Droplet discharge device and droplet discharge method

The integration of machine learning and imaging in electrostatic inkjet printing technologies stabilizes droplet ejection by adapting to substrate patterns, ensuring uniform droplet discharge and pattern formation.

JP2024162906APending Publication Date: 2024-11-21SIJTECHNOLOGY INC
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023078885
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-11
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Inkjet printing technologies using electrostatic ejection methods face challenges in uniformly ejecting droplets due to variations in patterns on the substrate, leading to inconsistent droplet sizes and formation.

Method used

A droplet ejection method utilizing machine learning to generate electrostatic droplet ejection conditions based on substrate patterns, incorporating imaging and machine learning models to adjust parameters like discharge voltage, frequency, and nozzle distance, ensuring stable droplet ejection regardless of substrate patterns.

Benefits of technology

Stable and uniform droplet ejection is achieved, allowing for consistent pattern formation on substrates using electrostatic methods, with the ability to adjust conditions dynamically and instantaneously based on pattern recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024162906000001_ABST
    Figure 2024162906000001_ABST
Patent Text Reader

Abstract

To provide a droplet discharge device which performs electrostatic droplet discharge stably without being affected by a pattern at the substrate side.SOLUTION: A droplet discharge method according to one embodiment includes: imaging a pattern on a substrate; obtaining first imaging data corresponding to the pattern; and applying the first imaging data to a machine learning model to generate droplet discharge conditions. The droplet discharge method may further include generating the machine learning model by machine learning targeted at the first imaging data obtained in advance.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a droplet ejection device and a droplet ejection method. [Background technology]

[0002] In recent years, inkjet printing technology has been applied to industrial processes. One example is the manufacturing process of color filters for liquid crystal displays. Conventionally, so-called piezo-type heads that eject droplets by mechanical pressure or vibration have been widely used as inkjet printing technology, but electrostatic ejection inkjet heads that can eject finer droplets have been attracting attention. Patent Document 1 discloses an electrostatic ejection inkjet recording device. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 10-34967 Summary of the Invention [Problem to be solved by the invention]

[0004] On the other hand, the patterns provided on the base material are various and not uniform. In the case of the piezoelectric method, droplets can be discharged without being affected by the pattern on the substrate side, but in the case of droplet discharge using the electrostatic method, droplets of a uniform size cannot be discharged (patterns cannot be formed uniformly) because the droplet discharge conditions differ between areas with patterns and areas without patterns.

[0005] SUMMARY OF THE PRESENTLY PREFERRED EMBODIMENTS Accordingly, an object of the present invention is to stably perform droplet ejection using an electrostatic method without being affected by the pattern on the substrate side. [Means for solving the problem]

[0006] According to one embodiment of the present invention, there is provided a droplet ejection method including imaging a pattern on a substrate, imaging the pattern on the substrate, obtaining first imaging data corresponding to the pattern, applying the first imaging data to a machine learning model to perform machine learning, and generating droplet ejection conditions using an electrostatic method.

[0007] The droplet ejection method may further include generating the machine learning model by machine learning using first imaging data acquired in advance.

[0008] The droplet ejection method may further include imaging the shape of droplets ejected from an electrostatic droplet ejection nozzle, and the machine learning model may be a learning model that has learned the correspondence between the pattern and the shape of the droplets.

[0009] In the droplet discharging method, the machine learning model may be a learning model that has learned a correspondence relationship between the pattern and the conductivity of the substrate.

[0010] In the above-described droplet ejection method, the machine learning model may be a learning model that has learned a correspondence relationship between the pattern and at least one of a shape of an electrostatic droplet ejection nozzle and humidity.

[0011] In the droplet discharging method, the machine learning model may be a learning model that has learned a correspondence relationship between the pattern and a contact angle of the droplet with respect to the substrate.

[0012] In the above droplet ejection device, the droplet ejection conditions may include at least one of an ejection voltage, a frequency, and a distance between the substrate and the electrostatic droplet ejection nozzle.

[0013] According to one embodiment of the present invention, there is provided a droplet ejection device including an imaging unit that images a pattern on a substrate, and a control unit that acquires first imaging data corresponding to the pattern, applies the first imaging data to a machine learning model to perform machine learning, and generates droplet ejection conditions using an electrostatic method.

[0014] In the droplet ejection device, the control unit may generate the machine learning model by machine learning targeting first imaging data acquired in advance.

[0015] The above-mentioned droplet ejection device may further include a second imaging unit that images the shape of droplets ejected from an electrostatic droplet ejection nozzle, and the machine learning model may be a learning model that has learned the correspondence between the pattern and the shape of the droplets.

[0016] In the droplet ejection device, the machine learning model may be a learning model that has learned a correspondence relationship between the pattern and the conductivity of the substrate.

[0017] In the above droplet ejection device, the machine learning model may be a learning model that has learned a correspondence relationship between the pattern and at least one of a shape of an electrostatic droplet ejection nozzle and humidity.

[0018] In the droplet ejection device, the machine learning model may be a learning model that has learned a correspondence relationship between the pattern and a contact angle of the droplet with respect to the substrate.

[0019] In the above droplet ejection device, the droplet ejection conditions may include at least one of an ejection voltage, a frequency, and a distance between the substrate and an electrostatic droplet ejection nozzle. Effect of the Invention

[0020] By using one embodiment of the present invention, droplet ejection by an electrostatic method can be stably performed without being affected by the pattern on the substrate side. [Brief description of the drawings]

[0021] [Figure 1] 1 is a schematic diagram of a droplet ejection device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a functional block diagram of a control unit according to an embodiment of the present invention. [Diagram 3]4 is a droplet ejection condition data set according to an embodiment of the present invention. [Figure 4] FIG. 4 is a flow diagram of droplet ejection according to an embodiment of the present invention. [Diagram 5] FIG. 4 is a flow diagram of a learning process according to an embodiment of the present invention. [Figure 6A] FIG. 2 is a schematic diagram of imaging data. [Figure 6B] FIG. 2 is a schematic diagram of imaging data. [Figure 7A] FIG. 2 is a schematic diagram of imaging data. [Figure 7B] FIG. 2 is a schematic diagram of imaging data. [Figure 7C] FIG. 2 is a schematic diagram of imaging data. [Figure 8A] FIG. 2 is a schematic diagram of imaging data. [Figure 8B] FIG. 2 is a schematic diagram of imaging data. [Figure 8C] FIG. 2 is a schematic diagram of imaging data. [Figure 9] FIG. 4 is a flow diagram of droplet ejection according to an embodiment of the present invention. [Figure 10] FIG. 4 is a flow diagram of droplet ejection according to an embodiment of the present invention. [Figure 11] 1 is a schematic diagram of a droplet ejection device according to an embodiment of the present invention. [Figure 12] FIG. 2 is a functional block diagram of a control unit according to an embodiment of the present invention. [Figure 13] 1 is a schematic diagram of a droplet ejection device according to an embodiment of the present invention. [Figure 14A] 2 is a schematic enlarged top view of a portion of the multi-nozzle head 151. FIG. [Figure 14B] 2 is a schematic cross-sectional view showing an enlarged portion of a multi-nozzle head 151. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Hereinafter, each embodiment of the invention disclosed in this application will be described with reference to the drawings. However, the present invention can be embodied in various forms without departing from the gist of the invention, and the present invention should not be interpreted as being limited to the description of the embodiments exemplified below.

[0023] In the drawings referred to in this embodiment, the same parts or parts having similar functions are given the same or similar symbols (symbols consisting of A, B, or -1, -2, etc., suffixed to a number) and their repeated explanation may be omitted. Also, the dimensional ratios in the drawings may differ from the actual ratios for the convenience of explanation, and some components may be omitted from the drawings.

[0024] Furthermore, in the detailed description of the present invention, when defining the positional relationship between a certain component and another component, "above" and "below" do not only mean a component being located directly above or below another component, but also include a component being located between the other components, unless otherwise specified.

[0025] First Embodiment (1-1. Configuration of the Droplet Discharge Device 100) FIG. 1 is a schematic diagram of a droplet ejection device 100 according to one embodiment of the present invention.

[0026] The droplet ejection device 100 includes a control unit 110 , a storage unit 115 , a power supply unit 120 , a drive unit 130 , a droplet ejection unit 140 , an imaging unit 160 , and a substrate holding unit 190 .

[0027] The control unit 110 includes a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other arithmetic processing circuit. The control unit 110 controls the discharge process of the droplet discharge unit 140 using a preset droplet discharge program.

[0028] The storage unit 115 functions as a database for storing a droplet ejection program and various information used in the droplet ejection program. The storage unit 115 is implemented by a memory, an SSD, or a memory-capable element.

[0029] The power supply unit 120 is connected to the control unit 110, the storage unit 115, the drive unit 130, the droplet discharge unit 140, and the substrate holding unit 190. The power supply unit 120 applies a voltage to the droplet discharge unit 140 based on a signal input from the control unit 110. In this example, the power supply unit 120 applies a pulsed voltage to the droplet discharge unit 140. Note that the voltage is not limited to a pulsed voltage, and a constant voltage may be constantly applied. The voltage applied from the power supply unit 120 to the droplet discharge unit 140 causes droplets to be discharged from a tip 141a of a droplet discharge nozzle 141 described later in the direction of the substrate 200 (third direction D3).

[0030] The driving unit 130 is composed of driving members such as a motor, a belt, and gears. The driving unit 130 moves the droplet discharge unit 140 (droplet discharge nozzle 141) relatively (in this example, in a first direction D1 or a second direction D2; the first direction D1 and the second direction D2 intersect (are perpendicular)) with respect to the substrate holding unit 190 based on an instruction from the control unit 110. In this way, the driving unit 130 controls the relative positional relationship between the substrate 200 and the droplet discharge unit 140 when the device is in use (when discharging droplets).

[0031] The driving unit 130 may suitably fix the droplet discharge unit 140 (droplet discharge nozzle 141) and move the substrate 200. The driving unit 130 may be used in combination with a position adjustment mechanism such as a goniostage to finely adjust the droplet discharge nozzle 141.

[0032] The droplet ejection unit 140 includes a droplet ejection nozzle 141 and an ink tank 143. An electrostatic inkjet nozzle is used for the droplet ejection nozzle 141. Therefore, the droplet ejection nozzle is also called an electrostatic droplet ejection nozzle. The inner diameter of the tip 141a of the droplet ejection nozzle 141 is several hundred nm to 20 μm, preferably 1 μm to 15 μm, and more preferably 5 μm to 12 μm.

[0033] The droplet discharge nozzle 141 has a glass tube, and an electrode 145 is provided inside the glass tube. In this example, a thin tungsten wire is used for the electrode 145. Note that the electrode 145 is not limited to tungsten, and nickel, molybdenum, titanium, gold, silver, copper, platinum, or the like may be provided.

[0034] The droplet discharge nozzle 141 of the droplet discharge part 140 is provided perpendicular to the surface of the substrate 200 (the upper surface of the substrate holding part 190).

[0035] The substrate holding unit 190 has a function of holding the substrate 200. In this example, a stage is used as the substrate holding unit 190. There is no particular limitation on the mechanism by which the substrate holding unit 190 holds the substrate 200, and a general holding mechanism is used. In this example, the substrate 200 is vacuum-adsorbed to the substrate holding unit 190. However, the present invention is not limited to this, and the substrate holding unit 190 may hold the substrate 200 using a fixture.

[0036] The substrate 200 refers to a member onto which droplets are discharged from the droplet discharge unit 140. In this example, a glass substrate is used for the substrate 200. The substrate 200 is not limited to a glass substrate. For example, it may be a metal plate or a plate-shaped member made of resin. Furthermore, it is not limited to a substrate, and may be a structure of other shape as long as it can be fixed. Furthermore, patterns P1, P2, and P3 formed of metal wiring or organic resin material are provided on the substrate 200. Furthermore, a counter electrode for droplet discharge may be provided on the substrate 200. At this time, a GND (Ground) potential may be applied to the substrate 200.

[0037] The imaging unit 160 captures an image of a pattern formed on the substrate 200. In this example, a CCD (Charge Coupled Device) type or CMOS (Complementary Metal Oxide Semiconductor) type image sensor (camera) is used for the imaging unit 160. Information acquired by the imaging unit 160 is sent to the control unit 110 and the storage unit 115.

[0038] (1-2. Functional block diagram of the control unit) Fig. 2 is a functional block diagram of the control unit 110. As shown in Fig. 2, the control unit 110 includes, as functional units, an acquisition unit 1101, a learning processing unit 1103, a generation unit 1105, and a discharge instruction unit 1107 in a program (a droplet discharge program) for realizing droplet discharge.

[0039] The acquiring unit 1101 has a function of acquiring imaging data captured by the imaging unit 160. The acquired imaging data is stored in the storage unit 115.

[0040] The learning processing unit 1103 has a function of performing machine learning based on a machine learning model generated using previously acquired imaging data (teacher data).

[0041] The generation unit 1105 has a function of generating droplet discharge conditions based on the learning results. Fig. 3 shows a droplet discharge condition data set 300. In this example, the droplet discharge condition data set 300 includes a pattern identifier 301, a distance 303 between the substrate and the droplet discharge nozzle, a discharge voltage 305, and a frequency 307 for applying a pulse voltage when discharging droplets.

[0042] The discharge instruction unit 1107 has a function of issuing instructions to the droplet discharge unit 140, the power supply unit 120, and the drive unit 130 so as to discharge droplets from the droplet discharge unit 140 based on the generated droplet discharge conditions.

[0043] (1-3.Droplet discharge method) Next, the droplet ejection method according to this embodiment will be described with reference to Figures 4 to 8. Figure 4 is a flow diagram S100 showing the droplet ejection method.

[0044] First, the imaging unit 160 images a pattern formed on the substrate 200 based on an instruction from the control unit 110 (step S101). When imaging the pattern, the entire pattern may be imaged at a wide angle, may be imaged by a scanning operation, or may be imaged partially. The captured imaging data is transmitted to the control unit 110. The acquisition unit 1101 acquires the imaging data (step S103), and the imaging data is stored in the storage unit 115.

[0045] At this time, the control unit 110 may obtain information such as the minimum pattern dimensions, shape, and the presence or absence of steps from the imaging data.

[0046] Next, the learning processing unit 1103 applies the acquired imaging data to a machine learning model to perform machine learning (step S105). The machine learning model is generated as follows.

[0047] Fig. 5 is a flow diagram S1050 of machine learning model generation in this embodiment. As shown in Fig. 5, imaging data for generating a machine learning model is acquired in advance (step S1051). Figs. 6A-B, 7A-C, and 8A-C are schematic diagrams of imaging data. As shown in Figs. 6A-B, 7A-C, and 8A-C, the imaging data acquired in advance is acquired so as to correspond to various patterns.

[0048] In the case of linear patterns, FIG. 6A shows a pattern 400a in which the intervals s between adjacent linear patterns 410 are different. Specifically, in the pattern 400a, a pattern 400a-1 has an interval s1. A pattern 400a-2 has an interval s2. A pattern 400a-3 has an interval s3. The intervals increase in the order of interval s1, interval s2, and interval s3. In this case, the shape of the pattern 400a may be the same. FIG. 6B shows a pattern 400b in which linear patterns 410 having different line widths L are formed. Specifically, in the pattern 400b, a pattern 400b-1 has a line width L1. A pattern 400b-2 has a line width L2. A pattern 400b-3 has a line width L3. The line width L increases in the order of line width L1, line width L2, and line width L3. In this case, the shape of the pattern 400b may be the same.

[0049] In the case of a circular pattern, FIG. 7A shows a pattern 400c in which circular patterns 420 having different diameters R are formed. Specifically, in the pattern 400c, a pattern 400c-1 has a circular pattern 420 with a diameter R1. A pattern 400b-2 has a circular pattern 420 with a diameter R2. A pattern 400b-3 has a circular pattern 420 with a diameter R3. The diameter R increases in the order of diameter R1, diameter R2, and diameter R3. FIG. 7B shows a pattern 400d in which the intervals s between adjacent circular patterns 420 are different. Specifically, in the pattern 400d, a pattern 400d-1 has an interval s1. A pattern 400d-2 has an interval s2. The intervals s increase in the order of interval s1, interval s2. The circular patterns 420 may be the same. FIG. 7C shows a pattern 400e having a complex circular pattern. Specifically, the pattern 400e-1 has a ring shape with a space in the center, the pattern 400e-2 has an ellipse shape, and the pattern 400e-3 has a shape in which two circles are partially joined together.

[0050] 8A, 8B, and 8C show more complicated patterns. In the pattern 400f, multiple patterns are partially overlapped. Specifically, in FIG. 8A, in the pattern 400f-1, two linear patterns 410 are overlapped at the overlapping portion 410fc. In the pattern 400f-2, three linear patterns 410 are overlapped at the overlapping portion 410fc. As a result, the patterns 400f-1 and 400f-2 have steps of different heights. FIG. 8B shows a star-shaped pattern 400g as a pattern having multiple concaves and convexes. FIG. 8C shows a pattern 400h having even more concaves and convexes.

[0051] Next, teacher data is generated based on the previously acquired imaging data (step S1053). At this time, teacher data may be generated arbitrarily based on input from the user. Specifically, teacher data may be generated by inputting the results of an actual discharge process. Furthermore, the teacher data may be one pattern and its corresponding subdivided droplet discharge conditions, or a group of patterns in a predetermined range and its corresponding droplet discharge conditions. In the case of one pattern and its corresponding subdivided droplet discharge conditions, more accurate droplet discharge conditions can be generated (set). In the case of a group of patterns in a predetermined range and its corresponding droplet discharge conditions, a wide range of droplet discharge conditions can be generated (set).

[0052] Next, machine learning is performed using the teacher data and the previously acquired imaging data (step S1055). For the machine learning, a known learning method such as backpropagation or genetic algorithm (GA) may be used. By repeating the machine learning, a machine learning model for generating conditions for discharging droplets (droplet discharging conditions) is generated (step S1057).

[0053] Returning to Fig. 4, machine learning is performed by applying the imaging data to the machine learning model, and as a result, the control unit 110 generates (outputs) droplet discharge conditions (step S107). At this time, as shown in Fig. 3, a droplet discharge condition data set is generated as the droplet discharge conditions, including a pattern identifier 301 corresponding to the acquired imaging data, a distance 303 between the substrate and the droplet discharge nozzle, a discharge voltage 305, and a frequency 307 for applying a pulse voltage when discharging droplets.

[0054] The control unit 110 instructs the droplet discharge unit 140 to discharge droplets according to the generated droplet discharge conditions. Finally, in response to the instruction, the droplet discharge unit 140 discharges droplets onto the substrate (step S109). This completes the droplet discharge method.

[0055] In this embodiment, the pattern on the substrate is used as an input parameter to set optimal droplet discharge conditions (parameters) according to the pattern. This allows droplet discharge by electrostatic method without being affected by the pattern, making it possible to form a uniform pattern. Therefore, by using this embodiment, droplet discharge by electrostatic method can be stably performed without being affected by the pattern on the substrate side.

[0056] In addition, when the droplet ejection conditions generated in this embodiment are subdivided conditions corresponding to one pattern, droplets can be ejected under optimal conditions for each ejection. In addition, when the droplet ejection conditions generated in this embodiment correspond to a group of patterns in a predetermined range, droplets can be ejected under optimal conditions for that range.

[0057] Furthermore, in the case of this embodiment, after an image of a pattern on a substrate is captured, droplet ejection conditions can be generated instantaneously, which makes it possible to eject droplets while dynamically controlling the ejection conditions.

[0058] <Second embodiment> In this embodiment, a droplet ejection method different from that of the first embodiment will be described. Specifically, an example will be described in which the conductivity of the substrate side is applied to a machine learning model to generate droplet ejection conditions. For the sake of explanation, the same configuration as that of the first embodiment will be omitted as appropriate.

[0059] 9 is a flow diagram S100A showing a droplet discharge method in this embodiment. In this embodiment, the acquisition unit 1101 may acquire conductivity data of the substrate (step S104) together with the imaging data (step S103). Then, the learning processing unit 1103 may apply the conductivity of the substrate side together with the imaging data to one machine learning model (step S105A). At this time, the machine learning model is a learning model that has been made to learn the correspondence between the pattern on the substrate and the conductivity of the substrate side.

[0060] In the case of electrostatic droplet ejection, if electric charges are accumulated on the substrate side (inside the substrate or pattern), the voltage to be applied to the droplet ejection nozzle may not be applied appropriately. In this case, even if other droplet ejection conditions (parameters) are appropriate, droplets cannot be ejected stably. Therefore, when the conductivity of the substrate side is low (in the case of an insulating substrate), it is desirable to control the distance between the substrate and the droplet ejection nozzle to be longer. The distance between the substrate and the droplet ejection nozzle may be affected by the size of the pattern. Therefore, by applying the conductivity (or resistivity) of the substrate (or pattern) to the machine learning model, it is possible to set optimal droplet ejection conditions, particularly the optimal distance between the substrate and the droplet ejection nozzle.

[0061] Therefore, by using this embodiment, droplet ejection by the electrostatic method can be performed more stably.

[0062] In this embodiment, an example is shown in which the conductivity on the substrate side is applied to one machine learning model together with the imaging data, but the present invention is not limited to this. For example, the conductivity on the substrate side may be applied to another machine learning model to perform machine learning and generate a correction coefficient corresponding to the conductivity. In this case, new droplet discharge conditions may be generated by combining the correction coefficient with the droplet discharge conditions generated by applying the pattern on the substrate to the machine learning model.

[0063] <Third embodiment> In this embodiment, a droplet discharge method different from that of the first embodiment will be described. Specifically, an example will be described in which the contact angle on the substrate side is applied to a machine learning model to generate droplet discharge conditions. For the sake of explanation, the same configuration as in the first embodiment will be omitted as appropriate.

[0064] Fig. 10 is a flow diagram S100B showing a droplet discharge method in this embodiment. As shown in Fig. 10, in this embodiment, the acquisition unit 1101 may acquire contact angle data on the substrate side (step S104B) together with the imaging data (step S103). The learning processing unit 1103 may apply the contact angle of the substrate together with the imaging data to one machine learning model (step S105B). At this time, the machine learning model is a learning model that has been made to learn the correspondence between the pattern on the substrate and the contact angle on the substrate side.

[0065] Depending on the contact angle of the substrate, a pattern of a predetermined size may not be formed on the substrate. Therefore, depending on the contact angle of the substrate, the generating unit 1105 may add a pre-treatment (liquid repellency treatment, lyophilic treatment) before droplet discharge to the droplet discharge conditions.

[0066] In this embodiment, an example is shown in which the contact angle on the substrate side is applied to one machine learning model together with the imaging data, but the present invention is not limited to this. For example, the contact angle on the substrate side may be applied to another machine learning model to perform machine learning, and a corresponding correction coefficient may be generated. In this case, new droplet discharge conditions may be generated by combining the correction coefficient with the droplet discharge conditions generated by applying the pattern on the substrate to the machine learning model.

[0067] Therefore, by using this embodiment, droplet ejection by the electrostatic method can be performed more stably.

[0068] <Fourth embodiment> In this embodiment, a droplet ejection device different from the first embodiment will be described. Specifically, a droplet ejection device having a second imaging unit will be described. Note that the same parts as those in the first embodiment will be omitted as appropriate.

[0069] Fig. 11 is a schematic diagram of a droplet ejection device 100C. As shown in Fig. 11, the droplet ejection device 100C includes a second imaging unit 170 in addition to the control unit 110, the storage unit 115, the power supply unit 120, the drive unit 130, the droplet ejection unit 140, the imaging unit 160, and the substrate holding unit 190.

[0070] The second imaging unit 170 can capture an image of the shape of droplets discharged from the droplet discharge nozzle 141. The captured imaging data (also referred to as second imaging data) is transmitted to the control unit 110. The acquisition unit 1101 acquires the second imaging data, and the learning processing unit 1103 may apply the second imaging data together with the first imaging data to one machine learning model to perform machine learning. In this case, the machine learning model is a learning model that has been made to learn the correspondence between the pattern on the substrate and the shape of the droplets.

[0071] In the case of this embodiment, since the dynamic shape of the droplet can be grasped, machine learning can be performed using information that has not been input as a parameter until now. This makes it possible to generate more appropriate droplet ejection conditions. Therefore, by using this embodiment, droplet ejection by the electrostatic method can be performed more stably.

[0072] In this embodiment, an example in which the second imaging data is applied to one machine learning model together with the first imaging data is shown, but the present invention is not limited to this. For example, the second imaging data (droplet shape) may be applied to another machine learning model to perform machine learning, and a corresponding correction coefficient may be generated. In this case, new droplet discharge conditions may be generated by combining the correction coefficient with the droplet discharge conditions generated by applying the pattern on the substrate to the machine learning model.

[0073] In the present embodiment, the imaging unit 160 and the second imaging unit 170 are provided separately, but the present invention is not limited to this. The imaging unit 160 and the second imaging unit 170 may be provided as an integrated unit.

[0074] <Fifth embodiment> In this embodiment, a droplet ejection device different from that of the first embodiment will be described. Specifically, an example in which the control unit 110 has an analysis unit as a functional unit will be described. For convenience of explanation, the same configuration as that of the first embodiment will be omitted as appropriate.

[0075] Fig. 12 is a functional block diagram of the control unit 110D. As shown in Fig. 13, the control unit 110D includes, as functional units, an acquisition unit 1101, a learning processing unit 1103, a generation unit 1105, and an ejection instruction unit 1107, as well as an analysis unit 1109, in a program (droplet ejection program) for realizing droplet ejection.

[0076] The analysis unit 1109 has a function of comparing and analyzing the shape or size of droplets discharged under the generated droplet discharge conditions with the shape or size of droplets set in advance, and determining the droplet discharge conditions. The analysis results may be applied to a machine learning model. This makes it possible to generate appropriate droplet discharge conditions. Therefore, by using this embodiment, droplet discharge by the electrostatic method can be performed more stably.

[0077] Sixth Embodiment In this embodiment, a droplet ejection device different from that of the first embodiment will be described. Specifically, an example in which a droplet ejection unit has a multi-nozzle head will be described. For convenience of explanation, the same configuration as that of the first embodiment will be omitted as appropriate.

[0078] (1-1. Configuration of Droplet Discharge Device 100E) 13 is a schematic diagram of a droplet ejection device 100 according to an embodiment of the present invention. As shown in FIG 13, the droplet ejection device 100E includes a control unit 110, a storage unit 115, a power supply unit 120, a drive unit 130, a droplet ejection unit 140E, an imaging unit 160, and a substrate holding unit 190.

[0079] Fig. 14A is a schematic top view of a portion of the multi-nozzle head 151. Fig. 14B is a schematic cross-sectional view of a portion of the multi-nozzle head 151. As shown in Figs. 13, 14A, and 14B, the droplet ejection unit 140E includes an ink tank 143 and a multi-nozzle head 151.

[0080] The multi-nozzle head 151 includes a plate portion 152 and a plurality of second droplet ejection nozzles 153. The multi-nozzle head 151 is used while being fixed to a mount and an attachment (not shown). The second droplet ejection nozzles 153 are electrostatic droplet ejection type inkjet nozzles.

[0081] The plate portion 152 is provided in a plate shape. In this example, the plate portion 152 extends in the first direction D1. The thickness of the plate portion 152 is set appropriately. In this example, the thickness of the plate portion 152 is not less than 10 μm and not more than 100 μm.

[0082] The second droplet discharge nozzles 153 are provided at an upper portion connected to one surface (lower surface) of the plate portion 152. The second droplet discharge nozzles 153 are arranged in a line in the first direction D1, but may also be arranged in the second direction D2 as appropriate. The second droplet discharge nozzles 153 have a tapered shape.

[0083] The plate portion 152 has a through hole 152o having an inner diameter r152o larger than the inner diameter r153a of the discharge port (tip portion 153a of the second droplet discharge nozzle 153) of the second droplet discharge nozzle 153 in a portion (overlapping portion) corresponding to the second droplet discharge nozzle 153. The inner diameter of the through hole of the plate portion 152 may be 1 μm or more and 100 μm or less. The inner diameter of the tip portion 153a of the second droplet discharge nozzle 153 may be several hundreds of nm or more and 50 μm or less, preferably 1 μm or more and 30 μm or less, and more preferably 5 μm or more and 20 μm or less. In this embodiment, a voltage can be applied to the electrode 155 provided on each second droplet discharge nozzle 153. The electrode 155 may be provided with tungsten, nickel, molybdenum, titanium, gold, silver, copper, platinum, or the like.

[0084] In the present embodiment, droplet discharge conditions may be generated for each nozzle according to the pattern on the substrate, which makes it possible to set appropriate droplet discharge conditions according to the presence or absence, shape, size, etc. of a pattern, and also to reduce the pattern formation time by discharging droplets using multiple nozzles.

[0085] In addition, in this embodiment, the shape of the droplet ejection nozzle may be applied to the machine learning model, which makes it possible to generate appropriate droplet ejection conditions depending on whether the nozzle is a single nozzle or a multi-nozzle.

[0086] In this embodiment, an example in which an electrode is provided on each second droplet ejection nozzle 153 has been shown, but the present invention is not limited to this. For example, a voltage may be applied directly to the second droplet ejection nozzle 153 and the plate portion 152 without the electrode 155, or a voltage may be applied to the ink stored in the second droplet ejection nozzle 153. Also, although an example in which a voltage is applied to the second droplet ejection nozzle 153, the plate portion 152, or the ink has been shown, a voltage may be applied to a jig (for example, a mount or attachment) that holds the multi-nozzle head 151.

[0087] (Modification) Within the scope of the concept of the present invention, a person skilled in the art may conceive of various modifications and alterations, and it is understood that such modifications and alterations also fall within the scope of the present invention. For example, those in which a person skilled in the art appropriately adds or deletes components, combines or modifies the designs of the above-mentioned embodiments, or adds, omits, or modifies the conditions of processes, are included in the scope of the present invention as long as they include the gist of the present invention.

[0088] (Variation 1) In the first embodiment of the present invention, an example is shown in which a substrate is imaged and image data is acquired, but the present invention is not limited to this. For example, shape data (design data) corresponding to a pattern formed on a substrate may be acquired and applied to a machine learning model.

[0089] (Variation 2) In the first embodiment of the present invention, an example was shown in which the droplet discharge conditions to be generated include the distance 303 between the substrate and the droplet discharge nozzle, the discharge voltage 305, and the frequency 307 at which a pulse voltage is applied when discharging droplets, but the present invention is not limited to this. For example, the droplet discharge conditions may appropriately include other parameters such as the scanning speed, the discharge amount, and the discharge position (coordinates).

[0090] (Variation 3) In one embodiment of the present invention, the droplet ejection conditions may be generated by applying at least one of the humidity and the diameter of the droplet ejection nozzle to a machine learning model. In this case, the machine learning model may be a learning model that has learned a correspondence relationship between a pattern on a substrate and at least one of the shape and humidity of an electrostatic droplet ejection nozzle. [Explanation of symbols]

[0091] 100 droplet ejection device, 110 control section, 115 memory section, 120 power supply section, 130 drive section, 140 droplet ejection section, 141 droplet ejection nozzle, 141a tip section, 143 ink tank, 145 electrode, 151 multi-nozzle head, 152 plate section, 152o through hole, 153 second droplet ejection nozzle, 153a tip section, 155 electrode, 16 0···imaging section, 170···second imaging section, 190···substrate holding section, 200···substrate, 300···droplet ejection condition data set, 301···pattern identifier, 303···distance, 305···ejection voltage, 410···linear pattern, 410fc···superimposing section, 420···circular pattern, 1101···acquisition section, 1103···learning processing section, 1105···generation section, 1107···ejection instruction section, 1109···analysis section

Claims

1. The pattern on the board is imaged, acquiring first imaging data corresponding to the pattern; Applying the first imaging data to a machine learning model to perform machine learning, and generating a droplet ejection condition by an electrostatic method. Droplet ejection method.

2. Further comprising generating the machine learning model by machine learning targeting first imaging data acquired in advance. The droplet ejection method according to claim 1 .

3. Imaging the shape of droplets discharged from an electrostatic droplet discharge nozzle; The machine learning model is a learning model that has learned the correspondence between the pattern and the shape of the droplet. The droplet ejection method according to claim 1 .

4. The machine learning model is a learning model that has learned a correspondence relationship between the pattern and the conductivity of the substrate. The droplet ejection method according to claim 1 .

5. The machine learning model is a learning model that learns a correspondence relationship between the pattern and at least one of a shape of an electrostatic droplet ejection nozzle and humidity. The droplet ejection method according to claim 1 .

6. The machine learning model is a learning model that learns a correspondence relationship between the pattern and a contact angle of a droplet with respect to the substrate. The droplet ejection method according to claim 1 .

7. The droplet ejection conditions include at least one of an ejection voltage, a frequency, and a distance between the substrate and an electrostatic droplet ejection nozzle. The droplet ejection method according to claim 1 .

8. an imaging unit that images a pattern on a substrate; a control unit that acquires first imaging data corresponding to the pattern, applies the first imaging data to a machine learning model to perform machine learning, and generates conditions for electrostatic droplet ejection; Including, Droplet ejection device.

9. The control unit generates the machine learning model by machine learning on first imaging data acquired in advance. The droplet ejection device according to claim 8.

10. a second imaging unit that images the shape of a droplet discharged from the electrostatic droplet discharge nozzle; The machine learning model is a learning model that has learned the correspondence between the pattern and the shape of the droplet. The droplet ejection device according to claim 8.

11. The machine learning model is a learning model that has learned a correspondence relationship between the pattern and the conductivity of the substrate. The droplet ejection device according to claim 8.

12. The machine learning model is a learning model that learns a correspondence relationship between the pattern and at least one of a shape of an electrostatic droplet ejection nozzle and humidity. The droplet ejection device according to claim 8.

13. The machine learning model is a learning model that learns a correspondence relationship between the pattern and a contact angle of a droplet with respect to the substrate. The droplet ejection device according to claim 8.

14. The droplet ejection conditions include at least one of an ejection voltage, a frequency, and a distance between the substrate and an electrostatic droplet ejection nozzle. The droplet ejection device according to claim 8 .

Citation Information

Patent Citations

  • Ink jet recorder

    JP1998034967A