WO2019005170A1 - Systems, methods, and apparatuses for implementing dynamic learning mask correction for resolution enhancement and optical proximity correction (opc) of lithography masks - Google Patents

Abstract

Methods, systems, and apparatuses for implementing dynamic learning mask correction include: creating a mask via a lithography process; performing a learning phase to identify a set of mask shape variables that change simulated intensity values of the mask and a simulated shape of the mask by iteratively moving mask segments; collecting data representing image contrast of the mask, mask critical dimension, and contour critical dimensions of the mask for each iteration of moving the mask segments; performing sensitivity analysis on the collected data to determine which mask segments correspond to an improved simulated image contrast of the mask when moved; selecting one or more mask segments that correspond to the improved simulated image contrast of the mask when moved; and applying Optical Proximity Correction for a model of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

Description

SYSTEMS, METHODS, AND APPARATUSES FOR IMPLEMENTING DYNAMIC LEARNING MASK CORRECTION FOR RESOLUTION ENHANCEMENT AND OPTICAL PROXIMITY CORRECTION (OPC) OF LITHOGRAPHY MASKS

5 CLAIM OF PRIORITY ϊ^ββθΐ] None.

COPYRIGHT NOTICE

[0002] A portion of the disclosure of this patent document contains material which is 10 subject to copyright protection. The copyright owner has no objection to the facsimile

reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent, and Trademark Office patent file or ^'records, but otherwise reserves all copyright rights whatsoever.

I S TECHNICAL FIELD

|0 03] The subject matter described herein relates generally to the field of semiconductor and electronics manufacturing, and more particularly, to systems, methods, and apparatuses for implementing dynamic learning mask correction for resolution enhancement and Optica! Proximity Correction (OPC) of lithography masks.

20

BACKGROUND

1000 J The subject matter discussed in the background section should not he assumed to be prior art merely as a result of its mention in the background section. Similar ly, a problem mentioned in the background section or associated with the s ubject matter of the background 5 section should not be assumed to have been previously recognized in the prior art. The subject matter in the background section merely represents different approaches, which in and of themselves may also correspond to embodiments of the claimed subj ect matter.

fOOOS] Conventional techniques utilizing Optical Proximity Correction (OPC) models are not sufficiently accurate as semiconductor manufacturing techniques necessitate increasingly 30 strict tolerances and smaller physical dimensions.

|0006j For advanced process nodes in semiconductor manufacturing, one of the major challenges is to control defects and yield to a le vel that is viable for high volume matrufacturjftg (HVM). To maintain the density scaling and reduce cell Ibotprint. process margins are ^'tightening which in turn causes systematic and random process variations resulting in the processes being more prone to defects.

{8607] When the physical dimensions of the manufactured semiconductors were greater , such inaccuracies and uncertainty were manageable due to the greater margins. However, as the physical size of these manufactured semiconductors is reduced further into the nanometer realm of manufacturing, it becomes necessary to operate using increasingly accurate models if manufacturing yields are to remain economically viable.

(0608} The present state of the art. may therefore benefit from the systems, methods, and apparatuses for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correclion (OPC) of lithography masks as described herein.

BRIEF BESCMPTIO OF THE DRAWINGS {0009] ^'Embodiments are illustrated by way of example, and not by way of limitation, and will be more fully understood with reference to the following detailed description when considered in connection with the figures in which:

{6010] Figure 1A depicts application of an Adaptive Local Intensity (All) corrector to OPC simulation results in accordance with described embodiments:

{©Oil] Figure IB describes a process flow implementing the Dynamic Learning

Corrector (DLC) process from Figure 1A in greater detail;

{6012] Figure 2 describes a decision process flow including a methodology for

determining which segment to utilize for mask optimization in accordance with described embodiments;

10013] Figure 3 describes an exemplary data set collected for a specified gauge having been measured or evaluated as having a simulated image contrast (SIC) in accordance with described embodiments;

{00 4] Figure 4 provides a schematic comparison simulation results util izing an

Adaptive Local Intensity (AL1) Corrector versus- a standard corrector, in accordance with described embodiments;

{6615] Figure 5 describes the application of an Adaptive Adjacent Intensity (ΑΑΪ) type corrector in accordance with described embodiments;

{0016] Figure 6 depicts an exemplar Artificial Neural Network (ANN) machine ^{earning algorithm in which the input variables predict the deterministic model error o process driven changes in critical dimension .i ccordance with described embodiments;

10017} Figure 7 is a schemati c of a computer system in accordance with described embodiments;

[6018] Figure 8 illustrates a semiconductor device (or an interposer) that includes one or more described embodiments;

(001 ) Figure 9 illustrates a computing device irs accor dance with one implementation of the invention; and

f0020j Figure 10 is a flow diagram illustrating a method for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks in accordance with described embodiments.

DETAILED DESCRIPTION

[6021] Described herein are systems, methods, and apparatuses for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks. For instance, in accordance with one embodiment there are means described for reducing Optical Proximity Correction (OPC) model error, wherein such means include: creating a mask via a lithography process; performing a learning phase to identify a set of variables that change simulated intensity values of the mask and a simulated shape of the mask by iterat vely moving mask segments; collecting data representing image contrast of the mask, mask critical dimension (CD), and contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments; perforating sensitivity analysis on the collected data to determine which mask segments correspond to a improved simulated image contrast of the mask when moved; selecting one or more mask segments that correspond to the improved simulated image contrast of the mask when moved; and applying Optical Proximity Correction (OPC) for a mode l of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

[6022] Optical Proximity Correction (OPC") models require accurate mask layout dimensions as input parameters. The greater the accuracy, tire more useful and accurate the resulting mode? will be for the semiconductor manufacturing processes.

10023] The methodologies described herein provide improvements to OPC model accuracy which may then be applied to a variety of lithography processes.

[6024] As circuit feature sizes shrink to enabl continued Moore's Law scaling, the accuracy requirements for the placement and size of individual wafer features continue to tighten. A. critical determiner of this wafer feature scaling is lithography fidelity, which is a direct result of proper feature sizing on the product photomasks. The size of mask features is directly determined by the OPC (Optical Proximity Correction) as determined by the OPC lithography model. As the Edge Placement Error (EPE) requirements tighten for next generation node manufacturing; such as 1.0»nanometer (rim) node and beyond, significant improvements are required in OPC modeling capability to enable both efficient process development cycles, and ultimately to enable correct feature sixes on product wafers to so as to attain proper circuit function.

f0025j Weak or otherwise insufficient image contrast leads to scumming and other image contrast related defects on the resist due to insufficient light intensity in certain areas. For instance, in areas of a photolithographic mask for which there is insufficient contrast, exposure of the mask followed by developer to etch away the exposed areas leaves undesirable residue which hi tarn contributes to process defects in the fabrication of functional silicon wafers (e.g., suc as CPUs, etc ). Mo ving the contours of the features c an help to impro ve image contrast and thus improve the light intensity in those problem areas, however, because the features must be placed with extreme precision, it is undesirable to alter or move those feature contours.

fOO j Prior solutions simply involve making the problematic features larger, such as those identified features resulting in insufficient light intensity due to weak contrast. Once the features are enlarged, they generally will develop better and eliminate the scumming problem. Unfortunately, such a solution consumes physical space on the semiconductor chip, thus reducing the area within which functional elements may be located. It is therefore necessary to have tighter controls on the placement and development of such features via the

photolithographic mask. More problematically still, as the contours of features are altered or the sizes of features are enlarged, there is an increased risk of creating additional defects due to features touching, thus resulting in a short or features being too distant and not connecting when they should, resulting in an open, either of which will result in defective product and therefore reduce yield for the manufacturing process.

{0027] Therefore, rather than enlarging features, the methodologies described herein seek to identify the most sensitive segments which respond to small changes resulting in greater image contrast rather than seeking to identif isolated or localized feature changes to increase contrast in the image.

|0028j By utilizing a dynamic contractor, which is described m greater detail below, machine learning sensitivity analysis is applied to identity where the most critical feature locations) are within the. photolithographic mask which may be manipulated to cause an increased contrast response within the mask. For instance, through the application of machine learning sensitivity analysis, Dynamic Learning Correctors (DLCs) are able to identify which segment of a mask is most sensiti ve and therefore corresponds to a largest impac t on the corrector, preferably with the slightest or smallest deviation from a contour target.

f0029] For instance, when performing an OPC (Optical Proximity Correction)

correction, polygons of the photolithographic mask are taken and the edges of such polygons are broken down into many much smaller segments. Those smaller segments are then perturbed or moved so as to hit or align to specified target contours, thus resulting in the images being located in the correct position,

f0030j By moving or altering those smaller segments via application of machine learning sensitivity analysis, OPC the optimization processing as described herein identifies which of those broken oat segments, when moved, produce a largest change in the contrast for the smallest change in the critical dimension (CD) of the target contour in question.

j¾ 3l] Because the optics are non-linear, the closest edge to a problematic defect site exhibiting weak or insufficient contrast may happen not to be the segment which provides the greatest increase in improving image contrast of the mask. However, because the physics of the opti cs and mask itself is incredibly complex, i t is not feasible for a human to identi fy through any realistic process which segments or contours should be altered to render the needed contrast improvement. There are billions and billions of different geometries rendering any simple rule for optimizatio wholly impractical Therefore, application of the machine learning sensitivity analysis provides a significantly more effective and intelligent algorithmic process by which to identify the region in the mask for which small changes will yield large and beneficial

improvements to the image contrast, while simul taneously reducing contour deviation to a minimum.

[0032} Resist patterns generated by lithographic processes are the results of complicated optical, chemical and physical phenomenon, which can be modeled based on optical image parameters and geometric parameters. Model predictions, which are represented in the form of contours, are generated based on the distorted image maps, which are numerically stable and efficient enough to be used for high volume manufacturing. To generate an accurate and yet numerically efficient model, machine learning algorithms such as Artificial Neural ^'Network (ANN) algorithms are employed in accordance with certain embodiments.

0033] The methodologies illustrated here as well as the systems and apparatuses which implement or embody such methodologies permit dynamic learning of mask corrections for resolution enhancement and improved OPC Model prediction beyond currently known

techniques. According to certain embodiments, such resolution enhancement and OPC model improvement techniques may be appl ied to any of Aerial I mage Correction, Machine Learning improvements. Optical Proximity Correction (OPC) of semi-physical models, and resolution mhancement techniques utilized within the semiconductor ntaxmfacturing process.

10034! The existing modeling software for OPC is deficient in terms- of the accuracy when new lithographic systems having tighter .margins and dimensions are utilized, resulting in unworkable errors inhibiting the newer technologies from being successfully scaled to high volume manufacturing. Conventional OPC models simply cannot predict with sufficient accuracy and precision to the smaller feature size and feature geometries associated new technologies.

f0035j improvements to the existing OPC models are needed such that the OPC models which drive manufacturing may he successfully utilized to attain sufficiently accurate pattern fidelity on the mask and on the manufactured silicon wafers.

10036! Use of the conventional models have been demonstrated to result in unacceptable yields from the fabrication proces due to the degree of uncertainty of such models, resulting in the features which are actually printed ultimately foiling within the layers due to the size of the features as printed which is a direct result of the inaccurate models. The position and size of the features of the mask are cri tical to the functioning of the circ uit produced and on the scale of nanometers the margin for error becomes incredi bly small thus necessitating model outputs with a- greater degree of precision. Printing deviances of tens of nanometers in the process which are not accounted for in the design will render an entire silicon wafer inoperable. Such a circuit simply will not work and will not function like a processor because such deviations cause joins and shorts which are not part of the circuit design.

|0037] The conventional OPC models need further adjustment and correction and improved contour correction or shifting means so as to move these printed features into the correct position with a greater degree of precision. Such improvements ensure functionally operable silicon and thus improve manufacturing yields and profitability for any given product.

|0038| In the following description, numerous specific details are set forth such as examples of specific systems, languages, components, etc.. in order to provide a thorough understanding of the various embodiments. It will be apparent, however, to one skilled in the art that these specific details need not he employed to practice the embodiments disclosed herein. In other instances, well known materials or methods have not been described in detail in order to avoid unnecessarily obscuring the disclosed embodiments.

[0639] In addition to various hardware components depicted in the figures and described herein, embodiments further include various operations which are described below. The operations described in accordance with such embodiments may be performed by hardware components or may be embodied in machine-executable instructions, which may be used to cause a general-purpose or special-purpose processor programmed with the instructions to perform the operations. Alternatively, the operations may be performed by a combination of hardware and software.

{0040] Any of the disclosed embodiments may he used alone or together with one another in any combination. Although various embodiments may have been partially motivated by deficiencies with conventional techniques and approaches, some of which are described or alluded to within the specification, the embodiments need not necessarily address or solve any of these deficiencies, but rather, may address only some of the deficiencies, address none of the deficiencies, or be directed toward different deficiencies and problems which are not directly discussed.

|0041 j Implementations of embodiments of the invention may be formed or carried out on a substrate, such as a semiconductor substrate, in one implementation, the semiconductor substrate may be a crystalline substrate formed using a bdk silicon or a stKcon-on-insulator substructure. In other implementations, the semiconductor substrate may be formed using alternate materials, which may or may not be combined with silicon, that include but are not limited to germanium, indium anttnionide, lead teiluride, indium arsenide, indium phosphide, gallium arsenide, indium gallium arsenide, gallium antimomde, or other combinations of group III-V or group IV materials. Although a few examples of materials from which the substrate may be formed are described, here, any material that may serve as a foundation upon which a semiconductor device may be built falls w ithin the spirit and scope of the present invention .

{0042] A plurality of transistors, such as raetal-oxide-seraiconductor field-effect transistors (MOSFET o simply MOS transistors), may be fabricated on the substrate. In various implementations of the invention, the MOS transistors may be planar transistors, nonplanar transistors, or a combination of both. Nonplanar transistors include FiiiFET transistors such as double-gate transistors and tri-gate transistors, and wrap-around or all-around gate transistors such as nanoribbon and nanowire transistors. Although the implementations described herein may illustrate only planar transistors, it should be noted that the invention may also be carried out using nonplanar transistors,

|8843} Each MOS transistor includes a gate stack formed of at least two layers, a gate dielectric layer and a gate electrode layer. The gate dielectric layer may include one layer or stack of layers. The one or more layers .may include silicon oxide, silicon dioxide fSiOs) and/or a high-k dielectric material. The high-fc dielectric material may include elements such as hafnium, silicon, oxygen, titanium, tantalum, lanthanum, aluminum, zirconium, barium, strontium, yttrium, lead, scandium, niobium, and zinc . Examples of high-k materials that may be used in the gate dielectric layer include, but are not limited to, hafnium oxide, hafnium silicon oxide, lanthanum oxide, lanthanum aluminum oxide, zirconium oxide, zirconium silicon oxide, taniakj.ro oxide, titanium oxide, barium strontium titanium oxide, barium titanium oxide, strontium- titanium oxide, tt ium oxide, aluminum oxide, lead scandium tantalum oxide, and lead zinc niobate. In some embodiments, an annealing process may be carried, out on the gate dielectric layer to improve its quality when a high-k material is used.

|0044| The gate electrode layer is formed on the gate dielectric layer and may consist of at least one P-type workfunction metal or N-type workfunction metal, depending on whether the transistor is to be a PMOS or an MMOS transistor. In some implementations, the gate electrode layer may consist of a stack of two or more metal layers, where one or more metal layers are workfunction metal layers and at least one metal layer is a fill metal layer.

{0045) For a PMOS transistor, metals thai may be used for the gale electrode include, but are not limited to, ruthenium, palladium, platinum, cobalt, nickel, and conductive metal oxides, e.g., ruthenium oxide. A P~ ype metal layer will enable the formation of a PMOS gate electrode with a orkfunction that is between about 4.9 eV and about 5.2 eV. For an MOS transistor, metals that may be used for the gate electrode include, but are not limited to, hafnium, zirconium, titanium, tantalum, aluminum, alloys of these metals, and carbides of these metals such as hafnium carbide, zirconium carbide, titanium carbide, tantalum carbide, and aluminum carbide. An N-type metal layer will enable the formation of an MOS gate electrode wi th a workfunction that is between about 3.9 eV and. about 4.2 eV.

{0046) In some implementations, die gate electrode may consist of a "U"-shaped structure that mcludes a bottom portion substantially parallel to the surface of the substrate and two side all portions that are substantially perpendicular to the top surface of the substrate. In another implementation, at least one of the metal layers that form the gate electrode may simply be a planar layer that is substantially parallel to the top surface of the substrate and does not include sidewall portions substantially perpendicular to the top surface of the substrate. In further implementations of the invention, the gate electrode may consist of a combination of U-shaped structures and planar. non- U-shaped structures. For example, the gate electrode may consist of one or more U-shaped metal layers formed atop one or more planar, non-U-shaped layers.

|8847j In some implementations of the invention, a pair of sidewall spacers may be formed on opposing sides of the gate stack that bracket the gate stack. The sidewall spacers may be formed from a material such as silicon nitride, silicon oxide, silicon carbide, silicon nitride doped with carbon, and silicon oxynitride. Processes for forming sidewall spacers are well know in the art and generally include deposition and etching process steps. In an alternate implementation, a plurality of spacer pairs may be used, for instance, two pairs, three pairs, or four pairs of sidewall spacers may be formed on opposing sides of the gate stack. f0O48| As is well known in the art, source and drain regions are formed wi thin the substrate adjacent to the gate stack of each MOS transistor. The source and drain regions are generally formed using either an implantation/diffusion process or an etching/deposition process. In the former process, dopants such as boron, aluminum, antimony, phosphorous, or arsenic may be ion-implanted into the substrate to form the source and drain regions. An annealing process that activates the dopants and causes them to diffuse further into the substrate typically follows the ion implantation process, in the latter process, the substrate may first be etched to form recesses at the locations of the source and drain regions. An epitaxial deposition process may then be carried out to fill the recesses with material that is used to fabricate the source and drain regions. In some implementations, the source and drain regions may be fabricated using a silicon alloy such as silicon germanium or silicon carbide. In some implementations the epitaxially deposited silicon alloy may be doped in situ with, dopants such as boron, arsenic, o phosphorous. In. further embodiments, the source and drain regions ma be formed using one or more alternate semiconductor materials such as germanium or a group I1I-V material or alloy. And in further embodiments, one or more layers of metal and/or metal alloys may be used to form the source and drain regions.

}0049] Due or more interlayer dielectrics (ILD) are deposited over the MOS transistors. The ILD layers may be formed using dielectric materials known for their applicability in integrated circuit structures, such as low-k dielectric materials. ^'Examples of dielectric materials that may be used include, but are not limited to, silicon dioxide (SiOa), carbon doped oxide

(CJDO), silicon nitride, organic polymers such as perfluorocyclobutane or

polytetrafluoroediylene, fluorosilicate glass (FSG), and organosilicat.es such as silsesquioxane, siloxane, or organosilicate glass. The ILD layers may include pores or air gaps to further reduce their dielectric constant.

jo«50] Figure J A depicts application of an Adaptive Local Intensity (ALT) corrector to

OPC simulation results 100 hi accordance with described embodiments.

[0051] More particularly, multiple stages are depicted including, the training stage 105, the decision iteration 1 10 phase and the continued correction 1 15 phase.

0052] At trai ning stage 105 each of the depicted gauges (1 and 2 } are assigned a corrector. As shown here, gauge 1 is assigned a corrector and the corrector is measured 120. Processing then records the measurement of the measured corrector 120 as well as measures nearby gauges 2 which was also assigned a corrector, thus resulting in measured nearby gauges 140,

{0053] After the training period, processing determines a most effective response mask segment for each gauge that measured a low SIC at the last iteration step in the training period. For instance, Figure 2 below details an exemplary procedure by which to determine the .most effective response segment, however, other algorithms may be utilized, to pick a. most effective segment or alternatively to select, multiple segments. During the decision iteration 1 10 phase, an allowed tolerance may be increased for feature contour deviation from target (e.g., thus loosening the required contour compliance to target) at the gauge associated with the correction segment to allow a larger response range for mask segments used in the correction than other segments.

f0054j Such a dynamic tolerance increase allows for optimized contour convergence while deviating only where necessary to improve SIC. Procedures to determine the size of the looser tolerance may involve, for example, loosening the tolerance of all responding gauges to a fixed value or computing the required mask shape change required to improve the SIC above a fixed or dynamic threshold, contrast.

|8055j Therefore, according to a particular embodiment, the decision iteration 1 10 phase includes at least determining a fixed iteration and a most effective response segment for each gauge. Here, gauge 1 at center block (b) as represented by element 130 exhibits a weak

Simulated Image Contrast (SIC) for iteration 2 as represented by element 150. Conversely, for this iteration, the most effective monitored gauge is gauge 3 as represented by element 145. Because all measurements are for Critical Dimensions (CD), the corrector which resulted in the most effecti ve gauge 3 as depicted by element 145 is therefore correspondingly applied to gauge 4 (element 155). thus assigning the most effective monitored gauge from gauge 3 (element 145) to correct for the Simulated image Contrast (SIC) of gauge 4 as represented by element 155. With the most effective gauge having been applied to correct for the weak Simulated image Contrast (SIC) of other gauges, processing now advances to the continued correction phase 115, |0056{ At the continued correction phase 1 15, the iterative processing for each of the remaining gauges to be requested request the Simulated image ContTasi (SIC) as represented by element 160, with the SIC bein requested from the original gauge of interest (gauge 2) which then responds in turn with its measured SIC.

|β857| continued correction phase 1 15 applies SIC optimization via the unique segments identified over the remaining iterations of the simulation. During the course of this correction, gauges of identified segments query the original gauge for which they were identified requesting SIC information from those original gauges of interest (element 160) so as to locate the SIC at the marginal or failing site. The correction approach described for the AO corrector is distinguished from the previously known methodologies due to its use of direct inter-gauge communication allowing segments to respond to simulated measurements at gauges to which they are not associated and further by utilizing machine learning approaches to determine the best segments to use for correction in the simulation without a priori knowledge. Consequently, previousiy known methodologies often result in larger simulated feature contour deviations from target than the Ail corrector methodologies utilized herein. For instance, refer to the corrected mask shape at element 465 of Figure 4 as described below which depicts these larger contour deviations.

|G0S8j Optical proximity correction (OPC) of lithography masks has become, a critical operation in pursuance of successfully writing nanometer scale features onto lithographic photo resists. Optical proximity correction (OPC) requires ensuring that not only the predicted feature contours (e.g. , boundaries of the exposed feature on the photo resist) are in the correct locati ons on the photo resist, but additionally requires that the intensity and contrast of the aerial image are sufficient to ensure a h igh-quality reproduction of the mask feature shapes within the photo resist,

|8059j High image contrast at the photo resist may be ensured by identifying low contrast areas and increasing the size of the mask feature, for instance, by utilizing phase shift masks, at the she where the low contrast is identified and sometimes also at nearby sites.

f0060j While such an approach does improve the image contrast, increasing the mask feature size has the unfortunate consequence of also increasing the size of the feature contours around the lo contrast site. As feature pitch continues to decrease, such contour deviations introduce device/lithograp ic liabilities in the fabrication process due to photo resist features interacting with unintended existing structures on the wafer and therefore reducing yields due to wafer failures caused by, for example, opens, shorts, or other non-compliant feature placement within the increasingly strict tolerances required.

{0061 j Therefore, a new Optical proximity correction (OPC) approach as described herein improves image contrast at the photo resist without significantly increasing the size of the feature contours, and therefore provides a benefit over those lithographic processes which do not employ such image correction schemes. According to certain embodiments, a machine learning algorithm is applied to dynamically identify (e.g., learn) the most effective local correction schemes , thus producing appropriate Dynamic Learning Correctors (DLCs).

{0662] Dynamic Learning Correctors (DLCs) are applied to the gauges in accordance with described embodiments to improve the resulting Simulated Image Contrast (SIC) in comparison to conventional approaches.

{0663] Application of the dynamic learning correctors improves the simulated image contrast, which in turn reduces the number of sites marginally subjected to or having a potential scumming risk and therefore improving any worst case image contrast. Scumming is the result of residue being left on the wafer being fabricated subsequent to processing. For instance, after an

I I imaged wafer has been processed (e.g., etched, ion implanted, etc.) the .remaining photo resist must be removed. esist stripping techniques iuctode, for example, wet stripping using organic or inorganic solutions such as an organic acetone stripper or alternatively applying dry (e.g., plasma) stripping. Unfortunately, certain strippers, such as acetone, tend to leave residues on the wafer and is thus undesirable for semiconductor processing.

|(K)64] Regardless, application of the Dynamic Learning Correctors (DLCs) permits accurate matching of both the number and distribution of predicted feature contour deviations from each of the target simulation and the original simulation.

fu065j According to one embodiment, a set of Dynamic Learning Correctors (DLCs) algorithms are provided, in which the set of Dynamic Learning Correctors (DLCs) are configured through the application of a. machine learning algorithm to iteratively improve throughout, the- cowrse-of ah iterative OPC simulation.. More particularly, tile Dynamic Learning Correctors (DLCs) iteratively learn how a predicted image contrast changes as a result of variation of the photolithographic mask and a variety of predicted wafer features of that, photolithographic mask.

{0066] Having learned the predicted image contrast information via the iterative processing, the Dynamic ^Learning Correctors (DLCs) are then utilized to automatically identify and target the most effective section of the photolithographic mask layout or features to be optimized so as to improve local contrast at the photo resist.

{0067] According to one embodiment, Dynamic Learnin Correctors (DLCs) first monitor a simulated contrast and photo resist response as a function of the mask's shape over a defined training period, represented here as training stage 105. At the end of the training stage 105 the Dynamic Learning Corrector (DLC) identifies which mask section optimizations have a most significant impact on the image contrast.

{0068] The sections identified as having a largest or most significant impact on the image contrast are then targeted by a corrector algorithm implemented by the decision iteration i 10 phase until an optimum correction is attained sufficient to improve image contrast.

{0069] Other mask sections which were not identified as being significant to the local image contrast, response may then be corrected usin alternative OPC correction methodologies so as to maintain the original feature contours of the mask, such that, the image contrast of the mask is improved while simultaneously minimizing deviations of the feature contours from their intended target.

{0070] In laboratory experimentation, the above described OPC simulations result in significant improvements over previously known techniques by providing overall, better correction solutions for image contrast issues a photolithographic mask's layout while dramatically reducing feature contour to target error. For instance, application of such techniques for a particular functional silicon die layout considerably reduced the total number of sites originally exhibiting marginal image contrast and significantly improved the contrast in many worst-case low contrast sites when compared with previously known techniques. Moreover, the experimental results were then validated against aon-DLC OPC solutions for an entire chip-set demonstrating validated contrast gains as desired while delivering equivalent performance in both positive and negative feature contour deviations as compared to the previously known techniques not utilizing the Dynamic Learning Corrector (DLC) as to improve image contrast as described herein.

|0071 j Figure 18 describes a process How 101 implementing the Dynamic Learning Corrector (DLC) process from Figure 1 A in greater detail.

j¾072] As shown here, elements .125 (A), 130 (B), and 135 (C) from Figure IB correspond to the same elements 125 (A), 130 (B), and 135 (C) from Figure 1A,

[0073] The process flow 101 starts 165 by performing initial simulatio steps to prepare the mask shape for learning by retrieving the simulated resist contours at approximately correct locations (block 170).

{0074] At block 175, corresponding to element 125 (A), processing then performs the learning simulation operations (e.g., training stage 105) by recordin and storing data related, to image contrast; mask, critical dimensions (CDs) and contour critical dimensions (CDs) while iterative corrections continue.

{0075] At block i SO, corresponding to element 130 (B), processing performs sensitivity analysis on the learning data by determining effectiveness of each segment to improve image contrast from the measured mask data,

10076} Next, at block 185, corresponding to element 135 (C), processing selects the most effective segments to use to respond to improved changes in contrast in targeted problem areas of the mask layout.

{0077} At block 1 0, processing then continues with the OPC simulation using only the most effective segments selected to respond to the image issues associated with So or marginal contrast.

{0078] Processing then ends at block 195.

[0079J Figure 2 describes a decision process flow 200 including a methodology for determining which segment to utilize for mask optimization.

{0080] Consider the geometry at element 125 (A) in which there are two side polygons attacking the longer trench polygon in the center and consequently prod ucing a weak image contrast for the long center trench. At the geometry within the center block at element 130 (B), processing seeks to improve the contrast fay adjustments to the- contours of the center trench, ^"In particular, processing perturbs the nearest neighbors to the weak SIC measured 150 in question residing between the three sets of gauges as depicted at both sides of the center trench . More specifically, the machine learning processing seeks to derive algorithm by which to determine which segment or set of segments should be moved so as to improve the image contrast.

(0081 ) According to one embodiment, processing records the mask Critical Dimension (CD), the intensity, and the contour Critical Dimension (CD), for every set of gauges. Thus, processing measures each Critical Dimension (CD) placement, measures the contour placement, and measures the mask Critical Dimension (CD) as well as the intensity at the weak image site corresponding to the weak SIC as measured at element 150.

10682] Having then identified which segment or pairs of segments causes the largest change in the intensity as its function of changing the mask CD it is then possible to modify the mask Critical Dimension (CD) in a way that improves image contrast and increases light intensity at the weak image site (element 150) while simultaneously altering the mask Critical Dimensions (CDs) as little as possible.

f 083j After a certain number of machine learning iterations (via training stage 105), processing then takes the captured and recorded mfoonatton from the training stage 105 and plots the information to determine which pair of segments will exhibit the strongest response to the image intensity for the smallest change the mask's Critical Dimension (CDs) as learned by the machine learning sensitivity analysis described above.

f00S4] According to certain embodiments, processing will then select a pair of segments having been identified as causing a greatest increase in image contrast with a smallest change in the mask's Critical Dimension (CDs). However, in alternative embodiments, processing additionally or alternatively plots the contour Critical Dimension (CDs) of the mask against the mask's Critical Dimension (CDs) so as to verify that as the contour CDs become larger the corresponding mask Critical Dimension (CDs) also become proportionally larger, thus providing a data validation for the output of the machine learning sensitivity analysis (e.g., training stage 105).

fOOSS] Such a validation scheme verifies that for any contour, modifications to that contour exhibit corresponding movements in the correct dimension within the mask's Critical Dimension (CDs) as well. According to certain embodiments, selecting a pair of segments for use in performing an OPC correction includes selecting a pair of segments corresponding to a small change in contour Critical Dimension (CDs) with a large or greatest increase in image contrast where the selected, pair of segments has been validated by mapping^, the perturbed contours against the mask's Critical Dimension (CDs) to perform the validation as described above.

10086! As depicted here, the decision process flow 200 starts at block 205 where processing advances first to block 210 where processing records, for each gauge during the training period, at least (1) a local simulated image contrast; (if) a local mask and simulated contour critical dimension (CD); and (iii) mask and simulated contour critical dimensions (CDs) for nearby gauges. This processing corresponds to the training period phase as depicted at element 225 (A).

f0087] Within the process flow 200, processing then advances to the decision iteration phase as represented by element 230 (B), within which processing first proceeds to block 215 where processing creates a linear least squares regression between a simulated image contrast of the gauge of interest and the mask critical dimensions of the nearby gauge being scored, thus representing "fit L"

|80&8j Processing next advances to block.220 where processing creates a linear least squares regression between a simula ted con tour critical dimension (CD) of the gauge of interest and the mask critical dimension (CD ) of the nearby gauge being scored, thus representing "fit 2."

{0089] Processing next advances to block 240 where processing assigns a score of zero

(0) to the gauge efficacy if the coefficient of determination for "fit V is less than 0.9.

|0 9 1| Processing next advances to block 245 where processing assigns a score of zero

(0) to the gauge efficacy if the coefficient of determination for "fit 2" is less than 0.9.

{0091] Processing next advances to block 250 where processing assigns a score of zero

(0) to the gauge efficacy if the slope of the regression curve at the current mask critical dimension (CD) for "fit 1 " is positive, whereas at block 255, processing assigns a score of zero to die gauge efficacy if the slope of the regression curve at the c urrent mask critical dimension

(CD) for "t t 2" is negative,

{0092] Otherwise, processing at block 260 will assign a. score to the gauge efficacy equal to the absolute value of the slope of the regression curve from "fit Γ" a the current mask critical dimension (CD).

f8S93} Continuing with the process flow 200, processin then advances to the continued correction phase as represented b element 235 (c), within which processing advances to decision point 265 where processing chooses or selects a segment associated with the gauge having the highest score for performing the further correction, subsequent to which processing then advances to block 270 where processing continues with the OPC simulation and image correcting processes.

{0094] The decision process flow 200 thus describes how the simulation determines the most effective correction for a gauge having been measured or evaluated as having a simulated image contrast (SIC) problem subsequent to the training period represented b -element 225 (A).

10095} Additional algorithms may additionally be applied to such processing, for instance, to further improve or alter the choice regarding which correcting segment is selected at the decision iteration phase as represented by element 230 (B).

100 6} Figure 3 describes an exemplary data set 300 collected for a specified gauge having been measured or evaluated as having a simulated image contrast (SIC) in accordance with described embodiments. In the example data set 300 depicted here, the "Next Gauge" 315 receives the highest score since corresponds to the regression curve with the largest slope at the current mask CD. The current gauge 310 evaluated and the previous gauge 305 are additionally depicted at Figure 3.

10097} Figure 4 provides a schematic comparison simulation results 400 utilizing an Adaptive Local Intensity (ALI) Corrector versus a standard corrector, in accordance with described ernbodi meats.

10098} As shown here at the left most block as represented by element 425 (A) there is an exem lary target geometry with an SIC issue 405.

10099] Within the middle block as represented by element 430 (B), processing utilizes a standard corrector via which processing identifies an SIC issue 455 and corrects to improve the image 10. As depicted, all nearby gauges 470 as represented, by the arrows, are utilized to identify the SIC issue 455 for the purposes of correcting the image, thus leading to the corrected mask shape 465 as represented by the hashing which extends past the original tall rectangle in the middle, and consequently resulting in a contour deviation from the target.460.

100100} Conversely, at the right most block as represented by element 435 (C), processing utilizes an Adaptive Local Intensity (ALI) corrector to correct for the SIC identified by nearby gauges with one gauge being utilized for the corrections as represented by element 415. With the ALI approach, an SIC is identified by all nearby gauges, however, one gauge is used for the correction 485 of the identified SIC issue 455 wh ic h in turn renders the correc ted mask shape 465 and notably, results in a significantly reduced contour deviation 480 from the target, especiall when compared with the standard corrector processing as applied at the center block as represented by element 430 (B).

fOOlO.i^'l Figure 5 describes the application 500 of an Adaptive Adjacent Intensity (AAI) type corrector in accordance with described embodiments.

100102} More particularly, at some initial iteration there is a gauge which requests the simulated image contrast (SIC) from its paired gauge corresponding to a specific target feature 510, If there is an SIC issue 505 (3) present, then for the remaining iterations, t he gauge 515 (2) will continue to request the SIC at its paired gauge 510 (1) and correct for the simulated image contrast (SIC) problem via -continued correction processing.

1001031 According to certain embodiments, neighboring segments are selected arid utilized for perturbing the contour critical dimensions as described above, however, in alternative embodiments, segments for non«ne!ghboring regions are analyzed and ultimately selected for improving the image contrast via an alternative Dynamic Learning Corrector (DLC) process which is described i greater detail below.

1001041 Due to the non-linearity of the optics it is not possible to know a priori which segment will have a greatest impact upon image contrast for a particular SIC issue 505, and that segment may not necessarily he a neighboring segment Therefore, the machine learning sensitivity analysis may be applied to collectively to identify segments on neighboring polygons (e.g. , non-neighboring segments) that exhibit a large impac t to the contrast of the SIC issue 505 at a particular target feature 5.10.

}¾ø tOSj Therefore, the machine learning sensitivity analysis seeks to identif polygons for which a small change to the contours of those polygons results in an increase in image contrast to the SIC issue 505 for the target feature 505 of a different polygon, thus utilizing non- neighboring segments and segments from a second polygon to render a change in image contrast to a target feature at a first polygon.

fU 106| According to another alternative embodiment, specific gauges to be manipulated are configurable and identified via a set of rules. With such an embodiment, the identified gauges are adjusted, for instance, pulled in closer to a particular feature, so as to respond to a polygon within which an image contrast issue is identified, without changing contours of a main polygon identified via the rule set or with minimal changes to the contours of the main polygon identified. In such a way, the image contrast may be improved without changing a main polygon identified by the rule set, just permitting greater flexibility and allowing for manual intervention to the decision making of the machine learning intensity analysis when necessary.

0 107J Accordin to certain embodiments, the Dynamic Learning Corrector (DLC) process includes calculating learnings from a photolithographic mask a the same location as a .geometry of interest and then creating an algorithm history of the mask iterations observed based on the calculated learnings and the machine learning intensity analysis, thus providing a dynamic learning process which changes over time based on the observations. According to such embodiments, the Dynamic Learning Corrector (DLC) process may therefore farther include fine-tuning a specific mask before advancing that particular mask into a production phase.

fOOlOSl Because every mask is unique, it is not appropriate to apply a standardized set of OPC corrections to the different masks, however, by fine-tuning such masks based on

observations collected through the machine teaming intensity analysis it is possible to improve never before seen masks prior to advancing them to production without having to re-run all of the machine learning .intensity analysis from scratch fat every different mask, which could take an infeasible quantity of computer processing power as well as delay the release of a new mask otherwise ready for production.

{00109} In accordance with described embodiments, processing which includ.es the application of an Adaptive Adjacent intensity (AAJ) corrector includes use of a set of dynamic learning corrector (DLC) algorithms that leam during the course of an iterative OPC simulation how the predicted image contrast changes as a result of variation of the mask and predicted wafer features.

100110} The DLCs described herein consist of two types of OPC correction algorithms, either of which may be utilized separately or in conjunction with one another to improve simulated image contrast (SIC) at the photo resist while minimizing changes in the feature contour placements to improve pattern fidelity. Noted above are the Adaptive Adjacent Intensity (AAJ) type correctors. Adaptive Local Intensity (ALI) type correctors are described in greater detail below.

{001111 Although the examples provided here focus specifically on feature pattern contours in th photo resist; the methodologies are nevertheless extendable for use with feature contours simulated based any post-lithographic processing step. As described herein, predicted feature sizes will are referred to by the "feature contours" that are intended to hit a specified drawn target, rather than the more specific "post-develop photo resist feature contour."

{00112} Distinct from the Adaptive Adjacent Intensity (AAI) type correctors, the Adaptive Local intensity (ALi) type correctors provide a mechanism to identify low SIC sites during an iterative OPC simulation and respond with only the most effective mask section changes necessary so as to minimize feature contour deviations from the target feature 5.10. Adaptive Local intensity (ALI) type correctors are implemented by dividing the mask target shape into segments that are then moved to optimize the mask shaped based on local simulations of the photo resist responses such as the SIC and the feature contou placement.

|001131 Local O C simulations at a region defined by a gauge on the feature contour target (e.g., a measurement node) associated with each segment of the mask is performed by iteraiively simulating responses around the gauge 515 (2) and perturbing the segment position in response.

{00114} Previously utilized OPC techniques utilized gauges that operate in isolation of one another and then report simulated results to the segment or segments associated to the isolated gauge, rather than facilitating inter-gauge communication as described herein with respect to the AAI and ALi type correctors. fOOi lSJ OPC methodologies described herein therefore improve upon prior known solutions by enabling and utiiMng connmtnications between gauges, thus^■facilitating the gauges to communicate with one another and determine the best correction approach based on a learning phase.

{00116} The correction formulation for the ALE corrector as depicted at Figure 1 A above, involves three fundamental operations, information collection during a training period (at element 125 (A), a decision making phase at element 130 (B) to determine the most effective mask correction scheme base on the data collected at (A), and final mask correction via conttmied correction at element 135 (C) to meet image (SIC) and contour tolerances.

JOOIl 7} During the training phase, each gauge to whic the AO corrector is assigned records the measured local SIC along with the local measured feature contours and mask shape. During this training period no correction, for the SIC is_. performed and the segments ate constrained to responding to .feature contours. Along wit this information, local feature contours and mask shape are recorded for nearest neighbor gauge locations as well Such an operation therefore utilizes the described inter -gauge commimication during the OPC optimization processing. According to certain embodiments, the ALI connector implementation records information from only its nearest neighbor gauges while in alternative embodiments, the ALI connector may instead record general training information can from any arbitrary distance from the gauge of interest. Such distances may be configurable as part of a tuning or optimization process for the OPC simulation and optimization processing.

fOOI.lSJ Broadly speaking, the Dynamic Learning Corrector (DLC) provides a mechanism for an attacker neighbor feature that worsens SIC on a victim feature to respond to help the victim feature. More specifically, application of either of the DLC corrector algorithms facilitates inter-gauge communication between gauges on different target features 510 (1) so as to improve the overall image contrast.

00119| Use of the DLC AAI corrector type algorithms is therefore similar to the ALI corrector technique described above in which one gauge is associated to many gauges on other features. Application of an ALI connector may therefore use a similar machine learning approach.

00120] Association between gauges on nearby features is done through the existing target search gauge association methodology. Once gauge association between features is complete, the corrector operates by requesting the SIC from each gauge's associated target feature 51.0 (1 ) gauge on a specified iteration after some fixed run-in period. If an SIC issue 505 (3) is identified, such as a low SIC, the the corrector is executed for the -remaining iterations of the simulation. [091211 Optionally, the AA.T corrector can be configured to change the tolerance of gauges identifying a low SIC as measured by their associated gauge. This allows the corrector to only act where needed to respond to the SIC issue identified within a given tolerance, and as a result, allows for possibly better feature contour convergence elsewhere.

{99122] Procedures to determine the size of the Looser tolerance ma involve, for example, loosening the tolerance of all responding ga uges to a fixed value or computing the required mask shape change required to improve the SIC above some fixed or dynamic threshold contrast.

|09t23] As with the AAI type connector approach, the Adaptive Local Intensity (ALl) corrector formulation is also ^'distinct from prior known methodologies as it specifically utilizes inter-gauge communication of SIC information and therefore allows mask segments to respond to a weak image In a feature to which the segment is no associated.

]09124] Figure 6 depicts an exemplary Artificial ^'Neural Network (ANN) 605 machine learning algorithm in which the input variables predict the deterministic model error or process driven changes in critical dimension in accordance with described embodiments.

{00125] More particularly, there is depicted on the left an initial OPC contour 620 represented by the bold line and outermost ellipse and then a second contour is further depicted via the inner ellipse showing the Scanning Electron. Microscope (SE ) contour 625 representing fabrication data as taken from an SEM image.

{00126] On the right is a neural network 605 having as input several input image parameters 610 which are calculated using an OPC model and the reselling output from the neural network is a contour shift prediction 61 . In accordance with one embodiment the image input parameters 610 used to train the neural network 605 are simulated optical parameters representing the entirety of the feature set for use as the input image parameters 10,

{00127] Any specific layout will have a deterministic OPC model error. Neural networks, such as Artificial Neural Network (ANN) 605, are able to approximate any arbitrary formula. Therefore, in accordance with described embodiments, the neural network 605 is trained to describe the relationship between the layout and the respective OPC model error, usin the image input parameters 10 to enable the neural network 60S to learn about the layout, A sufficient quantity of image input parameters 63.0 are provided so as to enable such learning.

[09128] A simulated optical image from OPC modeling is obtained from the convolution of the layout and an optical transfer function. According to described embodiments, image parameters are defined via optical images that capture information about the layout, which is then used to train the neural network 605 to describe the deterministic relationship between the layout and conventional OPC model error. {06129] Through she a pplication of machine learning techniques using, for instance, the depicted Artificial Neural Network (ANN) 605 as well as Convoktional Neural Networks (CNN), it is possible to train OPC models capable of compensating for discrepancies between simulated semi-physical models and actual physical measurements taken from SEM images. Using such compensation, the OPC models are made to yield improved predictive results with reduced EPE at various process steps, resulting in greatly improved image contrast which will in turn help with meeting the technology requirement of that node, such as placement of features within design tolerances.

{00136] Within the context of machine learning, an Artificial Neural Network (ANN) 605 is a computational approach which is based on a large collection of .neural units loosely modeling the way a biological brain solves problems with large clusters of biological neurons connected by axons. Each neural unit is connected with many others, and links can be enforcing or inhibitory in their effect on. the activation state of connected neural units,. Such systems are self-learning and trained rather than explicitly programmed and they excel in areas where the solution or feature detection is difficult to express in a traditional computer program.

{60131] A so-called convolutional neural network (CNN, or ConvNet) is a type of feedforward artificial neural, network i» which the connectivity pattern between its neurons is inspired by the organization of the animal visual cortex. Individual cortical neurons respond to stimuli in a restricted region of space known as the receptive field. The receptive fields of different neurons partially overlap such that they tile the visual field. The response of an individual neuron to stimuli within its receptive field can be approximated mathematically by a convolution operation.

{00132] According to certain embodiments, geometric information from the layout is additionally utilized to directly train the neural network to predict the amount of model error associated with the geometry. Once trained, the neural network 605 provides a more accurate model which can then be implemented by shifting the initial OPC contour 620 by an amount predicted by neural network 605.

{60133] The OPC" model provides a forward function which connects what is on the mask to what is on the wafer. Software algorithms provide a basic physics solution to this problem, but the solution requires many approximations which thus operates as a source of inaccuracies. Described embodiments therefore reformulate the problem in such a way that a list of features is provided to the neural network for the purposes of teaming. Through training based on such input features, such as those derived from the depicted input image parameters, the neural network learns what adaptations are necessary to conform the base OPC models to known physical models. These adaptations are output, as the contour shift predictions 615 and result in a semi-physical model which permits formulas and conⁿectio s by which the trained neural network 605 describes the differences between the incoming base OPC model and the observed physical realities.

|00134| Because the model, the structures, the features, and the mask is so complex it simply is not practical to manually determine what function or adaptations to the incoming base OPC" model axe necessary to conform that model to the reality as observed in the SEM imagery taken from fabricated silicon wafers utilizing the base OPC model. Use of the neural network 605 to apply machine learning therefore provides a significant advantage as the neural network is leveraged to determine the complexity and learn the necessary adaptations.

|00!35} In accordance with certain embodiments, the neural network 605 predicts the deterministic model error or process driven changes in critical dimensions for the size and position of feaiutes. According to a particular embodiment, the neural network provides contour fitting to the SEM contour 625 image representing the actually observed physical outputs from a fabrication process utilizing the base OPC model which provides the initial OPC contour 620.

1001361 i accordance with certain embodiments, the initial OPC contour 620 is a result of the software algorithms which provides a prediction of the physical space via a serai-physical model According to such an embodiment, SEM image data is then collected for the patterns to generate or determine the SEM contour 625 representing the fab data for actual physical samples of fabricated physical silicon wafers generated using the initial OPC base model.

{00137} According to such an embodiment, the delta between the initial OPC contour

620 and the SEM con tour 625 is the determined model error of the base OPC model. The neural network 605 is trained to be able to predict the determined model error of the base OPC model so as to output an improved model using the contour shift, prediction 615 provided by the trained neural network 605,

{00138} Figure 7 is a schematic of a computer system 700 in accordance with described embodiments. The computer system 700 (also referred to as the electronic system 700) as depicted can embody means for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) o lithograph masks, according to any of the several disclosed embodiments and their equivalents as set forth in. this disclosure. The computer system 700 may be a mobile device such as a net-book computer. The computer system 700 may be a mobile device such as a wireless smartphone or tablet. The computer system 700 ma be a desktop computer. The computer system 700 may be a hand-held reader. The computer system 700 may be a server system. The computer system 700 may be a supercomputer or high-performance computing system,

{00.139} In accordance with one embodiment, the electronic system 700 is a computer system that includes a system bus 720 to electrically couple the various components of the electronic system 700. The system b s 720 is a single bus or any combination of bosses according to various embodiments. The electronic system 700 includes a voltage source 730 that provides power to the integrated circuit 710. In some embodiments, the voltage source 730 supplies current to the integrated, circuit 7 i 0 through the system bus 720.

{001401 Such ait integrated circuit 710 is electrically coupled to the system bus 720 and includes any circuit, or combination of circuits according to art embodiment. In an embodiment, the integrated circ uit 710 includes a processor 7 ! 2 that can be of any type. As used herein, the processor 712 may mean any type of circuit such as, but not limited to, a microprocessor, a microcontroller, a graphics processor, a digital signal processor, or another processor. In an embodiment, the processor 712 includes, or is coupled with, electrical devices having gradient eocapsulant protection, as disclosed herein.

180141 J In accordance with one embodiment, SRAM embodiments are found in memory caches of the processor. Other types of circuits that, can be included in the integrated circuit 710 are a custom circuit or an application-specific integrated circuit (ASIC), such as a

communications circuit 714 for use in wireless devices such as cellular telephones, smart phones, pagers, portable computers, two-way radios, and similar electronic systems, or a communications circuit for servers, in an embodiment, the integrated, circuit 710 includes on-die memory 716 such as static random-access memory (SRAM). In an embodiment, the integrated circuit 710 includes embedded on-die memory 716 such as embedded dynamic random-access memory (eDRAM).

{001421 " accordance with one embodiment, the integrated circuit 10 is complemented with a subsequent integrated circuit.71 1. Useful embodiments include a dual processor 713 and a dual communications circuit 715 and dual on-die memory 717 such as SRAM, in accordance with one embodiment, the dual integrated circuit 710 includes embedded on-die memory 717 such as eDRAM.

{00143} In one embodiment, the electronic system 700 also includes an external memory 740 that in turn may includ one or more memory elements suitable to the particular application, such as a main memory 742 in the form of RAM, one or more hard drives 744, and/or one or more drives that handle removable media 746, such as diskettes, compact, disks (CDs), digital variable disks (DVDs), flash memory drives, and other removable media known in the art. The external memory 740 may also be embedded memory 748 such as the first die in a die stack, according to an embodiment.

{001441 h* accordance with one embodiment, the electronic system 700 also includes a display device 750 and an audio output 760. in one embodiment, the electronic system 700 includes an input device 770 such as a controller thai may be a keyboard, mouse, trackball, game controller, microphone, voicenrecognMon device, or any- other input device that inputs information into the electronic system 700. In an embodiment, an input device 770 is a camera. In an embodiment, an input device 770 is a digital sound recorder, in an embodiment, an input device 770 is a camera and a digital sound recorder.

{001451 As shown herein, the integrated circuit 7] 0 can be. implemented in a number of different embodiments, including means for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks for a semiconductor substrate package, according to any of the several disclosed embodiments and their equi alents, an electronic system, a computer system, one or more methods of fabricating an integrated circuit, and one or more methods of fabricating an electronic assembly that includes a package substrate or a semiconductor package having therein means for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks, according to any of the several disclosed embodiments as set forth herein in the various embodiments and their art-recognized equivalents. The elements, materials, geometries, dimensions, and sequence of operations can all be varied to sui t particular I/O coupling requirements including arra contact count, array contact configuration for a microelectronic die embedded in a processor mounting substrate according to an of the se veral disclosed package substrates and semiconductor packages having means for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks for semiconductor substrate package embodiments and their equivalents. A foundation substrate 798 may be included, as represented by die dashed line of Figure 7. Passive devices 799 may also be included, as is also depicted in Figure 7.

(001 6} Figure $ illustrates a semiconductor device 800 (or an i terposer) that includes one or more described embodiments. The interposer 800 is an intervening substrate used to bridge a first substrate 802 to a second substrate 804. The first substrate S02 may be, for instance, an integrated circuit die. The second substrate 804 may be, for instance, a memory module, a computer motherboard, or another integrated circuit die. Generally, the purpose of an interpose §00 is to spread connection to a wider pitch or to reroute a connection to a different connection. For example, an interposer 800 may couple an integrated circuit die to a ball grid array (BGA) 806 that can subsequently be coupled to the second substrate 804. In some embodiments, the first and second substrates 802/804 are attached to opposi ng sides of the interposer 800, In other embodiments, the first and second substrates 802/804 are attached to the same side of the interposer 800. And in further embodiments, three or more substrates are interc nnected by way of the interposer 800. f00147| The interposer 800 may be formed of an epoxy resin, a fiberglass-reinforced epoxy resin, a ceramic material, or a polymer material such as polyimide. to further

implementa tions, the interposer may be formed of alternate rigid or flexible materials that may include the same materials described above for use in a semiconductor substrate, such as silicon, germanium, and other group Ill-V and group IV materials,

(00148J The interposer may include metal · interconnects 808 and vias 810, including but not limited to through-silicon vias (TSVs) 812. The interposer 800 may farther include embedded devices 814, including both passive and active devices. Such devices include, but are not limited to, capacitors, decoupling capacitors, resistors, inductors, fuses, diodes, transformers, sensors, and electrostatic discharge (ESD) devices. More complex devices such as radio- frequency (RF) devices, power amplifiers, power management devices, antennas, arrays, sensors, and MEMS devices may also be formed on the interposer 800. In accordance with described embodiments, apparatuses or processes disclosed herein may be used_, in the fabrication of interposer 800.

1001 1 Figure 9 illustrates a computing device 900 in accordance with one

implementation of the invention. The computing device 900 houses a board 902. The board 902 may include a number of components, including but not limi ted to a processo 904 and at least one communication: chip 906. The processor 904 is physically and electrically coupled to the board 902. in some implementations the at least one communication chip 906 is also physicall and electrically coupled to the board 902. to further implementations, the communication chip 906 is part of the processor 904.

fOOlSOf Depending on its applications, computing device 900 may include other components that may or may not be physically and electrically coupled to the board 902. These other components include, but are not limited to, volatile memory (e.g., DRAM), non-volatile memory (e.g.. ROM), flash memory, a. graphics processor, a digital signal processor, a crypto processor, a chipset an antenna, a display, a touchscreen display, a touchscreen controller, a battery, an audio codec, a video codec, a power amplifier, a global positioning system (GPS) device, a compass, an accelerometer, a gyroscope, a speaker, a camera, and a mass storage devic (such as hard disk drive, compact disk (CD), digital versatile disk (DVD), and so forth).

00151} The communication chip 906 enables wireless communications for the transfer of data to and from the computing device 900. The term "wireless" and its derivatives may be used to describe circuits, devices, systems, methods, techniques, communications channels, etc., that, may communicate data through the use of modulated electromagnetic radiation through a non-solid medium. The term does not imply that the associated devices do not contain any wires, although in some embodiments they might not. The communication chip 906 may implement any of a number of wireless standards or protocols, including but .not limited to Wi-Fi (ΪΕΕΕ 8 )2.1 1 familyK WiMAX (IEEE 802.16 family), IEEE 802.20, long term evolution (LTE), Ev- DO, HSPA+, HSDP.A+, HSUPA+ EDGE, GSM, GPRS, CDMA, TDMA, DECT, Bluetooth, derivatives thereof, as well as any other wireless protocols that axe designated as 3G, 4G, 5G, and beyond. The computing device 900 may include a ploxality of communication chips 906. For instance, a first communication chip 906 may be dedicated to shorter range wireless

communications such as Wi-Fi and Bluetooth arid a second communication chip 906 may be dedicated to longer range wireless communications such as GPS, EDGE, GPRS, CDMA,

WiMAX, LTE, Ev-DO, and others.

|00!52} The processor 904 of the computing device 900 includes an integrated circuit die packaged within me processor 904. In some implementations of the invention, the integrated circuit die of the processor includes one or more devices, such as MOS-FET transistors built in accordance with implementations- of the invention. The term "processor" .may refer to any device or portion of a device that processes electronic da ta from registers and/or memory to transform that electronic data into other electronic data that may be stored in registers and/or memory.

{001531 Tbe communication, chip 906 also includes an integrated circuit die packaged within the communication chip 906. In accordance with another implementation of the invention, the mtegrated circuit die of the communication ^'chip- includes one or more devices, such as .MOS- FET transistors built in accordance with implementations of the invention.

|0<M54| In further iirtplementations, another component housed within the computing device 900 may contain an integrated circuit die that includes one or more devices, such as

MOS-FET transistors built in accordance with implementations of the invention.

100155J In various implementations, the computing device 900 may be a laptop, a netbook, a notebook, an uitrabook, a smartphone, a tablet, a personal digital assistant (PDA), an ultra, mobile PC. a mobile phone, a desktop computer, a server, a printer, a scanner, a monitor, a set-top box, an entertainment control unit, a digital camera, a portable music player, or a digital video recorder, to further implementations, the computing device 900 may be any other electronic device that processes data.

{001561 Figure 10 is a flow diagram illustrating method 1000 for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks in accordance with described embodiments. Some of the blocks and/or operations listed below are optional in accordance with certain embodiments. The numbering of the blocks presented is for the sake of clarity and is not intended to prescribe an order of operations in which th various blocks must occur. Additionally, operations from method 1000 may be utilized in a variety of combinations. j 00 57} At block 1005, the .method .1000 for implementing dynamic learning mask correction for resolution enhancement and Optical Proximity Correction (OPC) of lithography masks operates via the following processes.

{00158] At block 1010, the method includes creating a mask via a lithography process. {00159] At block 1015, the method includes performing a Seaming phase to identify a set of mask shape variables that change simulated intensity values of the mask and a simulated shape of the mask by iterative!}' moving mask segments.

{00160] At block 1.020, the method includes collecting data representing image contrast of the mask, mask critical dimension (CD ), and contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments.

{001611 At Mock 1025, the method includes performing sensitivity analysis on the collected data to determine which mask segments correspond to an improved simulated image contrast of the mask when mo ved,

{001621 At block 1030, the method includes selecting one or more mask segments that correspond to the improved simulated image contrast of the mask when moved.

{001631 At block 1035. the method includes applying Optical Proximity Correction (OPC) for a model of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when .moved.

{00164} While the subject matter disclosed herein has been described by way of example and in terms of the specific embodiments, it is to be understood that the cl aimed embodiments are not limited to the explicitly enumerated embodiments disclosed. To the contrary, the disclosure is intended to cover various modifications and similar arrangements as would be apparent to those skilled in the art. Therefore, the scope of the appended claims should be accorded the broadest interpretation so as to encompass all such modifications and similar arrangements. It is to be understood that, the above description is intended to be illustrative, and not restrictive. Many other embodiments will be apparent to those of skill in the art upon reading and understanding the above description. The scope of the disclosed subject matter is therefore to be determined in reference to the appended claims, along with the Ml scope of equivalents to which such claims are entitled.

{00165] it is therefore in accordance with the described embodiments, that:

{0 166| According to one embodiment there is a method for reducing Optical Proximity Correction (OPC) .model error, in which the method includes: simulating a mask for a corresponding lithography process; performing a learning phase to identify a set of mask shape variables that, change simulated intensity values of the mask and a simulated shape of the mask by !teratively moving mask segments; collecting data representing image contrast of the mask, mask critical dimension (CD), and contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments; performin sensitivity analysis on the collected date to determine which mask segments correspond to an improved simulated image contrast of the mask when moved; selecting one or more mask segments that correspond to the improved simulated image contrast of the mask when moved ; and applying Optical Proximity Correction (OPC) for a model of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

J00167] According to another embodiment of the method, performing the learning phase includes two sub-phases including an initialization sub-phase and an iteration sub-phase; and in which the initialization sub-phase includes performing an initial OPC simulation to prepare the shape of the mask for learning the set of mask shape variables by retrieving simulated photolithographic resist contours derived from the mask,

|00168] According to another embodiment of the method, the initialization sub-phase provides rudimentary correction to the shape of the mask prior to iteratively moving the mask segments to learn the set of mask shape variables.

|00I6 1 According to another embodiment of the method, the iteration sub-phase inc ludes recording the retrieved simulated photolithographic resist contours deri ved from the mask and associated mask shape changes for each segments; and in which iteratively movin mask segments d uring the performin of the teamin phase includes iterati vely moving the segments in a collective, independent, or random manner to leant how the resist contours and simulated image contrast respond to altering the position of the moved segments and respond to changes in the mask critical dimensions.

|0Θ170] According to another embodiment, the method further includes; collecting slope plot data representing how the simulated intensity profil e of the mask change in relation to the mask critical dimension (CD) and mask segment positions; and in which performing the sensitivity analysis on the collected data includes fitting the collected plot data to the mask critical dimension (CD) to determine, for each moved mask segment, whether the moved mask segments contributes to improved simulated image contrast.

100171] According to another embodiment, the method further includes: collecting slope plot data representing how the simulated intensity values of the mask change in relation to the mask critical dimension (CD); and in which the method further includes fitting the plot data to the .mask critical dimension (CD) by: creating a linear least squares regression between the simulated image contrast for a location of interest (gauge) and the mask critical dimension (CD) of a nearby gauge being scored as a first response fi tting of the plot data; creating a first linear least squares regression between the simulated image contrast for a gauge of interest and the mask critical dimension (CD) of a nearby gauge being scored as a first fitting; creating a second linear least sqisares regression between a simulated contour criticat dimension (CD) of the gauge of interest and the mask critical dime nsion (CD) of the nearb ga uge being scored as a second fitting; assigning a reduced score or a score of zero (0) to a gauge efficacy variable if a coefficient of determination from the first fitting is less than a default value of 0.9 or less than a configurable threshold or function value; assigning a reduced score or a score of zero (0) to the gauge efficacy variable if the coefficient o ^'determination from the second fitting is less than the default value of 0.9 or less than the configurable threshold or -function value; assigning a reduced sc ore or a score of zero (0) to the gauge effi cacy variable when either (i) the slope of the regression curve at the mask critical dimension (CD) for the first fitting is positive or (ii) the slope of the regression curve at the mask critical dimension (CD) for the second fitting is negative; and assigning a score to the gauge efficacy variable equal to an absolute value of the slope of the regression -curve from the first fitting for the mask critical dimension (CD) when both (i) and (ii) are false.

1001721 According to another embodiment of the method, selec ting the one or more mask segments that correspond to the improved simulated image contrast of the mask when mask segment moved includes: selecting a mask segment which, when moved, corresponds to the greatest increase i simulated image contrast and the smallest change to any contour critical dimensions (CDs) of the mask.

{09173} According to another embodiment of the method, selecting the one or more mask segments thai correspond to the improved simulated image contrast of the mask when moved includes: selecting a mask segment f om a first polygon which, when moved, corresponds to the greatest increase in simulated image contrast for a gauge within a second, polygon, different than the first polygon.

{00174} According to another embodiment of the method, selecting the one or more mask segments that correspond to the improved simulated image contrast of the mask when moved includes: selecting a mask segment associated with a gauge evaluated as having a simulated image contrast (SIC) below a threshold which, when moved, corresponds to a increase in the simulated image contrast (SIC) for tire gauge in excess of a threshold and which, when moved, the mask segment results in a change to a simulated contour critical dimension (CD) of the gauge and a change to the mask critical dimension (CD) below a threshold.

{00175} According to another embodiment, the method further includes: iteiatively moving mask segments and evaluating for each moved mask segment if a simulated image contrast for a resist feature of interest (target feature) increases; and repeating {he moving of mask segments in different amounts and collecting the simulated image contrast data at the target feature; and evaluating which mask segment movement corresponds to the greatest increase in the simulated image contrast at the target feature with a deviation of simulated contour critical dimension (CD) of the target feature from a targeted contour critical dimension (CD) of the target feature below a threshold.

[00176} According to another embodiment of the method , applying the Optical

Proximity Correction (OPC) for the model of the mask using the one or more se lected mask segments reduces or eliminates scumming or other image contrast related defects of a photo resist exposed via the OPC corrected model of the mask.

100177} According to another embodiment, the method further includes: fabricating a new mask using the OPC corrected mode! of the mask; exposing a photo resist using the new mask, in which the photo resist exhibits the mask critical dimension (CD) and the contour critical dimensions (CDs) within tolerances for patterning; and etching the exposed photo resist, in whic the etched photo resist is defect free and residue tree at a silicon surface beneat the etched photo resist,

[00178] According to another embodiment, the method further includes: fabricating a physical silicon wafer using the OPC corrected mask, the physical silicon wafer having a plurality of features embodied therein as defined by the OPC corrected mask.

00179} According to another embodiment, the method further includes: creating a semi- physical simulation framework of the mask using an existing semi-physical model of the lithography process used to create the mask, the semi-physical model specifying optical intensity values representing the plurality of features of the mask; and in which iteralive!y moving mask segments includes shifting contours of the plurality of features of the mask as represented by the optical intensity values within the semi-physical model and collecting the data representing image contrast of the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments to determine if the movement of a mask segment corresponds to an increase in simulated image contrast (SIC ) and whether changes to the mask critical dimension (CD) and the contour critical dimensions (CDs) of the mask fells within a specified tolerance.

[00180] According to another embodiment, the method further includes: training a neural network to describe a relationship between the changes to the segments of the mask and the change in simulated image contrast (SIC) for the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask, in which the neural network includes an Artificial Neural Network (ANN) to algorithmkatly represent the relationship between the changes to the segments of the mask and the change in simulated image contrast (SIC) for the mask, the mask criti cal dimension (CD), and the contour critical dimensions (CDs) of the mask: training the neural network to output offset predictions for the selected one or mote mask segments that correspood to the improved, simulated image contrast of the mask when mo ved, in. which the output offset predictions are to he applied to a new semi-physical simulation via Optical Proximity Correction (OPC); and fabricating a new OPC corrected mask from the new semi-physical simulation.

| 0181 J According to yet another embodiment, there is a system to reduce Optical Proximity Correction (OPC) model error, in which the system includes: a processor and. a memory; a mask created via a lithography process; machine learning logic to execute via the processor and the memory, in which the machine learning logic is to perform a learning phase to identify a set of mask shape variables that change simulated intensity values of the mask and a simulated shape of the mask by tteratively moving mask segments; the machine learning logic to collect and store data representing image contrast of the mask, mask, critical dimension (CD), and contour critical dimensions (CDs) of the mask for each, iteration of moving the mask segments; the machine learning logic to perform sensitivity analysis on the collected data to determine which mask segments correspond to an improved simulated image contrast of the mask when moved; the machine learning logic to select one or more mask segments that correspond to the improved simulated image contrast of the mask when moved; and an Optical ftoximity

C orrec tion (OPC) simul ator to apply OPC correction to a model of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

{00.182] According to another embodiment of the system, performing the learning phase includes two sub-phases including an initialization sub-phase and an iteration sub-phase: and in which the initialization sub-phase includes performing an initial simulation to prepare the shape of the mask for learning the set of mask shape variables by retrieving simulated

photolithographic resis contours derived from the mask.

100183] According to another embodiment of the system, the machine learning logic to select the one or more mask segments that correspond to the improved simulated image contrast of the mask when moved includes the machine learning logic to select a mask segment which, when moved, corresponds to a greatest increase i simulated image contrast and a smallest change to any contour critical dimensions (CDs) of the mask.

{00184] According to another embodiment of the system, machine learning logic to apply the Optical Proximi ty Correction (OPC) for the model of the mask using the one or more selected mask segments reduces or eliminates scumming or other image related resist defect , of a photo resist exposed via the OPC corrected model of the mask.

{00185] Accordin to another embodiment, the system further includes: a new mask fabricated using the OPC^■corrected model of the mask; a photo resist exposed using the new m sk, in which the photo resist exhibits the mask critical dimension (CD) and the contour critical dimensions (CDs) within tolerances for patterning; and in which etching the exposed photo resist results in a defect free and residue free silicon surface beneath the etched photo resist.

{60186} According to another embodiment, the system further includes: an existing semi -physical mode l of the mask created using phy sical parameters of the lithography process used to create the mask, the semi-physical model specifying optical intensity values representing the plurality of features of the mask; and in which the machine learning logic to iteratively move the mask segments includes the machine learning logic to shift mask segments, thereby changing contours of the plurality of features of the mask as represented by the optical intensity values within the semi-physical model and collect the data representing image contrast of the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask for each itera tion of -moving the mask segments to determine if the mo vement of a mask segment corresponds to art increase in simulated image contrast (SIC) and whether changes to the mask critical dimension (CD) and the contour critical dimensions (CDs) of the mask falls within a specified tolerance.

fO tS?! According to a particular embodiment, there is a non-transitor computer readable storage media having instructions stored thereupon that, when executed by a processor, the instructions cause the processor to perform operations for reducing Optical Proximity

Correction (OPC) model error, in which operations include: creating a mask for a lithography process; performing a learning phase to identify a set. of mask shape variables that change simulated intensity values of the mask and a simulated shape of the mask by iterativel moving mask segments; collecting data representing image contrast of the mask, mask critical dimension (CD), and contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments; performing sensitivity analysis on the collected data to determine which mask segments correspond to an improved simulated image contrast of the mask when moved;

selecting one or more mask segments that correspond to the improved simulated image contrast of the mask when moved; and applying Optical Proximity Correction (OPC) for a model of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

{06188} According to another embodiment of the non-transitory computer readable storage media, performing the learning phase includes two sub-phases including an initialization sob-phase and an iteration sub-phase; in which the initialization sub-phase includes performing an initial simulation to prepare the shape of the mask for learning the set of mask shape variables by retrieving simulated photolithographic resist contours derived from the mask and iteratively moving mask segments to reduce initial model error; in which the iniiiafe tion sub-phase provides a rudimentary correction to the shape of the ffiasfc. prior iterati vely moving the mask segments to learn the set of mask shape variables; in which the iteration sub-phase includes recording the retrieved simuiated photolithographic resist contours derived from the mask into the segments; and in which iterativel moving mask segments during the performing of the learning phase includes iterati vely moving the segments in a collective, independent, or random manner to learn how the mask changes responsive to altering the position of the moved segments and responsive to changes in the mask critical dimensions,

{00189] According to another embodiment of the non- transitory computer readable storage media, selecting the one or more .mask segments that correspond to the improved simulated image contrast of the mask when moved includes: selecting a mask segment which, when moved, corresponds to a greatest increase in. simulated image contrast and a smallest change to any con tour critical dimensions (CDs) of the mask.

Claims

SYSTEMS, METHODS, AND APPARATUSES FOR IMPLEMENTING DYNAMIC LEARNING MASK CORRECTION FOR RESOLUTION ENHANCEMENT AND OPTICAL PROXIMITY CORRECTION (OPC) OF LITHOGRAPHY MASKS CLAIMS

What is claimed is;

S , A method for reducing Optical Proximity Correction (OPC) mode! error, whereto the me&od comprises:

simulating a mask for a corresponding lithography process;

performing a. learning phase to identify a set of mask shape variables that change simulated intensity values of the m ask and a simulated shape of the mas k by iteratively moving mask segments;

collecting data representing image contrast of the mask, mask critical dimension (CD), and contour critical dimensions (CDs) of tire mask for each iteration of moving the mask segments;

performing sensitivity analysis on the collected data to determine which mask segments

correspond to an improved simulated image contrast of the mask when moved;

selecting one or more mask segments that, correspond to the improved simulated image contrast f the mask when moved; and

apply ing Optical Proximity Correction ( OPC) for a model of the m ask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when mo ved,

2, The method of claim. 1 :

wherein performing the learning phase comprises two sub-phases including an initialization, sub- phase and an iteration sub-phase: and

wherein the initialization sub-phase includes performing an initial OPC simulation to prepare the shape of the mask for learning the set of mask shape variables by retrieving simulated photolithographic resist contours derived from the mask.

3, The method of claim 2, wherein the initialization sub-phase provides a rudimentary correction to the shape of the mask prior to iteratively moving the mask segments to learn the set of mask shape variables.

4, The method of claim 2;

wherein the iteration sub-phase comprises recording the retrieved simulated photolithographic resist contours derived from the mask and associated mask shape changes for eac segments; and

wherein itera&vely moving mask segments during the performing of the learning phase

comprises iteratively moving the segments in a collective, independent, or random manner to leara how the resist contours and simulated image contrast respond to altering the position of the moved segments and respond to changes in the mask critical dimensions,

5. The method of claim I :

wherein collecting the data further comprises collecting slope plot data representing how the simulated intensity profile of the mask change in relation to the mask critical dimension (CD) and mask segment positions; and

wherein performing the sensitivity analysis on the collected data comprises fitting the collected plot dat to the mask critical dimension (CD) to determine, for each moved mask segment, whether the moved mask segments contributes to improved simulated image contrast.

6. The me thod of c laim 1 :

wherein collecting the data further comprises collecting slope plot data representing how the simulated intensity values of the mask change in relation to the mask critical dimension. (CD); and

wherein the method further comprises fitting the plot data to the mask critical dimension (CD) by:

creating linear least squares regression between the simulated image contrast for a location of interest (gauge) and the mask critical dimension (CD) of a nearby gauge being scored as a first response fitting of the plot data;

creating first linear least squares regression between the simulated image contrast for a gauge of interest and the mask critical dimension (CD) of a nearby gauge being scored as a first fitting;

creating a second linear least squares regression between a simulated contour critical dimension (CD) of the gauge of interest and the mask critical dimension. (CD) of the nearby gauge being scored as a second fitting;

assigning a reduced score or a score of zero (0) to a gauge efficacy variable if a coefficient of determination from the first fitting is less than a default value of 0.9 or less than a configurable threshold or function value;

assigning a reduced score or a score of zero (0) to the gauge efficacy variable if the coefficient of determination from the second fitting is less than the default value of 0.9 or less than die configurable threshold or function value; assigning a reduced score or a score of zero (0) to the gauge efficacy variable when either (i) the slope of the regression curve at the mask critical dimension- (CD) for the first fittin is positive or (ii) the slope of the regression curve at the mask critical dimension (CD) for the second fitting is negative; and

assigning a score to the gauge efficacy variable equal to an absolute value of the slope of the regression curve from the first fitting for the mask critical dimension (CD) when both (i) and (it) are false.

7. The method of claim l_f wherein selecting the one or more mask segmen ts that correspond to the improved simulated image contrast of the mask when mask segment moved comprises:

selecting a mask segment which, when moved, corresponds to the greatest increase in simulated image contrast and the smallest change to any contour critical dimensions (CDs) of the mask.

8. The method of claim I, wherein selecting the one or more mask segments that correspond to the improved simulated image contrast of the mask when moved comprises:

selecting a mask segment from a first polygon which, when moved, corresponds to the greatest increase in simulated image contrast for a gauge within a second polygon, different than the first polygon,

9. The method of c laim 1, wherein selecting the one or more mask segments that correspond to the improved simulated image contrast of the mask when moved comprises:

selecting a mask segment associated with a gauge evaluated as having a simulated image contrast (SIC) below a threshold which, when moved, corresponds to a increase in the simulated image contrast (SIC) for the gauge in excess of a threshold and which, when moved, the mask segment results in a change to a simulated contour critical dimension (CD) of the gauge and a change to the mask critical dimension (CD) below a threshold.

10. The method of claim 1, wherein iterattvely moving mask segments comprises:

iterativeiy moving mask segments and evaluating for each moved mask segment if a simulated image contrast for a resist feature of interest (target feature) increases; and

repeating the movin of mask segments in different amounts and collecting the simulated imag contrast data at. the target, feature; and

evaluating which mask segment movement corresponds to the greatest increase in the simulated image contrast at the target feature with a deviation of simulated contour critical

dimension (CD) of the target feature from a targeted contour critical dimension (CD) of the target feature below a. threshold.

1 1. The method of claim 1, wherein applying the Optical Proximity Correction (OPC) for the model of the mask using the one or more selected mask segments reduces or eliminates scumming or other image contrast related defects of a photo resist exposed via the OPC corrected model of the mask.

12. The method of claim \ , further comprising:

febricating a new mask using the OPC corrected model of the mask;

exposing a photo resist using the new mask, wherein the photo resist exhibits the mask critical dimension (CD) and the contour critical dimensions (CDs) within tolerances for patterning; and

etching the exposed photo resist, wherein the etched photo resist is defect free and residue fee at a silicon surface beneath die etched photo resist,

13. The method of claim 1, further comprising:

fabricating a. physical silicon wafer using the OPC corrected mask, the physical silicon wafer having a plurality of features embodied therein as defined by the OPC corrected mask.

14. The method of claim 1, further comprising:

creating a semi-physical simulation f amework of the mask using an existing semi-physical model of the lithography process used to create the mask, the semi-physical model specifying optica! intensity values representing the plurality of features of the mask; and wherein iterative!y moving mask segments comprises shifting contours of the plurality of

features of the mask as represented fay the optical intensity values within the semi- physical model and collecting the data representing image contrast of the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask for eac iteration of moving the mask segments to determine if the movement of a mask segment corresponds to an increase in simulated image contrast (SIC) and whether changes to the mask critical dimension (CD) and the contour critical dimensions (CDs) of the mask falls within a specified tolerance.

15. The method of claim 1, further comprising:

training a neural network to describe a relationship, between the changes to the segments of the mask and the change in simulated image contrast (SIC) for the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask, wherein the neural network comprises an Artificial ^'Neural Network (ANN) to algoritiimieally represent the relationship between the changes to the segments of the mask and the change in simulated image contrast (SIC) for the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask;

training the neural network to output offset predictions for the selected one or more mask

segments that correspond to the improved simulated, i mage contrast of the mask when moved, wherein the output offset predictions are to be applied to a new semi-physical simulation via Optical Proximity Correction (OPC); and

fabricating a new OPC corrected mask from the new semi-physical simulation.

16. A system to reduce Optical Proximity Correction (OPC) model error, wherein the system comprises:

a processor and a memory;

a mask created via a lithography process;

machine learning logic to execute via the processor and the memory, wherein the machine

learning logic is to perform a learning phase to identify a set of mask shape variables that change simulated intensity values of the mask and a simulated shape of the mask by iterative !y moving mask segments;

the machine- learning logic to collect and store data representing image contrast of the mask, mask critical dimension (CO), and contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments;

the machine learning logic to perform sensitivity analysis on the collected data to determine which mask segments correspond to an improved simulated image contrast of the mask when moved;

the machine learning logic to select one or more mask segments that correspond to the improved simul ated Image contrast of the mask when moved; and

an Optical Proximity Correction (OPC) simulator to apply OPC correction to a model of the mask using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

17. The system of claim 16:

wherein performing the learning phase comprises two sub-phases including an initialization sub- phase and an iteration sub-phase; and

wherei the iratialization sub-phase includes performing an initial simulation to prepare the shape of the mask for learning the set of mask shape variables by retrieving simulated photolithographic resist contours derived from the mask.

18. The system of claim 16, wherein the machine learning logic to select the one or more mask segments thai correspond to the improved simulated image contrast of the mask when moved comprises the machine learning logic to select a mask segment which, when moved, corresponds to a greatest increase in simulated image contrast and a smallest change to any contour critical dimensions (CDs) of the mask.

.

19. The system of claim 16, wherein the machine learning logic to apply the Optical Proximity Correctio (OPC) for the mode! of the mask using the one or more selected mask segments reduces or eliminates scumming or other image related resist defect, of a photo resist exposed via the OPC corrected model of the mask.

20. The system of claim 16, further comprising:

a new mask fabricated using the OPC corrected model of the mask;

a photo resist exposed using the new mask, wherein the photo resist exhibits the mask critical dimension (CD) and the contour critical dimensions (CDs) within tolerances for patterning; and

wherein etching the exposed photo resist results in a defect free and residue tree silicon surface beneath the etched photo resist.

21. The system of claim 16, further comprising:

an existing semi-physical model of the mask created rising physical parameters of the

lithography process used to create the mask, the semi-physical, model specifying optical intensity values representing the plurality of features of the mask; and

wherei the machine learning logic to iteratively move the mask segments comprises the

machine learning logic to shift mask segments, thereby changing contours of the plurality of features of the mask as represented by the optica ! intensity values within t he semi- physical model and collect the da ta representing image c ontrast of the mask, the mask critical dimension (CD), and the contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments to determine if the movement of a mask, segment corresponds to art increase in simulated image contrast (SIC) and whether changes to the mask criticai dimension (CD) and the contour critical dimensions (CDs) of the mask mils within a specified tolerance.

22. Non-transitory computer readable storage media having instructions stored thereupon that, when executed by a processor, the instructions cause the processor to perform operations for reducing Optical Proximity Correction (OPC) model error, wherein operations comprise:

creating a mask for a lithography process;

performing a learning phase to identify a set of mask shape variables that change simulated intensity values of the mask and a simulated shape of the mask by iteratively moving mask segments;

collecting data representing image contrast of the mask, mask critica l dimension (CD), and contour critical dimensions (CDs) of the mask for each iteration of moving the mask segments;

performing sensitivity analysis on the collected data to determine which mask segments

correspond to an improved simulated image contrast of the mask when moved; selecting one or more mask segments that correspond to the improved simulated image contrast of the mask, when moved; and

applying Optical Proximity Correction (OPC) for a model of the mask, using the one or more selected mask segments that correspond to the improved simulated image contrast of the mask when moved.

.

23. The non-transitor computer readable storage media of claim 22:

wherein performing the learning phase comprises two sub-phases including an initialization sub- phase and an iteration sub-phase;

wherein the initialization sub-phase includes performing an initial simulation to prepare the shape of the mask for learning the set of mask shape variables by retrieving simulated photolithographic resist contours derived from the mask and iterativeiy moving mask segments to reduce initial model error;

wherein the initialization sub-phase provides a rudimentary correction to the shape of the mask prior iterativeiy moving the .mask segments to learn the set of mask shape variables; wherein the iteration sub-phase comprises recording the retrieved simulated photolithographic resist contours derived from the mask into the segments; and

wherein iterativeiy moving mask segments during the performing of the learning phase

comprises iterativeiy moving the segments in a collective, independent, or random manner to learn how the mask changes responsive to altering the position of the moved segments and responsive to changes in the mask critical dimensions.

24. The non-transitory computer readable storage media of claim 22, wherein selecting the one or more mask segments that, correspond to the improved simulated image contrast of the mask when moved comprises;

selecting a mask segment which, when moved, corresponds to a greatest increase in simulated image contrast, and a smallest change to any contour critical dimensions (CDs) of the mask.