CN107797391A - Optical adjacent correction method - Google Patents

Optical adjacent correction method Download PDF

Info

Publication number
CN107797391A
CN107797391A CN201711070108.8A CN201711070108A CN107797391A CN 107797391 A CN107797391 A CN 107797391A CN 201711070108 A CN201711070108 A CN 201711070108A CN 107797391 A CN107797391 A CN 107797391A
Authority
CN
China
Prior art keywords
mrow
msub
fragment
neural network
optical
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201711070108.8A
Other languages
Chinese (zh)
Other versions
CN107797391B (en
Inventor
时雪龙
赵宇航
陈寿面
李铭
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai IC R&D Center Co Ltd
Original Assignee
Shanghai Integrated Circuit Research and Development Center Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Integrated Circuit Research and Development Center Co Ltd filed Critical Shanghai Integrated Circuit Research and Development Center Co Ltd
Priority to CN201711070108.8A priority Critical patent/CN107797391B/en
Publication of CN107797391A publication Critical patent/CN107797391A/en
Application granted granted Critical
Publication of CN107797391B publication Critical patent/CN107797391B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03F—PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR
    • G03F7/00—Photomechanical, e.g. photolithographic, production of textured or patterned surfaces, e.g. printing surfaces; Materials therefor, e.g. comprising photoresists; Apparatus specially adapted therefor
    • G03F7/70—Microphotolithographic exposure; Apparatus therefor
    • G03F7/70425—Imaging strategies, e.g. for increasing throughput or resolution, printing product fields larger than the image field or compensating lithography- or non-lithography errors, e.g. proximity correction, mix-and-match, stitching or double patterning
    • G03F7/70433—Layout for increasing efficiency or for compensating imaging errors, e.g. layout of exposure fields for reducing focus errors; Use of mask features for increasing efficiency or for compensating imaging errors
    • G03F7/70441—Optical proximity correction [OPC]
    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03F—PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR
    • G03F1/00—Originals for photomechanical production of textured or patterned surfaces, e.g., masks, photo-masks, reticles; Mask blanks or pellicles therefor; Containers specially adapted therefor; Preparation thereof
    • G03F1/36—Masks having proximity correction features; Preparation thereof, e.g. optical proximity correction [OPC] design processes
    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03F—PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR
    • G03F7/00—Photomechanical, e.g. photolithographic, production of textured or patterned surfaces, e.g. printing surfaces; Materials therefor, e.g. comprising photoresists; Apparatus specially adapted therefor
    • G03F7/70—Microphotolithographic exposure; Apparatus therefor
    • G03F7/70483—Information management; Active and passive control; Testing; Wafer monitoring, e.g. pattern monitoring
    • G03F7/70491—Information management, e.g. software; Active and passive control, e.g. details of controlling exposure processes or exposure tool monitoring processes
    • G03F7/70508—Data handling in all parts of the microlithographic apparatus, e.g. handling pattern data for addressable masks or data transfer to or from different components within the exposure apparatus

Abstract

The present invention provides a kind of optical adjacent correction method, including：The edge decomposition of layout patterns is taken a little be used as environment adjacent sensing point in each segment into multiple fragments using optical near-correction algorithm；Calculate the imaging signal value set { S of the environment adjacent sensing point of each fragment1, S2, S3... Sn, n is positive integer；By each imaging signal value set { S1, S2, S3... SnInput as a neural network model, the neural network model exports the optimum displacement (Δ X, Δ Y) of each fragment；Each fragment is moved to obtain the optimal figure of the layout patterns according to the optimum displacement (Δ X, Δ Y) of each fragment respectively.Optical adjacent correction method provided by the invention can obtain the optimal figure of layout patterns.

Description

Optical adjacent correction method
Technical field
The present invention relates to technical field of semiconductors, more particularly to a kind of optical adjacent correction method.
Background technology
Optical near-correction (OPC) is the technology commonly used in semiconductor chip manufacture.Optical near-correction (OPC) is to pass through Mask plate figure is modified, maximum possible solves the problems, such as litho pattern deformation.Traditional OPC algorithm is in makeover process In only consider that each fragment of the mask plate figure error of itself determines the modification amount of each step, without considering mask plate figure phase The interaction of adjacent fragment, this algorithm formerly enter used by photoetching process under strong spatial coherence imaging lighting condition, Convergence rate is slow, or even can not restrain.At present, in 14nm, 10nm, cooperate with light shield from light source in 7nm node photoetching processes excellent The imaging optical illumination changed (SMO) and drawn all has the spatial coherence of very high level, therefore, between adjacent pattern or fragment Interaction become stronger.
In order to solve this problem, there are some OPC algorithms to consider the interaction between multiple fragments, but it is such OPC algorithm is computationally usually slow, because the interaction matrix between each fragment must iteration be more each time Newly.In addition, although these OPC algorithms and the OPC solutions that draw also comply with marginal position error (EPE) tolerance, but from Lithographic process window angle, which is seen, to be not optimal.
Therefore, non-iterative mode how is utilized to provide optimum position, and the OPC solutions provided for each fragment It is optimal from lithographic process window angle, is urgent problem to be solved.
The content of the invention
The defects of present invention is in order to overcome above-mentioned prior art to exist, there is provided a kind of optical adjacent correction method, to utilize Non-iterative mode provides optimum position for each fragment, and the OPC solutions provided are from lithographic process window angle Optimal.
According to an aspect of the present invention, there is provided a kind of optical adjacent correction method, including：Calculated using optical near-correction The edge decomposition of layout patterns into multiple fragments, is taken a little be used as environment adjacent sensing point in each segment by method；Calculate every Imaging signal value set { the S of the environment adjacent sensing point of individual fragment1, S2, S3... Sn, n is positive integer；By each imaging letter Number value set { S1, S2, S3... SnInput as a neural network model, the neural network model exports each fragment Optimum displacement (Δ X, Δ Y)；Each fragment is moved to obtain according to the optimum displacement (Δ X, Δ Y) of each fragment respectively Obtain the optimal figure of the layout patterns.
Alternatively, it is described by each imaging signal value set { S1, S2, S3... SnAs the defeated of a neural network model Enter, before the step of neural network model exports optimum displacement (the Δ X, Δ Y) of each fragment, including：Obtain test chart Shape set, each resolution chart row in resolution chart set is handled as follows：Will using optical near-correction algorithm The edge decomposition of resolution chart takes a little be used as environment adjacent sensing point in each segment into multiple fragments；Calculate each piece Imaging signal value set { the S of the environment adjacent sensing point of section1, S2, S3... Sn}；Calculated and tested using mask optimized algorithm The optimum shape of figure；The optimum displacement (Δ X, Δ Y) of each fragment is calculated with reference to the optimum shape of the resolution chart；By institute State the imaging signal value set of the environment adjacent sensing point of each fragment of each resolution chart in resolution chart set {S1, S2, S3... SnInput as neural network model, by each of each resolution chart in the resolution chart set Output of the optimum displacement (Δ X, Δ Y) of the fragment as neural network model, train the parameter of the neural network model.
Alternatively, the neural network model is multilayer perceptron neutral net.
Alternatively, the output of the multilayer perceptron neutral net calculates according to equation below：
Wherein, { wij, { ωj,To train the parameter that the multilayer perceptron neutral net obtains, yjFor The hidden unit of the hidden layer of the multilayer perceptron neutral net, M is the quantity of hidden unit, and M is positive integer, pjFor institute State the median in multilayer perceptron neural computing.
Alternatively, the parameter of the multilayer perceptron neutral net is trained using back-propagation algorithm.
Alternatively, the neural network model is BRF neutral nets.
Alternatively, the output of the BRF neutral nets calculates according to equation below：
Wherein, { wjAnd σ be the parameter for training the BRF neutral nets to obtain,For using training data point as A series of RBFs at center, M are the quantity of hidden unit.
Alternatively, the parameter of the BRF neutral nets is obtained using K mean cluster algorithm and recursive least squares.
Alternatively, the imaging signal value set { S1, S2, S3... SnCalculated according to equation below：
Wherein, M be light shield face on point (x, y) optical transport function, KiIt is the eigenfunction of equation below：
∫∫W(x1',y1'；x2',y2')Ki(x2',y2')dx2'dy2'=αiKi(x1',y1'),
Wherein, αiIt is the characteristic value of as above equation, W (x1', y1’；x2', y2') be calculated as follows：
W(x1',y1'；x2',y2')=γ (x2'-x1',y2'-y1')P(x1',y1')P*(x2',y2'),
Wherein, γ (x2’-x1', y2’-y1') it is point (x on light shield face1', y1') and point (x2', y2') light field be mutually concerned with The factor, P (x1', y1') be optical imaging system impulse response function, P*It is P conjugation.
Alternatively, it is described to be gone back the edge decomposition of layout patterns into the step of multiple fragments using optical near-correction algorithm Including：Multiple fragments are divided into one kind or multiclass in salient angle fragment, re-entrant angle fragment, tag wire fragment, general line fragment, it is described The classification of fragment also serves as the input of the neural network model.
Compared with prior art, advantage of the invention is that：
1) neural network model obtains the optimum displacement of domain fragment, can avoid during OPC to the mutual of fragment Acting matrix iteration, so can Fast Convergent to save the OPC calculating time.
2) neural network model is enabled using the result of calculation training neural network model of mask optimized algorithm According to mask optimized algorithm output optimum displacement, and OPC is efficiently completed, realizes OPC optimal solutions.
Brief description of the drawings
Its example embodiment is described in detail by referring to accompanying drawing, above and other feature and advantage of the invention will become It is more obvious.
Fig. 1 shows the flow chart of optical adjacent correction method according to embodiments of the present invention.
Fig. 2 shows the schematic diagram of neural network model according to embodiments of the present invention.
Fig. 3 and Fig. 4 is shown based on schematic diagram of the geometric description as pattern characteristics vector.
Fig. 5 and Fig. 6 shows the schematic diagram of layout patterns segmentation according to embodiments of the present invention.
Fig. 7 shows according to embodiments of the present invention based on schematic diagram of the ambient measurements as pattern characteristics vector.
Embodiment
Example embodiment is described more fully with referring now to accompanying drawing.However, example embodiment can be with a variety of shapes Formula is implemented, and is not understood as limited to example set forth herein；On the contrary, these embodiments are provided so that the disclosure will more Fully and completely, and by the design of example embodiment comprehensively it is communicated to those skilled in the art.Described feature, knot Structure or characteristic can be incorporated in one or more embodiments in any suitable manner.
In addition, accompanying drawing is only the schematic illustrations of the disclosure, it is not necessarily drawn to scale.Identical accompanying drawing mark in figure Note represents same or similar part, thus will omit repetition thereof.Some block diagrams shown in accompanying drawing are work( Can entity, not necessarily must be corresponding with physically or logically independent entity.These work(can be realized using software form Energy entity, or these functional entitys are realized in one or more hardware modules or integrated circuit, or at heterogeneous networks and/or place These functional entitys are realized in reason device device and/or microcontroller device.
In order to provide optimal location for each fragment using non-iterative mode, and make the OPC solutions of offer from Lithographic process window angle is optimal, and inventor is once used using input of the geometric description of domain as neural network model Characteristic vector obtains optimum displacement.Fig. 3 and Fig. 4 be may be referred to illustrate that geometric description is special as the input of neural network model Sign vector.As shown in figure 3, light shield face is divided into many units (such as 5nm X 5nm are a unit 301), each unit 301 In value { 1 or 0 } depend on how many layout patterns 302 " weight " be located in this unit.In further embodiments, may be used As shown in figure 4, sampled using equidistant concentric circles 303 to produce input feature value.
But the treatment effeciency of these characteristic vectors based on pure geometric description is more low：1) based on pure geometric description The quantity of the element of input feature value is very big, and this causes computational efficiency low；2) nerve is used as using simple geometric description The input feature value of network model, in order that model reaches certain precision, training pattern needs more training datas, because It is more for the quantity of free parameter；3) input feature value of neural network model, output response are used as using simple geometric description Mapping function meeting nonlinearity becomes big, while is also nonmonotonic, and this needs more complicated neural network model to be mapped to catch The essence of function, model is also set to reduce the ability predicted extensively in itself.
In order to improve the efficiency of neural network model, another characteristic vector as input is inventor provided.Can be with Understand, the optimal correction position of each fragment of domain depends entirely on its environment adjacent, then the key issue of characteristic vector Can how effectively describe the environment adjacent of a point.It is appreciated that the interaction strength of the neighbouring fragment of domain and The interaction strength of adjacent patterns depends not only on their geometric distance, and depends on image-forming condition, such as numerical aperture Set with imaging lighting condition.Therefore, it may be considered that a kind of optical ruler module set based on imaging lighting condition To obtain characteristic vector, rather than geometry module is based purely on to obtain characteristic vector.
The optical ruler module set based on imaging lighting condition can be asked for using imaging equation, such as utilize portion The Hopkins imaging formulas divided under coherent light illumination condition setting：
Wherein, γ (x2-x1, y2-y1) be on object plane, i.e., on light shield face, 2 points of (x1, y1) and (x2, y2) light field it is mutual The dry factor, the mutual coherence factor of light field are to be set to determine by imaging lighting condition.P(x-x1, y-y1) it is optical imaging system Impulse response function, it is determined by optical system pupil function.More particularly, P (x-x1, y-y1) it is because on object plane In (x1, y1) point, the light source disturbance of a unit amplitude and zero phase and on the imaging plane gone point (x, y) complex amplitude.M (x1, y1) it is point (x on object plane1, y1) optical transport function.Variable with asterisk refers to the conjugation of variable, for example, P*It is P Conjugation, M*It is M conjugation.
According to Mercer theorems, above-mentioned equation can change into simpler form：
Wherein,Represent convolution algorithm, { αiAnd { KiBe below equation characteristic value and eigenfunction：
∫∫W(x1',y1'；x2',y2')Ki(x2',y2')dx2'dy2'=αiKi(x1',y1')
W(x1',y1'；x2',y2')=γ (x2'-x1',y2'-y1')P(x1',y1')P*(x2',y2')
Above-mentioned simpler form equation represents that partially coherent imaging system can resolve into a series of coherence imaging systems, Coherence imaging system is separate.Although partially coherent image formation system decomposition is also had into a series of method of coherence imaging systems A lot, but the above method is optimal method, commonly referred to as optimal Coherent decomposition.In other words, eigenfunction set { KI(x, Y) } it is the optimal optics yardstick based on the setting of imaging lighting condition.To this end it is possible to use, function set { KI(x, y) } under Measured value is imaged sets of signal values { S1, S2, S3... Sn(n is positive integer) be used as neural network model input feature value. For example, Fig. 7 is may refer to, by the imaging signal value set { S of the environment adjacent sensing point 305 of a fragment of layout patterns 3021, S2, S3... SnAs neural network model input feature value, (Fig. 7 is only schematically, to be not intended to the guarantor that limitation is invented Protect scope).
According to above-mentioned principle, the present invention provides a kind of optical adjacent correction method, referring to Fig. 1.Fig. 1 shows 4 steps altogether：
Step S110：Using optical near-correction algorithm by the edge decomposition of layout patterns into multiple fragments, at each Taken in section and be a little used as environment adjacent sensing point.
Specifically, the segmentation of layout patterns refers to Fig. 5 and Fig. 6 (layout patterns 302 is divided into multiple fragments 304).In certain embodiments, multiple fragments can be divided into salient angle fragment, re-entrant angle fragment, tag wire fragment, general line piece One kind or multiclass in section.For in general edge line fragment, if necessary, the fragment length for being also based on them enters One step is divided into subclass.In other specific embodiments, environment adjacent sensing point is located on line segment and is the midpoint of each fragment.
Step S120：Calculate the imaging signal value set { S of the environment adjacent sensing point of each fragment1, S2, S3... Sn}。
Specifically, imaging signal value set { S1, S2, S3... SnCalculating can enter by the formula provided in above-mentioned principle Row calculates.
Step S130：By the imaging signal value set { S1, S2, S3... SnInput as a neural network model, The neural network model exports the optimum displacement (Δ X, Δ Y) of each fragment.
Specifically, before step S130, it may include following steps：
Resolution chart set is obtained, each resolution chart in resolution chart set is handled as follows：
Using optical near-correction algorithm by the edge decomposition of resolution chart into multiple fragments.Phase is chosen in each segment Adjacent environment detection point.
Imaging signal value set { the S of the environment adjacent sensing point of each fragment is calculated by above-mentioned principle1, S2, S3... Sn}。
The optimum shape of resolution chart is calculated using mask optimized algorithm.Wherein, mask optimized algorithm is for example Can be source mask optimization (Source Mask Optimization, SMO).
The optimum displacement (Δ X, Δ Y) of each fragment is calculated with reference to the optimum shape of the resolution chart.
After each resolution chart in resolution chart set has been handled, by each institute in the resolution chart set State the imaging signal value set { S of the environment adjacent sensing point of each fragment of resolution chart1, S2, S3... SnAs nerve The input of network model, by the optimum displacement (Δ of each fragment of each resolution chart in the resolution chart set X, Δ Y) output as neural network model, train the neural network model.In further embodiments, can resolution chart A resolution chart in set carries out parameter training after having handled using the data of the resolution chart.
Further, the environment adjacent sensing point of each fragment is (xj,, yj) when, its imaging signal value set is { S1 (xj,, yj),S2(xj,, yj),…..Sn(xj,, yj), the optimum displacement of the corresponding fragment obtained is (Δ X (xj,, yj), Δ Y (xj,, yj)).It is { S from imaging signal value set1(xj,, yj),S2(xj,, yj),…..Sn(xj,, yj) arrive (Δ X (xj,, yj), Δ Y(xj,, yj)) mapping function can be nonlinear, but it can be dull.
It can mathematically be expressed as：
ΔX(xj,, yj)=fx ({ S1(xj,, yj),S2(xj,, yj),…..Sn(xj,, yj)})；
ΔY(xj,, yj)=fy ({ S1(xj,, yj),S2(xj,, yj),…..Sn(xj,, yj)})
Due to Δ X (xj,, yj),ΔY(xj,, yj) for { S1(xj,, yj),S2(xj,, yj),…..Sn(xj,, yj) dependence Can be nonlinear, also, the concrete form of this mapping function is unknown, so above-mentioned steps are the nerves in 130 Network model can be multilayer perceptron neutral net (as shown in Figure 2) or RBF (RBF) network.
In a specific embodiment, neural network model can be multilayer perceptron neutral net.In multilayer perceptron Neutral net, the operation of the hidden unit of hidden layer are the weighted sums of input, are then exported by activation primitive. In the present embodiment, the output of the multilayer perceptron neutral net calculates according to equation below：
Wherein, { wij, { ωj,To train the parameter that the multilayer perceptron neutral net obtains, yjFor (as shown in Figure 2 202,201 be the input of multilayer perceptron neutral net to the hidden layer of the multilayer perceptron neutral net.203 For the output of multilayer perceptron neutral net) hidden unit, M is the quantity of hidden unit, and M is positive integer, pjTo be described Median in multilayer perceptron neural computing.In this embodiment, the parameter of multilayer perceptron neutral net is available Back-propagation algorithm is trained.
In another specific embodiment, neural network model can be BRF neutral nets.It is hidden in BRF neutral nets The operation for hiding unit is a series of RBFs centered on training data point.Output is the weighting of all RBFs With.In the present embodiment, the output of BRF neutral nets calculates according to equation below：
Wherein, { wjAnd σ be the parameter for training the BRF neutral nets to obtain,For using training data point as A series of RBFs at center, M are the quantity of hidden unit.In this embodiment, the parameter of BRF neutral nets can profit Obtained with K mean cluster algorithm and recursive least squares.
In some change case of above-described embodiment, as described in step S110, multiple fragments via classification, then may be used Input using the grouped data of multiple fragments as neural network model.Meanwhile in the training process of neural network model, Each fragment is classified, and the input using the grouped data of fragment as neural network model, it is accurate that model is obtained with training The parameter of the higher neural network model of true rate.
Step S140：Each fragment is moved to obtain according to the optimum displacement (Δ X, Δ Y) of each fragment respectively Obtain the optimal figure of the layout patterns.
One or more embodiments of the invention is only schematically described above, without prejudice to before present inventive concept Put, different change case is all within protection scope of the present invention.
Compared with prior art, advantage of the invention is that：
1) neural network model obtains the optimum displacement of domain fragment, can avoid during OPC to the mutual of fragment Acting matrix iteration, so can Fast Convergent to save the OPC calculating time.
2) neural network model is enabled using the result of calculation training neural network model of mask optimized algorithm According to mask optimized algorithm output optimum displacement, and OPC is efficiently completed, realizes OPC optimal solutions.
Those skilled in the art will readily occur to the disclosure its after considering specification and putting into practice invention disclosed herein Its embodiment.The application is intended to any modification, purposes or the adaptations of the disclosure, these modifications, purposes or Person's adaptations follow the general principle of the disclosure and including the undocumented common knowledges in the art of the disclosure Or conventional techniques.Description and embodiments are considered only as exemplary, and the true scope of the disclosure and spirit are by appended Claim is pointed out.

Claims (10)

  1. A kind of 1. optical adjacent correction method, it is characterised in that including：
    The edge decomposition of layout patterns is taken into some conducts in each segment into multiple fragments using optical near-correction algorithm Environment adjacent sensing point；
    Calculate the imaging signal value set { S of the environment adjacent sensing point of each fragment1, S2, S3... Sn, n is positive integer；
    By each imaging signal value set { S1, S2, S3... SnInput as a neural network model, the neutral net Model exports the optimum displacement (Δ X, Δ Y) of each fragment；
    Each fragment is moved to obtain the layout patterns according to the optimum displacement (Δ X, Δ Y) of each fragment respectively Optimal figure.
  2. 2. optical adjacent correction method as claimed in claim 1, it is characterised in that described by each imaging signal value set {S1, S2, S3... SnInput as a neural network model, the neural network model exports the optimum displacement of each fragment Before the step of (Δ X, Δ Y), including：
    Resolution chart set is obtained, each resolution chart in resolution chart set is handled as follows：
    The edge decomposition of resolution chart is taken into some conducts in each segment into multiple fragments using optical near-correction algorithm Environment adjacent sensing point；
    Calculate the imaging signal value set { S of the environment adjacent sensing point of each fragment1, S2, S3... Sn}；
    The optimum shape of resolution chart is calculated using mask optimized algorithm；
    The optimum displacement (Δ X, Δ Y) of each fragment is calculated with reference to the optimum shape of the resolution chart；
    The imaging of the environment adjacent sensing point of each fragment of each resolution chart in the resolution chart set is believed Number input of the value set { S1, S2, S3 ... Sn } as neural network model, by each survey in the resolution chart set The optimum displacement (Δ X, Δ Y) for attempting each fragment of shape trains the neutral net mould as the output of neural network model The parameter of type.
  3. 3. optical adjacent correction method as claimed in claim 1, it is characterised in that the neural network model is Multilayer Perception Device neutral net.
  4. 4. optical adjacent correction method as claimed in claim 3, it is characterised in that the multilayer perceptron neutral net it is defeated Go out and calculated according to equation below：
    <mrow> <msub> <mi>p</mi> <mi>j</mi> </msub> <mo>=</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </msubsup> <msub> <mi>w</mi> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> <msub> <mi>S</mi> <mi>i</mi> </msub> <mo>,</mo> </mrow>
    <mrow> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>1</mn> <mo>+</mo> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mi>a</mi> <mo>(</mo> <mrow> <msub> <mi>P</mi> <mi>j</mi> </msub> <mo>-</mo> <msubsup> <mi>p</mi> <mi>j</mi> <mn>0</mn> </msubsup> </mrow> <mo>)</mo> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>,</mo> </mrow>
    <mrow> <mi>&amp;Delta;</mi> <mi>x</mi> <mi> </mi> <mi>o</mi> <mi>r</mi> <mi>&amp;Delta;</mi> <mi>y</mi> <mo>=</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </msubsup> <msub> <mi>&amp;omega;</mi> <mi>j</mi> </msub> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>,</mo> </mrow>
    Wherein, { wi,j, { ωj,To train the parameter that the multilayer perceptron neutral net obtains, yjTo be described The hidden unit of the hidden layer of multilayer perceptron neutral net, M is the quantity of hidden unit, and M is positive integer, pjTo be described more Median in layer perceptron neural network calculating.
  5. 5. optical adjacent correction method as claimed in claim 4, it is characterised in that the ginseng of the multilayer perceptron neutral net Number is trained using back-propagation algorithm.
  6. 6. optical adjacent correction method as claimed in claim 1, it is characterised in that the neural network model is BRF nerves Network.
  7. 7. optical adjacent correction method as claimed in claim 6, it is characterised in that the output of the BRF neutral nets according to Equation below calculates：
    <mrow> <msub> <mi>&amp;Phi;</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mover> <mi>S</mi> <mo>&amp;RightArrow;</mo> </mover> <mo>,</mo> <msub> <mover> <mi>S</mi> <mo>&amp;RightArrow;</mo> </mover> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <msqrt> <mrow> <mn>2</mn> <mi>&amp;pi;</mi> </mrow> </msqrt> </mfrac> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> <msup> <mi>&amp;sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>|</mo> <mo>|</mo> <mover> <mi>S</mi> <mo>&amp;RightArrow;</mo> </mover> <mo>-</mo> <msub> <mover> <mi>S</mi> <mo>&amp;RightArrow;</mo> </mover> <mi>j</mi> </msub> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mo>,</mo> </mrow>
    <mrow> <mi>&amp;Delta;</mi> <mi>X</mi> <mi> </mi> <mi>o</mi> <mi>r</mi> <mi>&amp;Delta;</mi> <mi>Y</mi> <mo>=</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </msubsup> <msub> <mi>w</mi> <mi>j</mi> </msub> <msub> <mi>&amp;Phi;</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mover> <mi>S</mi> <mo>&amp;RightArrow;</mo> </mover> <mo>,</mo> <msub> <mover> <mi>S</mi> <mo>&amp;RightArrow;</mo> </mover> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> </mrow>
    Wherein, { wjAnd σ be the parameter for training the BRF neutral nets to obtain,For centered on training data point A series of RBFs, M are the quantity of hidden unit.
  8. 8. optical adjacent correction method as claimed in claim 7, it is characterised in that the parameter of the BRF neutral nets utilizes K Means clustering algorithm and recursive least squares obtain.
  9. 9. the optical adjacent correction method as described in any one of claim 1 to 8, it is characterised in that the imaging signal values collection Close { S1, S2, S3... SnCalculated according to equation below：
    <mrow> <msub> <mi>S</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <mo>|</mo> <msub> <mi>K</mi> <mi>i</mi> </msub> <mo>&amp;CircleTimes;</mo> <mi>M</mi> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>,</mo> </mrow>
    Wherein, M be light shield face on point (x, y) optical transport function, KiIt is the eigenfunction of equation below：
    ∫∫W(x1',y1'；x2',y2')Ki(x2',y2')dx2'dy2'=αiKi(x1',y1'),
    Wherein, αiIt is the characteristic value of as above equation, W (x1', y1’；x2', y2') be calculated as follows：
    W (, x1',y1'；x2',y2')=γ (x2'-x1',y2'-y1')P(x1',y1')P*(x2',y2')
    Wherein, γ (x2’-x1', y2’-y1') it is point (x on light shield face1', y1') and point (x2', y2') the mutual coherence factor of light field, P(x1', y1') be optical imaging system impulse response function, P*It is P conjugation.
  10. 10. the optical adjacent correction method as described in any one of claim 1 to 8, it is characterised in that described to utilize optical adjacent Correcting algorithm also includes the edge decomposition of layout patterns into the step of multiple fragments：
    Multiple fragments are divided into one kind or multiclass in salient angle fragment, re-entrant angle fragment, tag wire fragment, general line fragment, it is described The classification of fragment also serves as the input of the neural network model.
CN201711070108.8A 2017-11-03 2017-11-03 Optical proximity correction method Active CN107797391B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711070108.8A CN107797391B (en) 2017-11-03 2017-11-03 Optical proximity correction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711070108.8A CN107797391B (en) 2017-11-03 2017-11-03 Optical proximity correction method

Publications (2)

Publication Number Publication Date
CN107797391A true CN107797391A (en) 2018-03-13
CN107797391B CN107797391B (en) 2020-04-24

Family

ID=61549025

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711070108.8A Active CN107797391B (en) 2017-11-03 2017-11-03 Optical proximity correction method

Country Status (1)

Country Link
CN (1) CN107797391B (en)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108665060A (en) * 2018-06-12 2018-10-16 上海集成电路研发中心有限公司 A kind of integrated neural network for calculating photoetching
CN108828899A (en) * 2018-06-01 2018-11-16 上海华力集成电路制造有限公司 Connect the optical proximity correction method of aperture layer
CN109143796A (en) * 2018-10-26 2019-01-04 中国科学院微电子研究所 Determine the method, apparatus and model training method, device of photolithography light source
CN109459912A (en) * 2018-12-29 2019-03-12 上海华力集成电路制造有限公司 A kind of OPC method for sorting out aperture layer with feature vector
CN111538213A (en) * 2020-04-27 2020-08-14 湖南大学 Electron beam proximity effect correction method based on neural network
CN112219270A (en) * 2018-06-05 2021-01-12 科磊股份有限公司 Active learning for defect classifier training
CN112230508A (en) * 2020-10-30 2021-01-15 上海华力微电子有限公司 Optical proximity correction method
CN112415855A (en) * 2019-08-22 2021-02-26 长鑫存储技术有限公司 OPC correction method and OPC correction device
CN112859509A (en) * 2019-11-28 2021-05-28 长鑫存储技术有限公司 OPC correction method and OPC correction device
CN113168086A (en) * 2021-03-19 2021-07-23 长江存储科技有限责任公司 System and method for designing a photomask
CN113759657A (en) * 2020-06-03 2021-12-07 中芯国际集成电路制造(上海)有限公司 Optical proximity correction method
CN115598937A (en) * 2022-12-13 2023-01-13 华芯程（杭州）科技有限公司（Cn） Photoetching mask shape prediction method and device and electronic equipment
US11755814B2 (en) 2019-01-29 2023-09-12 Asml Netherlands B.V. Method and apparatus for layout pattern selection
CN117389108A (en) * 2023-11-30 2024-01-12 全芯智造技术有限公司 Method, apparatus and medium for optical proximity correction
CN113759657B (en) * 2020-06-03 2024-05-03 中芯国际集成电路制造(上海)有限公司 Optical proximity correction method

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020051564A1 (en) * 1998-12-04 2002-05-02 Norbert Benesch Method and device for optically monitoring fabrication processes of finely structured surfaces in a semiconductor production
US6466314B1 (en) * 1998-09-17 2002-10-15 Applied Materials, Inc. Reticle design inspection system
CN104915528A (en) * 2015-07-15 2015-09-16 中国科学院微电子研究所 Layout design method, layout design system and manufacture method of integrated circuit
CN106777829A (en) * 2017-02-06 2017-05-31 深圳晶源信息技术有限公司 A kind of optimization method and computer-readable storage medium of integrated circuit mask design
WO2017171891A1 (en) * 2016-04-02 2017-10-05 Intel Corporation Systems, methods, and apparatuses for modeling reticle compensation for post lithography processing using machine learning algorithms

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6466314B1 (en) * 1998-09-17 2002-10-15 Applied Materials, Inc. Reticle design inspection system
US20020051564A1 (en) * 1998-12-04 2002-05-02 Norbert Benesch Method and device for optically monitoring fabrication processes of finely structured surfaces in a semiconductor production
CN104915528A (en) * 2015-07-15 2015-09-16 中国科学院微电子研究所 Layout design method, layout design system and manufacture method of integrated circuit
WO2017171891A1 (en) * 2016-04-02 2017-10-05 Intel Corporation Systems, methods, and apparatuses for modeling reticle compensation for post lithography processing using machine learning algorithms
CN106777829A (en) * 2017-02-06 2017-05-31 深圳晶源信息技术有限公司 A kind of optimization method and computer-readable storage medium of integrated circuit mask design

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108828899A (en) * 2018-06-01 2018-11-16 上海华力集成电路制造有限公司 Connect the optical proximity correction method of aperture layer
CN112219270A (en) * 2018-06-05 2021-01-12 科磊股份有限公司 Active learning for defect classifier training
CN108665060B (en) * 2018-06-12 2022-04-01 上海集成电路研发中心有限公司 Integrated neural network for computational lithography
CN108665060A (en) * 2018-06-12 2018-10-16 上海集成电路研发中心有限公司 A kind of integrated neural network for calculating photoetching
CN109143796A (en) * 2018-10-26 2019-01-04 中国科学院微电子研究所 Determine the method, apparatus and model training method, device of photolithography light source
CN109459912A (en) * 2018-12-29 2019-03-12 上海华力集成电路制造有限公司 A kind of OPC method for sorting out aperture layer with feature vector
US11755814B2 (en) 2019-01-29 2023-09-12 Asml Netherlands B.V. Method and apparatus for layout pattern selection
CN112415855B (en) * 2019-08-22 2022-04-12 长鑫存储技术有限公司 OPC correction method and OPC correction device
CN112415855A (en) * 2019-08-22 2021-02-26 长鑫存储技术有限公司 OPC correction method and OPC correction device
CN112859509B (en) * 2019-11-28 2022-04-12 长鑫存储技术有限公司 OPC correction method and OPC correction device
CN112859509A (en) * 2019-11-28 2021-05-28 长鑫存储技术有限公司 OPC correction method and OPC correction device
CN111538213A (en) * 2020-04-27 2020-08-14 湖南大学 Electron beam proximity effect correction method based on neural network
CN111538213B (en) * 2020-04-27 2021-04-27 湖南大学 Electron beam proximity effect correction method based on neural network
CN113759657A (en) * 2020-06-03 2021-12-07 中芯国际集成电路制造(上海)有限公司 Optical proximity correction method
CN113759657B (en) * 2020-06-03 2024-05-03 中芯国际集成电路制造(上海)有限公司 Optical proximity correction method
CN112230508A (en) * 2020-10-30 2021-01-15 上海华力微电子有限公司 Optical proximity correction method
CN113168086A (en) * 2021-03-19 2021-07-23 长江存储科技有限责任公司 System and method for designing a photomask
CN115598937B (en) * 2022-12-13 2023-04-07 华芯程(杭州)科技有限公司 Photoetching mask shape prediction method and device and electronic equipment
CN115598937A (en) * 2022-12-13 2023-01-13 华芯程（杭州）科技有限公司（Cn） Photoetching mask shape prediction method and device and electronic equipment
CN117389108A (en) * 2023-11-30 2024-01-12 全芯智造技术有限公司 Method, apparatus and medium for optical proximity correction
CN117389108B (en) * 2023-11-30 2024-03-12 全芯智造技术有限公司 Method, apparatus and medium for optical proximity correction

Also Published As

Publication number Publication date
CN107797391B (en) 2020-04-24

Similar Documents

Publication Publication Date Title
CN107797391A (en) Optical adjacent correction method
CN107908071B (en) Optical proximity correction method based on neural network model
CN109976087B (en) Method for generating mask pattern model and method for optimizing mask pattern
US11119401B2 (en) Simulating near field image in optical lithography
Ghaseminezhad et al. A novel self-organizing map (SOM) neural network for discrete groups of data clustering
WO2020038254A1 (en) Image processing method and apparatus for target recognition
CN111310407A (en) Method for designing optimal feature vector of reverse photoetching based on machine learning
US11366952B2 (en) Method, apparatus and electronic device for Hessian-free photolithography mask optimization
US20210150347A1 (en) Guided training of machine learning models with convolution layer feature data fusion
CN107070867B (en) Network flow abnormity rapid detection method based on multilayer locality sensitive hash table
JPH07220995A (en) Light intensity distribution simulating method
US20210089923A1 (en) Icospherical gauge convolutional neural network
CN110426914A (en) A kind of modification method and electronic equipment of Sub-resolution assist features
CN113496247A (en) Estimating an implicit likelihood of generating a countermeasure network
WO2022016802A1 (en) Physical feature map- and dcnn-based computation method for machine learning-based inverse lithography technology solution
US20230375916A1 (en) Inverse lithography and machine learning for mask synthesis
CN109544610A (en) A kind of method for registering images based on convolutional neural networks
KR20180036239A (en) Method of optimizing a mask using pixel based learning
CN112036381B (en) Visual tracking method, video monitoring method and terminal equipment
CN110377921A (en) A kind of failure determination threshold value calculation method based on the more cell spaces of central symmetry
CN103325130A (en) Geometric iteration image fitting method based on T spline
CN114065693A (en) Method and system for optimizing layout of super-large-scale integrated circuit structure and electronic equipment
CN115995042A (en) Video SAR moving target detection method and device
Latypov et al. Probability prediction of EUV process failure due to resist-exposure stochastic: applications of Gaussian random fields excursions and Rice's formula
US11281839B2 (en) Method, apparatus and electronic device for photolithographic mask optimization of joint optimization of pattern and image

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant