debris from the droplets 128 is built based on and indicative of the shape and orientation of the planarized droplets, and the plasmafied pulse and plasma profile of the droplets 128. Thus, the debris from the droplets 128 will accumulate on the inner surfaces 131 and 135 in a pattern. Some areas can have no debris, while other areas have a large amount of debris.

[0084] The vibration sensors 118 detect the location of the accumulation of debris on the surfaces 131 and 135. When a piece of debris impacts an area on the surface 131 or 135, the collector 114 or the housing 124 will vibrate at that location. The vibration sensor 118 at the corresponding location on the outer surface 133 or 137 will sense the vibration and generate a sensor signal indicative of the vibration. The sensor signal is communicated to the control system 116.

[0085] Based on the sensor signals received from the vibration sensors 118, the control system 116 can determine the locations of accumulated debris on the inner surfaces 131 and 135. The control system 116 can direct cleaning operations to those specific locations to remove the debris. The cleaning operations can include flowing a cleaning fluid to those locations where debris is accumulated. The cleaning fluid can include hydrogen, or other cleaning fluids that can help break down and remove the debris. The cleaning fluid can be provided through channels, holes, or vents at the edges of the collector 114. These channels, holes, or vents can be selectively opened. Other ways of cleaning the debris can be used without departing from the scope of the present disclosure.

[0086] As previously described, the scattering pattern of debris on the inner surfaces 131 and 135 is indicative of the shape and orientation of the planarized droplet, as well as the plasma profile of the droplet 128 after receiving a plasma pulse. For example, if the plasma pulse is concentrated at the front edge or the back edge rather than the center of the planarization surface, only some portions of the droplet 128 can become plasma-ized, resulting in a large amount of debris. Thus, the scattering pattern of debris from the droplet 128 is indicative of the portion of the droplet 128 that was plasma-ized. The scattering pattern can also be indicative of the shape and orientation of the planarized droplet 128 before being illuminated by the plasma pulse.

[0087] The control system 116 can generate a 3D model of the plasma-ized droplet based on the sensor signals received from the vibration sensors 118. This 3D model can be indicative of how the planarization pulses and the plasma pulses interact with the droplet. The control system 116 can utilize a machine learning model to connect the relationship between the characteristics of the plasma pulses and the planarization pulses and the characteristics of the droplet 128. The control system 116 can then determine adjustments that can be made to improve plasma-ization and EUV light generation.

[0088] FIG. 2C is a bottom view of the collector 114 according to some embodiments FIG. 2A and FIG. 2B is a bottom view of the collector 114 according to some embodiments FIG. 2C shows the distribution of the vibration sensors 118 on the outer surface 133 of the collector 114. As can be seen from FIG. 2C The vibration sensors 118 are distributed at many locations on the outer surface 133 of the collector 114. In some embodiments, the vibration sensors 118 can be more densely distributed at locations near the aperture 129, and less densely distributed at locations near the outer edges of the collector 114. It can be desirable to more densely distribute the vibration sensors 118 at locations where more debris can be received.

[0089] FIG. 2D is a bottom view of the collector 114 according to some embodiments FIG. 2A and FIG. 2B is a bottom view of the collector 114 according to some embodiments FIG. 2DA distribution of vibration sensors 118 on an outer surface 133 of the collector 114 is shown. The collector 114 can be formed of individual mirror units 140. As shown from the perspective of the collector 114, the vibration sensors 118 are distributed in a grid pattern on the outer surface 133 of the collector 114. The vibration sensors 118 can be placed at each mirror unit 140. FIG. 2D As can be seen, the vibration sensors 118 are distributed in a grid pattern on the outer surface 133 of the collector 114. A respective vibration sensor 118 can be placed at each mirror unit 140.

[0090] In some embodiments, the mirror units 140 are defined or bounded by pinch joints 142. The pinch joints 142 can correspond to locations where a wash or cleaning fluid can be flowed onto adjacent mirror units 140. The wash fluid can be selectively flowed onto those mirror units 140 that have accumulated contamination debris. The pinch joints can likewise be used to flow a wash or cleaning fluid onto mirror units 140 that are expected to accumulate debris based on the known characteristics of the planarization pulses and plasma pulses and the liquid droplets. Various schemes can be utilized to distribute the vibration sensors on the bottom surface 133 of the collector 114 without departing from the scope of the present disclosure.

[0091] FIG. 3 is a flowchart of a process for operating an EUV lithography system according to some embodiments. The method 300 can utilize the processes, structures, and components previously described with respect to FIGS. 1-2D FIG. 1. At 302, the method 300 includes irradiating a liquid droplet with a laser pulse. One example of a liquid droplet is the liquid droplet 128 of FIG. 2A and FIG. 2B FIG. 1. One example of a laser pulse is the laser pulse 130 shown in FIG. 2A and FIG. 2B FIG. 1. At 304, droplet debris is scattered from the liquid droplet. At 306, the method 300 includes sensing the debris impact with a vibration sensor. One example of a vibration sensor is the vibration sensor 118 of FIG. 4A FIG. 1. At 308, the method 300 includes performing signal processing on a sensor signal generated by the vibration sensor. At 310, the method 300 includes determining a characteristic of a plasma evolution of the liquid droplet 128 based on the sensor signal. At 312, the method 300 includes generating a 3D plasma model and a time-resolved plasma model based on the sensor signal. At 314, the method 300 includes identifying a parameter adjustment to improve plasma generation. At 316, the method 300 includes adjusting a plasma generation parameter. Further details regarding each of these steps are provided in subsequent figures.

[0092] FIGS. 1-2D is an illustration of a liquid droplet debris impacting a collector 114 of an EUV light generation system according to some embodiments. The collector 114 can include the characteristics described with respect to FIG. 4A FIG. 1. FIG. 3 is an illustration of a liquid droplet debris impacting a collector 114 of an EUV light generation system according to some embodiments. The collector 114 can include the characteristics described with respect to FIG. 4AFIG. 4 illustrates an example of step 304 of method 300 of FIG. 3, but FIG. 4 can be applied to other methods and processes without departing from the scope of the present disclosure.

[0093] In FIG. 4A In the example of FIG. 4, droplet 128 (not shown) has received a plasmatized laser pulse. A plasma is generated from droplet 128. Droplet particles 146 and free electrons 148 are ejected from plasmatized droplet 128. As previously described, the ejection or scattering pattern can be indicative of the properties of the plasma and the system parameters used to generate the plasma. Droplet particles 146 and free electrons 148 travel toward inner surface 131 of collector 114. Droplet particles 146 and free electrons 148 impact inner surface 131 of collector 114. Contamination debris 150 on inner surface 131 illustrates the accumulation of droplet particles 146 at a particular location on inner surface 131 of collector 114.

[0094] The impact of droplet particles 146 and free electrons 148 causes vibrations 149 to propagate through collector 114 from inner surface 131 to outer surface 133. Vibrations sensors 118 coupled to outer surface 133 sense vibrations 149. Vibrations sensors 118 generate sensor signals indicative of the vibrations. Vibrations sensors 118 directly below the impact site will sense stronger vibrations than vibrations sensors 118 on either side. The sensor signals from these vibrations sensors 118 will be indicative of how close each vibrations sensor 118 is to the impact site. Thus, the sensor signals are indicative of the location of the impact site of contamination debris 150.

[0095] The sensor signals generated by the individual vibrations sensors 118 can be indicative of the time of flight of droplet particles 146 and free electrons 148. The distance between the droplet impact site and each vibrations sensor 118 is known. The speed of propagation of vibrations through collector 114 is also known. Thus, the sensor signals from vibrations sensors 118 will each sense the impact vibrations from debris particles 146 at slightly different times. The timing of the sensor signals, together with the known timing of the plasmatized pulse, is indicative of the time of flight of droplet particles 146 prior to impact. The time of flight is indicative of the velocity of droplet particles 146. The velocity of droplet particles 146 is indicative of the energy of droplet particles 146. The energy of droplet particles 146 is in turn indicative of the properties of the plasma generated from the droplet.

[0096] FIG. 1The cross-sectional nature of the view shows that the collector 114 includes a cap layer 152. The cap layer 152 can include a heat resistant and light transmissive material. This enables EUV light to pass through the cap layer 152 so that the EUV light can be reflected by other layers beneath the cap layer 152. The collector 114 can also include a multilayer structure 154. The multilayer structure 154 can include alternating silicon and molybdenum layers. Alternatively, the multilayer structure 154 can include other types of layers. The multilayer structure 154 can include layers that both transmit light and reflect light. The collector 114 can also include a substrate 156 beneath the multilayer structure 154. The substrate 156 can include a relatively thick layer of silicon or another suitable material. The substrate 156 can include a highly reflective material to ensure that all EUV light that passes through the cap layer 152 will be reflected toward the scanner 108 (see FIG. 4B ). The bottom of the substrate 154 corresponds to the outer surface 133 of the collector 114. The collector 114 can have other structures and materials without departing from the scope of the present disclosure.

[0097] In some embodiments where the droplet 128 is tin, the droplet particles 146 correspond to positively charged tin particles. The positively charged tin particles can include ionized tin atoms or groups of ionized tin atoms. The droplet particles 146 can include other materials based on the material of the droplet 128.

[0098] FIG. 4B is a graph 400 illustrating sensor signals provided by one or more vibration sensors 118, according to some embodiments. The vibration sensors 118 generate a voltage in response to receiving vibrations from a debris impact. Each vibration sensor 118 can record time of flight and amplitude information associated with a debris impact. The sensor signals can have other forms or characteristics without departing from the scope of the present disclosure. FIG. 5 The illustrated sensor signals can correspond to a superposition of sensor signals from each of a plurality of vibration sensors 118.

[0099] FIGS. 1-2B is a block diagram illustrating a vibration sensor 118 coupled to a signal processor 158, according to some embodiments. The signal processor 158 can be part of the control system 116 (see FIG. 5 ). Alternatively, the signal processor 158 can be separate from the control system 116. FIG. 5 and FIG. 3 may be particularly relevant to step 308 of the method 300 of FIG. 5 , but FIG. 5 and FIG. 4B may be relevant to other systems, methods, and processes without departing from the scope of the present disclosure.

[0100] Vibration sensor 118 is coupled to signal processor 158. Specifically, vibration sensor 118 provides sensor signals to signal processor 158. Signal processor 158 receives the sensor signals and performs simple signal processing and analysis on them. Based on the signal analysis and processing, signal processor 158 generates output data 160. Output data 160 may correspond to sensor data.

[0101] The signal processor 158 can initially receive voltage-based sensor signals, for example, FIG. 6 The signal processor 158 can then perform signal processing techniques on the sensor signal. The signal processor 158 can filter out background noise from the sensor signal. The signal processor can perform frequency distribution analysis of the sensor signal. The signal processor 158 can perform amplitude analysis of the sensor signal. The signal processor 158 can perform phase deviation analysis. The signal processor 158 can perform velocity analysis of the sensor signal. The signal processor 158 can perform superposition analysis on the sensor signal, for example, by superimposing sensor signals from two or more vibration sensors and then performing signal analysis on the superimposed sensor signal. The signal processor 158 can perform deconvolution analysis on the sensor signal. The signal processor 158 outputs output data 160. The output data 160 can include the various types of analysis data described above regarding the sensor signal. The analysis data 160 can include other types of analysis data without departing from the scope of this disclosure.

[0102] In some embodiments, another part of the signal processor 158 or control system 116 may receive and retain environmental information. The environmental information may be referred to as EUV light generation parameter data, or may be a subset of EUV light generation parameter data. The environmental information may include information related to the materials of the droplet 128, collector 114, and shell 124. Material information may include elemental composition, mass, density, thickness, sound velocity, natural resonant frequency, melting point, boiling point, electronic configuration, allotropy, ionization energy, van der Waals radius, crystal structure, Young's modulus, shear modulus, bulk modulus, and other information related to the various materials constituting the collector 114 and shell 124 within the droplet 128.

[0103] The environmental information can include information related to the laser 112. The laser information can include pulse duration, pulse-to-pulse position separation, pulse-to-pulse delay time, beam stability, beam energy, beam phase, beam profile, beam caustic, EUV energy, wavefront information, and other information associated with the laser 112. This information can include laser-to-droplet position associated with a flattening pulse, and laser-to-flattening droplet position associated with a flattening droplet 128. Some laser information can be determined based on other parameters through a machine learning process, as will be set forth in greater detail below.

[0104] The environmental information can include information related to the droplet 128. This information can include droplet velocity, droplet size, droplet frequency, droplet temperature, droplet material, and other information related to the droplet 128.

[0105] The environmental information can include thermal information. The thermal information can include temperature within the EUV light generation chamber 104, temperature of heaters associated with the EUV light generation chamber 104, collector surface temperature, air temperature, flow inlet temperature, substrate temperature, heat of fusion, heat of vaporization, molar heat capacity, thermal conductivity, thermal expansion, etc. Although not shown in the figures, the EUV light generation chamber 114 can include various flow inlets, director vanes, mass flow controllers, and other mechanisms for selectively flowing fluids into the EUV light generation chamber 104.

[0106] The environmental information can include flow information. The flow information can include center cone flow information, umbrella cone flow information, perimeter flow information, flash flow information, their flow information, mass flow control limit information, mass flow control resolution, backflow pressure, or other types of information related to the flow of fluids into and out of the EUV light generation chamber 104. The environmental information can also include pressure information, e.g., vacuum pressure information, chamber pressure, vapor pressure, or other information related to the pressure within the EUV light generation chamber 104.

[0107] Various types of environmental information can be stored in or provided to the signal processor 158 to assist in signal processing. The signal processor 158 can take into account various types of environmental information when analyzing or processing sensor signals received from the vibration sensor 118.

[0108] FIG. 6 A plurality of plots including sensor signals from one or more vibration sensors 118 according to some embodiments. FIGS. 1-5 The description of FIG. 1 is referenced with respect to FIG. 6The described components, processes, and systems. Plot 600 is the raw sensor signal from the vibration sensor 118 during the plasma generation process. At time to, a flattening laser pulse from the laser 112 impacts the droplet 128. At time ti, a plasmafying pulse from the laser 112 impacts the flattened droplet 128. At time t2 and near thereto, plasma evolution of the plasmafied droplet 128 occurs. Further, near time t2, droplet debris 150 scatters from the plasmafied droplet. At time t3, the interaction has ended. As will be set out in more detail below, plots 602, 604, and 606 illustrate components of the raw sensor signal corresponding to each event.

[0109] Plot 604 illustrates the component of the raw sensor signal from the vibration sensor 118 that is based on the flattening laser pulse from the laser 112 impacting the droplet 128. There is some delay between the flattening pulse at time to and the sensing of the flattening pulse. Plot 604 illustrates that shortly after the flattening pulse impact at time to, the vibration sensor 118 detects some vibrations associated with the shockwave from the flattening pulse. The vibrations from the flattening pulse persist in the signal until before time t2.

[0110] Plot 602 illustrates the component of the raw sensor signal from the vibration sensor 118 that is based on the plasmafying pulse from the laser 112 impacting the droplet 128 at time ti. The shockwave from the plasmafying pulse is significant in the signal 602 until approximately time t3. The amplitude of the vibrations from the shockwave of the plasmafying pulse is greatest in the component of the raw signal.

[0111] Plot 606 illustrates the component of the raw sensor signal from the vibration sensor 118 that is based on the particle debris 150 impacting the collector 135 or the inner surface 131 of the housing 124, as the case can be. Some debris results from the impact of the flattening pulse. Most of the debris results from the plasmafying pulse. The signals of plots 602, 604, and 606 can be extracted from the raw sensor signal of plot 600 by the signal analyzer 658. The various signals and signal components can have different forms than shown without departing from the scope of the present disclosure. FIG. 7 Other forms are shown without departing from the scope of the present disclosure.

[0112] FIG. 6 Plot 700 of sensor signal intensity in the frequency domain according to some embodiments. Plot 700 represents FIGS. 1-6The Fourier transform of the raw sensor signal of plot 600. Thus, plot 700 represents a frequency domain representation of the signal from vibration sensor 118 that results from laser 112 hitting droplet 128 and droplet debris 150 hitting collector 114. Plot 700 can include a true signal-to-noise ratio as well as anomalous frequency peaks 702 that can be caused by system instability. Noise can be identified and filtered out through machine learning. Anomalous peaks 702 can be used to diagnose system problems by comparison to data from a large database. Anomalous peaks can be caused by environmental vibrations, inherent frequencies of hardware components, or other factors. Thus, converting the time domain sensor signal from vibration sensor 118 to a frequency domain signal can be very beneficial in diagnosing system problems as described above. FIG. 8 The signal generated from the sensor signal of vibration sensor 118 by the components, systems, and processes described. Signal processor 158 can perform a Fourier transform on the time domain signal of plot 600 to generate the frequency domain signal of plot 700.

[0113] The Fourier transform helps to distinguish frequency distributions. The frequency distribution can include a true signal-to-noise ratio as well as anomalous frequency peaks 702 that can be caused by system instability. Noise can be identified and filtered out through machine learning. Anomalous peaks 702 can be used to diagnose system problems by comparison to data from a large database. Anomalous peaks can be caused by environmental vibrations, inherent frequencies of hardware components, or other factors. Thus, converting the time domain sensor signal from vibration sensor 118 to a frequency domain signal can be very beneficial in diagnosing system problems as described above.

[0114] In the long term, time domain signals and frequency domain signals can be collected and stored. Various time domain signals and frequency domain signals can be tagged according to the quality of EUV light generation with which they are associated. As will be set forth in greater detail below, this data can be used in a machine learning process that can help to determine root causes and correct abnormal conditions to improve laser plasma conditions.

[0115] FIG. 8 is an illustration of collector 114 according to some embodiments. FIG. 8 and FIG. 3 The description of method 300 of FIG. 8 may be particularly relevant to step 310 of method 300 of FIG. 8 The principles illustrated with respect to FIG. 2A may be implemented in other systems, components, and processes without departing from the scope of the present disclosure. In FIG. 2B and FIG. 8 the collector 114 is shown in a different orientation than FIG. 2A In fact, FIG. 2B and FIG. 8 the collector 114 of can have the same orientation as

[0116] FIG. 8 A calibration process that can help to determine plasma evolution characteristics of droplet 128 is illustrated. FIG. 8 A known point 162 above collector 114 is illustrated. Known point 162 will have a specific set of position coordinates (x, y, z) in a coordinate system. FIG. 8A primary focus 164 associated with the collector 114 is also shown. As previously mentioned, the collector 114 can be a parabolic or elliptical mirror having a primary focus. The primary focus 164 corresponds to the primary focus of the collector 114. FIG. 8 The locations of the three vibration sensors 118 are also shown. The locations of the vibration sensors 118 are shown on the inner surface 131 of the collector 114. However, in practice, the vibration sensors are located at the corresponding locations on the outer surface 135 of the collector 114.

[0117] A collision test or calibration procedure can be performed on the vibration sensors 118 and the collector 114. First, a first known mass is launched from the known point 162 toward the collector 114 at a velocity V. The first mass impacts the inner surface 131 of the collector 114. A vibration wave propagates from the impact location to the various vibration sensors 118. The vibration arrives at each vibration sensor 118 after a respective time period.

[0118] Next, a second known mass is launched from the primary focus 164 toward the collector 114 at a velocity V. The second mass impacts the inner surface 131 of the collector 114. A vibration wave propagates from the impact location to the various vibration sensors 118. The vibration arrives at each vibration sensor 118 after a respective time period.

[0119] The signal difference between the sensor signals of the various vibration sensors caused by the impacts of the first mass and the second mass can be used to calibrate the vibration sensors 118 at defined locations in the space. In one example, a calibration matrix or parameter space can be constructed by launching masses from various locations. The masses can vary, the velocities can vary, the release locations can vary, and the target impact locations can vary. The vibration sensor signals for each impact can be recorded. The calibration matrix can include various masses and velocities in the parameter space. A reference data table can include the differences in launch target locations.

[0120] The characteristics of the sensor signals according to these calibration data can be used as reference data to benchmark the sliding distance between the plasma location and the primary focus for a real-time mother system setup. Thus, the control system 116 can utilize the calibration data to determine characteristics of the plasma evolution.

[0121] The system resolution improves with the number of vibration sensors. While three vibration sensors 118 are shown, in practice, more than three vibration sensors 118 can be used, such as FIGS. 2A-2D three vibration sensors 118 are shown, in practice, more than three vibration sensors 118 can be used, such as FIG. 8 is shown.

[0122] With continued reference to FIG. 9Another method can be utilized to determine the position of the debris on the collector 114. In this example, instead of emitting particles from a known point 162, particles can be emitted towards the collector from a plurality of unknown or unrecorded positions. When the debris hits the inner surface 131 of the collector 114, each small piece of the droplet debris can be conceptualized as a point source, which becomes a point vibration source. The vibration waves propagate through the collector 114 and reach each vibration sensor 118 at different respective times based on their respective distance from the hitting position. From the time difference of the arrival of the vibration signal from the hit to each vibration sensor 118, the position of the hit can be inferred. For example, the shorter the time it takes for the vibration signal to reach a specific vibration sensor 118, the closer the impact position is to that vibration sensor. With three or more vibration sensors 118 of known positions, an exact impact position can be derived based on the different elapsed times of the arrival of the vibration signal to the vibration sensors 118. The plasma position in space can also be calculated by considering the time difference between the excitation flattening pulse and the excitation plasma pulse to determine the impact position and the impact time. Because the speed of light is constant for both laser pulses, the time difference between the excitation laser pulses can be converted to a travel distance of the droplet 128, as the velocity of the droplet 128 is known. Thus, this procedure can be used to determine the plasma position.

[0123] FIG. 9 is an illustration of a collector 114 according to some embodiments. FIG. 3 and the respective description can specifically relate to FIG. 9 step 310 of the method 300 of FIG. 9 the principles of can be implemented according to other components, systems and procedures without departing from the scope of the present disclosure. FIG. 2D the collector 114 of can correspond to FIG. 2D the collector 114 of. Specifically, the collector 114 can be made of mirror units 140 which are separated from each other by shrink joints 142. The mirror units 140 are arranged in a grid, as FIG. 9 is shown.

[0124] FIG. 9 droplet particles 146 and free electrons 148 which have been ejected or scattered from the plasma droplet 128 are shown. The droplet particles 146 correspond to the debris which is deposited on the inner surface 131 of the collector 114. As there is a vibration sensor 118 for each mirror unit 140, the position of each debris deposit can be determined based on the sensed vibrations. For smaller sized mirror units 140 and a correspondingly larger number of vibration sensors 118, the resolution of the debris position detection can be improved. The principles of can be utilized to determine the orientation of the bottom surface of the flattening droplet 128, as the debris scattering occurs according to the orientation of the bottom surface of the flattening droplet 128. FIG. 10A the principles of can be utilized to determine the orientation of the bottom surface of the flattening droplet 128, as the debris scattering occurs according to the orientation of the bottom surface of the flattening droplet 128.

[0125] Shrink fitting 142 can be used to insert an inlet. The inlet allows cleaning fluid or wash fluid to flow to the location of debris 150. The inlet can be selectively controlled so that the cleaning fluid flow is directed only to those locations where debris has accumulated.

[0126] FIG. 10A The process for constructing a two-dimensional representation of plasma evolution according to some embodiments is illustrated. FIG. 3 And the corresponding description can be specifically related to FIG. 10A Method 300 is related to step 310, but FIG. 10A The principles shown can be used in conjunction with other components, systems and processes without departing from the scope of this disclosure.

[0127] At position 1002, the plasma-enhanced laser pulse 130 impacts the bottom of the planarized droplet 128. Therefore, the planarized droplet 128 has been impacted by the planarization pulse. In the example of Figure 1000, the plasma-enhanced pulse 130 is well aligned with the planarized droplet 128. Good alignment corresponds to the plasma-enhanced pulse 130 being concentrated at the center of the bottom surface of the planarized droplet 128.

[0128] At 1004, a vibration sensor coupled to collector 114 senses vibrations from planarization pulses, plasma pulses, and impacts from fragments of droplet 128. Step 1004 shows a top view of collector 114. Vibration sensors 118 are divided into groups 118a, 118b, 118c, and 118d based on the intensity of the vibrations sensed by the vibration sensors. FIG. 10B In the example, vibration sensor 118a, located closer to the center of collector 114, senses the strongest vibration. Vibration sensor 118b, slightly farther from the center than vibration sensor 118a, senses moderately strong vibration, but not as strong as sensor 118a. Vibration sensor 118c, slightly farther from the center than vibration sensor 118b, senses weak vibration. Vibration sensor 118d, slightly farther from the center than vibration sensor 118c, senses little or no vibration. This pattern indicates that debris scattering and laser pulse shock waves are concentrated near opening 129.

[0129] At 1006, the spatial density distribution of vibration is reconstructed on the inner surface 131 of collector 114. The spatial density distribution indicates the vibration intensity distribution on the surface of the collector based on the sensor signals received from each set of vibration sensors 118a-d in step 1004.

[0130] At 1008, a 2D plasma evolution model 170 is reconstructed based on the spatial density of the neutral point at the location of the plasma. More specifically, as shown in step 1006, a plasma evolution is generated based on the reconstruction of the spatial density distribution on the collector surface. The plasma evolution reconstruction indicates dense plasma regions and sparse plasma regions in the plasmatized droplet 128.

[0131] FIG. 10B A process 1010 for constructing a two-dimensional representation of a plasma evolution is shown in accordance with some embodiments. FIG. 10A The process 1010 is the same as the process 1000 FIG. 10B except that in FIG. 10A the initial alignment of the plasmatization laser pulses 130 with the flattened droplet 128 is poor, whereas in FIG. 10B the alignment is good. The only difference in the use of the process and steps in FIG. 10B is the different reference numbers used to avoid confusion.

[0132] At 1012, the plasmatization laser pulses 130 impact the bottom of the flattened droplet 128. Thus, the flattened droplet 128 has been impacted by the flattening pulses. In the example of FIG. 10B the alignment of the plasmatization pulses 130 with the flattened droplet 128 is poor. In FIG. 10A the plasmatization pulses 130 impact the front edge of the bottom surface of the flattened droplet 128, rather than the center of the bottom surface of the flattened droplet 128.

[0133] At 1014, the vibration sensors coupled to the collector 114 sense vibrations from the impact of the flattening pulses, the plasmatization pulses, and from debris from the droplet 128. Step 1014 shows a top view of the collector 114. The vibration sensors 118 are divided into groups 118a, 118b, 118c, and 118d based on the intensity of the vibrations that the vibration sensors sense. The vibration sensors 118a sense vibrations more intensely than the vibration sensors 118b. The vibration sensors 118b sense vibrations more intensely than the vibration sensors 118c. The vibration sensors 118c sense vibrations more intensely than the vibration sensors 118d. FIG. 10B The difference between FIG. 10B is that the vibration sensors 118a that sense vibrations intensely are not as neatly concentrated around the aperture 129 in FIG. 10A as they are in FIG. 10B . In the example of FIG. 10C the vibration sensors 118a closest to the center of the collector 114 sense vibrations most intensely.

[0134] At 1016, a spatial density distribution of the vibrations is reconstructed on the inner surface 131 of the collector 114. The spatial density distribution is indicative of the intensity distribution of the vibrations on the surface of the collector based on the sensor signals received from the groups of vibration sensors 118a-d in step 1014.

[0135] At 1018, a 2D plasma evolution model 170 is reconstructed based on the spatial density of the neutral points at the location of the plasma. More specifically, as shown in step 1016, the plasma evolution is generated based on the reconstruction of the spatial density distribution on the collector surface. The plasma evolution reconstruction is indicative of dense plasma regions and sparse plasma regions in the plasmatized droplet 128. The plasma evolution reconstruction is indicative of the shape and density of the plasma at times tl, t2, t3, and t4.

[0136] FIG. 10B A process 1020 for constructing a two-dimensional representation of the plasma evolution is shown in accordance with some embodiments. FIG. 10B The process 1020 is the same as the process 1010 FIG. 10C except that in FIG. 2D the collector 114 includes separate mirror units 140 and contraction joints 142 in accordance with FIG. 9 and FIG. 10C The different reference numbers for the processes and steps are used only to avoid confusion. FIG. 10C

[0137] At 1022, the plasmatization laser pulse 130 impinges on the bottom of the planarization droplet 128. Thus, the planarization droplet 128 has been impinged by a planarization pulse. In FIG. 10C the example, the plasmatization pulse 130 is poorly aligned with the planarization droplet 128. In FIG. 11 the plasmatization pulse 130 is concentrated on the leading edge of the bottom surface of the planarization droplet 128, rather than the center of the bottom surface of the planarization droplet 128.

[0138] At 1024, vibration sensors coupled to the collector 114 sense vibrations from the impingement of the planarization pulse, the plasmatization pulse, and from debris from the droplet 128. Step 1024 shows a top view of the collector 114. The vibration sensors 118 are divided into groups 118a, 118b, 118c, and 118d based on the intensity of the vibrations sensed by the vibration sensors. The vibration sensors 118a sense vibrations more intensely than the vibration sensors 118b. The vibration sensors 118b sense vibrations more intensely than the vibration sensors 118c. The vibration sensors 118c sense vibrations more intensely than the vibration sensors 118d.

[0139] ​At 1026, a spatial density distribution of the vibrations is reconstructed on the inner surface 131 of the collector 114. The spatial density distribution is indicative of the intensity distribution of the vibrations on the surface of the collector based on the sensor signals received from the sets of vibration sensors 118a-d in step 1024.

[0140] At 1028, a 2D plasma evolution model 170 is reconstructed based on the spatial density of the neutral points at the location of the plasma. More specifically, as shown in step 1026, the plasma evolution is generated based on the reconstruction of the spatial density distribution on the surface of the collector. The plasma evolution reconstruction is indicative of dense plasma regions and sparse plasma regions in the plasmatized droplets 128. The plasma evolution reconstruction represents the shape and density of the plasma at times tl, t2, t3, and t4.

[0141] FIG. 11 A process 1100 for generating a 3D holographic and time-resolved plasma model is shown in accordance with some embodiments. FIG. 3 The corresponding description can be specifically related to step 312 of the method 300 of FIG. 11 but the principles of FIGS. 10A-10C can be utilized in accordance with other processes, systems, and components without departing from the scope of the present disclosure.

[0142] At 1102, a plurality of plasma evolution models 170 is generated. Each plasma evolution model 170 can be generated in accordance with the process described with respect to FIGS. 10A-10C At 1102, a plurality of plasma evolution models 170 is generated. Each plasma evolution model 170 can be generated in accordance with the process described with respect to FIGS. 12A-12H The principles described with respect to

[0143] At 1104, the plurality of plasma evolution models 170 is combined into a 3D plasma model 172. The 3D plasma model 172 is 3D in the sense that it includes two spatial axes and one time axis. The two spatial axes correspond to the axes of the 2D plasma evolution models 170. The time axis is generated because each 2D plasma model 170 represents a different time instant.

[0144] 3D plasma model 172 indicates plasma center 174, main focus 164, offset distance 176 of the plasma from main focus 164, plasma-ization pulse position 178, and plasma edge 180. The various positions can be given in Euclidean coordinates and have a time component. For example, plasma center 174 can be given at time ti early in the evolution of the plasma. Plasma edge 180 can be given at time t2 later in the evolution of the plasma. Debris position 150 can also be included at time t3 much later than time t2. Various other processes can be utilized to generate 3D holographic and time-resolved plasma models without departing from the scope of the present disclosure.

[0145] FIGS. 12A-12H is a plot of a 3D plasma model and corresponding flattening and plasma-ization pulses in accordance with some embodiments. FIGS. 12A-12H Diagnoses related to 3D plasma model and associated parameters are shown. FIG. 3 The principles associated with FIGS. 12A-12H Step 314 of method 300 can be related, but the principles associated with FIG. 12A can be used in conjunction with other systems, processes, and components without departing from the scope of the present disclosure.

[0146] FIG. 12A A 3D plasma model 172 associated with generating EUV light from droplet 128 is shown. FIG. 12B A flattening laser pulse 130a and a plasma-ization laser pulse 130b for flattening and plasma-izing droplet 128 are also shown. 3D plasma model 172 represents good laser targeting and plasma generation. Flattening pulse 130a is concentrated on droplet 128. Plasma-ization pulse 130b is concentrated on impinging the entire flattened droplet 128. There is no flattened droplet 128 protruding outside of plasma-ization pulse 130b. In addition, the flattened droplet has only a relatively small tilt relative to the horizontal. All of these parameters result in good plasma formation, good EUV light generation, and low debris scattering.

[0147] FIG. 12B A 3D plasma model 172 associated with generating EUV light from droplet 128 is shown. FIG. 12CA flattening laser pulse 130a and a plasma formation laser pulse 130b are also shown for flattening and plasma formation of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma formation. The flattening droplet 128 is too large. The plasma formation pulse 130b cannot hit the entire flattening droplet 128. Some portions of the flattening droplet 128 protrude from the plasma formation pulse 130b. The portions of the droplet 128 that are not plasma formed cause a large amount of debris on the collector 114 and the enclosure 124. The portions of the flattening droplet 128 that are not hit by the plasma formation pulse 130b are not plasma formed.

[0148] FIG. 12C A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12D A flattening laser pulse 130a and a plasma formation laser pulse 130b are also shown for flattening and plasma formation of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma formation. The flattening droplet 128 is tilted at a very large angle relative to the horizontal. The larger the angle, the smaller the contact area with the plasma formation pulse 130b. This results in a large amount of debris.

[0149] FIG. 12D A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12E A flattening laser pulse 130a and a plasma formation laser pulse 130b are also shown for flattening and plasma formation of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma formation. The plasma formation pulse 130b is not concentrated on the flattening droplet 128. This results in poor plasma evolution and an increase in the amount of debris and contamination.

[0150] FIG. 12E A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12F A flattening laser pulse 130a and a plasma formation laser pulse 130b are also shown for flattening and plasma formation of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma formation. A piece of debris 182 has broken off from the droplet 128. In addition to increasing the amount of debris, this also results in the unintended generation of EUV light.

[0151] FIG. 12G A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12GA flattening laser pulse 130a and a plasmafication laser pulse 130b are also shown for flattening and plasmafication of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma generation. The flattening pulse 130a is not focused on the droplet 128. This results in poor plasma evolution and an increased amount of debris and contamination.

[0152] FIG. 12G A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12H A flattening laser pulse 130a and a plasmafication laser pulse 130b are also shown for flattening and plasmafication of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma generation. The flattening pulse 130a does not have enough energy to flatten the droplet 128. When the plasmafication pulse 130b hits the droplet 128, the droplet 128 is not fully plasmafied. This results in poor plasma evolution and an increased amount of debris and contamination.

[0153] FIG. 12H A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12I A flattening laser pulse 130a and a plasmafication laser pulse 130b are also shown for flattening and plasmafication of the droplet 128. The 3D plasma model 172 represents poor laser targeting and plasma generation. The flattening pulse 130a is not properly focused on the droplet 128. This results in the flattened droplet 128 tilting in an undesirable direction. The angular tilt causes the shockwave to partially progress in the direction of the incoming droplet 128. This will destabilize the incoming droplet and the EUV light generation will also be destabilized. This results in poor plasma evolution and an increased amount of debris and contamination.

[0154] FIG. 12I A 3D model of the EUV light generation chamber 104 and the droplet 128 hit by the plasmafication pulse 130b is shown. In this example, the flattened droplet 128 is too large. Some portions of the flattened droplet 128 protrude outside of the plasmafication pulse 130b. The result is that debris 146 broken off from the non-plasmafied portions of the droplet 128 reflect off of the collector 114 and travel from the EUV light generation chamber towards the scanner 108 (see FIG. 1 ). If any of the debris 146 hits a reticle within the scanner 108, this can be very problematic. If the debris 146 hits a reticle, the lithography process can be ruined. In addition, the reticle can need to undergo a time consuming and very expensive cleaning process to remove the debris 146 from the reticle. FIG. 12J

[0155] FIG. 12I ​A 3D model of the EUV light generation chamber 104 and the droplet 128 being hit by the plasmafication pulse 130b is shown. In FIG. 12K In the example of FIG. 3, a satellite 182 has broken off from the droplet 128. As a result, debris 146 is scattered from the droplet 128. The debris 146 forms a debris deposit 150 on the inner surface 135 of the enclosure 124. This can also result in a debris deposit 150 on the inner surface 131 of the collector 114, degrading the collector 114.

[0156] FIG. 12D A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown. FIG. 12K A flattening laser pulse 130a and a plasmafication laser pulse 130b used to flatten and plasmaficate the droplet 128 are also shown. FIG. 3 The corresponding description in FIG. 3 can specifically relate to steps 314 and 316 of the method 300 of FIG. 2. FIG. 12K The 3D plasma model 172 represents poor laser targeting and plasma generation. The plasmafication pulse 130b is not concentrated on the flattened droplet 128. This results in poor plasma evolution and an increased amount of debris and contamination.

[0157] After the vibration sensor 118 has sensed the vibration, the sensor data, along with other parameters associated with the plasma generation, are passed to an analysis model of the control system 116. The analysis model has been trained with a machine learning process to identify parameters that can be adjusted to improve the quality of the plasma generation and the EUV light generation. The analysis model analyzes the various data and generates parameter adjustment data that indicates recommended parameter adjustments to improve the plasma and EUV light generation. In FIG. 12L In the example of FIG. 3, the parameter adjustment data indicates that the flattening pulse should be adjusted to increase the energy of the flattening pulse. This can include a specific energy increase of, for example, 3 mJ or other value.

[0158] FIG. 12K A 3D plasma model 172 associated with the generation of EUV light from the droplet 128 is shown after the recommended parameter adjustments according to FIG. 12L FIG. 12L A flattening laser pulse 130a and a plasmafication laser pulse 130b used to flatten and plasmaficate the droplet 128 are also shown. In FIG. 13 In FIG. 3, the flattening pulse 138 is concentrated on the droplet 128 and has sufficient energy to flatten the droplet 128. The plasmafication pulse 130b is also concentrated on the droplet 128. The droplet 128 has a low inclination relative to the horizontal. The result is a high quality plasma evolution and EUV light generation.

[0159] FIG. 14 ​is a graph 1300 showing separation distance versus energy of a flattening pulse for planarizing pulses and plasma pulses according to some embodiments. The graph 1300 shows an upper control limit 1302 and a lower control limit 1304. In some embodiments, separation distances and pulse energies that fall between the upper control limit 1302 and the lower control limit 1304 result in good laser targeting and good plasma generation. The graph 1300 also shows a line 3006 that is intermediate between the upper control limit 1302 and the lower control limit 1304. The graph 1300 shows a plurality of data points 1308 that fall between the upper control limit 1302 and the lower control limit 1304. The graph 1300 shows a single data point 1310 that is above the upper control limit. The data point 1310 corresponds to a process for which parameter adjustment data will be generated to ensure that the next process falls between the upper control limit 1302 and the lower control limit 1304.

[0160] In some embodiments, the separation distance is between 4 pm and 12 pm. In some embodiments, the energy of the flattening pulse ranges between 0.00088 and 0.0168 atomic units. Other separation distances and energies can be utilized without departing from the scope of the present disclosure.

[0161] FIG. 14 is a flowchart of a method 1400 for operating an EUV light generation system according to some embodiments. FIG. 3 may be particularly relevant to steps 314 and 316 of the method 300 of FIG. 14 but the principles of the method 300 of FIG. 5 may be used with other processes, systems, and components without departing from the scope of the present disclosure.

[0162] At 1402, a big data mining process is performed to collect historical EUV light generation parameter data related to a large number of previously performed EUV light generation processes. The historical EUV light generation parameters can include sensor data from vibration sensors, laser pulse parameters related to the energy, timing, and occupancy of flattening pulses and plasma pulses, and droplet parameters related to the size, velocity, temperature, and spacing of droplets. The historical EUV light generation parameters can include environmental data, including material information, thermal information, fluid information, pressure information, and other environmental information, such as those mentioned with respect to FIG. 15 The historical EUV light generation parameter data can also include flag data indicating whether each EUV light generation process was satisfactory.

[0163] At 1404, an analysis model is trained with a machine learning process based on historical sensor data and environmental data from the data mining process of step 1402. The machine learning process trains the analysis model to generate parameter adjustment data. The parameter adjustment data indicates an adjustment to one or more EUV light generation parameters that will result in improved EUV light generation. Further details regarding the machine learning process will be set forth with respect to FIG. 15 elaborated.

[0164] Example adjustments can include adjusting laser alignment if the sensor data indicates poor targeting. If the sensor data indicates droplet satellites, the adjustment can include adjusting the velocity of the droplets 128. If the sensor data indicates poor size of the planarization droplet, the adjustment can include reducing the planarization pulse energy. If the sensor data indicates an undesirable tilt angle of the planarization droplet, the adjustment can include adjusting the distance of the laser sensor to the droplet sensor. If the sensor data indicates poor planarization pulse timing, the adjustment can include adjusting the planarization pulse firing time. If the sensor data indicates insufficient energy of the planarization pulse, the adjustment can include increasing the planarization pulse power. If the sensor data indicates low plasma pulse energy, the adjustment can include adjusting the plasma pulse power. If the sensor data indicates planarization droplet direction, the adjustment can include adjusting the Y-axis position of the planarization laser pulse. If the sensor data indicates undesirable droplet debris on the collector or housing, the adjustment can include adjusting the Z-axis position of one or both of the plasma laser and the planarization laser. If the sensor data indicates collector contamination, the adjustment can include adjusting the planarization droplet angle by adjusting the position of the planarization laser. Other adjustments can be made for these and other issues indicated by the sensor data without departing from the scope of the present disclosure.

[0165] After training the analysis module with the machine learning process in step 1404, the analysis module is ready to improve the functioning of the EUV light generation system. At 1406, the method 1400 includes generating a plasma and EUV light by irradiating a droplet with a flattening pulse and a plasmafying pulse. At 1408, the method 1400 includes generating a vibration sensor signal based on plasma generation shockwave and debris impact. At 1410, the method 1400 includes generating sensor data by performing signal analysis on the sensor signal. At 1412, the method 1400 includes providing the sensor data and environmental data related to the current plasma generation process to the analysis model. At 1414, the analysis model analyzes the sensor data and environmental data. At 1416, the analysis model generates parameter adjustment data. At 1418, the control module adjusts the EUV light generation parameters based on the parameter adjustment data. The process then returns to 1406, where a plasma EUV light is generated with the adjusted parameters. The process can repeat continuously until the sensor data indicates satisfactory EUV light generation.

[0166] FIG. 15 is a block diagram of a control system 116 according to some embodiments. FIG. 1 The control system 116 of FIG. 1 is one example of a control system 116. According to some embodiments, the control system 116 is configured to control the operation of an EUV light generation system, such as FIG. 2A the EUV light generation system 100 of FIG. 2B and FIG. 15 the EUV light generation system 200. The control system 116 utilizes machine learning to adjust the parameters of the EUV light generation system. The control system 116 can adjust the parameters of the EUV light generation system to maintain high quality EUV light generation.

[0167] In one embodiment, the control system 116 includes an analysis model 190 and a training module 192. The training module 192 trains the analysis model 190 with a machine learning process. The machine learning process trains the analysis model 190 to predict future EUV light generation quality and select parameters of an EUV light generation process that will result in high EUV light generation quality. Although the training module 192 is shown as separate from the analysis model 190, in practice, the training module 192 can be part of the analysis model 190. Although FIG. 15 the description primarily targets generating EUV parameter adjustments, the principles of FIG. 5 can be used to train the analysis model 190 to predict debris contamination on the collector 114 or the enclosure 124, and adjust the stream flow parameters or the EUV light generation parameters to reduce the debris contamination.

[0168] The control system 116 includes or stores training set data 194. The training set data 194 includes historical EUV light generation quality data 196 and historical EUV light generation parameter data 198. The historical EUV light generation quality data 196 indicates whether each historical EUV light generation process was satisfactory. The historical EUV light generation parameter data 198 includes data about process conditions or parameters during the EUV light generation processes associated with the historical EUV light generation quality data. As will be set out in greater detail below, the training module 192 utilizes the historical EUV light generation quality data 196 and the historical EUV light generation parameter data 198 to train the analysis model 190 through a machine learning process.

[0169] In one embodiment, the historical EUV light generation quality data 196 includes data indicating the quality of the EUV light generation processes. For example, during operation of a semiconductor manufacturing facility, thousands or millions of semiconductor wafers can be processed over the course of several months or years. A corresponding large number of EUV light generation processes are performed while processing the wafers. The EUV light generation processes are performed to generate light for EUV lithography processes. The historical EUV light generation quality data 196 includes the EUV light generation quality of these EUV light generation processes, or a selected time period of the performance of EUV light generation.

[0170] In one embodiment, the historical EUV light generation parameter data 196 includes various process conditions or parameters during the EUV light generation processes associated with the historical EUV light generation quality data 196. Thus, for each EUV light generation quality value in the historical EUV light generation quality data 196, the historical EUV light generation parameter data 198 can include process conditions or parameters that existed during the time period associated with that EUV light generation quality value. The historical EUV light generation parameter data 198 can include sensor data from vibration sensors, as well as environmental data related to laser parameters, droplet parameters, temperature parameters, material parameters, flow flow parameters, and pressure parameters, as previously described. The historical EUV light generation parameter data 198 can include environmental data of the types described above with respect to the process conditions data 192. FIG. 5

[0171] In one embodiment, the training set data 194 links the historical EUV light generation quality data 196 with the historical EUV light generation parameter data 198. In other words, each EUV light generation quality value in the historical EUV light generation quality data 196 is linked to process condition data associated with the EUV light generation process. In this way, the historical EUV light generation quality values are labels for the machine learning process. As will be set out in greater detail below, the labeled training set data can be utilized in a machine learning process to train the analysis model 190 to generate recommended parameter adjustment data to improve future EUV light generation processes. ​

[0172] In one embodiment, the analysis model 190 includes a neural network. The neural network can include a random forest network or other type of neural network. The training of the analysis model 190 will be described with respect to a neural network. However, other types of analysis models or algorithms can be used without departing from the scope of the present disclosure. The training module 192 trains the neural network using the training set data 194 through a machine learning process. During the training process, the neural network receives historical EUV light generation parameter data 198 as input from the training set data. During the training process, the neural network outputs predicted EUV light generation quality data. The predicted EUV light generation quality data predicts the EUV light generation quality that would result from the historical EUV light generation parameter data. The training process trains the neural network to generate predicted EUV light generation quality data. The training process also trains the neural network to generate recommended parameter adjustment data to improve the EUV light generation quality.

[0173] In one embodiment, the neural network includes a plurality of neural layers. Each neural layer includes neurons that define one or more internal functions. The internal functions are based on weight values associated with the neurons of each neural layer of the neural network. During training, for each set of historical EUV light generation parameter data, the control system 116 compares the predicted EUV light generation quality data to actual historical EUV light generation quality data associated with those process conditions. The control system generates an error function that indicates how well the predicted EUV light generation quality data matches the historical EUV light generation quality data. The control system 116 then adjusts the internal functions of the neural network. As the neural network generates predicted EUV light generation quality data based on the internal functions, adjusting the internal functions will cause different predicted EUV light generation quality data to be generated for the same set of historical EUV light generation parameter data. Adjusting the internal functions can cause the predicted EUV light generation quality data to result in a greater error function (a worse match to the historical EUV light generation quality data 196) or a lesser error function (a better match to the historical EUV light generation quality data 196).

[0174] After adjusting the internal functions of the neural network, the historical EUV light generation parameter data 198 is passed to the neural network again, and the analysis model 190 generates predicted EUV light generation quality data again. The training module 192 compares the predicted EUV light generation quality data to the historical EUV light generation quality data 196 again. The training module 192 adjusts the internal functions of the neural network again. This process is repeated over a number of iterations of monitoring the error function and adjusting the internal functions of the neural network until a set of internal functions is found that causes the predicted EUV light generation quality data to match the historical EUV light generation quality data 196 over the entire training set.

[0175] At the start of the training process, the predicted EUV light generation quality data will likely not match the historical EUV light generation quality data 196 very closely. However, as the training process iterates multiple times by adjusting the internal functions of the neural network, the error function will decrease until a set of internal functions is found that makes the predicted EUV light generation quality data match the historical EUV light generation quality data 196. Identifying the set of internal functions that makes the predicted EUV light generation quality data match the historical EUV light generation quality data 196 corresponds to the completion of the training process.

[0176] In one embodiment, the analysis model 190 includes two neural networks coupled together in an encoder-decoder configuration. The encoder neural network is trained through the training process described above to generate predicted EUV light generation quality. The decoder network is trained to receive the predicted EUV light generation quality and reproduce historical EUV light generation parameter data 198 that generated the predicted EUV light generation quality.

[0177] The training of the decoder neural network is similar to that of the encoder neural network. The decoder neural network comprises multiple neural layers as described above for the encoder neural network. The decoder neural network receives EUV light generation quality values ​​as input and generates historical predicted EUV light generation parameters as output. The training process utilizes historical EUV light generation parameter data 198 as labels. For each EUV light generation quality value, the decoder neural network generates predicted EUV light generation parameters. The predicted EUV light generation parameters are compared with the historical EUV light generation parameter data, and an error function is generated. The internal function of the decoder neural network is adjusted iteratively until the decoder neural network can generate predicted EUV light generation parameter data that matches the historical EUV light generation parameter data 198 within a tolerance range.

[0178] In one embodiment, after the analysis model 190 has been trained, it can be used to generate a set of recommended process conditions that will produce improved EUV light generation. For example, current EUV light generation process conditions or parameters are provided to the encoder neural network of the analysis model 190. The encoder neural network generates a prediction of future EUV light generation quality based on the current EUV light generation process conditions or parameters. If the predicted future EUV light generation quality is below a selected threshold, a higher EUV light generation quality value can be provided to the decoder neural network. The decoder neural network then generates a set of recommended EUV light generation parameter adjustments to produce this higher EUV light generation quality. The control system 116 can then adjust the operation of the various components of the EUV light generation system to implement the recommended EUV light generation parameter adjustments.

[0179] In one embodiment, the control system 116 includes a processing resource 202, a memory resource 204, and a communication resource 206. The processing resource 202 can include one or more controllers or processors. The processing resource 202 is configured to execute software instructions, process data, make parameter control decisions, perform signal processing, read data from memory, write data to memory, and perform other processing operations. The processing resource 202 can include a physical processing resource 202 located at a site or facility of the EUV light generation system. The processing resource can include a virtual processing resource 202 that is remote from the site or facility of the EUV light generation system. The processing resource 202 can include a cloud-based processing resource, including processors and servers accessed via one or more cloud computing platforms.

[0180] In one embodiment, the memory resource 204 can include one or more computer-readable memories. The memory resource 204 is configured to store software instructions associated with the functionality of the control system and its components, including but not limited to the analysis model 190. The memory resource 204 can store data associated with the functionality of the control system 116 and its components. The data can include training set data 194, current process condition data, and any other data associated with the operation of the control system 116 or any of its components. The memory resource 204 can include a physical memory resource located at a site or facility of the EUV light generation system. The memory resource can include a virtual memory resource that is remote from the site or facility of the EUV light generation system. The memory resource 204 can include a cloud-based memory resource accessed via one or more cloud computing platforms.

[0181] In one embodiment, the communication resource can include resources that enable the control system 116 to communicate with components associated with the EUV light generation system. For example, the communication resource 206 can include wired and wireless communication resources that enable the control system 116 to receive sensor signals from vibration sensors and control devices of the EUV light generation system, such as planarization and plasma formation lasers, droplet generators, and flow flow devices. The communication resource 206 can enable the control system 116 to communicate with remote systems. The communication resource 206 can include or can facilitate communication via one or more networks, such as wired networks, wireless networks, the Internet, or an intranet. The communication resource 206 can enable components of the control system 116 to communicate with each other.

[0182] In one embodiment, the analysis model 190 is implemented by the processing resource 202, the memory resource 204, and the communication resource 206. The control system 116 can be a distributed control system with components and resources and locations that are remote from each other and from the EUV light generation system.

[0183] In some embodiments, the control system 116 includes information about FIG. 16 The signal processor 158 is described. In this example, the vibration sensor 118 provides a sensor signal to the control system 116. The signal processor 158 then performs signal processing on the sensor signal and generates the sensor data previously described. In some embodiments, the signal processor 158 may be external to the control system 116.

[0184] FIGS. 1-15 This is a flowchart of a method 1600 for operating an EUV lithography system according to some embodiments. Method 1600 can utilize information regarding... FIG. 1 The system, components, and processes are described. At 1602, the method includes performing a photolithography process by generating extreme ultraviolet light in an extreme ultraviolet light generation chamber. An example of an extreme ultraviolet light generation chamber is... FIG. 2A The extreme ultraviolet light generating chamber 104. At 1604, the method includes generating sensor signals using multiple vibration sensors coupled to the extreme ultraviolet light generating chamber. An example of a vibration sensor is... FIGS. 1-16 The vibration sensor 118. At 1606, the method includes analyzing the sensor signal. At 1608, the method includes adjusting the parameters of extreme ultraviolet light generation based on the sensor signal.

[0185] FIG. 2A This is a flowchart of a method 1700 for operating an EUV lithography system according to some embodiments. Method 1700 can utilize information regarding... FIG. 2A The system, components, and processes are described. At 1702, the method includes coupling multiple vibration sensors to a collector mirror of an extreme ultraviolet lithography system. An example of a vibration sensor is... FIG. 2A Vibration sensor 118. An example of a collector reflector is... FIG. 2A The collector mirror 114. At 1704, method 1700 includes generating extreme ultraviolet light by irradiating the droplet with a laser in the extreme ultraviolet light generation chamber. An example of a droplet is... ​ The droplet is 128. An example of an extreme ultraviolet light generation chamber is... ​ The extreme ultraviolet light generation chamber. At 1706, the method includes generating a sensor signal indicating the distribution of debris from the droplets on the collector mirror using a vibration sensor. At 1708, method 1700 includes adjusting one or more parameters of the extreme ultraviolet generation based on the distribution of the debris.

[0186] In some embodiments, a method includes performing a photolithography process by generating extreme ultraviolet (EUV) light in an EUV generation chamber, and generating sensor signals using a plurality of vibration sensors coupled to the EUV generation chamber. The method includes analyzing the sensor signals and adjusting parameters for EUV light generation based on the sensor signals.

[0187] In some embodiments, a method includes coupling a plurality of vibration sensors to a collector mirror of an extreme ultraviolet lithography system and generating extreme ultraviolet light by irradiating a droplet with a laser inside an extreme ultraviolet generation chamber. The method includes generating, with the vibration sensors, sensor signals indicative of a distribution of debris from the droplet on the collector mirror and adjusting one or more parameters of the extreme ultraviolet generation based on the distribution of the debris.

[0188] In some embodiments, an extreme ultraviolet lithography system includes a droplet generator configured to output a droplet into an extreme ultraviolet generation chamber and a laser configured to generate a plasma from the droplet by irradiating the droplet with a laser. The system includes a collector mirror configured to focus extreme ultraviolet light emitted from the plasma and a plurality of first vibration sensors coupled to the collector mirror.

[0189] The foregoing has outlined rather broadly the features of several embodiments in order that the detailed description can be better understood. Those skilled in the art will appreciate that they can readily use the disclosure as a basis for designing or modifying other processes and structures for carrying out the same purposes and / or for implementing the same advantages of the embodiments introduced herein. Those skilled in the art will realize that such equivalent constructions do not depart from the spirit and scope of the disclosure and that they can make various changes, substitutions and alterations herein without departing from the spirit and scope of the disclosure.

[0190] Example 1 is a method of operating an extreme ultraviolet lithography system, comprising: performing a lithography process by generating extreme ultraviolet light in an extreme ultraviolet generation chamber; generating sensor signals with a plurality of vibration sensors coupled to the extreme ultraviolet generation chamber; analyzing the sensor signals; and adjusting a parameter of the extreme ultraviolet light generation based on the sensor signals.

[0191] Example 2 is the method of Example 1, wherein generating extreme ultraviolet light comprises: generating extreme ultraviolet light by generating a plasma from a droplet inside the extreme ultraviolet generation chamber.

[0192] Example 3 is the method of Example 2, wherein the plurality of vibration sensors are coupled to a collector mirror of the extreme ultraviolet generation chamber.

[0193] Example 4 is the method of Example 3, further comprising: determining a characteristic of the plasma based on the sensor signals.

[0194] Example 5 is the method of Example 4, wherein generating the plasma comprises: irradiating the droplet with a laser, wherein adjusting the parameter of the extreme ultraviolet light generation comprises adjusting a parameter of the laser based on the sensor signals.

[0195] Example 6 is the method of Example 4, wherein adjusting the parameter of the extreme ultraviolet light generation comprises adjusting a parameter of the droplets.

[0196] Example 7 is the method of Example 4, further comprising detecting a distribution of droplet debris on the collector mirror based on the sensor signals; and determining a characteristic of the plasma from the distribution of droplet debris.

[0197] Example 8 is the method of Example 7, wherein determining the characteristic of the plasma comprises generating a model of the plasma.

[0198] Example 9 is the method of Example 7, wherein irradiating the droplets with a laser comprises irradiating each droplet with a first laser pulse; and irradiating each droplet with a second laser pulse, wherein the sensor signals are indicative of a shockwave from the first laser pulse and a shockwave from the second laser pulse.

[0199] Example 10 is the method of Example 9, further comprising determining a characteristic of the plasma based on the shockwave from the first laser pulse and the shockwave from the second laser pulse.

[0200] Example 11 is the method of Example 3, further comprising generating sensor data based on the sensor signals; passing the sensor signals to an analysis model trained with a machine learning process; generating, with the analysis model, parameter adjustment data indicative of a parameter of the extreme ultraviolet light generation to be adjusted; and adjusting the parameter of the extreme ultraviolet light generation in response to the parameter adjustment data.

[0201] Example 12 is a method of operating an extreme ultraviolet lithography system, comprising: coupling a plurality of vibration sensors to a collector mirror of an extreme ultraviolet lithography system; generating extreme ultraviolet light by irradiating droplets with a laser within an extreme ultraviolet light generation chamber; generating, with the vibration sensors, sensor signals indicative of a distribution of debris from the droplets on the collector mirror; and adjusting one or more parameters of the extreme ultraviolet light generation based on the distribution of debris.

[0202] Example 13 is the method of Example 12, further comprising storing historical extreme ultraviolet light generation data corresponding to data from a plurality of previously performed extreme ultraviolet light generation processes; and training, with a machine learning process, an analysis model with the historical extreme ultraviolet light generation data to generate parameter adjustment data for adjusting parameters of an extreme ultraviolet light generation process.

[0203] Example 14 is the method of example 13, further comprising: generating sensor data from the sensor signals; providing the sensor data to the analytical model; generating parameter adjustment data with the analytical model; and adjusting extreme ultraviolet light generation parameters according to the parameter adjustment data.

[0204] Example 15 is the method of example 12, further comprising: performing a Fourier transform on the sensor signals.

[0205] Example 16 is an extreme ultraviolet lithography system, comprising: a droplet generator configured to output a droplet into an extreme ultraviolet light generation chamber; a laser configured to generate a plasma from the droplet by irradiating the droplet with a laser; a collector mirror configured to focus extreme ultraviolet light emitted from the plasma; and a plurality of first vibration sensors coupled to the collector mirror.

[0206] Example 17 is the extreme ultraviolet lithography system of example 16, further comprising: a housing cone coupled to the collector mirror, the housing cone and the collector mirror collectively defining the extreme ultraviolet light generation chamber; and a plurality of second vibration sensors coupled to the housing cone.

[0207] Example 18 is the system of example 16, wherein the plurality of first vibration sensors are arranged in a grid on a backside of the collector mirror.

[0208] Example 19 is the system of example 16, further comprising: a control system configured to receive sensor signals from the first vibration sensors, analyze the sensor signals, and adjust parameters of one or both of the droplet generator and the laser based on the sensor signals.

[0209] Example 20 is the system of example 19, wherein the control system comprises an analytical model trained with a machine learning process and configured to analyze the sensor signals.

Claims

1. A method for operating an extreme ultraviolet lithography system, comprising: Photolithography is performed by generating extreme ultraviolet light in an extreme ultraviolet light generation chamber, wherein generating extreme ultraviolet light includes generating extreme ultraviolet light by generating plasma from droplets in the extreme ultraviolet light generation chamber; Sensor signals are generated using multiple vibration sensors coupled to the extreme ultraviolet light generation chamber, wherein the multiple vibration sensors are coupled to the collector reflector of the extreme ultraviolet light generation chamber; Analyze the sensor signals; and The parameters for extreme ultraviolet light generation are adjusted based on the sensor signals. The method further includes: The distribution of droplet fragments on the collector mirror is detected based on the sensor signal; and The characteristics of the plasma are determined based on the distribution of the droplet fragments.

2. The method according to claim 1, wherein, Generating the plasma includes irradiating the droplet with a laser, wherein adjusting the parameters of the extreme ultraviolet light generation includes adjusting the parameters of the laser based on the sensor signal.

3. The method according to claim 1, wherein, Adjusting the parameters for generating extreme ultraviolet light includes adjusting the parameters for the droplets.

4. The method according to claim 1, wherein, Determining the characteristics of the plasma includes: a model for generating the plasma.

5. The method according to claim 1, wherein, Irradiating the droplet with a laser includes: Irradiate each droplet with the first laser pulse; and Each droplet is irradiated with a second laser pulse, wherein the sensor signal indicates the shock wave from the first laser pulse and the shock wave from the second laser pulse.

6. The method according to claim 5, further comprising: The characteristics of the plasma are determined based on the shock waves from the first laser pulse and the shock waves from the second laser pulse.

7. The method according to claim 1, further comprising: Sensor data is generated based on the sensor signals; The sensor signals are then transmitted to an analytical model trained using a machine learning process. The analytical model is used to generate parameter adjustment data, which indicates the parameters generated by the extreme ultraviolet light that are being adjusted. as well as The parameters for generating extreme ultraviolet light are adjusted in response to the parameter adjustment data.

8. A method for operating an extreme ultraviolet lithography system, comprising: Multiple vibration sensors are coupled to the collector mirror of the extreme ultraviolet lithography system; Extreme ultraviolet light is generated by irradiating droplets with a laser in an extreme ultraviolet light generation chamber; Using the vibration sensor, a sensor signal is generated indicating the distribution of fragments from the droplet on the collector mirror; as well as One or more parameters for extreme ultraviolet light generation are adjusted based on the distribution of the fragments.

9. The method according to claim 8, further comprising: Store historical extreme ultraviolet (EUV) generation data corresponding to data from multiple previously executed EUV generation processes; as well as The historical extreme ultraviolet (EUV) light generation data is used to train an analysis model through a machine learning process to generate parameter adjustment data, which is used to adjust the parameters of the EUV light generation process.

10. The method of claim 9, further comprising: Sensor data is generated from the sensor signals; The sensor data is provided to the analysis model; The analytical model is used to generate parameter adjustment data; as well as Adjust the extreme ultraviolet light generation parameters according to the parameter adjustment data.

11. The method of claim 8, further comprising: Perform a Fourier transform on the sensor signal.

12. An extreme ultraviolet lithography system, comprising: A droplet generator is configured to output droplets into an extreme ultraviolet light generation chamber; A laser is configured to generate plasma from the droplet by irradiating the droplet with a laser. The collector mirror is configured to focus extreme ultraviolet light emitted from the plasma; as well as Multiple first vibration sensors are coupled to the collector reflector. The extreme ultraviolet lithography system further includes: The control system is configured to receive sensor signals from the first vibration sensor, analyze the sensor signals, and adjust the parameters of one or both of the droplet generator and the laser based on the sensor signals. The control system is further configured to detect the distribution of droplet fragments on the collector mirror based on the sensor signals, and to determine the characteristics of the plasma based on the distribution of the droplet fragments.

13. The extreme ultraviolet lithography system according to claim 12, further comprising: The outer cone is coupled to the collector mirror, and the outer cone and the collector mirror together define the extreme ultraviolet light generation chamber; as well as Multiple second vibration sensors are coupled to the housing cone.

14. The system according to claim 12, wherein, The plurality of first vibration sensors are arranged in a grid on the back side of the collector reflector.

15. The system according to claim 12, wherein, The control system includes an analysis model that is trained using a machine learning process and configured to analyze the sensor signals.

Images