A focused ion beam etching method and apparatus

Abstract

The present application relates to a focused ion beam etching method and device. The method comprises: determining a processing target of FIB etching processing; globally optimizing in a multi-dimensional process parameter space according to the processing target of FIB etching processing to obtain a process parameter of FIB etching processing, comprising: inputting initial in-situ characterization data and process parameters in the multi-dimensional process parameter space into a prediction model to obtain a predicted processing result under the processing condition of the process parameters; obtaining a reward value through a reward function according to the processing result and the determined processing target of FIB etching processing; obtaining a process parameter for the next round of circulation through an optimization algorithm according to the reward value and the corresponding process parameter, so as to realize global optimization in the multi-dimensional process parameter space and obtain the process parameter corresponding to the maximum reward value. The etching method can efficiently and accurately give the process parameter corresponding to the processing target, and improve the consistency and yield of the processing product quality.

Description

Technical Field

[0001] This invention relates to the field of micro-nano fabrication technology, and in particular to a focused ion beam etching method and apparatus. Background Technology

[0002] Focused ion beam (FIB) plays an irreplaceable role as a high-precision micro- and nano-fabrication technique in fields such as semiconductor device modification, transmission electron microscopy (TEM) sample preparation, three-dimensional tomographic imaging, and quantum device development. Its principle is to use a high-energy ion beam (typically Ga+) to exfoliate, deposit, or implant materials, thereby achieving the construction of submicron and even nanoscale structures.

[0003] Currently, existing process optimization mainly focuses on geometric parameters such as etching depth, width, sidewall perpendicularity, and bottom surface roughness. Operators observe the processing results using a scanning electron microscope (SEM) and typically adjust equipment parameters manually based on the shape, size, and material of the target structure, combined with their own processing experience. These parameters are then manually corrected and optimized through repeated experiments to achieve the desired geometric morphology. In this approach, the setting of FIB etching parameters relies heavily on the operator's experience, thus limiting the processing quality. Furthermore, for complex multilayer heterostructures or novel materials highly sensitive to ion beams, parameter tuning is time-consuming, and the success rate is difficult to guarantee, leading to a waste of expensive equipment time and labor costs.

[0004] Furthermore, although high-end spherical aberration transmission electron microscopes can be used to perform atomic-resolution characterization of processed samples, the feedback from this atomic-resolution characterization of processed samples is severely delayed and disconnected from the previous processing steps. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a focused ion beam etching method and apparatus, which has the advantages of efficiently and accurately providing the process parameters corresponding to the processing target to be achieved for FIB etching, thereby improving the consistency and yield of processed product quality.

[0006] A focused ion beam etching method includes the following steps:

[0007] Determine the processing objectives for FIB etching; Based on the processing objectives of FIB etching, global optimization is performed in a multi-dimensional process parameter space to obtain the process parameters of FIB etching. The multi-dimensional process parameter space consists of beam current intensity, accelerating voltage, and residence time. Based on the obtained FIB etching process parameters, perform the FIB etching process; The process involves global optimization within a multi-dimensional process parameter space to obtain the process parameters for FIB etching, based on the processing objective of the FIB etching process. This includes: inputting initial in-situ characterization data and process parameters from the multi-dimensional process parameter space into a prediction model to obtain the predicted processing result under the given process parameter conditions; obtaining a reward value through a reward function based on the predicted processing result and the determined processing objective of the FIB etching process; and obtaining the process parameters for the next iteration through an optimization algorithm based on the reward value and the corresponding process parameters. Ultimately, this process achieves global optimization within the multi-dimensional process parameter space, obtaining the process parameters corresponding to the maximum reward value as the process parameters for the FIB etching process.

[0008] The present invention predicts future processing results under certain process parameter conditions and evaluates the merits of the process parameters based on the predicted results. This guides the optimization algorithm to find the optimal combination of process parameters that best meets the processing objectives. This method can provide accurate processing parameter schemes within seconds, significantly shortening the process development cycle. It is efficient and accurate, and also greatly reduces the reliance on operator experience, lowers the cost of talent training and the threshold for equipment use. It effectively avoids common problems in manual trial and error, such as over-cutting, sidewall tilting, and surface damage, and improves the consistency and yield of processed products.

[0009] Furthermore, the optimization algorithm is a Bayesian optimization algorithm, a genetic algorithm, a particle swarm optimization algorithm, or a deep reinforcement learning algorithm.

[0010] Furthermore, the reward function is a linear combination function that sets a penalty term based on the features corresponding to the processing target of the FIB etching process.

[0011] Furthermore, the processing target includes processing requirements for microscopic processing indicators, and the prediction model is trained by fusing in-situ characterization features and corresponding non-in-situ microscopic characterization features, as shown below:

[0012] in, Indicates in-situ characterization features. These are non-in-situ microscopic characterization features. This is a vector of input process parameters including ion beam current intensity, accelerating voltage, and residence time. To predict the processing results, output a state vector. This is a deep learning feature fusion regression model obtained after training.

[0013] Furthermore, the prediction model is trained using the following methods: First, standard structure etching is performed using a FIB (Fiber Embedded Ink) device to obtain multiple sets of in-situ characterization data of the FIB process and corresponding non-in-situ microscopic characterization data, thereby establishing multiple sets of multi-dimensional physical correlation datasets. Then, feature extraction is performed on the multi-dimensional physical correlation datasets to extract the process parameter features of the FIB process. Next, the characterization features extracted from the multi-dimensional physical correlation datasets and their corresponding process parameter features are spliced ​​and fused in a multi-dimensional feature space, and joint inference is performed through a regression network to finally train the prediction model.

[0014] Furthermore, representational features are extracted from in-situ and out-of-situ microscopic representational data using two parallel convolutional neural networks or visual Transformers, respectively.

[0015] Furthermore, the method also includes the following steps: acquiring in-situ characterization data of FIB etching in real time, and predicting the processing results in real time using a prediction model based on the in-situ characterization data and the corresponding FIB etching process parameters. If the predicted processing results show an evolution trend that exceeds the set processing target, an adaptive correction mechanism is adopted to adjust the process parameters of FIB etching.

[0016] Furthermore, the adaptive correction mechanism, based on the real-time predicted processing results and the required processing objectives, optimizes the multi-dimensional process parameter space under the set process parameter boundary conditions using a reward function and a local optimization algorithm, and re-maximizes the reward function.

[0017] Furthermore, the micro-processing parameters include the ultimate non-destructive layer tolerance and / or stoichiometry.

[0018] In addition, the present invention also provides a focused ion beam etching apparatus, which performs FIB etching processing using the above-described focused ion beam etching method.

[0019] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the focused ion beam etching method described in this invention; Figure 2 This is a schematic diagram of the training process of the multimodal prediction model of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention more readily understood by those skilled in the art, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] refer to Figure 1, Figure 1 A schematic flowchart of the focused ion beam etching method of the present invention is shown. The focused ion beam etching method of the present invention includes the following specific steps: S1. Determine the processing objectives of FIB etching.

[0023] In this invention, the processing target of focused ion beam etching is first determined, that is, the processing requirements that the processing indicators of interest in FIB etching need to be achieved are determined. The processing indicators can be macroscopic processing indicators and / or microscopic processing indicators, such as processing depth, tolerance of the ultimate non-destructive layer, stoichiometry, etc. Those skilled in the art can select processing indicators and set specific parameters according to specific processing requirements.

[0024] In one specific embodiment, the substrate material to be processed is determined to be GaAs, the target depth to be achieved after etching is 100nm, the tolerance of the ultimate non-destructive layer is <1.8nm, and the Ga:As ratio is 0.98.

[0025] S2. Based on the processing objectives of FIB etching, global optimization is performed in the multi-dimensional process parameter space to obtain the process parameters for FIB etching.

[0026] The multidimensional process parameter space consists of beam current intensity, accelerating voltage, and residence time. Specifically, this step involves: inputting initial in-situ characterization data and process parameters from the multidimensional process parameter space into a prediction model to obtain the predicted processing result under the given process parameter conditions; obtaining a reward value through a reward function based on the predicted processing result and the determined FIB etching processing target; obtaining the next round of process parameters through an optimization algorithm based on the reward value and the corresponding process parameters; inputting the next round of process parameters and initial in-situ characterization data into the prediction model, and obtaining the next process parameters using the same method described above, thus iteratively obtaining process parameters, ultimately achieving global optimization within the multidimensional process parameter space, and obtaining the process parameters corresponding to the maximum reward value, i.e., the optimal process parameters, as the process parameters for FIB etching processing. The initial in-situ characterization data represents the initial surface morphology of the sample before processing; that is, before processing, based on the initial substrate state reflected by the initial in-situ characterization data, the final result after executing the corresponding process parameters is predicted, thereby finding the optimal process parameters. In a specific embodiment, the initial in-situ characterization data is initial in-situ contrast image data.

[0027] The optimization algorithm can be a Bayesian optimization algorithm, a genetic algorithm, a particle swarm optimization algorithm, or a deep reinforcement learning algorithm (such as PPO (Proximal Policy Optimization) or DQN (Deep Q-Network)). The Bayesian optimization is more efficient in small sample and high-dimensional space search.

[0028] The process parameters corresponding to the maximum reward value can be represented as follows: (1) in, Represents the process parameters in the multidimensional process parameter space. Represents the optimal process parameters, reward function A linear combination function that sets a penalty term based on the processing parameters corresponding to the processing target of FIB etching.

[0029] The solution of this invention, through a prediction model, a reward function, and an optimization algorithm, enables constraint-aware global search within a multi-dimensional process parameter space. It quickly inversely derives the optimal initial etching process parameters that achieve the maximum reward and can further generate a recommended parameter optimization report. It not only predicts the future processing results corresponding to the processing indicators under the processing parameter conditions, but also evaluates the merits of the process parameters based on the predicted processing results, thereby guiding the optimization algorithm to find the process parameter combination that best meets the processing objectives.

[0030] In one specific embodiment, the prediction model can predict the final etching depth, sub-nanometer surface damage layer thickness, and local stoichiometry after processing. Meanwhile, the reward function is a linear combination of a depth penalty term, a damage penalty term, and a composition penalty term, expressed as follows: (2) in, 、 、 These represent the actual processing results predicted by the multimodal prediction model. , , These are the defined target processing depth, target limit non-destructive layer tolerance, and target stoichiometry, respectively. and All are empirical weighting coefficients; For the asymmetric soft-hard constraint piecewise function targeting damage characteristics, when the predicted damage layer thickness exceeds the limit value, i.e. At that time, high penalties are imposed to achieve forced blocking.

[0031] The prediction model is a multimodal prediction model, which is trained by fusing in-situ characterization features (modal one), corresponding non-in-situ microscopic characterization features (modal two), and process parameters. It establishes a mapping relationship between multidimensional microscopic features and ion beam etching process parameters, thereby enabling accurate prediction of macroscopic and / or microscopic processing indicators using in-situ characterization data and corresponding process parameters, as shown below: (3) in, Indicates in-situ characterization features. These are non-in-situ microscopic characterization features. This is a vector of input process parameters including ion beam current intensity, accelerating voltage, and residence time. To predict the processing result, a state vector is output, and this vector reflects the data of the processing index corresponding to the previously determined processing target. This is a deep learning feature fusion regression model obtained after training.

[0032] refer to Figure 2 The pre-training method for the multimodal prediction model is as follows: First, standard structures (such as rectangular grooves, circular holes, TEM films, etc.) are etched using FIB equipment to obtain multiple sets of in-situ characterization data of FIB processing and corresponding non-in-situ microscopic characterization data. Multiple sets of multi-dimensional physical correlation datasets are then established. The in-situ characterization data of FIB processing and the corresponding non-in-situ microscopic characterization data form one set. In these two types of data, the in-situ characterization data (such as real-time contrast images acquired by the probe) reflects the evolution of macroscopic contrast, morphology, material density, and local amorphization during the processing, and has real-time characteristics. The non-in-situ microscopic characterization data (such as atomic-level damage images acquired by high-resolution electron microscopy) reflects the lagging, high-precision true values ​​of the microscopic physical baseline. In the training and construction of the predictive model, non-in-situ microscopic characterization data serves as a high-quality supervisory signal (standard answer) to guide the predictive model in learning the nonlinear mapping relationship between in-situ surface features (such as contrast) and the actual microscopic processing results (such as damage) of the underlying and internal layers. This allows for the elimination of reliance on lagging line characterization in practical processing applications, enabling accurate prediction of invisible internal microscopic processing quality (such as damage layer thickness and compositional changes) solely based on the acquired in-situ characterization data and corresponding process parameters. In this invention, macroscopic and microscopic performance values ​​can be accurately predicted using only "morphology" and "process," breaking down the interdisciplinary mapping barrier between "morphology-performance-process" and providing a decision-making basis for obtaining optimal process parameters before processing, as well as subsequent closed-loop monitoring and dynamic correction.

[0033] In this invention, the in-situ characterization data and the corresponding non-in-situ microscopic characterization data are adjusted according to the type of processing indicators of interest in the actual processing task, and the acquisition devices for the corresponding in-situ characterization data and non-in-situ microscopic characterization data can also be selected according to the data to be acquired.

[0034] In a specific embodiment, the in-situ characterization data is in-situ contrast feature data, specifically high-depth TIF images of FIB secondary electron yield contrast features and / or ion-induced secondary ion mass spectrometry (SIMS) contrast feature data, providing the initial surface morphology and the real-time contrast and morphology evolution during processing, which can correspond to the macroscopic etching depth index and serve as the input of the prediction model; the non-in-situ microscopic characterization data is non-in-situ atomic resolution damage layer and stoichiometry feature data, serving as the supervision signal for training the prediction model, specifically HAADF (High-Angle Annular Dark-Field STEM) / LAADF (Low-Angle Annular Dark-Field STEM) atomic resolution damage layer and stoichiometry information data obtained by non-in-situ spherical aberration lens electron microscopy (Cs-TEM), and the atomic resolution damage layer and stoichiometry can adopt an EMD matrix format that includes both. In this embodiment, the prediction model is jointly trained by a large number of "in-situ contrast images" and "non-in-situ microscopic physical reference true values", so that the prediction model learns to accurately predict the macroscopic etching depth, as well as the atomic-level damage and composition changes that occur in the bottom layer and inside, based on the evolution law of in-situ contrast.

[0035] Of course, in other embodiments, non-in-situ microscopic characterization data can also be obtained through other characterization methods such as X-ray photoelectron spectroscopy (XPS) as needed.

[0036] Then, based on multiple sets of the aforementioned multi-dimensional physical association datasets and the corresponding process parameters of FIB processing, a multimodal prediction model is constructed. Specifically: Feature extraction is performed on the multi-dimensional physical association dataset to extract process parameter features for FIB processing. In this step, two parallel convolutional neural networks (CNNs) are used to extract representational features from the in-situ representation data and the non-in-situ microscopic representation data, respectively. These representational features reflect the data of the indicators involved in the processing target. Simultaneously, a linear layer is used to extract features of the process parameters (including ion beam intensity, accelerating voltage, and residence time) corresponding to each set of multi-dimensional physical association datasets during FIB processing. Alternatively, in other embodiments, a Vision Transformer (ViT) can be used instead of a CNN for feature extraction. CNNs are superior in local feature extraction, while ViT performs better in capturing global contextual information; both can achieve the purpose of feature extraction.

[0037] Next, the three types of features extracted above are concatenated and fused in a multi-dimensional feature space, and joint inference is performed through a regression network to optimize the training of the multimodal prediction model. Specifically, for macroscopic processing indicators (such as etching depth), the prediction model is considered successfully trained when the difference between the predicted value obtained by the prediction model and the set standard value is less than a set threshold; and for microscopic processing indicators (such as tolerance of the ultimate non-destructive layer and stoichiometry), the prediction model is considered successfully trained when the difference between the predicted value obtained by the prediction model and the data obtained from non-in-situ microscopic characterization data is less than a set threshold.

[0038] S3. Perform FIB etching based on the obtained FIB etching process parameters.

[0039] In this step, the FIB etching process is performed according to the process parameters obtained in step S2.

[0040] This invention abandons the traditional single-modal manual trial-and-error approach, adopting a multi-modal prediction model that integrates in-situ macroscopic information and non-situ microscopic information. Furthermore, by combining reward functions and optimization algorithms, it achieves etching processing from "target analysis" to "parameter back-calculation." This not only effectively utilizes non-situ microscopic characterization data, such as high-resolution data from spherical aberration transmission electron microscopy (especially low-angle annular dark-field imaging), as high-quality monitoring signals to guide the optimization of macroscopic FIB processing parameters, but also eliminates the need to debug parameters from scratch for the target processing structure. It can provide accurate processing parameter schemes within seconds, reducing the parameter exploration time from hours or even days to minutes, significantly shortening the process development cycle. It is highly efficient and accurate. In addition, it greatly reduces reliance on operator experience, lowers talent training costs and equipment usage barriers, and effectively avoids common problems in manual trial-and-error such as over-etching, sidewall tilting, and surface damage, improving the consistency and yield of processed products.

[0041] S4. Acquire in-situ characterization data of FIB etching process in real time, and make real-time prediction of processing results and adjust process parameters based on the in-situ characterization data.

[0042] In this invention, to further improve processing accuracy, after issuing the processing command, the system enters an intelligent closed-loop control state, acquiring in-situ characterization data of the FIB etching process in real time. Based on this in-situ characterization data and the corresponding FIB etching process parameters, a multi-modal prediction model is used to predict the processing result in real time. If the predicted processing result shows an evolution trend greater than the set processing target, an adaptive correction mechanism is automatically adopted to adjust the FIB etching process parameters, achieving real-time closed-loop monitoring and dynamic correction. In a specific embodiment, the in-situ characterization data in this step can be a high-depth TIF image of the secondary electron yield contrast characteristics acquired by the FIB in-situ probe.

[0043] The adaptive correction mechanism, based on real-time predicted processing results and the desired processing objectives, uses a reward function and a local optimization algorithm to infer the optimal process parameters. Specifically, it employs a local optimization algorithm (such as a local Bayesian optimization algorithm, a local genetic algorithm, a local particle swarm optimization algorithm, or a local deep reinforcement learning algorithm) to optimize in a multi-dimensional process parameter space under stricter process parameter boundary conditions, using the current process parameters as a benchmark. This re-maximizes the reward function, enabling hot updates of the process parameters, which are then sent to the FIB control system. The boundary conditions are selected based on a mapping between the predicted exceedance feature type and the process parameter sensitivity. Different process parameters have different weights on the final etching result. For example, the accelerating voltage is the most crucial physical factor determining the high-energy ion implantation depth and the thickness of the microcrystalline lattice damage layer, while beam intensity and residence time have a greater impact on the macroscopic etching rate and morphology.

[0044] In one specific embodiment, the actual damaged layer was detected. There exists a limit to the tolerance of lossless layers. When the evolution trend is observed, an adaptive correction mechanism is adopted to forcibly block high-voltage injection, dynamically reduce the accelerating voltage to a safe threshold of <15KV, and rematch the appropriate beam intensity and residence time to obtain the process parameters corresponding to the maximum value of the reward function. This ensures the macroscopic morphology depth and stoichiometry while minimizing microscopic physical damage.

[0045] The solution of this invention provides real-time monitoring of FIB etching, further realizing a fully closed-loop intelligent processing system that enables controllable morphology, predictable damage, and reverse parameter calculation from "target analysis" to "parameter back-calculation" and then to "dynamic correction," thereby further improving processing accuracy and product quality.

[0046] Furthermore, the present invention also relates to a focused ion beam etching apparatus, which performs FIB etching processing using the above-described focused ion beam etching method.

[0047] Compared to existing technologies, the solution of this invention aligns and integrates the in-situ macroscopic feedback of FIB processing with the real physical processing results of non-in-situ microscopic processing through a deep learning framework. This not only breaks down the interdisciplinary mapping barriers between "morphology-performance-process", but also transforms the process knowledge scattered in personal experience into storable, computable, and transferable model data, forming a core process database for enterprises. Furthermore, through the model's learning and training on effective and successful processing samples, it can back-calculate the optimal etching process parameters for the target processing structure in a short time, significantly shortening the process development cycle. This is highly efficient and accurate, greatly reducing reliance on operator experience, improving the stability of product processing quality, lowering talent training costs and equipment usage thresholds, and effectively avoiding common problems in manual trial and error such as over-etching, sidewall tilting, and surface damage, thus improving processing consistency and yield. Furthermore, the closed-loop monitoring based on dynamic hot updates breaks through the limitations of traditional equipment that relies solely on post-process offline characterization, enabling real-time early warning of microscopic damage trends during the processing. It also utilizes local optimization algorithms for rapid and adaptive parameter correction and hot updates, significantly improving the yield of extreme processes and minimizing the processing damage layer while ensuring morphological accuracy.

[0048] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and the present invention also intends to include these modifications and variations.

Claims

1. A focused ion beam etching method, characterized in that, Includes the following steps: Determine the processing objectives for FIB etching; Based on the processing objectives of FIB etching, global optimization is performed in a multi-dimensional process parameter space to obtain the process parameters of FIB etching. The multi-dimensional process parameter space consists of beam current intensity, accelerating voltage, and residence time. Based on the obtained FIB etching process parameters, perform the FIB etching process; The process involves global optimization within a multi-dimensional process parameter space to obtain the process parameters for FIB etching, based on the processing objective of the FIB etching process. This includes: inputting initial in-situ characterization data and process parameters from the multi-dimensional process parameter space into a prediction model to obtain the predicted processing result under the given process parameter conditions; obtaining a reward value through a reward function based on the predicted processing result and the determined processing objective of the FIB etching process; and obtaining the process parameters for the next iteration through an optimization algorithm based on the reward value and the corresponding process parameters. Ultimately, this process achieves global optimization within the multi-dimensional process parameter space, obtaining the process parameters corresponding to the maximum reward value as the process parameters for the FIB etching process.

2. The method according to claim 1, characterized in that, The optimization algorithm is a Bayesian optimization algorithm, a genetic algorithm, a particle swarm optimization algorithm, or a deep reinforcement learning algorithm.

3. The method according to claim 1 or 2, characterized in that, The reward function is a linear combination function that sets a penalty term based on the characteristics corresponding to the processing target of FIB etching.

4. The method according to claim 3, characterized in that, The processing target includes processing requirements for microscopic processing indicators, and the prediction model is trained by fusing in-situ characterization features and corresponding non-in-situ microscopic characterization features, as shown below: in, Indicates in-situ characterization features. These are non-in-situ microscopic characterization features. This is a vector of input process parameters including ion beam current intensity, accelerating voltage, and residence time. To predict the processing results, output a state vector. This is a deep learning feature fusion regression model obtained after training.

5. The method according to claim 4, characterized in that, The prediction model is trained using the following methods: First, standard structure etching is performed using a FIB (Fiber Embedded Ink) device to obtain multiple sets of in-situ characterization data of the FIB process and corresponding non-in-situ microscopic characterization data, thereby establishing multiple sets of multi-dimensional physical correlation datasets. Then, feature extraction is performed on the multi-dimensional physical correlation datasets to extract the process parameter features of the FIB process. Next, the characterization features extracted from the multi-dimensional physical correlation datasets and their corresponding process parameter features are spliced ​​and fused in a multi-dimensional feature space, and joint inference is performed through a regression network to finally train the prediction model.

6. The method according to claim 5, characterized in that, The representational features in in-situ representational data and in-situ microscopic representational data can be extracted by using two parallel convolutional neural networks or visual Transformers.

7. The method according to any one of claims 1-6, characterized in that, The process also includes the following steps: real-time acquisition of in-situ characterization data of FIB etching, and real-time prediction of the processing results based on the in-situ characterization data and the corresponding FIB etching process parameters through a prediction model. If the predicted processing results show an evolution trend greater than the set processing target, an adaptive correction mechanism is adopted to adjust the process parameters of FIB etching.

8. The method according to claim 7, characterized in that, The adaptive correction mechanism, based on the real-time predicted processing results and the required processing objectives, optimizes the multi-dimensional process parameter space under the set process parameter boundary conditions using a reward function and a local optimization algorithm, and re-maximizes the reward function.

9. The method according to any one of claims 4-8, characterized in that, The micro-processing parameters include the ultimate non-destructive layer tolerance and / or stoichiometry.

10. A focused ion beam etching apparatus, wherein the apparatus performs FIB etching using the method described in any one of claims 1-9.

Images