Plasma chamber with multi-stage rotationally adjusted crossflow

Abstract

To provide a plasma processing chamber with one or more sidewalls having a rotating, regulated crossflow.SOLUTION: A plasma processing chamber 100A has a support surface 114 within one or more sidewalls to hold a workpiece, a first gas injector 118A along the one or more sidewalls to inject a first gas flow 124A generally parallel to and across a surface of the workpiece in a first direction, a first pump port 120A along the one or more sidewalls generally opposite the first gas injector to exhaust the first gas flow, a second gas injector 118B along the one or more sidewalls to inject a second gas flow 124B generally parallel to and across a surface of the workpiece in a second direction, and a second pump port 120B along the one or more sidewalls generally opposite the second gas injector to exhaust the second gas flow.SELECTED DRAWING: Figure 1A

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Non-provisional Patent Application No. 17 / 023,186, filed September 16, 2020, the entire contents of which are incorporated herein by reference. [Technical field]

[0002] FIELD OF THE DISCLOSURE Embodiments of the present disclosure relate to the field of semiconductor processing, and more particularly to a plasma chamber including a rotating regulated cross-flow. [Background technology]

[0003] During plasma etching, deposition, or other processing, a workpiece such as a semiconductor wafer is inserted into a sealed plasma reactor chamber and gas is injected into the chamber above the wafer and then pumped out of the chamber. The plasma chamber often comprises either (1) a parallel plate capacitively coupled plasma (CCP) source, where one electrode has the workpiece on its surface facing the plasma and the other electrode has an array of gas injection holes (a showerhead) on its surface facing the plasma, or (2) an inductively coupled plasma (ICP) or microwave source that includes a radio-frequency (RF) window generally opposite and facing the workpiece and an array of gas injection holes in or near the window.

[0004] In the axisymmetric gas flow approach described above, pressure and concentration gradients cause processing differences between the center and edge of the workpiece. In addition, extraneous plasmas can be generated in the gas injection holes due to proximity to high density plasmas or disruption by high electric fields, resulting in non-uniformities that change over time. More specifically, gas injection holes are typically formed in a slab of material such as silicon or silicon carbide. Energetic ions impinging on the edge of the hole can deform or chip the hole over time. The deformed holes can result in high intensity plasmas that disrupt the slab, necessitating the showerhead to be changed after a certain time (e.g., 600 hours). In some applications, the showerhead cost can account for approximately $15 of the cost of a semiconductor wafer. Summary of the Invention

[0005] Embodiments disclosed herein include a plasma processing chamber with one or more sidewalls, an internal support surface within the one or more sidewalls for holding a workpiece, a first gas injector along the one or more sidewalls for injecting a first gas flow in a first direction generally parallel to and across a surface of the workpiece, A first pump port along one or more sidewalls generally opposite the first gas injector discharges a first gas flow, a second gas injector along one or more sidewalls injects a second gas flow in a second direction generally parallel to and across the surface of the workpiece, and a second pump port along one or more sidewalls generally opposite the second gas injector discharges the second gas flow.

[0006] Disclosed herein are embodiments of a method for implementing a rotational gas crossflow in a plasma processing chamber, and a non-transitory computer readable medium having stored thereon software instructions that, when executed by a processor, perform the following steps: and a non-transitory computer readable medium that causes a processor to rotate a gas cross-flow in a plasma processing chamber by performing the following steps: during a first phase, the steps include injecting a first gas flow in a first direction generally parallel to and across a surface of the device by a first gas injector at a first location along one or more sidewalls of the plasma processing chamber and exhausting the first gas flow from the plasma processing chamber by a first pump port at a second location along the one or more sidewalls generally opposite the first gas injector; during a second phase, the steps include injecting a second gas flow in a second direction generally parallel to and across a surface of the device by a second gas injector at a third location along the one or more sidewalls and exhausting the second gas flow from the plasma processing chamber by a second pump port at a fourth location along the one or more sidewalls generally opposite the second gas injector.

[0007] Embodiments disclosed herein include a plasma processing chamber with one or more sidewalls. A support within the one or more sidewalls for holding a workpiece. A first gas injector is at a first location along the one or more sidewalls and a first pump port is at a second location along the one or more sidewalls generally opposite the first gas injector. A second gas injector is at a third location along the one or more sidewalls and a second pump port is at a fourth location along the one or more sidewalls generally opposite the second gas injector. A multi-stage rotational cross-flow operation includes at least a first stage and a second stage. The first stage includes injecting a first gas flow generally parallel to and across a surface of the workpiece in a first direction by the first gas injector and exhausting the first gas flow by the first pump port. A second step includes injecting a second gas flow in a second direction generally parallel to and across the surface of the workpiece with a second gas injector and exhausting the second gas flow with a second pump port. [Brief description of the drawings]

[0008] [Figure 1A] FIG. 2 illustrates a top view of a plasma processing chamber having a multi-stage rotating cross-flow operation according to an embodiment. [Figure 1B] 1 is a cross-sectional view of a plasma processing chamber according to an embodiment. [Figure 1C] 1 is a cross-sectional view of a plasma processing chamber according to an embodiment. [Figure 2A] FIG. 1 is a schematic semi-transparent perspective view of a three-stage rotating crossflow plasma processing chamber according to an embodiment. [Figure 2B] FIG. 2 is a schematic top view of a three-stage rotating crossflow plasma processing chamber according to another embodiment. [Figure 2C] 1 illustrates a timing diagram of a three-stage rotating crossflow operation performed by a plasma processing chamber. [Figure 2D] 1 illustrates a perspective view of the top of the chamber lid, showing the gas delivery system described above, according to one embodiment. [Figure 2E] 1 shows an oblique cross-sectional view of a plasma chamber according to an embodiment. [Figure 2F] FIG. 2 is a perspective view of a vacuum chamber having a pump port formed therein according to an embodiment; [Figure 2G] FIG. 2 is a perspective view of a vacuum chamber having a pump port formed therein according to an embodiment; [Figure 2H] 1 is a cross-sectional view of a vacuum chamber having a pump port formed therein according to an embodiment; [Figure 2I] FIG. 1 is a semi-transparent perspective view of an exemplary inductively coupled plasma (ICP) chamber with a three-stage rotating crossflow according to one embodiment. [Figure 2J] FIG. 1 is a semi-transparent perspective view of an exemplary inductively coupled plasma (ICP) chamber with a three-stage rotating crossflow according to one embodiment. [Figure 2K] FIG. 1 is a semi-transparent perspective view of an exemplary inductively coupled plasma (ICP) chamber with a three-stage rotating crossflow according to one embodiment. [Figure 3A]FIG. 2 illustrates a top view of a plasma processing chamber having a four-stage rotating crossflow according to an embodiment. [Figure 3B] FIG. 1 illustrates a four-stage rotational crossflow operation according to one embodiment. [Figure 3C] FIG. 13 illustrates a four-stage rotating cross-flow operation, according to a further aspect of the disclosed embodiment, where gas is intentionally injected non-uniformly at the center and edges and exhausted at the opposite ports. [Figure 3D] FIG. 13 illustrates a four-stage rotating cross-flow operation, according to a further aspect of the disclosed embodiment, where gas is intentionally injected non-uniformly at the center and edges and exhausted at the opposite ports. [Figure 3E] FIG. 1 illustrates one stage of a multi-stage rotational cross-flow operation, according to one embodiment, where at least a portion of the gas flow is directed to the side of the workpiece, rather than 100% cross-flow across the workpiece. [Figure 3F] FIG. 1 illustrates a stage of a multi-stage cycle in which the gas flow is directed across the work piece using a narrower pump point, according to one embodiment. [Figure 4A] FIG. 1 illustrates a top view of rotating gas flow in a three-stage rotating crossflow plotted over time for every 60°, according to one embodiment. [Figure 4B] FIG. 1 illustrates a top view of rotating gas flow in a three-stage rotating crossflow plotted over time for every 60°, according to one embodiment. [Figure 4C] FIG. 1 illustrates a top view of rotating gas flow in a three-stage rotating crossflow plotted over time for every 60°, according to one embodiment. [Diagram 5] 1 illustrates a cross-sectional view of a portion of a wafer including stacked memory devices that may be processed by a plasma processing chamber including a rotating gas crossflow, according to one embodiment. [Figure 6] Referring now to FIG. 6, a block diagram of a processing tool that utilizes machine learning (ML) models is shown according to one embodiment. [Figure 7A]FIG. 1 is a flow diagram illustrating a process for generating an ML model, according to one embodiment. [Figure 7B] FIG. 1 is a flow diagram illustrating a process for generating an ML model, according to one embodiment. [Figure 8] FIG. 1 illustrates a flow diagram showing a process for developing a process recipe using an ML model, according to one embodiment. [Figure 9] FIG. 1 shows a flow diagram illustrating a process for baselining a processing tool according to one embodiment. [Figure 10] In one embodiment, a schematic diagram of a machine in the form of an exemplary computer system is shown in which a set of instructions may be executed to cause said machine to perform any one or more of the methods described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] The disclosed embodiments relate to a plasma chamber with a rotating regulated cross-flow. In the following description, numerous specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present disclosure. It will be apparent to one of ordinary skill in the art that the embodiments of the present disclosure may be practiced without these specific details. In other instances, well-known aspects, such as the fabrication of integrated circuits, have not been described in detail so as not to unnecessarily obscure the embodiments of the present disclosure. Furthermore, it should be understood that the various embodiments illustrated in the figures are for illustrative purposes only and are not necessarily drawn to scale.

[0010] Conventional plasma chambers (i.e., CCP or ICP) typically inject gas axially symmetrically onto the workpiece through gas injection holes located symmetrically above or around the workpiece. As mentioned above, the axially symmetric gas flow can result in pressure and concentration gradients, which can destroy the gas hole entrances and cause non-uniformities in the workpiece. In other words, as wear occurs in gas holes in regions that are in the high density, high electric field (|E|) plasma region, the hole geometry can change and as the plasma penetrates, the hole can change the local plasma properties in its vicinity. In addition, the geometry change can result in changes in local gas flow rates and velocities. This requires relatively frequent replacement of the showerhead, which increases the cost of the workpiece.

[0011] Correspondingly, embodiments disclosed herein relate to a plasma chamber (e.g., CCP or ICP) with a multi-stage rotating regulated gas cross-flow for etching, deposition, or other material processing. The plasma processing chamber includes two or more gas injectors and two or more pump ports along the sidewall. In a first stage, one of the gas injectors is forced to flow gas in one direction, generally parallel to and across the surface of the workpiece or device, and the gas is then exhausted through the pump port. In a second stage, the gas flow is rotated using the other gas injector, and a gas flow is forced in a different direction, generally parallel to and across the surface of the workpiece, and the gas is then exhausted through the other pump port. In other embodiments, a gas injection valve connected to the gas injector and / or a throttle valve connected to the pump port may be used to regulate the rotating gas flow.

[0012] A plasma processing chamber with rotating, regulated gas cross-flow eliminates the need for a showerhead (and gas injection holes) in the high density, high electric field (|E|) plasma region, thus avoiding a source of plasma non-uniformity. The disclosed embodiments prevent plasma formation in the gas injection holes due to breakdown from close proximity to the high density plasma or high electric field, resulting in non-uniformity and plasma characteristics that change over time. The disclosed embodiments avoid high center-to-edge pressure and concentration gradients that cause center-to-edge process differences. To minimize plasma non-uniformity, pressure distribution can be tailored across the plasma volume. Additionally, the disclosed embodiments eliminate low gas velocity regions (i.e., center of the workpiece) for uniform reactant material and by-product removal.

[0013] 1A-1C are diagrams illustrating an embodiment of a plasma processing chamber of a plasma reactor having a multi-stage rotating cross-flow operation. FIG 1A is a top view of a plasma processing chamber having a multi-stage rotating cross-flow operation according to one embodiment. FIG 1B and FIG 1C show cross-sectional views of different embodiments of the plasma processing chamber.

[0014] 1A and 1B, the plasma processing chamber 100A has one or more chamber sidewalls 112 and a support surface 114 for holding a workpiece 116 (e.g., a semiconductor wafer) for processing. The plasma processing chamber 100 may be used to perform a variety of processes, such as etching, deposition, surface treatment, or material modification, on the workpiece 116 by dispersing gases within the chamber. For example, the plasma processing chamber 100A may include, but is not limited to, a plasma etch chamber, a plasma enhanced chemical vapor deposition chamber, a physical vapor deposition chamber, an ion implantation chamber, an atomic layer deposition (ALD) chamber, an atomic layer etch (ALE) chamber, or other vacuum processing chamber suitable for fabricating various devices.

[0015] In the illustrated embodiment, one or more sidewalls 112 surround a processing region 110 in which a workpiece 116 (e.g., a wafer or substrate) is processed. In the illustrated example, the plasma processing chamber 100A is shown with an axisymmetric shape (e.g., cylindrical) that results in a single cylindrical sidewall 112. However, in other embodiments, the plasma processing chamber 100A may have any other shape that also results in a single sidewall 112, such as an ellipse, or may be shaped as a square or rectangle, in which case the plasma processing chamber 100A would have four sidewalls.

[0016] According to a disclosed embodiment, the plasma processing chamber 100 includes at least two gas injectors 118A and 118B (collectively referred to as gas injectors 118) and at least two pump ports 120A and 120B (collectively referred to as pump ports 120) located generally along the sidewall(s) 112. In one embodiment, the gas injectors are formed in openings through a liner in the sidewall 112. The plasma processing chamber 100A can be configured to use the gas injectors 118 and pump ports 120 to rotate the gas flow 124 laterally across the workpiece 116 to provide a multi-stage rotational cross-flow operation. In one embodiment, the multi-stage rotational cross-flow operation includes at least a two-stage cycle, and can further include a three-stage cycle, a four-stage cycle, ... where gases in each stage are injected from one side of the plasma processing chamber 100A and exhausted generally from the opposite side. As used herein, the phrase "located generally along the sidewall(s)" is intended to explain that any of the gas injectors 118 and / or pump ports 120 may be located within the sidewall, or may be horizontally in contact with or adjacent to the sidewall, or may be located within the perimeter region of the top of the chamber or within the perimeter region of the bottom of the chamber.

[0017] Rotating the gas flow laterally across the workpiece 116 may result in improved control of gas velocities and pressure gradients, resulting in better process uniformity across and between wafers.

[0018] 1B, the plasma processing chamber 100A further includes a chamber lid 104 on the sidewall 112. A support pedestal 108 may include a support surface 114 on which a workpiece 116 rests. In an embodiment, the support pedestal 108 and support surface 114 are fixed and not rotatable, and the workpiece 116 secured to the support surface 114 does not rotate during processing. In one embodiment, the workpiece 116 is electrostatically secured to the support surface 114. In another embodiment, the support surface 114 is axially movable for plasma gap adjustment or wafer transfer. A processing region 110 in the plasma processing chamber 100A is defined by the area between the chamber lid 104, the support pedestal 108 (and support surface 114), and the sidewall 112. The chamber floor 106 is below the sidewall 112, and the chamber floor 106 is below the processing region 110. The support pedestal 108 is below the chamber lid 104 and above the chamber floor 106 and is surrounded by sidewalls 112. In an embodiment, the chamber lid 104 and the support surface 114 may be separated by a distance of about 50 mm to 90 mm. In one embodiment, the plasma processing chamber 100A is a parallel plate capacitively coupled plasma (CCP) process chamber in which a first electrode 105 is above the workpiece 116. A second electrode is included within the support pedestal 108 at a location 113 below the support surface 114. In one embodiment, the first electrode 105 is connected to an RF source having a power in the range of 200 to 10,000 watts and a frequency in the range of 40 to 200 MHz. In one embodiment, the second electrode is connected to ground. A plasma is generated above the wafer and between the two electrodes. In one embodiment, the workpiece 116 is electrostatically clamped to the support surface 114 by one or more clamping electrodes within or below the support surface 114. In an embodiment, the workpiece 116 is coupled to a bias electrode (e.g., at a low RF frequency in the range of 0.1-20 MHz) for additional plasma control during processing. The power to the first electrode 105 can be pulsed during processing to pulse the generated plasma.

[0019] In one embodiment, the workpiece 116 may include any substrate commonly used in a semiconductor manufacturing environment. For example, the workpiece may include a semiconductor wafer. In one embodiment, the semiconductor material may include, but is not limited to, silicon or III-V semiconductor materials. In some embodiments, the semiconductor wafer may be a semiconductor-on-insulator (SOI) substrate. Typically, semiconductor wafers have standard dimensions (e.g., 200 mm, 300 mm, 450 mm, etc.). However, it should be understood that the workpiece 116 may have any dimensions. Embodiments may also include workpieces including non-semiconductor materials, such as glass or ceramic materials. In one embodiment, the workpiece 116 may include circuits or other structures fabricated using semiconductor processing equipment. In yet another embodiment, the workpiece 116 may include a reticle or other lithography mask object.

[0020] 1A and 1B show an example of a two-stage cycle rotating cross-flow operation. In a first stage, a gas injector 118A injects a first gas flow 124A generally parallel to and across a surface of the workpiece 116 in a first direction and has opposing pump ports 120A along one or more side walls 112 generally opposite the first gas injector 118A for exhausting the gas flow 124A. In a second stage, a gas injector 118B injects a second gas flow 124B generally parallel to and across a surface of the workpiece 116 in a second direction and has opposing pump ports 120B along one or more side walls 112 generally opposite the gas injector 118B for exhausting the gas flow 124B. In an embodiment, the direction of the second gas flow 124B is different from the direction of the first gas flow 124A. In one embodiment, substantially parallel means within a range of about 0° to 15°, and generally opposite means within a range of about 0° to 30°.

[0021] Thus, gas injector 118A and opposing pump port 120A form one gas injector-pump port pair, and gas injector 118B and opposing pump port 120B form a second gas injector-pump port pair. In one embodiment, gas injectors 118A and 118B can each include an array of individual gas injectors, as shown in FIG. 1A. In alternative embodiments, gas injectors 118A and 118B each include only a single vent gas injector. In some embodiments, gas injector 118A includes an array of individual gas injectors and gas injector 118B is a single vent gas injector, or vice versa.

[0022] As shown in FIG. 1A, each gas injector-pump port pair (i.e., a gas injector and an opposing pump port) is symmetrically positioned along a sidewall 112 of the plasma processing chamber 100A along a horizontal plane generally parallel to the orientation of the workpiece 116. Any number of gas injectors 118 and pump ports 120 may be provided. To ensure equal gas distribution, roughly speaking, one gas injector-pump port pair is offset from the position of the adjacent injector-pump port pair by an angle equal to a total of 360 degrees divided by the number of injector-pump port pairs. For example, if there are two injector-pump port pairs, the injector-pump port pairs are offset from each other by 180 degrees (360 degrees / 2). If there are three injector-pump port pairs, the injector-pump port pairs are offset by 120 degrees (FIGS. 2A and 2B), and so on. In some embodiments, the span of the gas injector is less than the span of the corresponding pump port, as shown. In other embodiments, the span of the gas injector is the same as the span of the corresponding pump port. In other embodiments, the span of the gas injector is greater than the span of the corresponding pump port. Gas can be injected through gas injector openings of various geometries, such as holes, slots, etc., and the various gas injectors can have the same or different geometries and sizes.

[0023] In some embodiments, the number of gas injectors 118 and pump ports 120 are equal, while in other embodiments the number of gas injectors 118 and pump ports 120 may differ. In some embodiments, one pump port is associated with a corresponding gas injector, as shown. In other embodiments, an array of pump ports is associated with a corresponding gas injector.

[0024] 1B, the gas injector 118 is located within an opening in the sidewall 112 in the processing region 110. For example, the opening can be located in a liner of the sidewall 112. In one embodiment, the opening in the sidewall 112 is vertically located between the chamber lid 104 and the substrate support pedestal 108. In the illustrated embodiment, the opening in the sidewall 112 is adjacent to the bottom of the chamber lid 104.

[0025] In one embodiment, the location of the pump port 120 may be vertically offset from the location of the gas injector 118 by a distance approximately equal to the distance between the bottom of the chamber lid 104 and the top of the support pedestal 108 along a vertical plane generally parallel to the orientation of the support pedestal 108. In this embodiment, the pump port 120 may be located in a cavity between the sidewall 112 and the support pedestal 108 and above the chamber floor 106. In other embodiments, the pump port 120 may be located in an additional opening in the sidewall 112 somewhere between the chamber lid 104 and the chamber floor 106. In other embodiments, gas may be injected from an outer perimeter region at the top of the chamber and / or exhausted from an outer perimeter region at the bottom of the chamber and still flow substantially parallel to the workpiece over the workpiece processing region.

[0026] As mentioned above, the plasma processing chamber 100A of the disclosed embodiments injects gases across the workpiece 116 generally parallel to the workpiece 116. This contrasts with the typically axisymmetric top-down gas flow injection from a "showerhead" electrode in a CCP source reactor, and the radial outward / downward gas injection from a nozzle array near a central axis in an ICP source or microwave source reactor. Additionally, instead of pump ports or pump plenums located axisymmetrically around the periphery of the workpiece, in embodiments gases are preferentially exhausted from the side of the workpiece generally opposite the injection side.

[0027] In an embodiment, the gas flow 124 of each cross-flow stage can be switched on and off to control the rotation of the gas flow. In another embodiment, instead of switching the gas flow 124 on and off, a modulation function can be applied to the flow rate of the gas flow 124 from the gas injector 118 and / or the outlet conductance (or pressure) produced by the pump port 120 to approach open / closed states or to ramp between states using a modulation function such as a sinusoidal function. As shown in FIG. 1B, the flow rate of one or both of the first gas flow 124A and the second gas flow 124B can be adjusted using one or more gas injection valves 122A and 122B (e.g., piezoelectric valves) connected to the gas injectors 118A and 118B, respectively. In an embodiment, the gas injection valves 122A and 122B are connected to one or more gas sources 126, which can inject a single gas or a mixture of different gases into the processing region 110 during each rotation stage. In one embodiment, a constant total gas flow can be applied by the gas injectors 118 to smoothly and sequentially inject gas flows across different sides of the workpiece 116 in a complete cycle, which can be repeated as necessary.

[0028] Additionally, in some embodiments, one or more pump ports 120 may be adjusted. For example, the conductance (pressure) of the pump ports may be adjusted using respective pressure control valves 127A and 127B of pump ports 120A and 120B. Also shown are pump ports 120A and 120B coupled to one or more pumps 132 to vent gas. In the illustrated example, pressure control valve 127A of pump port 120A is shown in a closed position and pressure control valve 127B is shown in an open position to vent first gas flow 124A. Pressure control valves 127A and 127B are smoothly operated between two states of conductance or pressure, which are repeated in a similar sequence as gas injectors 118A and 118B. In one embodiment, pressure control valves 127A and 127B include throttle valves.

[0029] The plasma chamber 100A can inject various types of process gases. Exemplary process gases can include: i) dielectric etch gases including one or more of C4F8, C4F6, C3F6, CH2F2, C3H2F4; ii) deposition gases including one or more of CH4, C2H2; iii) additional gases for co-flow etching or deposition including one or more of Ar, N2, O2, He, Kr, Xe, COS; iv) semiconductor material etch deposition gases including one or more of SiCl4, SiCH2Cl2; v) hydride-based deposition gases including one or more of BH3, AlH3, GaH3, NH3; vi) oxide material etch deposition gases including one or more of SiCl4, SiCH2Cl2, O2; and vii) annealing gases including one or more of NH3, N2, Ar.

[0030] In some embodiments, the plasma processing chamber 100A may further include sensors 131 and systems for sensitive and real-time measurement and monitoring of process chamber conditions including gas flow, velocity, pressure, temperature, etc. Particular embodiments may include capacitive wall sensors, on-chip or off-chip thermal sensors, pressure sensors, and / or integrated sensors (capacitive and thermal sensors) on substrates such as ceramic or glass or silicon or flexible substrates. In some embodiments, sensors may be distributed throughout the chamber to monitor chamber conditions at various locations, which may then be correlated with overall process performance such as etch rate, etch non-uniformity, particle generation, process drift, and pressure uniformity. In one embodiment, multiple pressure sensors or an array of pressure sensors may be distributed throughout the chamber to provide data regarding gas flow during processing (e.g., turnover rate, uniformity, velocity).

[0031] 1B further shows that the plasma processing chamber 100A can be connected to a controller 140, which itself can be connected to a user interface 142. In some embodiments, the controller can be connected to the gas injection valve 122, the pressure control valve 127, the gas source 126, the pump 132, and the sensor 131 to control the operation of the plasma processing chamber 100A. A user can set process parameters and monitor the operation of the plasma processing chamber 100A through the user interface 142 via the controller 140.

[0032] The multi-stage structure of the plasma processing chamber allows for a variety of configuration options. For example, FIG. 1C shows a cross-sectional view of a plasma processing chamber 100B in an embodiment including a top-down gas flow in addition to one or more pairs of gas injectors 118 and pump ports 120 that provide a side-to-side gas flow. In this embodiment, the chamber lid 104 can be configured with a showerhead plate 128 (for simplicity, the controller and UI in FIG. 1B are not shown). The showerhead plate 128 can have a central manifold 129 and one or more outer manifolds 130 for distributing gas into the processing region 110 along with the gas distributed by the gas injectors 118A and 118B. The showerhead plate 128 can be used to introduce additional gas into the chamber with a vertical velocity component, but the injection of gas from one side by the gas injector 118A and the exhaust on the other side of the workpiece 116 by the pump port 120B results in a horizontal component of the gas velocity across most of the workpiece 116. Similarly, the pump port 120 may be located in the sidewall 112 or on the top or bottom of the chamber, however, the pump port 120 is generally directly opposite the injection side. Thus, although there may be a component of gas velocity exiting in a vertical direction, the gas velocity is generally horizontal in the region above the workpiece 116 and parallel to the workpiece 116.

[0033] 2A-2C are diagrams illustrating a plasma processing chamber of a plasma reactor having a three-stage rotating cross-flow operation according to one embodiment. Fig. 2A is a schematic semi-transparent perspective view of a three-stage rotating cross-flow plasma processing chamber. Fig. 2B is a schematic top view of a three-stage rotating cross-flow plasma processing chamber according to another embodiment.

[0034] 2A, a plasma processing chamber 200A having three-stage rotational cross-flow operation is similar to the embodiment shown with respect to FIGS. 1A-1C in that the chamber 200 includes a sidewall 212 surrounding a workpiece 216. However, in addition to the two gas injectors 218A and 218B and the two opposing pump ports 220A and 220B, the plasma processing chamber 200 further includes a gas injector 218C and an opposing pump port 220C located on generally opposite sides of the sidewall 212 for exhausting the gas flow. The gas injector 218A and the opposing pump port 220A form one gas injector-pump port pair, the gas injector 218B and the opposing pump port 220B form a second gas injector-pump port pair, and the gas injector 218C and the opposing pump port 220C form a third gas injector-pump port pair. (Gas injectors 218A-218C are collectively referred to as gas injectors 218, and pump ports 220A-220C are collectively referred to as pump ports 220).

[0035] In this embodiment, the gas injectors 218 each include a single vent in the sidewall 212 as shown. In one embodiment, the gas injectors 218 are symmetrically positioned about the central axis of the plasma processing chamber 200, and the pump ports 220 are symmetrically positioned about the central axis of the plasma processing chamber 200, as shown. In a three-stage rotating crossflow embodiment including three injector-pump port pairs, the injector-pump port pairs are offset from one another by 120° (360° / 3). More specifically, the gas injectors 218 are positioned approximately 120° apart from one another, and the pump ports 220 are positioned 120° apart from one another. The pump ports 220 laterally distributed among the spaced apart gas injectors 218 are similarly vertically offset from the gas injectors 218.

[0036] 2B shows a top view of plasma processing chamber 200B including an array of individual gas injectors, referred to as gas injector array 218D, distributed around the circumference of sidewall 212. Also shown are three gas injector valves 122A-122C and three pressure control valves 127A-127C (one for each pump port 120) (see FIG. 1B). A set of smaller gas injectors (such as four injectors as shown) in gas injector array 218 can be adjusted by a later one of gas injector valves 122A-122C to generate gas flow in various directions across workpiece 216. The gas flow is then exhausted by one of the pump ports controlled by a corresponding one of pressure control valves 127A-127C generally opposite the gas injector valve 122A-122C that adjusts it. In this case, in one embodiment, the span of the gas injectors is greater than the span of the corresponding pump port, resulting in a slightly convergent flow (eg, flow 299) to a relatively narrow pump port.

[0037] 2C shows a timing diagram of the three-stage rotary crossflow operation implemented by plasma processing chamber 200B in more detail. The timing diagram assumes that there are three gas injection valves 122 (GV1, GV2, GV3) and three pressure control valves 127 (PV1, PV2, PV3). The X-axis represents time, and the Y-axis represents i) the gas valve opening at the bottom, the pump port closing at the middle, and the chamber pressure measured by the Baratron (pressure gauge) at the top.

[0038] A controller can be connected to plasma processing chamber 200 and configured to control gas inlet valves 122A-122C and pressure control valves 127A-127C. The controller initiates a first phase by fully opening GV1 to 100% and partially opening GV2 and GV3, for example, to about 2-5%. During the first phase, PV1 is opened, PV2 and PV3 are closed, and the chamber pressure is between 1 mT and 500 mT.

[0039] Near the transition between the first and second stages, GV1 begins to close and GV2 opens fully to 100%, rotating the direction of gas flow and initiating the second stage. GV1 and GV3 are partially open, about 2-5%. During the second stage, the controller opens PV2 and keeps PV1 and PV3 closed. The chamber pressure can remain between 1 mT and 500 mT in some embodiments, or between 10 mT and 200 mT in other embodiments.

[0040] Near the transition between the second and third stages, GV2 is ramped down and GV3 is opened to 100% to rotate the gas flow direction and begin the third stage. GV1 and GV2 are partially open, approximately 2-5%. In the third stage, the controller opens PV3 and keeps PV1 and PV2 closed. This completes a three-stage cycle, which can be repeated as necessary. As shown, a relatively constant chamber pressure is maintained during the three gas flow stages. In one embodiment, sequential opening and closing of GV1, GV2, and GV3 can effectively create a rotational gas flow that can mimic the rotation of the wafer. In one embodiment, one full rotation of the gas flow is performed at a rate within a range of approximately 100 ms to 10 sec.

[0041] There may be several different variations between the gas flow stages and cycles. In other words, each parameter controlling the operation of the plasma processing chamber may vary between stages and cycles. For example, the time to complete a full cycle may be the same or different between different cycles. The time to complete a stage may be the same or different within a cycle, and may be the same or different between different cycles. The direction of gas flow rotation (e.g., clockwise, counterclockwise) may be the same or different within a cycle stage, may be discontinuous, and may be the same or different between cycles. The gas flow speed may be the same or different within a cycle stage, or may be the same or different between cycles. The opening of the gas valves and the time the gas valves are open may be the same or different within a cycle stage, or may be the same or different between cycles. The opening of the pressure control valves and the time the pressure control valves are open may be the same or different within a cycle stage, or may be the same or different between cycles. For example, in one embodiment, the rotation is performed at one rate for a first portion of the process and then decelerated to a second rate for a second portion of the process. In one embodiment, the rotation is performed at one rate for a first portion of the process and then accelerated to a second rate for a second portion of the process. In one embodiment, the rotation is fast for a first portion of a rotation cycle and slow for a second portion of the rotation. In one embodiment, the rotation is slow for a first portion of a rotation cycle and accelerated for a second portion of the rotation. The rotation speed can be varied within a cycle or from cycle to cycle to compensate for process non-uniformities. In other embodiments, the direction is changed between clockwise and counterclockwise within a cycle, between cycles, or between sets of cycles. Similarly, in embodiments, the gas flow rates between the first, second, and third stages can be changed within a cycle, between cycles, or between sets of cycles.

[0042] FIG. 2D is a perspective view of the top of the chamber lid 104, illustrating the gas delivery system described above. In one embodiment, the gas delivery system 225 includes an array of gas inlet valves 122, where each of the gas inlet valves 122 is located above the chamber lid 104 and symmetrically arranged around the circumference of the chamber lid 104. In the illustrated embodiment, the gas delivery system 225 includes six gas inlet valves 122, although the specific number may vary, such as two or more. The upper side of each gas inlet valve 122 can be connected to a gas line assembly 250 arranged in a spoke and hub formation, where the hub is connected to the gas source 126 shown in FIG. 1B and FIG. 1C. The lower side of the gas inlet valve 122 can be connected to a respective set of circulating gas lines 252. Each set of circulating gas lines 252 can be connected to one or more gas injectors 118. In the particular embodiment shown, there are six sets of circulating gas lines 252 with four inlets each connected to a gas injector 118 for a total of 24 inlets.

[0043] In an embodiment, the gas inlet valve 122 may include an analog variable conductance fast gas valve that allows for fast response without excessive pressure spikes that could cause gas ignition or arcing or make the RF match control difficult to follow. Examples of gas inlet valves include commercially available Swagelok eDE valves and Fujikin piezo valves. The Swagelok eDE valve has a 15-20 msec open / close time, is well suited for air / vacuum sealing, and can have a life of 40M cycles. The Fujikin piezo valve has a proportional flow rate, a 10 msec open / close time, and can have a life of 40M cycles or more depending on usage. Both can provide a gas flow of up to 2.5 slm at an upstream pressure of 400T.

[0044] 2E shows an oblique cross-sectional view of the plasma chamber, illustrating the connection of the circulating gas line 252 to the gas injector 118. Also, in one embodiment, the sidewall 112 can include an outer sidewall 112A and an inner sidewall 112B (or liner), with the gas injector formed in the space between the outer sidewall 112A and the inner sidewall 112B, and gas is injected from the circulating gas line 252 through an opening in the inner sidewall 112B.

[0045] 2F-2G are perspective views of a vacuum chamber having a pump port 120 formed therein, and FIG. 2H is a cross-sectional view of a vacuum chamber having a pump port 120 formed therein. In an embodiment, the vacuum chamber 275 is under a dynamic vacuum controlled by a pump 132 (FIGS. 1B and 1C). In one embodiment, the vacuum pressure can range from 1 mT to 500 mT. In one embodiment, the chamber floor 106 includes an upper chamber floor 106A and a lower chamber floor 106B, and the pump port 120 is formed in a cavity in the vacuum chamber 275 between the upper chamber floor 106A and the lower chamber floor 106B. The pump ports 120 are also shown symmetrically disposed about the support pedestal 108.

[0046] An actuator 277 is connected to the pressure control valve 127 to control each of the pump ports 120. FIG. 2H shows that one of the actuators 277 raises and lowers the corresponding pressure control valve 229 in the cavity of each pump port 120, thereby opening and closing the pump ports 120. FIG. 2F shows that in one embodiment, the pressure control valve 229 can include a single integral body for sealing the associated port, while FIG. 2G shows that in another embodiment, the pressure control valve 229 is divided into one or more adjacent sections (two in this case), each controlled by a corresponding actuator 277. In one embodiment, referring to FIG. 2H, the left pressure control valve 127 is lowered (open) and the right pressure control valve 127 is raised (closed). In FIGS. 2F and 2G, all pressure control valves are shown in the closed position.

[0047] 2I-2K show an exemplary inductively coupled plasma (ICP) chamber with a three-stage rotating crossflow according to one embodiment. As shown in FIG. 2I, the ICP chamber 280 includes an electrode 282 in the form of a planar multi-spiral coil adjacent to the chamber lid (not shown). The electrode 282 includes an RF-driven post 286 and may include three grounded ends 284 along its maximum radius. FIG. 2J shows gas injectors 288 symmetrically positioned around the periphery of the top of the chamber. In one embodiment, the gas injectors 288 may have 60° wide inlets with 60° wide spaces between them. FIG. 2K shows pump ports 290 symmetrically positioned around the periphery of the bottom of the chamber, each pump port 290 located directly 180° opposite one of the gas injectors 288.

[0048] 3A-3F are top views of a plasma processing chamber having four-stage rotational crossflow operation according to one embodiment. Figure 3A illustrates a plasma processing chamber 300 that may be rectangular in shape with four sidewalls 312. Each of the four sidewalls 312 includes one of four gas injector arrays 318A-318D and one of four opposing pump ports 320A-320D.

[0049] FIG. 3B illustrates a four-stage rotational cross-flow operation. Through a four-stage cycle, gas is pumped into each of the four side walls 312 and exhausted from the opposite side. The conductance of each pump port 320A-320D can be adjusted using a high-speed individual throttle valve. Stage 1 shows a first gas flow from left to right. Stage 2 shows a clockwise rotation for a second gas flow from top to bottom. Stage 3 shows a clockwise rotation for a third gas flow from right to left. And stage 4 shows a clockwise rotation for a fourth gas flow from bottom to top. In one embodiment, each stage can last about 0.5 seconds to 2 seconds depending on the application.

[0050] 3C and 3D illustrate a four-phase rotating cross-flow operation, according to further aspects of the disclosed embodiment, where gas is intentionally injected non-uniformly at the center and edge, and exhausted at the opposite ports. In this embodiment, the individual gas injectors in each of the gas injector arrays 318A-318D can be switched on and off by the gas injector valves 122 or can have flow rates regulated by the gas injector valves 122. FIG. 3C shows an example of four stages of center-to-edge gas flow, where in each stage, the gas flow injected from the center gas injector in each of the gas injector arrays 318A-318D has a higher flow rate relative to the edge gas injectors in the gas injector arrays 318A-318D. 3D shows four example stages of edge-to-center gas flow, where the gas flows injected from the edge gas injectors in each of the individual gas injectors in the gas injector arrays 318A-318D have a higher flow rate relative to the center gas injectors in the gas injector arrays 318A-318D. Such non-uniform center and edge gas injection in the disclosed embodiments can be purposefully varied and controlled over time to control process uniformity of the workpiece. In one embodiment, the relative center and edge flows of one or more gas injectors are changed during a cycle, between cycles, or between sets of cycles.

[0051] FIG. 3E shows one stage of a multi-stage (e.g., four-stage) rotating cross-flow operation, where at least a portion of the gas flow is directed to the side of the workpiece, rather than 100% cross-flow across the workpiece. In this extreme case, the opposing pump port is closed, but the side pump port is open, minimizing gas flow and velocity across the center of the workpiece. This process can be used to control uniformity. In an embodiment, diverted gas flows as shown in FIG. 3E are used for the entire process, or only for a portion of a cycle, or for one or a smaller set of cycles in a process scheme. In an embodiment, the diverted gas flows are rotated around the chamber for one or more cycles.

[0052] FIG. 3F shows a stage of a multi-stage cycle in which gas flow is directed across the workpiece using narrower pump points. As in FIG. 3C, the gas flow from the central gas injectors in each gas injector array has a higher flow rate relative to the end gas injectors in the gas injector array, with the opposing pump ports open and the other pump ports closed. In a further embodiment, narrower pump ports than those in the above embodiment force gas flow across the central region of the workpiece. In this embodiment, for a typical 300 mm wafer chamber, the smaller pump ports can have dimensions of 3.5 inches wide by (1 / 2) by (14 inches) long centerline radial arc length, and the larger single pump port can have dimensions of 3.5 inches wide by 14 inches long centerline radial arc length. In general, the pump port should be dimensioned or sized appropriately for the flow conductance of the process application, with a port width opening narrow enough to promote uniform "cross flow" over the wafer from the gas injection side to the pump port side of the chamber.

[0053] 4A-4C are top views of the rotating gas flow in a three-stage rotating crossflow plotted over time at 60° intervals. The arrows represent vectors indicating velocity magnitude and the contours represent pressure gradients. Snapshots of gas flow at 0°, 60°, 120°, 180°, 240°, and 300° are shown. The graph in Figure 4C shows that the pressure at the gas injector and pump ports over time is relatively stable across the three stages.

[0054] The exemplary operations shown in Figures 4A-4C can be used individually or, more likely, in combination over repeated cycles to maximize process uniformity. This tuning capability using gas injection at the peripheral boundary of the high-density plasma region and / or gas exhaust out of the high-density plasma region as control inputs without introducing geometric discontinuities (gas injection holes) allows the formation of a uniform plasma with minimal drift or change over time due to etching, wear, or coating on the exposed surfaces facing the plasma, i.e., the electrode / showerhead, including gas holes and gas nozzles. The use of a rotating regulated cross-flow can enable process uniformity control from the peripheral boundary of the chamber.

[0055] Reactive ion etching As an exemplary application, the plasma processing chamber can be used to perform precision reactive ion etching during semiconductor manufacturing.

[0056] FIG. 5 illustrates a cross-sectional view of a portion of a wafer including a stacked memory device that may be processed by a plasma processing chamber with rotating gas cross-flow, according to an embodiment. In one embodiment, an intermediate structure of a stacked memory device being fabricated is shown. In one embodiment, the intermediate structure 400 includes a 3D-NAND structure, including a substrate 402, an alternating layer stack 404 on the substrate 402, an inter layer dielectric (ILD) layer 406 on the alternating layer stack 404, and a mask layer 408 on the ILD layer 406. The alternating layer stack 404 may include interposed insulating layers 404A and 404B (e.g., silicon nitride, silicon oxide, etc.). Examples of the ILD layer 406 may include spin-on-glass (SOG), SOC, and SiON.

[0057] The mask layer 408 may define the pattern of an integrated circuit with the pattern used to guide the deposition of material or removal of material from the wafer in subsequent patterning steps. In this example, reactive ion etching is performed by a plasma processing chamber to remove material between some of the openings in the mask layer 408 to form openings 410 through the ILD layer 406 and the alternating layer stack 404 to the substrate 402, where the intersections of the openings 410 and the metal layer 404a may ultimately form memory cells. The gas flow injected by the plasma processing chamber (as described above) may be customized to control the uniformity of the etch depth and the uniformity of the aspect ratio (depth to width) of the openings 410. In one embodiment, one or more openings 410 may be etched to have a first aspect ratio through the ILD layer 406 and a second aspect ratio through the alternating layer stack 404. In an embodiment, one or more openings 410 may have a variable aspect ratio, referred to as a bowing, through the alternating layer stack 404 as shown. In one embodiment, the openings 410 may be etched to have a high aspect ratio of greater than 8: 1, 9: 1, or 10: 1. In an embodiment, one or more of the openings 410 may also have a variable etch depth.

[0058] In embodiments, 3D-NAND ion etch applications may include pillar etch, slit etch, peri-contact etch, stepped contact etch, and cell contact-1 etch as described above. In embodiments, aspect ratio, etch depth, and bowing characteristics may be parameters monitored by the machine learning model, as described below.

[0059] USING MACHINE LEARNING (ML) MODELS TO CONTROL A PLASMA PROCESSING CHAMBER WITH MULTI-STAGE ROTATING CROSSFLOW Configuring the above-mentioned plasma processing chamber to provide a desired outcome on a workpiece (e.g., wafer) requires a process recipe that includes a complex combination of several different, individually controllable process parameters (i.e., knobs). Examples include total gas flow mixture, gas pressure (mTorr), gas flow ramp open time (msec), gas flow time (msec), gas flow ramp close time (msec), etc.

[0060] To develop a process recipe for high volume manufacturing (HVM), process engineers rely on their experience and expertise to identify a baseline recipe that can provide a rough approximation of the desired outcome on the wafer. A design of experiment (DoE) is then created around the baseline recipe that relies on processing a set of wafers (or coupons) to see how the knobs work together. Process engineers can interpret the DoE results to further refine the baseline recipe, and additional DoE can be performed to reach the desired outcome on the wafer. Such an iterative process is time- and resource-intensive.

[0061] In addition, once a final process recipe has been developed, chamber drift during multiple iterations of processing for different wafers can result in changes in on-wafer outcomes. Chamber drift can be the result of erosion of consumable parts of the chamber, degradation of components (sensors, lamps, etc.), deposition of by-product films on surfaces, etc. Correspondingly, additional tuning is required even after an extensive recipe development process.

[0062] Thus, recipe development and chamber baselining are time- and resource-intensive. In particular, the process space available for tuning and optimizing a given process is so large that it is virtually impossible to empirically explore the entire process space within a reasonable time frame. Furthermore, due to the interactions between process parameters and their impact on process performance, it is extremely difficult to predict the combined effect of simultaneously changing multiple process parameters by manually scanning one process parameter at a time.

[0063] A second aspect of the disclosed embodiments includes a semiconductor manufacturing tool that utilizes one or more machine learning (ML) models to control a plasma processing chamber having a multi-stage rotating cross-flow. The ML models can also be used to develop process recipes and / or process devices and workpieces. The ML models can link input process parameters to device outputs.

[0064] In one embodiment, a method for controlling a process includes querying an ML model to control the timing of rotation of gas flows. In one embodiment, a method for formulating a semiconductor manufacturing process recipe includes selecting one or more device outcomes and querying an ML model to obtain a process recipe recommendation suitable for obtaining the device outcomes when processed by a plasma processing chamber having a multi-stage rotating cross-flow. This may be referred to as feed forward process tuning. In one embodiment, the method may further include performing a design of experiments (DoE) on a set of wafers to validate the process recipe recommended by the ML model. The DoE measurements may be used to make feedback process adjustments to modify the process recipe for future wafers.

[0065] Additionally, the ML model can be updated while a wafer is being processed in the chamber as on-tool performance becomes available, and then the process recommendations can be updated or the recipe can be actively modified, which may be referred to as "on the fly" or real-time process adjustments.

[0066] A recipe change can include modifying the recipe within a step, for example increasing the rotation frequency of the gas flow when etching the top of the wafer and decreasing the rotation frequency as the gas reaches the bottom, or vice versa. Another example is an updated machine learning model modifying input parameters during one rotation, for example slightly changing the etch depth at the beginning and end of a gas flow rotation when processing the stacked memory device of FIG. 5. The updated ML model can accurately track chamber drifts, allowing modifications to be made to the process recipe without extensive DoE of physical wafers or relying solely on the experience and knowledge of process engineers.

[0067] Correspondingly, embodiments disclosed herein leverage the use of ML models to query the entire process space without the need to process physical wafers in an extensive Design of Experiments (DoE), thus significantly reducing recipe development time and resources.

[0068] An ML model can be a model of a process space generated from a combination of statistical and physical models. As used herein, "process space" can refer to a multi-dimensional process space that maps process parameters to one or more device outcomes on a wafer. Process parameters, sometimes called knobs, are variables that can be controlled to control the process. For example, the knobs or process parameters may include, but are not limited to, any combination of temperature, RF source power, bias power, gas pressure (mTorr), gas flow ramp open time (msec), gas flow time (msec), gas flow ramp close time (msec), gas flow rates in various gas injectors, gas composition in various injectors, gas flow rates to various injectors, gas flow rotation frequency, gas flow composition frequency, gas flow rate / velocity (pressure gradient), gas flow direction, gas rotation phase, electron / plasma density, plasma density gradient, electron temperature, ion current density, plasma potential, sheath electric field potential, sheath electric field tilt angle, sheath electric field z component, mass fraction, flux, and ion current density relative to the workpiece.

[0069] A device outcome may refer to a measurable characteristic of a feature on a wafer after processing. For example, selected device outcomes may include any combination of feature profile, layer thickness, thickness uniformity, layer material composition, composition uniformity, porosity, film stress, process uniformity between chambers in a facility (e.g., chamber matching), wafer-to-wafer uniformity, and uniformity between different wafer lots. During an etch process, selected device outcomes may further include any combination of etch rate, center-to-edge etch or uniformity, circumferential uniformity of etch rate, etch feature uniformity (commonly represented by top-to-bottom critical dimension (CD)), tilt, bow, and mask remaining. That is, device outcomes are not limited to outcomes on a single wafer. Each point in the process space may represent a set of process parameter values ​​and the resulting device outcome (or outcomes) produced by that set of process parameters.

[0070] In one embodiment, a statistical model of the ML model can be constructed using DoE of actual wafers to input a portion of the process space. An algorithm can then be used to extrapolate the remainder of the process space. The physical model is based on real-world physical and chemical interactions that occur in the process chamber. Simulations of the physical and chemical interactions in the process chamber over a range of different process parameters can be used to generate the physical model. In one embodiment, the physical model is integrated with the statistical model to provide the ML model. For example, the physical model can be used to fill gaps in the statistical model and / or to validate extrapolated data points.

[0071] 6, a block diagram of a processing tool 600 utilizing a machine learning ML model is shown, according to one embodiment. The processing tool 600 includes tool hardware 640 corresponding to the plasma processing chamber described above, a machine learning model server 620, a front-end server 660, and a control server 650.

[0072] In one embodiment, the ML model server 620 may include a statistical model 625 and a physical model 627. The statistical model 625 and the physical model 627 may be communicatively connected to a database 630 for storing input data (e.g., sensor data, model data, measurement data, etc.) used to build and / or update the statistical model 625 and the physical model 627.

[0073] In one embodiment, the statistical model 625 is generated from the physical DoE and uses interpolation to provide an extended process space model. The processed physical wafer can provide a mapping between process parameters and specific device outcomes. The physical DoE can also be used to identify interactions between various process parameters. After the physical wafer data (metrology data, sensor data, process parameter data, etc.) is provided, the interpolation is used to fill gaps in the process space. In one embodiment, data such as metrology data can be obtained using an external tool communicatively connected to the ML model server 620 by a data link (e.g., a wired or wireless data link). The interpolation can be performed using any suitable algorithm. The algorithm can include, but is not limited to, neural networks, deep learning, or any other known techniques used for regression analysis (e.g., linear regression, partial least squares regression, Gaussian, polynomial, convolutional neural networks for regression, regression trees, and others).

[0074] In one embodiment, the statistical model 625 may be provided as a module that is sold or licensed for use with a processing tool; that is, the physical DoE for the statistical model 625 may be performed by the processing tool manufacturer. In another embodiment, the statistical model 625 may be generated by performing a physical DoE in situ. In yet another embodiment, a generic statistical model 625 may be provided by the tool manufacturer, and subsequent physical DoE may be performed in situ to provide a calibration of the statistical model 625 to more closely model the particular processing tool being investigated.

[0075] In one embodiment, the physics model 627 may be generated using real-world physical and chemical relationships. For example, physical and chemical equations for various interactions within a process chamber may be used to construct the physics model. The physics model 627 may also utilize the chamber geometry or other chamber settings to improve the accuracy of the physics model 627. The physics model 627 may be the result of a simulation of the physical and chemical interactions within a process tool across a number of different process parameters. The physics model 627 may be a module that is sold or licensed for use with a process tool.

[0076] In one embodiment, the physical model 627 and the statistical model 625 may be able to reference each other (as indicated by the arrows). The cross-referencing between the two models 627 and 625 allows for validation of each model and for filling in gaps in the individual models. In one embodiment, the physical model 627 and the statistical model 625 may be combined to provide a more robust ML model.

[0077] As shown, the ML model server 620 may be integrated with the processing tool 600. For example, the ML model server 620 may be communicatively connected to the front-end server 660 by a network connection, as indicated by the arrow. In other embodiments, the ML model server 620 may be external to the processing tool 600. For example, the ML model server 620 may be communicatively connected to the processing tool 600 via an external network, or the like.

[0078] In one embodiment, the front-end server 660 may include a user interface 665 for the ML model server 620. The user interface 665 provides an interface for a process engineer to utilize the ML modeling to perform various operations, such as recipe development or chamber baselining, as described in more detail below. In one embodiment, the user interface 665 may correspond to the user interface 142 of FIG. 1B.

[0079] The control server 650 may include a smart monitoring and control block 655. The smart monitoring and control block 655 may include modules for providing diagnostics and other monitoring of the processing tool 600. The modules include, but are not limited to, health checks, sensor drift, fault recovery, and leak detection. The smart monitoring and control block 655 may receive as input data from various sensors implemented in the tool hardware 640. The sensors may include standard sensors 647 that are typically present in the semiconductor manufacturing tool 600 to enable the operation of the tool 600. The sensors may also include modeling sensors 645 that are added to the tool 600. The modeling sensors 645 provide additional information necessary to build highly detailed ML models. For example, the modeling sensors may include virtual sensors and / or witness sensors. The virtual sensors may utilize data obtained from two or more physical sensors to perform interpolation and / or extrapolation to provide additional sensor data not obtainable from the physical sensors. In a particular example, a virtual sensor can utilize an upstream pressure sensor and a downstream pressure sensor to calculate the flow rate through a portion of a process tool, such as a gas cartridge. In general, the modeling sensors can include any type of sensor, such as, but not limited to, a pressure sensor, a temperature sensor, a gas concentration sensor, etc. In one embodiment, the smart monitoring and control block 655 can provide data used by the ML model server 620. In other embodiments, output data from the various modeling sensors 645 can be provided directly to the ML model server 620. In one embodiment, the control server 650 can correspond to the controller 140 of FIG. 1B.

[0080] 7A, a flow diagram illustrating a process for generating an ML model according to one embodiment is shown. In one embodiment, input from modeling DoE 715 is input to statistical model engine 724. Modeling DoE 715 may include processing of several physical wafers. The DoE 715 may include various data sources that are provided to statistical model engine 724. For example, metrology data 716 acquired during or after processing of the wafer may be provided to statistical model engine 724. In addition, sensor data 217 from sensors in the processing tool may be provided to statistical model engine 724. Process parameter data 718 (i.e., values ​​of various process parameters during processing of the wafer) may also be provided to statistical model engine 724.

[0081] In one embodiment, the statistical model engine 724 may be implemented as hardware and / or software suitable for analyzing various data sources and outputting the statistical model 725. The statistical model engine 724 may utilize machine learning based on neural networks or any other known technique used for regression analysis (e.g., linear regression, partial least squares regression, Gaussian, polynomial, convolutional neural networks for regression, regression trees, and others) to interpolate a larger process space than can be obtained from the physical DoE data alone.

[0082] In one embodiment, a physics model engine 726 is used to generate the physics model 727. In one embodiment, the physics model engine 726 may be implemented as hardware and / or software. The physics model engine 726 receives as input the chamber settings and real-world physical and chemical reaction equations. The physics model engine 726 may perform simulations of the physical and chemical interactions within the processing tool across a number of different process parameters to build the physics model 727. In this manner, changes to the processing parameters that modify the physical and / or chemical reactions within the processing tool may be mapped to expected device outcomes.

[0083] In one embodiment, the statistical model 725 and the physical model 727 are used as inputs for the generation of the ML model 728. For example, the statistical model 725 and the physical model 727 can be inputs for the ML model engine 729. The ML model engine 729 processes the physical model 727 and the statistical model 725 to output the ML model 728. In some embodiments, the physical model 727 can be used to derive some physical measurements that cannot be measured, and the output of the physical model 727 can be considered as an additional input to the statistical model. In such a situation, the ML model engine 729 adds information from the physical model 727 to the statistical model 725 to provide the ML model 728. Thus, the ML model 728 allows for the validation of individual points in the process space using the two models 725 and 727, providing a more complete process space that can be individually tailored to a given processing tool. However, in some embodiments, the physical model 727 and the statistical model 725 can be standalone models according to the output. That is, in some embodiments, the statistical model 725 and the physical model 727 may not be integrated into the ML model.

[0084] In one embodiment, the ML model may also be considered as another instance of the statistical model 725. For example, in FIG. 7B, the physical model 727 output by the physical model engine 726 may be used as an input to the statistical model engine 724. Thus, the statistical model engine 724 has an additional input to generate the statistical model 725 that includes information from the physical model 727. Notably, the statistical model engine 724 may already include data from the physical model 727, and the use of the ML model engine to generate the ML model may not be necessary in all embodiments.

[0085] 8, a flow diagram illustrates a process 870 for developing a process recipe using an ML model according to one embodiment. A target process recipe is a process recipe having a set of process parameters that results in a desired device outcome on a wafer. In one embodiment, the process 870 may begin with operation 871 which includes determining a desired device outcome. In one embodiment, the device outcome may be a wafer device dimension, material composition, etc. For example, the device outcome may include a layer thickness, thickness uniformity across a wafer, material composition of a layer, or material composition uniformity of a stacked memory device as shown in FIG. 5.

[0086] In one embodiment, process 870 may continue with operation 872, which includes querying the ML model to select a set of process parameters. In one embodiment, the ML model may be a model of the process space generated from a combination of a statistical model and a physical model. The statistical model may be generated using DoE of actual wafers as described above. The physical model may be based on real world physical and chemical reaction equations. For example, the physical model may be generated from a simulation of physical and chemical interactions in a processing tool across multiple different processing parameters. In one embodiment, the ML model may cover the entire process space available to the processing tool.

[0087] The ML model allows for the identification of stable process recipes without relying solely on the experience and knowledge of the process engineer. Instead, a baseline recipe can be selected from the ML model's process space that is expected to produce device outcomes that closely match the target device outcomes.

[0088] In one embodiment, process 870 may continue with operation 873 with running a small DoE to validate the model recommendations. Due to the accuracy of the ML model, a small DoE (e.g., 20 wafers or less) may be all that is needed to validate the model recommendations. In one embodiment, the DoE may be designed by a process engineer. In another embodiment, the DoE may be designed using the ML model.

[0089] In one embodiment, process 870 may continue with operation 874, which includes measuring results of the DoE wafer with one or more metrology tools. The metrology data may be used to verify that targeted device outcomes are achieved on the wafer.

[0090] In one embodiment, process 870 may continue with operation 875, which includes determining whether the desired device outcome has been achieved. If the desired device outcome has been achieved, the process proceeds along operation 876, where the process is complete. If the desired device outcome has not been achieved, the process may return or feed back to operation 872. In one embodiment, data from the small DoE may be fed back to the ML model to update the ML model. For example, if the process iteratively returns to operation 872, the DoE performed in operation 873 may be designed based on knowledge of what is missing in the ML model (e.g., for a particular process or plasma chamber) and based on additional knowledge learned from the DoE performed in the previous cycle. The updated ML model may then be queried to provide a second baseline recipe. In this way, even if the first iteration is not successful, the process can still quickly converge on a suitable recipe without the need for extensive DoE and wasting resources.

[0091] 9, a flow diagram is shown illustrating a process 980 for baselining a processing tool according to one embodiment. In one embodiment, the baseline process can be useful to account for chamber drift during processing of wafers in the processing tool. In one embodiment, the baseline process can be performed at any desired frequency. For example, process 980 can be performed every lot, every planned maintenance (PM) event, or when a processed wafer has a device outcome outside of a specified range.

[0092] In one embodiment, process 980 may begin with operation 981, which includes performing limited DoE of wafers using external metrology to baseline chamber performance. In one embodiment, the limited DoE may include 20 wafers or less. The limited DoE may utilize a recorded process recipe as a baseline. The external metrology may include any metrology suitable for determining device outcomes of processed wafers. For example, in the case of an oxidation process, ellipsometry may be used to investigate film thickness and film thickness uniformity across the wafer.

[0093] In one embodiment, process 980 may continue with operation 982, which includes adding device outcome and other metrology data to the ML model. The additional data added to the ML model may be referred to as a calibration data set. The calibration data set is used to update the ML model so that it more accurately reflects the current state of the processing tool. For example, process 580 may include operation 583, which includes adjusting the model predictions to account for the particular chamber conditions. That is, the process space of the ML model is updated to closely match the conditions of the processing tool being investigated.

[0094] In one embodiment, the ML model can be a model of the process space generated from a combination of statistical and physical models. The statistical model can be generated using DoE of actual wafers as described above. The physical model can be based on real world physics and chemical reaction equations. For example, the physical model can be generated from a simulation of physical and chemical interactions in a processing tool, such as a plasma processing chamber with a rotating cross-flow, across a number of different process parameters. In one embodiment, the ML model can cover the entire process space available to the processing tool.

[0095] In one embodiment, process 980 can continue with operation 984, which includes predicting optimized process parameters to achieve desired wafer outcomes for wafers processed successively in the chamber. The optimized process parameters can be selected after the ML model is updated to include the calibration data set. Correspondingly, the new process recipe provides wafer parameters that result in wafer outcomes that more closely match the targeted values ​​despite changes in chamber conditions. In this manner, chamber drift can be monitored and accounted for to maintain tight process windows and improve uniformity, repeatability, and yield. Correspondingly, unscheduled tool downtime is reduced because the process recipe can be precisely adjusted to account for the chamber drift. Additionally, when PM is performed, performing process 980 can provide shorter recovery times, thereby improving tool utilization.

[0096] In one embodiment, the ML model can be further used to provide continuous (or near-continuous) modification of the process recipe to account for chamber drift. For example, wafer data and process data obtained during processing of device wafers can be utilized to update the ML model. That is, a dedicated DoE may not be required to provide a calibration data set. Wafer data from the device wafers can be obtained for all wafers or for a subset of wafers being processed.

[0097] Such an embodiment may include providing an ML model of the processing tool. The ML model may include statistical and physical models similar to the ML models described above. In one embodiment, the process may begin with a recipe being executed in the processing tool to process a first wafer. After processing the first wafer, wafer data from the first wafer and process data from the processing tool related to the execution of the recipe may be acquired. In one embodiment, the wafer data may include metrology data such as, but not limited to, thickness, thickness uniformity, and profile. In one embodiment, the process data may include data obtained from sensors in the processing tool and / or tool setting information. In one embodiment, the wafer data and the process data are provided to the ML model and an updated ML model is generated. In one embodiment, the updated ML model is used to generate a modified recipe to account for chamber drift in the processing tool. The embodiment may then include executing the modified recipe in the processing tool to process a second wafer. Although the processing of a single first wafer is described above, it will be appreciated that multiple first wafers may be processed before an updated ML model is generated. In such an embodiment, multiple sets of wafer data and process data can be used to generate an updated ML model.

[0098] FIG. 10 illustrates a schematic diagram of a machine according to an exemplary embodiment of a computer system 1000 in which a set of instructions may be executed to cause the machine to perform any one or more of the methods described herein. In alternative embodiments, the machine may be connected (e.g., networked) to other machines in a local area network (LAN), an intranet, an extranet, or the Internet. The machine may operate in the role of a server or a client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), a tablet PC, a set-top box (STB), a web access, a server, a network router, a switch, or a bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by the machine. Additionally, although a single machine is shown, the term "machine" should also be understood to include any collection of machines (e.g., computers) that individually or cooperatively execute a set of instructions (or multiple sets of instructions) to perform any one or more of the methods described herein.

[0099] The exemplary computer system 1000 includes a processor 1002, a main memory 1004 (e.g., read only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM), etc.), a static memory 1006 (e.g., flash memory, static random access memory (SRAM), MRAM, etc.), and a secondary memory 1018 (e.g., a data storage device), which communicate with each other via a bus 1030.

[0100] Processor 1002 represents one or more general-purpose processing devices, such as a microprocessor, a central processing unit, or the like. More specifically, processor 1002 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. Processor 1002 may also be one or more special purpose processing devices, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, or the like. Processor 1002 is configured to execute processing logic 1026 for performing the operations described herein.

[0101] The computer system 1000 may further include a network interface device 1008. The computer system 1000 may also include a video display device 1010 (e.g., a liquid crystal display (LCD), a light emitting diode display (LED), or a cathode ray tube (CRT)), an alphanumeric input device 1012 (e.g., a keyboard), a cursor control device 1014 (e.g., a mouse), and a signal generating device 1016 (e.g., a speaker).

[0102] The secondary memory 1018 may include a machine-accessible storage medium (or, more specifically, a computer-readable storage medium) 1032 having stored thereon one or more sets of instructions (e.g., software 1022) that embody any one or more of the methods or functions described herein. This software 1022 may also reside, completely or at least partially, within the main memory 1004 and / or within the processor 1002 during execution by the computer system 1000, with the main memory 1004 and the processor 1002 also constituting machine-readable storage media. The software 1022 may also be transmitted or received over the network 1020 via the network interface device 1008.

[0103] While in an exemplary embodiment, the machine-accessible storage medium 1032 is shown as a single medium, the term "machine-readable storage medium" should be interpreted to include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) on which one or more sets of instructions are stored. The term "machine-readable storage medium" should also be interpreted to include any medium capable of storing or encoding a set of instructions for execution by a machine that cause the machine to perform any one or more of the methods of the present disclosure. Thus, the term "machine-readable storage medium" should be interpreted to include, but is not limited to, solid-state memory, optical media, and magnetic media.

[0104] In accordance with one embodiment of the present disclosure, a machine-accessible storage medium has stored thereon instructions that cause a data processing system to perform a method for processing a wafer using insights from an ML model and / or a method for updating or building an ML model.

[0105] An embodiment of a plasma chamber having a rotating regulated cross-flow is disclosed.

[0106] Exemplary embodiment 1: A plasma processing chamber includes one or more sidewalls. A support surface within the one or more sidewalls holds a workpiece. A first gas injector along the one or more sidewalls injects a first gas flow generally parallel to and across a surface of the workpiece in a first direction. A first pump port along the one or more sidewalls generally opposite the first gas injector exhausts the first gas flow. A second gas injector along the one or more sidewalls injects a second gas flow generally parallel to and across the surface of the workpiece in a second direction. A second pump port along the one or more sidewalls generally opposite the second gas injector exhausts the second gas flow.

[0107] Exemplary embodiment 2: The plasma processing chamber of embodiment 1, wherein the plasma processing chamber is configured to rotate the first and second gas flows laterally across the workpiece from one or more sidewalls using first and second gas injectors and first and second pump ports to provide a multi-stage rotational cross-flow operation, the multi-stage rotational cross-flow operation including at least a two-stage cycle.

[0108] Exemplary embodiment 3: The plasma processing chamber of embodiment 1, wherein the shape of the one or more sidewalls is cylindrical, elliptical, square, or rectangular.

[0109] Exemplary embodiment 4: The plasma processing chamber of embodiment 1, wherein the first gas injector and the second gas injector are located within openings in the one or more sidewalls.

[0110] Exemplary embodiment 5: a chamber lid on one or more side walls; a support pedestal including a support surface, the support pedestal being below the chamber lid and above the chamber floor and surrounded by one or more sidewalls; a processing region defined by an area between the chamber lid, the support pedestal, and the one or more sidewalls; 5. The plasma processing chamber of embodiment 4, further comprising:

[0111] Exemplary embodiment 6: The plasma processing chamber of embodiment 5, wherein a first gas injector and a second gas injector are located in one or more sidewalls between a chamber lid and the pedestal.

[0112] Exemplary embodiment 7: A plasma processing chamber as described in embodiment 5, wherein the positions of the first pump port and the second pump port are vertically offset from the positions of the first gas injector and the second gas injector by a distance approximately equal to the distance between the bottom surface of the chamber lid and the support pedestal.

[0113] Exemplary embodiment 8: The plasma processing chamber of embodiment 5, wherein the first pump port and the second pump port are present within the cavity between the one or more sidewalls and the support pedestal and above the chamber floor.

[0114] Exemplary embodiment 9: The plasma processing chamber of embodiment 5, wherein the first pump port and the second pump port are located within an additional opening in one or more sidewalls between the chamber lid and the chamber floor.

[0115] Exemplary embodiment 10: The plasma processing chamber of embodiment 1, wherein the first gas flow and the second gas flow are switched on / off to control the rotation of the gas flows.

[0116] Exemplary embodiment 11: The plasma processing chamber of embodiment 1, further comprising an adjustment function applied to a flow rate of at least one of the first gas flow and the second gas flow, or applied to an exit conductance caused by at least one of the first pump port and the second pump port.

[0117] Exemplary embodiment 12: The plasma processing chamber of embodiment 11, wherein the regulating feature comprises one or more gas injection valves for regulating a flow rate of at least one of the first gas flow and the second gas flow.

[0118] Exemplary embodiment 13: A plasma processing chamber as described in embodiment 12, wherein the one or more gas injection valves are coupled to one or more gas sources such that a single gas or a mixture of various types of gases is injected into the processing region during each rotation stage.

[0119] Exemplary embodiment 14: A plasma processing chamber as described in embodiment 12, wherein the first gas injector and the second gas injector apply a constant total gas flow to smoothly and continuously inject the gas flow across different sides of the workpiece within a complete cycle.

[0120] Exemplary embodiment 15: The plasma processing chamber of embodiment 1, further comprising one or more throttle valves for adjusting a pump port conductance or pressure of at least one of the first pump port and the second pump port.

[0121] Exemplary embodiment 16: A plasma processing chamber as described in embodiment 15, wherein one or more throttle valves operate smoothly between two states of conductance or pressure, and the operation is repeated in a similar sequence with the first gas injector and the second gas injector.

[0122] Exemplary embodiment 17: The plasma processing chamber of embodiment 1, further comprising a top-down gas flow.

[0123] Exemplary embodiment 18: A plasma processing chamber as described in embodiment 1, wherein the first gas injector and first pump port comprise a first injector-pump port pair, and the second gas injector and second pump port comprise a second gas injector-pump port pair, and the position of the first injector-pump port pair is offset 180° from the position of the second injector-pump port pair along a plane approximately parallel to the orientation of the workpiece.

[0124] Exemplary embodiment 19: The plasma processing chamber of embodiment 18, further comprising a top-down gas flow.

[0125] Exemplary embodiment 20: The plasma processing chamber of embodiment 18, further comprising a third gas injector and an opposing third pump port, providing a third injector-pump port pair and a three-stage rotational cross-flow operation.

[0126] Exemplary embodiment 21: A plasma processing chamber as described in embodiment 20, wherein the first pair of injector pump ports, the second pair of injector pump ports, and the third pair of injector pump ports are offset by 120 degrees from each other.

[0127] Exemplary embodiment 22: The plasma processing chamber of embodiment 20, wherein the first gas injector, the second gas injector, and the third gas injector are positioned approximately 120° apart from one another, the first pump port, the second pump port, and the third pump port are positioned 120° apart from one another, and the first pump port, the second pump port, and the third pump port are laterally distributed between the first gas injector, the second gas injector, and the third gas injector.

[0128] Exemplary embodiment 23: The plasma processing chamber of embodiment 20, further comprising a fourth gas injector and an opposing fourth pump port, providing a fourth injector-pump port pair and four-stage rotational cross-flow operation.

[0129] Exemplary embodiment 24: A plasma processing chamber as described in embodiment 23, wherein the position of each gas injector pump port pair along the annular sidewall is offset from the position of an adjacent injector pump port pair by an angle equal to a total of 360 degrees divided by the number of injector pump port pairs.

[0130] Exemplary embodiment 25: The plasma processing chamber of embodiment 1, wherein at least one of the first gas injector and the second gas injector includes a single vent in one or more sidewalls.

[0131] Exemplary embodiment 26: The plasma processing chamber of embodiment 1, wherein the first gas injector and the second gas injector comprise a gas injector array of individual gas injectors.

[0132] Exemplary embodiment 27: A plasma processing chamber as described in embodiment 26, wherein individual gas injectors are distributed around the circumference of one or more side walls, and the set of individual gas injectors is adjusted by one or more gas injection valves to generate gas flows in various directions across the workpiece.

[0133] Exemplary embodiment 28: The plasma processing chamber of embodiment 1, wherein at least one of the first gas injector and the second gas injector comprises a gas injector array of individual gas injectors.

[0134] Exemplary embodiment 29: A plasma processing chamber as described in embodiment 28, further comprising a center-to-edge gas flow, wherein at least a first gas flow or a second gas flow injected from a center gas injector among the individual gas injectors in the gas injector array has a higher flow rate relative to an end gas injector in the gas injector array.

[0135] Exemplary embodiment 30: A plasma processing chamber as described in embodiment 28, further comprising an edge-center gas flow, wherein at least a first gas flow or a second gas flow injected from an end gas injector among the individual gas injectors in the gas injector array has a higher flow rate relative to a central gas injector in the gas injector array.

[0136] Exemplary embodiment 31: A plasma processing chamber as described in embodiment 28, further comprising at least four gas injector arrays and opposing pump ports, wherein at least the first gas flow or the second gas flow is directed to the side of the workpiece rather than across the workpiece by closing the opposing pump ports and opening a lateral pump port among the pump ports.

[0137] Exemplary embodiment 32: The plasma processing chamber of embodiment 1, wherein the plasma processing chamber is used for performing reactive ion etching during semiconductor manufacturing.

[0138] Exemplary embodiment 33: A method of implementing a rotational gas crossflow in a plasma processing chamber. During a first phase, the steps include injecting a first gas flow in a first direction generally parallel to and across a surface of the device by a first gas injector at a first location along one or more sidewalls of the plasma processing chamber, and exhausting the first gas flow from the plasma processing chamber by a first pump port at a second location along the one or more sidewalls generally opposite the first gas injector. During a second phase, the steps include injecting a second gas flow in a second direction generally parallel to and across the surface of the device by a second gas injector at a third location along the one or more sidewalls, and exhausting the second gas flow from the plasma processing chamber by a second pump port at a fourth location along the one or more sidewalls generally opposite the second gas injector.

[0139] Exemplary embodiment 34: The method of embodiment 33, further comprising querying a machine learning (ML) model to control the timing of the first gas flow and the second gas flow.

[0140] Exemplary embodiment 35: The method of embodiment 34, further comprising formulating a semiconductor manufacturing process recipe for the device by selecting one or more device outcomes and querying the ML model to obtain a process recipe recommendation suitable for obtaining the device outcomes when processed by a plasma processing chamber including a rotating gas crossflow.

[0141] Exemplary embodiment 36: The method of embodiment 35, further comprising performing a design of experiments (DoE) on the set of wafers to validate the process recipe recommended by the ML model.

[0142] Exemplary embodiment 37: The method of embodiment 35, further comprising receiving as a process recipe any combination of temperature, RF source power, bias power, gas pressure (mTorr), gas flow ramp open time (msec), gas flow time (msec), gas flow ramp close time (msec), gas flow rate in various gas injectors, gas composition in various injectors, gas flow rate to various injectors, gas flow rotation frequency, gas flow composition frequency, gas flow rate / velocity (pressure gradient), gas flow direction, gas rotation phase, electron / plasma density, plasma density gradient, electron temperature, ion current density, plasma potential, sheath electric field potential, sheath electric field tilt angle, sheath electric field z component, mass fraction of atomic O, O flux, and ion current density relative to the workpiece.

[0143] Exemplary embodiment 38: The method of embodiment 35, further comprising selecting as the device outcome any combination of feature profile, layer thickness, thickness uniformity, layer material composition, composition uniformity, porosity, film stress, process uniformity between chambers within a facility, uniformity between wafers, and uniformity between different wafer lots.

[0144] Exemplary embodiment 39: The method of embodiment 38, further comprising selecting any combination of the following as device outcomes during the etching process: etch rate, center-to-edge etching or uniformity, circumferential uniformity of etch rate, etch feature uniformity, tilt, bow, and mask remaining.

[0145] Exemplary embodiment 40: The method of embodiment 33, further comprising: baseline the plasma processing chamber by performing a limited design of experiments (DoE) of the wafer using external metrology to baseline the chamber performance. The wafer outcomes and metrology data from the limited DoE are added as a calibration data set to an ML model, which includes statistical and physical models. The model predictions are adjusted to account for the specific chamber and / or wafer conditions identified by the limited DoE. Optimized process parameters are predicted to achieve the desired wafer outcomes of the wafers processed in the plasma processing chamber.

[0146] Exemplary embodiment 41:Embodiments disclosed herein include a plasma processing chamber with one or more sidewalls. A support within the one or more sidewalls for holding a workpiece. A first gas injector is present at a first location along the one or more sidewalls and a first pump port is present at a second location along the one or more sidewalls generally opposite the first gas injector. A second gas injector is present at a third location along the one or more sidewalls and a second pump port is present at a fourth location along the one or more sidewalls generally opposite the second gas injector. A multi-phase rotating cross-flow operation includes at least a first phase and a second stage. The first stage includes injecting a first gas flow by the first gas injector generally parallel to and across a surface of the workpiece in a first direction and exhausting the first gas flow by the first pump port. A second step includes injecting a second gas flow in a second direction generally parallel to and across the surface of the workpiece with a second gas injector and exhausting the second gas flow with a second pump port.

[0147] Exemplary embodiment 42: The plasma processing chamber of embodiment 41, further comprising a first gas injection valve connected to the first gas injector, a second gas injection valve connected to the second gas injector, a first pressure control valve connected to the first pump port, and a second pressure control valve connected to the second pump port.

[0148] Exemplary embodiment 43: A plasma processing chamber as described in embodiment 42, further comprising a controller connected to the plasma processing chamber, the controller configured to initiate a first gas flow during the first stage by fully opening the first gas injection valve and partially opening the second gas injection valve, and to open the first pressure control valve and close the second pressure control valve.

[0149] Exemplary embodiment 44: The plasma processing chamber of embodiment 43, wherein the controller is further configured to rotate the direction of gas flow by beginning to close a first gas injection valve near a transition between a first stage and a second stage, fully opening the second gas injection valve and partially opening the first gas injection valve to begin a second stage, and to open a second pressure control valve and close the first pressure control valve.

[0150] Exemplary embodiment 44: A non-transitory computer readable medium having software instructions stored thereon, the software instructions, when executed by a processor, causing the processor to rotate a gas cross-flow in a plasma processing chamber by performing the following steps: during a first stage, the steps including injecting a first gas flow generally parallel to and across a surface of the device in a first direction by a first gas injector at a first location along one or more sidewalls of the plasma processing chamber, and exhausting the first gas flow from the plasma processing chamber by a first pump port at a second location along the one or more sidewalls generally opposite the first gas injector. During a second stage, the steps include injecting a second gas flow in a second direction generally parallel to and across a surface of the device by a second gas injector at a third location along the one or more sidewalls, and exhausting the second gas flow from the plasma processing chamber by a second pump port at a fourth location along the one or more sidewalls generally opposite the second gas injector.

[0151] Exemplary embodiment 46: The non-transitory computer-readable medium of embodiment 45, further comprising querying a machine learning (ML) model to control the timing of the first gas flow and the second gas flow.

[0152] Exemplary embodiment 47: The non-transitory computer readable medium of embodiment 46, further comprising formulating a semiconductor manufacturing process recipe for the device by selecting one or more device outcomes and querying the ML model to obtain a process recipe recommendation suitable for obtaining the device outcomes when processed by a plasma processing chamber including a rotating gas crossflow.

[0153] Exemplary embodiment 48: The non-transitory computer-readable medium of embodiment 47, further comprising running a design of experiments (DoE) on the set of wafers to validate the process recipe recommended by the ML model.

[0154] Exemplary embodiment 49: The non-transitory computer-readable medium of embodiment 47, further comprising receiving as a process recipe any combination of temperature, RF source power, bias power, gas pressure (mTorr), gas flow ramp open time (msec), gas flow time (msec), gas flow ramp close time (msec), gas flow rate in various gas injectors, gas composition in various injectors, gas flow rate to various injectors, gas flow rotation frequency, gas flow composition frequency, gas flow rate / velocity (pressure gradient), gas flow direction, gas rotation phase, electron / plasma density, plasma density gradient, electron temperature, ion current density, plasma potential, sheath electric field potential, sheath electric field tilt angle, sheath electric field z component, mass fraction of atomic O, O flux, and ion current density relative to the workpiece.

[0155] Exemplary embodiment 50: A non-transitory computer-readable medium as described in embodiment 47, further comprising selecting as device outcomes any combination of feature profile, layer thickness, thickness uniformity, layer material composition, composition uniformity, porosity, film stress, process uniformity between chambers within a facility, uniformity between wafers, and uniformity between different wafer lots.

[0156] Exemplary embodiment 51: A non-transitory computer readable medium as described in embodiment 50, further comprising selecting any combination of the following as device outcomes during the etching process: etching rate, center and edge etching or uniformity, circumferential uniformity of etching rate, uniformity of etch features, tilt, bow, and mask remaining.

[0157] Exemplary embodiment 52: The non-transitory computer readable medium of embodiment 45, further comprising: baseline the plasma processing chamber by performing a limited design of experiments (DoE) of the wafer using external metrology to baseline the chamber performance. The wafer outcomes and metrology data from the limited DoE are added as a calibration data set to an ML model, which includes statistical and physical models. The model predictions are adjusted to account for the specific chamber and / or wafer conditions identified by the limited DoE. Optimized process parameters are predicted to achieve the desired outcomes of the wafers processed in the plasma processing chamber.

Claims

1. 1. A plasma processing chamber comprising: one or more side walls; a support surface within the one or more side walls for holding a workpiece; and a first gas injector along the one or more sidewalls for injecting a first flow of gas in a first direction generally parallel to and across a surface of the workpiece; a first pump port along the one or more sidewalls generally opposite the first gas injector for discharging the first gas flow; a second gas injector along the one or more sidewalls for injecting a second flow of gas generally parallel to and across the surface of the workpiece in a second direction different from the first direction; a second pump port along the one or more sidewalls generally opposite the second gas injector for discharging the second gas flow; and A plasma processing chamber comprising:

2. 2. The plasma processing chamber of claim 1, wherein the plasma processing chamber is configured to rotate the first and second gas flows laterally from the one or more sidewalls across the workpiece using the first and second gas injectors and the first and second pump ports to provide a multi-stage rotational cross-flow operation, the multi-stage rotational cross-flow operation including at least a two-stage cycle.

3. The plasma processing chamber of claim 1 , wherein the first gas injector and the second gas injector are located within openings in the one or more sidewalls.

4. 2. The plasma processing chamber of claim 1, wherein the locations of the first and second pump ports are vertically offset from the locations of the first and second gas injectors.

5. 10. The plasma processing chamber of claim 1, wherein the first gas flow and the second gas flow are switched on and off to control gas flow rotation.

6. 10. The plasma processing chamber of claim 1, further comprising an adjustment function applied to a flow rate of at least one of the first gas flow and the second gas flow or to an exit conductance provided by at least one of the first pump port and the second pump port.

7. 10. The plasma processing chamber of claim 1, further comprising a third gas injector and an opposing third pump port providing a third injector-pump port pair and a three-stage rotational cross-flow operation.

8. The plasma processing chamber of claim 1 , wherein at least one of the first gas injector and the second gas injector includes a single vent in the one or more sidewalls.

9. The plasma processing chamber of claim 1 , wherein the first gas injector and the second gas injector comprise a gas injector array of individual gas injectors.

10. 1. A method for implementing a rotational gas crossflow in a plasma processing chamber, the method comprising: during a first stage, injecting a first gas flow in a first direction generally parallel to and across a surface of the device by a first gas injector at a first location along one or more sidewalls of the plasma processing chamber, and exhausting the first gas flow from the plasma processing chamber by a first pump port at a second location along the one or more sidewalls generally opposite the first gas injector; during a second stage, injecting a second gas flow in a second direction, different from the first direction, generally parallel to and across a surface of the device by a second gas injector at a third location along the one or more sidewalls, and exhausting the second gas flow from the plasma processing chamber by a second pump port at a fourth location along the one or more sidewalls generally opposite the second gas injector; A method comprising:

11. 11. The method of claim 10, further comprising querying a machine learning (ML) model to control timing of the first gas flow and the second gas flow.

12. selecting one or more device outcomes; Querying the ML model to obtain a recommendation for a process recipe suitable for obtaining the device outcome when processed by the plasma processing chamber including the rotating gas crossflow; 12. The method of claim 11, further comprising formulating a semiconductor manufacturing process recipe for the device by

13. 13. The method of claim 12, further comprising running a design of experiments (DoE) on a set of wafers to verify the process recipe recommended by the ML model.

14. 14. The method of claim 13, further comprising receiving as the process recipe any combination of temperature, RF source power, bias power, gas pressure (mTorr), gas flow ramp open time (msec), gas flow time (msec), gas flow ramp close time (msec), gas flow rates in various gas injectors, gas composition in various injectors, gas flow rates to various injectors, gas flow rotation frequency, gas flow composition frequency, gas flow rate / velocity (pressure gradient), gas flow direction, gas rotation phase, electron / plasma density, plasma density gradient, electron temperature, ion current density, plasma potential, sheath electric field potential, sheath electric field tilt angle, sheath electric field z-component, mass fraction of atomic O, O flux, and ion current density relative to the workpiece.

15. 1. A plasma processing chamber comprising: one or more side walls; supports within the one or more side walls for holding a workpiece; and a first gas injector at a first location along the one or more sidewalls; a first pump port at a second location along the one or more sidewalls generally opposite the first gas injector; and a second gas injector at a third location along the one or more sidewalls; a second pump port at a fourth location along the one or more sidewalls generally opposite the second gas injector; and A multi-stage rotary cross-flow operation, a first stage including injecting a first gas flow in a first direction generally parallel to and across a surface of the workpiece with the first gas injector and exhausting the first gas flow with the first pump port; a second stage including injecting a second gas flow with the second gas injector in a second direction different from the first direction generally parallel to and across the surface of the workpiece, and exhausting the second gas flow with the second pump port. A multi-stage rotating cross-flow operation including at least 1. A plasma processing chamber comprising:

16. 16. The plasma processing chamber of claim 15, further comprising a first gas injection valve connected to the first gas injector, a second gas injection valve connected to the second gas injector, a first pressure control valve connected to the first pump port, and a second pressure control valve connected to the second pump port.

17. and a controller coupled to the plasma processing chamber, the controller comprising: during the first stage, initiating the first gas flow by fully opening the first gas inlet valve and partially opening the second gas inlet valve; opening the first pressure control valve and closing the second pressure control valve; The plasma processing chamber of claim 16 configured to perform:

18. The controller: beginning to close the first gas inlet valve near a transition between the first stage and the second stage; rotating the direction of gas flow by fully opening the second gas inlet valve and partially opening the first gas inlet valve to begin a second stage; opening the second pressure control valve and closing the first pressure control valve; The plasma processing chamber of claim 17 , further configured to:

19. A non-transitory computer readable medium having stored thereon software instructions which, when executed by a processor, perform the following steps: during a first stage, injecting a first gas flow in a first direction generally parallel to and across a surface of the device by a first gas injector at a first location along one or more sidewalls of the plasma processing chamber, and exhausting the first gas flow from the plasma processing chamber by a first pump port at a second location along the one or more sidewalls generally opposite the first gas injector; during a second stage, injecting a second gas flow in a second direction generally parallel to and across the surface of the device by a second gas injector at a third location along the one or more sidewalls, and exhausting the second gas flow from the plasma processing chamber by a second pump port at a fourth location along the one or more sidewalls generally opposite the second gas injector; 46. ​​A non-transitory computer readable medium that causes the processor to rotate a gas cross-flow within a plasma processing chamber by performing steps including:

20. 20. The non-transitory computer readable medium of claim 19, further comprising querying a machine learning (ML) model to control timing of the first gas flow and the second gas flow.

Images