Sample analysis system, learning model generation method, and sample analysis method

Abstract

A sample analysis system is provided, which includes: a droplet device that intermittently introduces a sample into a measurement area set as plasma; a luminescence detection device that detects the luminescence of the measurement area with a detection timing of a predetermined period set in advance; an analysis device that analyzes the sample based on the detected luminescence, the analysis device including: a distribution calculation unit that calculates the distribution of light intensity corresponding to the detection timing, the position in the measurement area, and the wavelength component of the luminescence, that is, the time / space light intensity distribution, based on the detected luminescence; and a characteristic determination unit that calculates a characteristic quantity related to the sample characteristic that represents the nature of the sample from the time / space light intensity distribution, and determines the sample characteristic based on the characteristic quantity.

Description

Technical Field

[0001] The present invention relates to a sample analysis system, a learning model generation method, and a sample analysis method. Background Art

[0002] In an inductively coupled plasma (ICP) emission analysis device that uses plasma as an atomization source or an ionization source, a sample is supplied to a plasma source to be plasmatized (excited), and the composition of the sample is analyzed based on an emission spectrum obtained by wavelength decomposing light from the plasma.

[0003] In Japanese Patent Publication JP2002-5837A, a spectroscopic analysis device is proposed, which uses a spectrometer to split light from a plasma, detects it in a manner corresponding to the position on the plasma through a plurality of CCD (Charge Coupled Device) photodetectors, and detects the distribution of luminous intensity at each position on the plasma based on the detection result. Summary of the invention

[0004] However, in conventional ICP emission spectrometry including Japanese Patent Publication No. JP2002-5837A, the information that can be determined from the emission spectrum is limited to the constituent elements of the sample. That is, the emission spectrum obtained in conventional ICP emission spectrometry cannot be used to analyze the characteristics of the sample other than the constituent elements.

[0005] Therefore, an object of the present invention is to provide an analysis method capable of determining a wider range of sample characteristics other than constituent elements based on excitation light generated by plasma of the sample.

[0006] According to one embodiment of the present invention, there is provided a sample analysis system, which includes: a droplet device that intermittently introduces a sample into a measurement area set as plasma; a light emission detection device that detects the light emission of the measurement area at a detection timing of a predetermined period set in advance; and an analysis device that analyzes the sample based on the detected light emission. In addition, the analysis device includes: a distribution calculation unit that calculates the distribution of light intensity corresponding to the detection timing, the position in the measurement area, and the wavelength component of the light emission, that is, the time / space light intensity distribution, based on the detected light emission; and a characteristic determination unit that calculates a characteristic quantity related to the sample characteristic that expresses the nature of the sample from the time / space light intensity distribution, and determines the sample characteristic based on the characteristic quantity.

[0007] According to another aspect of the present invention, a method for generating a learned model is provided, which generates a learned model that takes a time / space light intensity distribution associated with a sample as input and a sample characteristic that expresses the nature of the sample as output. The method for generating a learned model includes a process of performing machine learning in a manner that takes a time / space light intensity distribution associated with a known sample as input and a known characteristic that expresses the nature of the known sample as output. In addition, by intermittently introducing a known sample into a measurement area set as plasma, the luminescence generated in the measurement area is detected at a detection timing of a predetermined period set in advance, and the time / space light intensity distribution is obtained as a distribution calculation of the light intensity corresponding to the detection timing, the position on the measurement area, and the wavelength component based on the detected luminescence.

[0008] According to another aspect of the present invention, a method for generating a learned model is provided, wherein the method generates a learned model that takes a time series image group related to a sample as input and a sample characteristic that expresses the properties of the sample as output. The method includes performing a process of mechanical learning by taking a time series image group related to a known sample as input and a known characteristic that expresses the properties of the known sample as output. In addition, the time series image group is obtained by intermittently introducing the known sample into a measurement area set as plasma and photographing the light generated in the measurement area at a predetermined shooting timing of a predetermined period set in advance.

[0009] According to another aspect of the present invention, a sample analysis method is provided, which is performed using the generated learned model. In the sample analysis method, the temporal / spatial light intensity distribution or time series image group obtained for an unknown sample is set as an input to the learned model, and the output data is determined as an unknown characteristic representing the nature of the unknown sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a diagram for explaining the configuration of a sample analysis system common to each embodiment of the present invention.

[0011] Figure 2 A diagram showing the structure of a droplet device.

[0012] Figure 3 A diagram for explaining a measurement area set as plasma.

[0013] Figure 4 This figure shows an example of the form of a time series image group generated by a high-speed camera.

[0014] Figure 5This is a block diagram illustrating the structure of the analyzer.

[0015] Figure 6 This is a diagram for explaining a method of calculating a characteristic value related to a particle size.

[0016] Figure 7 This is a diagram for explaining a method of calculating characteristic quantities related to particle structure.

[0017] Figure 8 This is a diagram showing an example of the distribution of the time spectrum of each droplet.

[0018] Fig. 9 A diagram for explaining the structure of a machine learning model that realizes a characteristic determination unit.

[0019] Fig.10 It is a diagram showing various time series images related to microplastics with a particle size of 3 μm and a particle size of 10 μm in Example 1.

[0020] Fig.11 This is a diagram showing the temporal light intensity distribution associated with microplastics with a particle size of 3 μm and a particle size of 10 μm in Example 1.

[0021] Fig.12 This is a diagram showing the particle size fraction of microplastics contained in the liquid sample of Example 2.

[0022] Fig.13 The diagram shows a group of time-series images acquired with respect to the liquid sample (containing separated Ag particles and Au particles) according to the second embodiment.

[0023] Fig.14 The diagram shows a group of time-series images acquired with respect to the liquid sample (containing core-shell particles composed of Ag and Au) according to Example 2.

[0024] Fig.15 Graph showing the temporal / spatial light intensity distribution of two liquid samples involved in Example 2. DETAILED DESCRIPTION

[0025] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0026] (First Embodiment)

[0027] Figure 1 1 is a diagram for explaining the structure of a sample analysis system 10 according to the present embodiment. As shown in the figure, the sample analysis system 10 includes a droplet device 12, a sample supply control device 14, and a plasma measurement unit 16 as a measurement device.

[0028] Figure 2The diagram shows the structure of the droplet device 12. The droplet device 12 transfers the liquid sample S to be measured in the form of sample droplets S d It is intermittently supplied to the plasma PL in the form of.

[0029] Here, the liquid sample S of this embodiment can be considered as a mixed liquid in which one or more particles p are mixed in a predetermined ratio in a liquid such as water. Hereinafter, the ratio of particles p contained in the mixed liquid constituting the liquid sample S is referred to as “particle concentration Co”. p In addition, at this particle concentration Co p The concept of includes known arbitrary definitions of concentrations such as mass concentration, molar concentration, and volume concentration.

[0030] In addition, in this specification, the particle p includes both single element particles formed by one element in any coupling form (crystalline structure or amorphous structure, etc.), and multiple element particles formed by multiple elements mutually coupled in any coupling form (solid solution, core shell, or support, etc.). In addition, the composition of the particle p refers to one element or multiple elements constituting the particle p. For example, in the case where the particle p is a polystyrene bead, the composition of the particle p is C (carbon) and H (hydrogen).

[0031] Back to Figure 2 The droplet device 12 includes a droplet head 20 and a plasma gun 24 .

[0032] The dropper 20 includes an opening and closing mechanism 20 a and a sample guide 20 b which are provided in this order from the upper side in the vertical direction (the negative direction of the Y axis).

[0033] The opening and closing mechanism 20a is connected to the sample storage container 26 disposed above the droplet device 12 via the sample supply capillary 26a. The opening and closing mechanism 20a is opened and closed so as to switch between a blocking state and a passing state of the liquid sample S maintained in a negative pressure state in the sample supply capillary 26a, thereby transferring the sample droplet S to the sample supply capillary 26a. d It is intermittently released into the sample guide 20b.

[0034] More specifically, the opening and closing mechanism 20a is operated with a predetermined opening and closing cycle ΔT f The sample droplet S is supplied to the sample guide 20b at a temperature of several Hz to several tens of kHz. d According to this structure, the liquid sample S maintained in a negative pressure state in the sample supply capillary 26a serves as the opening and closing mechanism 20a with the opening and closing cycle △T f The corresponding droplet diameter r d The sample droplet S dThe liquid is discharged drop by drop into the sample guide 20b.

[0035] In this embodiment, the opening and closing period ΔT f According to the particle concentration Co in the liquid sample S described above p The droplet diameter r d The switch cycle ΔT is set to an appropriate value. f The specific settings are described later.

[0036] The sample guide 20b is used to guide the sample droplets S released by the opening and closing mechanism 20a. d The sample guide 20b is a sample passage that guides the sample toward the plasma gun 24. More specifically, the sample guide 20b has an opening and closing mechanism 20a installed on its upper portion, and is configured in a cylindrical shape. The cylindrical shape has a sample droplet S formed inside the opening and closing mechanism 20a. d In addition, a carrier gas introduction path 20c is provided on the wall portion of the sample guide 20b.

[0037] The carrier gas introduction path 20c is used to introduce the sample droplets S d The carrier gas introduction path 20c is formed into a notch shape that obliquely intersects with the elongation direction (vertical direction) of the wall of the cylindrical sample guide 20b. That is, the carrier gas introduction path 20c is in a form that obliquely intersects with the wall of the sample guide 20b so that the carrier gas cag has a vertically downward flow direction component in the sample guide 20b.

[0038] Therefore, by flowing the carrier gas cag through the carrier gas introduction path 20c, the sample droplets S in the sample guide 20b can be appropriately moved by the flow of the carrier gas cag. d The light is guided toward the plasma PL.

[0039] In addition, the amount of the carrier gas cag introduced from the carrier gas introduction path 20c can be adjusted according to the sample droplet S. d The droplet diameter r d For example, the volume flow rate of the carrier gas cag can be set in the range of 0 to 1 liter / min. In addition, as the carrier gas cag, the sample droplet S is d From the viewpoint of providing a function of inducing plasma PL while not inhibiting stable generation of plasma PL, it is preferred to use an inert gas, and particularly preferably, argon gas.

[0040] In the structure of the droplet device 12 of the present embodiment, the plasma PL is generated at a position vertically below the droplet head 20. Therefore, the sample droplets S discharged from the opening and closing mechanism 20a into the sample guide 20b are d The sample droplet S moves vertically downward (in the positive direction of the Y axis) toward the plasma PL. d Since the sample droplets S are guided toward the plasma PL by the action of gravity, even without using the carrier gas cag, the sample droplets S can be appropriately moved. d Therefore, with the structure of the droplet device 12 of this embodiment, the volume flow rate of the carrier gas cag can also be set to a relatively low range (for example, 0 to 0.1 liter / min).

[0041] On the other hand, the plasma gun 24 is connected to the lower end of the sample guide 20b via a connector 28. In addition, the plasma gun 24 is provided with a coil 22 as a plasma generating unit for generating plasma PL. The plasma gun 24 functions as a guide path that has a supply path for various gases for stable generation and cooling of plasma PL and guides the sample droplets S from the sample guide 20b to the sample guide 20b. d Induction pathway to plasma PL.

[0042] More specifically, the plasma gun 24 includes a gun body 24 a , a sample capillary 24 b as a sample passage, and a cooling gas supply passage 24 c .

[0043] The upper end of the gun body 24a is connected to the lower end of the sample guide 20b in the dripper 20 via the connector 28. The gun body 24a is formed into a substantially cylindrical shape with a sample capillary 24b formed therein by a material such as quartz.

[0044] In addition, an auxiliary gas introduction pipe 24d is connected to the side wall of the gun body 24a, and the auxiliary gas introduction pipe 24d introduces auxiliary gas ag (intermediate gas) for making the generated plasma PL float from the gun body 24a around the sample guide 20b (the relatively upper area in the vertical direction).

[0045] More specifically, the assist gas introduction pipe 24d is connected to the gun body 24a so as to communicate with the space between the gun body 24a and the sample capillary 24b. Therefore, the assist gas ag introduced through the assist gas introduction pipe 24d flows toward the plasma PL through the space between the gun body 24a and the sample capillary 24b.

[0046] In addition, the amount of the auxiliary gas ag introduced from the auxiliary gas introduction pipe 24d can be arbitrarily set from the viewpoint of how much the distance from the lower end of the gun body 24a to the generation position of the plasma PL (the distance for making the plasma PL float) is set. For example, the volume flow rate of the auxiliary gas ag can be set within the range of 1 to 1.5 liters / min. In addition, as the auxiliary gas ag, from the viewpoint of achieving the function of making the plasma PL float from the gun body 24a while not hindering the stable generation of the plasma PL, it is preferable to use an inert gas with low reactivity, and argon gas is particularly preferable.

[0047] In addition, the sample capillary 24b is provided inside the gun body 24a in a manner extending from the lower end of the sample guide 20b to the lower end of the coil 22. More specifically, the upper end of the sample capillary 24b is connected to the inside of the sample guide 20b via the connector 28, and the lower end extends in the vertical up-down direction toward the vicinity of the upper end of the coil 22. In particular, the length of the sample capillary 24b is set to be greater than a predetermined value determined from the viewpoint of suppressing the influence on the dripper 20 caused by the magnetic field generated by the coil 22. Therefore, it is preferable to configure the sample capillary 24b in a manner such that the sum of the extension length of the sample guide 20b in the vertical direction and the length of the sample capillary 24b (i.e., the distance L from the lower end of the opening and closing mechanism 20a to the upper end of the coil 22) is, for example, several tens of cm.

[0048] In addition, a cooling gas supply path 24c for supplying cooling gas cg (cold gas) is formed on the outer periphery of the gun body 24a below the auxiliary gas introduction pipe 24d. In addition, the cooling gas cg is a gas that acts as a protective gas for the gun body 24a while blocking the plasma PL from being affected by the external gas.

[0049] In particular, the cooling gas supply path 24c is made of a material such as quartz and is formed into a cylindrical shape. The cooling gas supply path 24c ensures a gap between the cooling gas cg and the outer peripheral surface of the gun body 24a, and covers the substantially entire circumferential area of ​​the gun body 24a. In this way, since the cooling gas cg flows downward in the vertical direction in a manner of surrounding the gun body 24a and the sample capillary 24b therein from the periphery, a cooling function for the gun body 24a and the sample capillary 24b is achieved. In addition, the lower end of the cooling gas supply path 24c exceeds the lower end of the gun body 24a and further extends downward in the vertical direction. Therefore, since the cooling gas cg flows in a manner of surrounding the periphery of the plasma PL, a function of blocking the plasma PL from being affected by external gas is appropriately achieved.

[0050] In addition, the amount of cooling gas cg introduced into the cooling gas supply path 24c can be arbitrarily set while obtaining a balance with the amount of carrier gas cag according to the state of plasma PL. For example, the volume flow rate of cooling gas cg can be set in the range of 12 to 15 liters / min. In addition, as cooling gas cg, from the viewpoint of realizing the cooling function of the gun body 24a and the protection function of plasma PL while not hindering the stable generation of plasma PL, it is preferable to use an inert gas with low reactivity, and argon gas is particularly preferable.

[0051] In addition, the above-mentioned coil 22 is wound around the outer periphery of the cooling gas supply path 24c at a position near the lower end of the plasma gun 24. The coil 22 generates a magnetic field for generating plasma PL by receiving the supply of alternating current from a power supply device (not shown). Plasma PL is generated by the action of this magnetic field. In addition, by appropriately adjusting the frequency and amplitude of the alternating current supplied to the coil 22, the state of the plasma PL can be appropriately adjusted.

[0052] Next, return to Figure 1 , the configuration of the sample supply control device 14 and the plasma measurement unit 16 will be described. The sample supply control device 14 controls the sample droplet S implemented by the droplet device 12. d The sample supply control device 14 of this embodiment is composed of a droplet controller 30 and a pulse generator 32.

[0053] The droplet controller 30 controls the sample droplet S d The introduction timing (opening and closing cycle △T of the opening and closing mechanism 20a) f More specifically, the droplet controller 30 controls the opening and closing mechanism 20a by an opening and closing cycle ΔT specified by a synchronization signal generated in the pulse generator 32. f The applied voltage is adjusted by switching on and off.

[0054] The pulse generator 32 detects the frame rate of the high-speed camera 38 according to the detection unit time ΔT. u The sample droplet S from the droplet device 12 is generated. d The pulse generator 32 is outputted as a synchronization signal that synchronizes the supply timing of the plasma PL with the detection time t of the light emission of the plasma PL to the droplet controller 30. In addition, when the structure is required to be simplified compared with the measurement accuracy for simple analysis applications, the pulse generator 32 can be omitted as appropriate.

[0055] The plasma measuring unit 16 includes a quartz lens 33 , a spectrometer 34 , an image intensifier 36 as a detection device, a high-speed camera 38 , and an analyzer 39 .

[0056] The quartz lens 33 focuses the light emitted by the plasma PL toward the spectrometer 34. In particular, the quartz lens 33 is configured to focus the measurement base point P O0 To the end point P O2 The light emission of the plasma PL in the region between (hereinafter also simply referred to as “measurement region MA”) is focused on the spectrometer 34 .

[0057] The spectrometer 34 measures the wavelength component λ N The spectrometer 34 decomposes the light focused in the quartz lens 33 by light (N=1, 2, 3, ...). More specifically, the spectrometer 34 images the light focused in the quartz lens 33 in such a manner that the wavelength components λ1, λ2, λ3, ... are arranged along the Y-axis direction. The spectrometer 34 is, for example, composed of a grating corresponding to the required wavelength resolution. In addition, the wavelength resolution of the spectrometer 34 is preferably in the order of 1 / 100 nm, for example, 0.04 nm or less.

[0058] The image intensifier 36 generates a spatial distribution of the light while amplifying the light decomposed by the spectrometer 34. More specifically, the image intensifier 36 detects the light decomposed by the spectrometer 34 as a two-dimensional image corresponding to the spatial distribution (vertical position and horizontal position) of the measurement area MA.

[0059] The high-speed camera 38 continuously captures the two-dimensional image detected by the image intensifier 36 at a preset frame rate (e.g., tens of thousands to millions of fps), and generates a temporal / spatial optical aspect ratio I d (t, X, Y) is a time series image group I as information m (t). The high-speed camera 38 generates the time series image group I m (t) is stored in a predetermined storage area as digital image data. That is, in this embodiment, the detection unit time ΔT u The time (eg, several hundreds of ns to several hundreds of μs) is determined by the frame rate set in the high-speed camera 38 .

[0060] In particular, in this embodiment, it is preferred to detect the unit time ΔT u The opening and closing cycle ΔT of the opening and closing mechanism 20a f The frame rate of the high-speed camera 38 is set to be sufficiently smaller than that of the sample droplet S supplied from the droplet device 12. d In the process of being excited while moving in the measurement area MA, the process is photographed multiple times. More specifically, it is preferable to capture several to several thousand frames of images during the process. u The opening and closing cycle △T fThe frame rate is about 1 / 100 to 1 / 100000 times.

[0061] Next, the time series image group I obtained by the image intensifier 36 and the high-speed camera 38 is further analyzed in detail. m The generation of (t) is described below.

[0062] Figure 3 1 is a diagram schematically showing a measurement area MA of plasma PL set in the present embodiment. As shown in the figure, the measurement area MA is set to be a measurement area of ​​plasma PL from a measurement base point P O0 From the beginning to the end point P O2 The horizontal width and vertical length of the area are arbitrarily set between . Figure 3 In FIG. 1 , for reference, a sample droplet S moving downward in a substantially vertical direction is shown. d The vertical position where excitation (luminescence) starts, i.e., the luminescence starting point P O1 .

[0063] In this embodiment, the image intensifier 36 can detect the state of light based on the spatial distribution of the light along the sample droplet S in the measurement area MA. d The vertical position (vertical position Y) of the shifted light and the light distributed to each wavelength component λ according to the resolution of the spectrometer 34 are shown in FIG. N The position (horizontal position X) is composed.

[0064] Figure 4 The time series image group I generated by the high-speed camera 38 is shown as m (t) is a diagram showing an example of the morphology.

[0065] In this embodiment, the high-speed camera 38 generates a detection unit time ΔT u Multiple time series images i m_t (exist Figure 4 In the figure, five time series images i m_t0 ~i m_t4 ) constitutes a time series image group I m (t).

[0066] That is, at each detection unit time △T u Each time series image i m_t In the generated sequence (time series sequence), the detection timing (detection time t) can be assigned. m The image coordinates in the longitudinal direction of (t) correspond to the sample droplet S in the measurement area MA. dThe moving path of particle p, i.e., the vertical position Y, the time series image group I m The horizontal image coordinates of (t) correspond to the wavelength component λ N Therefore, hereinafter, the time series image group I is represented by the same symbols (X, Y) in the vertical direction (Y-axis direction) and the horizontal direction (X-axis direction) in the measurement area MA. m (t) is the image coordinate of the image.

[0067] like Figure 4 As shown, in the time series image group I m (t) includes the sample droplet S d The excitation light E of the liquid lq lp and the excitation light E from particle p p In particular, Figure 4 In the example shown, the excitation light E of the particle p to be detected p From the time series image i at the detection time t=t2 m_t2 In addition, in this time series image i m_t2 The time series images i at the subsequent detection times t=t3 and t4 m_t3 And the time series image i m_t4 , excitation light E p The shape has changed.

[0068] The present inventors have focused on this point and found that each time series image i m_t The excitation light E of particle p p The change in the morphology of is related to the properties (particle size, particle shape, and particle structure, etc.) related to the particles p contained in the liquid sample S in addition to the composition. That is, the inventors have come up with the following idea, that is, by m (t) By extracting a feature value FV related to properties other than composition related to particles p contained in the liquid sample S (hereinafter also referred to as "sample characteristics Ch"), it is possible to analyze the characteristics of the liquid sample S that cannot be analyzed in the existing ICP emission spectrometry analysis.

[0069] Furthermore, the present inventors have found that it is preferable to use the excitation light E of the particle p to be measured. p More reliably included in the time series image group I m (t) according to the particle concentration Co of the liquid sample S. p The opening and closing cycle ΔT of the opening and closing mechanism 20a is adjusted f .

[0070] More specifically, it can be imagined that the particle concentration Co of the liquid sample S is p When the relative high probability is high, the high-speed camera 38 can capture the sample droplet S moving in the measurement area MA with a higher probability. d Therefore, in the time series image group I m (t) is more likely to contain the excitation light E from particle p p On the other hand, when the particle concentration Co of the liquid sample S is p In relatively low cases, based on the opposite reason, in the time series image group I m (t), it is more difficult to include the excitation light E from particle p p .

[0071] Therefore, in this embodiment, from the viewpoint of further improving the calculation accuracy of the feature value FV for determining the sample characteristic Ch, the particle concentration Co of the liquid sample S is used. p The opening and closing cycle ΔT of the opening and closing mechanism 20a is adjusted f More specifically, the particle concentration Co p The lower the value, the larger the sample droplet S d The droplet diameter r d In a sample droplet S d The opening and closing period △T is increased in such a way that the particle p is included with a higher probability. f Set shorter.

[0072] In addition, it can be imagined that in the sample droplet S d In the case where the particle size, particle shape, and particle structure of the particles are different from each other, and the particle concentration Co p In the case of relatively high m_t In the above, the excitation light E of the particles p with different characteristics p Therefore, in this case, the excitation light E p From the perspective of overlap, in order to reduce the d The number of particles p contained in the f is set to be relatively short. In addition, the particle size in the present embodiment refers to the particle diameter of the particle p. In particular, assuming that the particle p is spherical, the particle size is equivalent to the diameter corresponding to the composition (type of atoms) constituting the sphere and the particle diameter determined according to the number of atoms. In addition, in addition to this, the particle size can also be a particle diameter determined by comparison with a measurement using a predetermined labeled particle.

[0073] Next, the structure of the analyzer 39 of the plasma measurement unit 16 is described. The analyzer 39 of this embodiment is composed of a controller including a CPU (Central Processing Unit) and other computing / control devices, RAM (Random Access Memory), ROM (Read Only Memory), SSD (Solid State Drive), or hard disk (magnetic storage device) and other storage devices, as well as a keyboard, mouse, touch panel, display, printer, and I / O port and other various input and output devices. In addition, the following Figure 5 The functions described in the above are implemented by the above-mentioned hardware and the program (software) stored in the storage device.

[0074] Figure 5 This is a block diagram for explaining the configuration (function) of the analyzer 39. As shown in the figure, the analyzer 39 includes an image analysis unit 40, a characteristic identification unit 41, and a known sample DB 42.

[0075] From the viewpoint of obtaining the feature value FV related to the above-mentioned sample characteristic Ch, the image analysis unit 40 analyzes the time series image group I generated by the high-speed camera 38. m (t) to perform analysis.

[0076] In particular, the image analysis unit 40 of this embodiment analyzes the time series image group I m (t), the time series image group I is calculated m (t) each time series image i m_t The wavelength component λ of the light contained in N The corresponding image coordinates X and the sample droplet S in the measurement area MA d The image coordinate Y corresponding to the position of m_t The corresponding detection time t as the time / space spectrum I d More specifically, the image analysis unit 40 calculates the time series images i m_t The sample droplet S contained in d The temporal / spatial spectrum I of the intensity of the excitation light E (corresponding to the signal intensity of the image pixel of the excitation light E) d (t, X, Y).

[0077] In addition, the image analysis unit 40 of the present embodiment appropriately calculates the time / space spectrum I in an arbitrary range of the image coordinate Y as needed. d The time / wavelength spectrum I obtained after integrating (t, X, Y) d1(t, X), the time / space spectrum I in any range of image coordinates X d The time / position spectrum I obtained after integrating (t, X, Y) d2 (t, Y), and the time / space spectrum I in any range of both the image coordinate Y and the image coordinate X d The time spectrum I obtained by integrating (t, X, Y) d3 (t).

[0078] Here, for the time / wavelength spectrum I d1 The integration range of the image coordinate Y when (t, X) is calculated is based on the time / wavelength spectrum I d1 (t, X) is preferably appropriately set so as to include a signal of the intensity of the excitation light E. More specifically, it is preferable to set the time series image i m_t The area affected by the noise light intensity component (backlight of the background, etc.) contained in (the area with relatively high S / N) is set as the integration range of the image coordinate Y.

[0079] In addition, preferably, the time / position spectrum I d2 The integration range of the image coordinate X when calculating (t, Y) is determined as, and constitutes the sample droplet S d Elements (especially, sample droplet S d The wavelength region corresponding to the particles p) contained in the sample droplet S d When an element is unclear, the element can be determined in advance using an existing analysis method and the corresponding wavelength region can be set as the integration range, or the region obtained by adding the wavelength regions corresponding to some conceivable candidate elements can be set as the integration range.

[0080] The characteristic determination unit 41 determines the time / space spectrum I calculated by the image analysis unit 40. d (t, X, Y) and calculate the above-mentioned feature value FV.

[0081] Here, the feature value FV of this embodiment is the time series image group I m (t) or time / space spectrum I d (t, X, Y) is a parameter of one variable or multiple variables, and is a parameter uniquely related to the sample characteristics Ch of the liquid sample S (composition, particle size, particle shape, and particle structure, etc.).

[0082] An example of calculation of the feature value FV will be described in detail.

[0083] Figure 6This is a diagram for explaining a method of calculating a feature value FV related to a particle size as a sample characteristic Ch.

[0084] Specifically, in Figure 6 , a time series image group I obtained for the liquid sample S1 and the liquid sample S2 is shown. m (t) The liquid sample S1 and the liquid sample S2 are composed of the same composition (element A) and the same particle structure (structure x), but they respectively include two types of particles p1 and particles p2 having different particle sizes (particle diameters). In particular, Figure 6 It can be imagined that the particle size involved in particle p1 is larger than the particle size involved in particle p2. m (t).

[0085] In addition, Figure 6 In the time series image group I m (t) shows the time period included in the detection unit ΔT u The seven time series images i are obtained at the detection time t0 to t7 of the interval m_t0 ~i m_t6 .

[0086] As shown in the figure, in the time series image group I of the liquid sample S1 including the particles p1 of relatively large particle size, m In (t), each time series image i at detection time t0 to t5 m_t0 ~i m_t5 At the determined image coordinate X1 (i.e., wavelength component λ A ) appears near the particle p1's excitation light E p1 In addition, the excitation light E p1 The intensity of increases during the detection time t0 to t3, reaches an extreme value at the detection time t3, decreases during the detection time t3 to t5, and almost disappears at the detection time t6.

[0087] On the other hand, in the time series image group I of the liquid sample S2 including the particles p2 of relatively small particle size, m In (t), each time series image i at detection time t0 to t3 m_t0 ~i m_t3 At the same wavelength component λ1 as in the case of the liquid sample S1, the excitation light E of the particle p2 appears near p2 In addition, the excitation light E p2 The extreme value of the intensity increases during the process from detection time t0 to t1, decreases during the process from detection time t1 to t2 with detection time t1 being the extreme value, and almost disappears at detection time t3.

[0088] Therefore, in the case of particles p1 and p2 composed of the same element and the same structure, each excitation light E p1 And the excitation light E p2 The wavelength components that appear are roughly the same, but the detection time t and image coordinate Y when the respective intensities reach extreme values ​​are different.

[0089] More specifically, the particle size calibration of the smaller particle p2 reaches the excitation light E p2 The detection time t1 of the extreme value of the intensity is related to the time when the particle p1 with a relatively large particle size reaches the excitation light E p1 The detection time t3 of the extreme intensity of the particle p2 is relatively early. p2 The image coordinate Y1 when the intensity reaches the extreme value and the excitation light E of the relatively large particle p1 p1 The image coordinate Y2 when the extreme intensity is reached is relatively small. The reason for this is that the smaller the particle size, the faster the excitation after reaching the plasma PL. Therefore, if the extreme intensity of the excitation light E and the detection time t when the extreme intensity is reached and the image coordinate Y (the sample droplet S on the measurement area MA) are investigated, d That is, in this embodiment, the extreme intensity of the excitation light E, the detection time t when the extreme intensity is reached, and the image coordinate Y become the feature quantity FV related to the particle size.

[0090] In particular, in this case, since the composition of the particles p1 and p2 is the same, the time / position spectrum I without the wavelength component λ (image coordinate X) as a variable is used. d2 (t, Y), or time spectrum I d3 (t), thereby simplifying the calculation of the feature value FV.

[0091] Next, a method of calculating a feature value FV related to a particle structure as a sample characteristic Ch will be described.

[0092] Figure 7 This is a diagram for explaining a method of calculating a feature value FV related to a particle structure.

[0093] Specifically, in Figure 7 , a time series image group I obtained for the liquid sample S3 including the particle p3 and the liquid sample S4 including the particle p4 is shown. m (t).

[0094] In particular, the particle p3 contained in the liquid sample S3 is composed of two elements A and B, and adopts a structure in which they are coupled as a solid solution. In more detail, the particle p3 adopts a structure in which the element A and the element B are coupled in a state of mutual mixing. In addition, the particle p4 contained in the liquid sample S4 is composed of the same two elements A and B as the particle p3, and they are coupled in a so-called core-shell structure. In more detail, the particle p4 adopts a structure in which the part composed of the element B becomes the core and the part composed of the element A is coupled as the shell covering the core. In addition, the particle p3 and the particle p4 have respective particle sizes that are substantially the same as each other.

[0095] As shown in the figure, particle p3 is composed of two elements A and B. Therefore, the excitation light E from particle p3 p3 The extreme intensities involved are at the wavelength component λ corresponding to element A. A and the wavelength component λ corresponding to element B B In addition, the excitation light E from the particle p4 which is also composed of two elements A and B p4 The extreme intensities involved are also in the wavelength component λ A The vicinity and wavelength component λ B Appears near .

[0096] On the other hand, in the particle p3 adopting the solid solution structure, the wavelength component λ A and the wavelength component λ B The respective excitation light E p3 The change processes of the image coordinates Y at each detection time t are substantially consistent with each other.

[0097] On the other hand, in the particle p4 having a core-shell structure, first, the wavelength component λ from the element A constituting the outer shell A The intensity begins to appear first (detection time t0 and image coordinate Y 4AS ). On the other hand, the wavelength component λ from the element B constituting the core B The intensity is slower than the detection time t0 and is related to the image coordinate Y 4A Smaller image coordinate Y 4B It begins to appear (detection time t2 and image coordinate Y 4BS ).

[0098] In addition, the intensity from the element A constituting the shell is detected at the time t3 and the image coordinate Y 4AP In contrast, the intensity of the element B constituting the core is at the detection time t5 and the image coordinate Y 4BP It reaches its extreme value.

[0099] In addition, the intensity from the element A constituting the shell is detected at the time t5 and the image coordinate Y 4AL In contrast, the intensity of the element B constituting the core is also detected at the subsequent detection time t6 and the image coordinate Y 4BL was detected.

[0100] Regarding this phenomenon, the present inventors consider that the element A constituting the outer shell is exposed to the plasma PL earlier than the element B constituting the core, and therefore, excitation proceeds relatively early.

[0101] Therefore, in Figure 7 The time series image group I shown m In (t), in a particle p composed of two or more elements, information capable of determining the coupling form (i.e., structure) of the above elements is included. m (t) or time / space spectrum I d The extreme intensity of the excitation light E included in (t, X, Y), the detection time t when the extreme intensity is reached, and the image coordinate Y become the feature quantity FV that can identify the particle structure.

[0102] Back to Figure 5 , the known sample DB42 is, and the known sample S _k Related sample characteristics Ch (known characteristics Ch _k ) is used as the relative value to the known sample S _k The feature value FV is calculated as _k Time / Space Spectrum I d_k In addition, the database is stored in a way that the known particles p _k Related feature quantity FV _k The calculation can also be performed by the same method as the above method.

[0103] Next, an example of a calculation algorithm in the characteristic determination unit 41 based on the calculation form of the feature value FV will be described. The characteristic determination unit 41 can determine the sample characteristic Ch of the liquid sample S by executing the following steps (I) to (III), for example.

[0104] (I) From the time / position spectrum I calculated in the image analysis unit 40 d2 (t, Y), calculate the excitation light E of the particle p contained in the liquid sample S p The region (region determined by the vertical position Y and the detection time t) of the extreme value intensity (hereinafter also referred to as the “first extreme value intensity”) is used as the feature quantity FV. Specifically, in the time / position spectrum I d2In (t, Y), the excitation light E from the liquid, which has been investigated in advance through experiments, is extracted. lq and the area outside the area affected by pollution, etc., as an area equivalent to the excitation light E p The region of the first extreme intensity.

[0105] (II) Referring to the known sample DB42, extract the excitation light E p The first extreme intensity region is consistent with the known particle p _k data.

[0106] (III) Determine the extracted known particle p _k The data contains the known characteristics Ch _k , as the sample characteristics Ch of the liquid sample S. Thereby, the sample characteristics Ch (composition, particle size, particle shape, particle structure, etc.) of the liquid sample S are determined.

[0107] In addition, the above steps (I) to (III) are an example of an algorithm for determining the sample characteristic Ch of the liquid sample S, and the specific form of the algorithm is not limited to the above example.

[0108] In addition, the sample characteristic Ch of the liquid sample S determined by the characteristic determination unit 41 can also be appropriately stored in the known sample DB 42 in a manner associated with the area of ​​the first extreme value intensity calculated as the feature value FV, and output to any output device (display screen, etc.) not shown.

[0109] Hereinafter, the configuration and operational effects of the sample analysis system 10 according to the present embodiment described above will be summarized and described.

[0110] The sample analysis system 10 of the present embodiment comprises: a droplet device 12, which intermittently introduces a sample (liquid sample S) into a measurement area MA set as a plasma PL; a light emission detection device (image intensifier 36 and high-speed camera 38), which detects the sample at a predetermined period (detection unit time ΔT u ) detection timing, i.e., detection time (t=t0, t1, t2, ...) to detect the luminescence generated in the measurement area MA; an analyzer 39 as an analysis device, which analyzes the sample droplet S according to the detected luminescence d Conduct analysis.

[0111] The analyzer 39 also includes a distribution calculation unit (image analysis unit 40) that calculates the detection time t, the position in the measurement area MA (image coordinate Y), and the wavelength component λ of the light emission. N (Image coordinates X) is set as the distribution of variables, that is, the time / space light intensity distribution (time series image group I m(t) or time / space spectrum I d (t, X, Y)); characteristic determination unit 41, which calculates the sample droplet S from the time / space light intensity distribution d The feature quantity FV related to the sample characteristic Ch of the property is obtained, and the sample characteristic Ch is determined based on the feature quantity FV.

[0112] Thus, from the above-mentioned time / space light intensity distribution, a characteristic quantity FV related to information representing sample characteristics Ch other than composition that is not included in the luminescence spectrum (luminescence intensity distribution corresponding to wavelength components) in the existing ICP luminescence spectrum analysis is obtained, and the sample characteristics Ch are determined from the characteristic quantity FV. That is, it is possible to analyze properties other than composition (elements) of the liquid sample S that is difficult to analyze by the existing ICP luminescence spectrum analysis.

[0113] In particular, the sample characteristics Ch as the analysis target include at least one of the elements of the particles p constituting the liquid sample S, the particle size, and the particle structure.

[0114] Thus, the sample analysis system 10 can be applied to the purpose of analyzing the properties of particles p contained in a certain product in various fields. In particular, the sample analysis system 10 of this embodiment can analyze the particles p in the form of the liquid sample S without performing a process such as separating the particles p in the liquid sample S. Therefore, the analysis can be performed without complicated processing.

[0115] In addition, the sample of this embodiment is a liquid sample S obtained by mixing the particles p with a predetermined liquid. In addition, the droplet device 12 has a device for dropping the liquid sample S into droplets (sample droplets S) of a desired diameter. d ) is introduced into the plasma PL. In addition, the analyzer 39 is configured to analyze the concentration of particles p contained in the liquid sample S (particle concentration Co) and to open and close the discharge port. p ) and adjust the opening and closing cycle △T in the opening and closing mechanism 20a f .

[0116] Thus, the particle concentration Co p From the perspective of improving the analysis accuracy, the sample droplet S d The number of particles p contained in .

[0117] In addition, the characteristic determination unit 41 calculates the first extreme value intensity, the detection timing (detection time t) when the first extreme value intensity is used, and the position on the measurement area (image coordinate Y) when the first extreme value intensity is used as the feature value FV, wherein the first extreme value intensity is the wavelength component λ N(image coordinate X) versus time / space spectrum I as time / space light intensity distribution d (t, X, Y) integration (time / position spectrum I d2 Furthermore, the characteristic determination unit 41 determines the particle size based on the characteristic amount FV.

[0118] This enables the use of time / space spectrum I d (t, X, Y) specifies one characteristic of the sample characteristic Ch, namely, a more specific form of the particle size.

[0119] In addition, the sample characteristic Ch includes the form of coupling of one or more elements constituting the particle p, that is, the particle structure. d2 The particle structure is determined by the extreme value intensity in (t, Y) and the feature value FV when the extreme value intensity is used as the detection time t.

[0120] This enables the use of time / space spectrum I d (t, X, Y) specifies one characteristic of the sample characteristic Ch, that is, a more specific form of the particle structure related to the particle p.

[0121] In addition, in this embodiment, the following sample analysis method is provided, that is, a sample (sample droplet S d ) is intermittently introduced into the measurement area MA set as the plasma PL, and a predetermined period (detection unit time △T u ) is detected by detecting the light generated in the measurement area MA at the detection timing, that is, the detection time (t=t0, t1, t2, ...), and the detection time t, the position in the measurement area MA (image coordinate Y), and the wavelength component λ of the light N (Image coordinates X) is set as the distribution of variables, that is, the time / space light intensity distribution (time series image group I m (t) or time / space spectrum I d (t, X, Y)) is calculated and the sample droplet S is represented from the time / space light intensity distribution. d The feature quantity FV related to the sample characteristic Ch of the property is obtained, and the sample characteristic Ch is determined based on the feature quantity FV.

[0122] Thus, from the above-mentioned time / space light intensity distribution, a characteristic quantity FV related to information representing sample characteristics Ch other than composition that is not included in the luminescence spectrum (luminescence intensity distribution corresponding to wavelength components) in the existing ICP luminescence spectrum analysis is obtained, and the sample characteristics Ch are determined from the characteristic quantity FV. That is, the following analysis method is realized, that is, an analysis method capable of analyzing properties other than composition (elements) of a liquid sample S that is difficult to analyze by the existing ICP luminescence spectrum analysis.

[0123] (Second Embodiment)

[0124] In the second embodiment, the following is described. In this embodiment, the following example is described, that is, as the sample characteristic Ch of the liquid sample S, the particle size distribution P of the particles p contained in the liquid sample S is calculated. d In particular, in this embodiment, an example is described in which a liquid sample S contains a plurality of particles p1, p2, p3, ..., p having a common composition but different particle sizes. m In the case of d In particular, the particle size distribution P of this embodiment is d It means that the particles p1, p2, p3, ..., p of various particle sizes contained in the liquid sample S m The distribution of the number of particles p1, p2, p3, ... m The respective particle concentration Co p1 、Co p2 、Co p3 ……Co pm distribution.

[0125] Specifically, the characteristic determination unit 41 of this embodiment determines the particle size distribution P according to the following steps: d .

[0126] (I) Regarding particles p1, p2, p3, ... m The liquid sample S is imaged with the image coordinate X on the time / space spectrum I described in the first embodiment. d (t, X, Y) is integrated to calculate the time / position spectrum I d2 In particular, the integration range of the image coordinate X is set to correspond to the integration range of each particle p1, p2, p3, ... m The wavelength components λ corresponding to each constituent element p1 , p2 ... pm The range of the sum of .

[0127] (II) Next, by placing any sample droplet S d The supply start timing of the sample liquid droplet S is used as a reference point for a predetermined measurement time. d From the measurement base point P on the measurement area MA O0 Arrived at the end point P O2 time) versus time / position spectrum I d2 (t, Y) is integrated (integrated by image coordinate Y), so that each droplet Dp1~Dp n Find the time spectrum I d3 (Dp).

[0128] Figure 8 To indicate the number of small drops Dp1 to Dp n Time spectrum in I d3 An example of the distribution of (Dp) is shown in Figure 1. In particular, Figure 8 In the figure, the horizontal axis represents the 1st to nth droplet times Dp1 to Dp n The vertical axis represents the droplet times Dp1 to Dp n Time spectrum in I d3 (Dp).

[0129] like Figure 8 As shown, each droplet is Dp1~Dp n Time spectrum in I d3 (Dp) in the determined droplet (in Figure 8 Dp2, Dp5, Dp n ) is larger than the others. That is, it can be seen that in the case of small droplets Dp2, Dp5, Dp n It contains particles p1 to p m The excitation light E p1 ~E pm The intensity components produced are Dp2, Dp5, Dp n The sample droplet S d contains particles p1, p2, p3...p m .

[0130] (III) To implement the particle size distribution P d Determination of each droplet Dp1 ~ Dp n Time spectrum in I d3 (Dp), the particles p1 to p m The sample droplet S d Time spectrum I d3 More specifically, the extraction exceeds the determination that the particles p1 to p mThe threshold value of the time spectrum I is determined from the perspective of the intensity component d3 (Dp) and the small droplet number Dp at this time. In particular, in this embodiment, the small droplet numbers Dp2, Dp5, and Dp n The time spectra I d3 (Dp2), I d3 (Dp5), I d3 (Dp n ), as the second extreme value intensity. In addition, in order to simplify the record, the following also records their distribution as extreme value I d3 (Dp2), I d3 (Dp5), I d3 (Dp n ).

[0131] (IV) By dividing the extracted small droplets Dp2, Dp5, Dp n The extreme value I d3 (Dp2), I d3 (Dp5), I d3 (Dp n ) and the known sample S stored in the known sample DB42 k Data (with particle size distribution P d The corresponding droplet times Dp and the extreme value I d3 Related data) are compared to determine the particle size distribution P of the liquid sample S d .

[0132] As described above, the characteristic determination unit 41 of the present embodiment calculates the time spectrum I as the time / space light intensity distribution. d3 (especially I d3 (Dp2), I d3 (Dp5), I d3 (Dp n )) and detection timing (especially small droplet times Dp2, Dp5, Dp n ) as the characteristic quantity FV, and determine the particle size distribution P of the liquid sample S d .

[0133] Hereinafter, the configuration and operational effects of the sample analysis system 10 according to the present embodiment described above will be summarized and described.

[0134] In the sample analysis system 10 of the present embodiment, the sample characteristic Ch includes particles p1, p2, p3, ..., p4, p6, p7, p8, p9, p10, p11, p12, p13, ..., p14, p15, p16, p17, p18, p19, p20, p21, p22, p30, ..., p40, p50, p60, p70, p80, p9 n In the case of each particle p1, p2, p3 ... p n The existence ratio (particle size distribution P d ). In addition, the characteristic determination unit 41 determines the second extreme value intensity (extreme value Id3 (Dp2), I d3 (Dp5), I d3 (Dp n )), and the detection timing when the second extreme intensity is used (small droplet times Dp2, Dp5, Dp n ) is calculated as the feature value FV, wherein the second extreme value intensity is, with the wavelength component λ N (image coordinate X) and the position on the measurement area MA (image coordinate Y) versus the time / space spectrum I d (t, X, Y) is integrated (time spectrum I d3 In addition, the characteristic determination unit 41 determines the extreme value intensity of each particle p1, p2, p3, ..., p according to the characteristic value FV. n Particle size distribution P d .

[0135] This enables the use of time / space spectrum I d (t, X, Y) determines one of the sample characteristics Ch, namely, the particle size distribution P d A further specific form.

[0136] In addition, in the present embodiment, an example is described in which each of the small droplet times Dp2, Dp5, and Dp6 of the predetermined detection timing is set. n And the extreme value I at this time d3 (Dp2), I d3 (Dp5), I d3 (Dp n ) as the characteristic quantity FV, and determine the particle size distribution P d However, the present invention is not limited to this, and it is also possible to adopt the method of only setting the number of small droplets Dp and the extreme value I of the predetermined detection timing. d3 (Dp) is used as the feature value FV, and the particle size distribution P is calculated. d That is, it includes a structure that can determine the particle size distribution P with a certain degree of accuracy by only one of the above parameters. d The information can be appropriately applied to analytical applications that require simplification of calculations rather than measurement accuracy.

[0137] In addition, from the perspective of achieving similar simplification of calculation, the following structure may be adopted, that is, m The excitation light E p1 ~E pm The extreme value position in the measurement area MA (time / position spectrum I d2 The value of the image coordinate Y that becomes the extreme value) is used as the feature quantity, and the particle size distribution P is calculated. d structure.

[0138] In the present embodiment, regarding a plurality of particles p1, p2, p3, ..., p having a common composition but different particle sizes, m Particle size distribution P d However, the present invention is not limited thereto, and by combining the method described in the first embodiment with the method of the present embodiment in a liquid sample S including particles p having different compositions (and / or particle structures), it is also possible to simultaneously perform the determination of the composition (and / or particle structure) of the particles p and the particle size distribution P. d analysis.

[0139] (Third Embodiment)

[0140] In the present embodiment, the processing in the characteristic identification unit 41 described in the first embodiment or the second embodiment is realized by a so-called artificial intelligence model.

[0141] Fig. 9 This is a diagram for explaining the structure of the learned model M that realizes the function of the characteristic identification unit 41.

[0142] As shown in the figure, the characteristic determination unit 41 of this embodiment is composed of a time / space spectrum I d The learned model M of this embodiment is composed of an artificial intelligence model, and is particularly composed of a neural network composed of an input layer, an intermediate layer, and an output layer. More specifically, the learned model M of this embodiment is composed of a deep learning (deep learning) model having two or more intermediate layers.

[0143] In addition, the learned model M is obtained by executing the transformation relative to the known sample S in the model before learning. k The measured time / space spectrum I d_k (t, X, Y) is set as the input layer and compared with the known sample S k Related known characteristicsCh _k The output layer is configured for machine learning.

[0144] Here, as described in the first embodiment or the second embodiment, in the time / space spectrum I d The information (feature value FV) related to the properties (composition, particle size, particle structure, etc.) of the particles p contained in the liquid sample S is included.

[0145] Therefore, when the learning model M obtained by the above-mentioned machine learning is installed in the analyzer 39, when the unknown sample Sun The time / space spectrum involved I d_un (t, X, Y) as input, it is possible to un The actual characteristics involved are perfectly consistent with the unknown characteristics Ch _u Set as output data.

[0146] Hereinafter, the configuration and operation effects of the present embodiment described above will be summarized and described.

[0147] In the sample analysis system 10 according to the present embodiment, the characteristic determination unit 41 is configured to obtain a time / space spectrum I d_k The learned model M is constructed with the sample characteristics Ch as input (input layer) and the sample characteristics Ch as output (output layer).

[0148] In addition, the learned model M is obtained by performing the k Correlated time / space spectrum I d_k (t, X, Y) is used as input, and the performance of the known sample S k The known characteristics of the properties of Ch _k The model M is learned to output the unknown sample S. un The time / space spectrum I d_un (t, X, Y) is set as input and represented on the unknown sample S un The unknown particle p contained in _un The unknown characteristics of the nature of Ch _u The analyzer 39 is operated in the output mode.

[0149] Thus, the function of the characteristic determination unit 41, that is, the function of determining the characteristic from the time / space spectrum I can be realized relatively easily by machine learning. d (t, X, Y) calculates the feature value FV related to the sample characteristic Ch, and based on this, determines the function of the sample characteristic Ch related to the liquid sample S.

[0150] In particular, as described in the first embodiment, the time / space spectrum I d (t, X, Y) is the time series image group I m (t) obtained, the time series image group I m (t) A large number of time series images i are obtained by taking pictures with a frame rate of tens of thousands to millions of fps by a high-speed camera 38. mTherefore, it can be imagined that the amount of calculation involved in determining the feature value FV and the sample characteristic Ch becomes huge. In addition, the sample characteristic Ch to be determined also includes multiple factors such as the composition of the particle p, the particle size, and the particle structure. On this basis, depending on the difference in these factors, the appearance of the intensity extreme value is based on the detection time t, the image coordinate Y (the sample droplet S d Therefore, it is conceivable that the calculation algorithm of the specific feature quantity FV is complicated and the calculation burden is heavy.

[0151] On the other hand, by executing the learned model M on the calculation logic in the characteristic determination unit 41 as in the present embodiment, the accuracy of the analysis is ensured while suppressing the increase in the calculation load.

[0152] In addition, in this embodiment, a method for generating a sample droplet S is provided. d Correlated time / space spectrum I d_un A method for generating a learned model M that takes (t, X, Y) as input and a sample characteristic Ch as output. The method for generating a learned model includes: k Correlated time / space spectrum I d_un (t, X, Y) is used as input, and the performance of the known sample S k The known characteristics of the properties of Ch _k In addition, by setting the known sample S k The plasma PL is intermittently introduced into the measurement area MA set as the plasma PL, and the detection unit time ΔT is set in advance. u The light emission generated in the measurement area MA is detected at the detection timing (detection time t = t0, t1, t2, ...), and the detected light emission is used as a correlation with the detection timing, the position on the measurement area MA (image coordinate Y), and the wavelength component λ N The distribution of light intensity corresponding to (image coordinate X) is calculated to obtain the time / space spectrum I d_un (t, X, Y).

[0153] This enables the analyzer 39 to analyze the unknown sample S with high accuracy. un The unknown characteristics involved Ch _u The learning model M is completed.

[0154] In particular, by adopting a deep learning model as the learned model M, it is possible to further improve the accuracy of calculating the feature value FV and determining the sample characteristic Ch.

[0155] In addition, the disclosure scope of this specification also includes Figure 1 The learned model M is independent of the structure of the sample analysis system 10 described in the examples and the like. That is, the disclosure scope of this specification includes the learned model M obtained by combining the learned model M with the known sample S k Correlated time / space spectrum I d_k (t, X, Y) is set as input (input layer), and the known sample S k The known characteristics of the properties of Ch _k It is obtained by performing machine learning by setting it as the output (output layer).

[0156] In addition, the scope of disclosure of this specification includes a sample analysis method, which is a sample analysis method using the learned model M and is a method of analyzing an unknown sample S. un The time / space spectrum I d_un (t, X, Y) is applied to the input (input layer) of the learned model M, and the output data (sample characteristics Ch) is determined as the unknown sample S un The unknown particle p contained in _un nature.

[0157] In addition, in this embodiment, the time / space spectrum I associated with the liquid sample S is d_k The above example describes an example of a learning model M and a method for generating the model M in which (t, X, Y) is used as input and the sample characteristic Ch is used as output. However, as an alternative, it is also possible to realize a method for generating a learning model M in which the digital image data, that is, the time series image group I is used as output. m (t) is used as input and the sample characteristic Ch is used as output. That is, the time series image group I is m (t) each time series image i m_t The input vector of a predetermined dimension composed of the variables (pixels, etc.) to be defined and the detection time t corresponding to each image contains the time / space spectrum I d_k (t, X, Y) similarly extracts the information of the feature quantity FV. Therefore, by m (t) is directly input to form a learned model M, thereby omitting the need for the time series image group I m (t) time / space spectrum I d_k Processing involved in the calculation of (t, X, Y).

[0158] More specifically, the scope of the disclosure of this specification includes generating a sample droplet S d Related time series image group I mA method for generating a learned model M that takes a sample characteristic Ch as an input and a learned model M as an output. In particular, in the method for generating a learned model, the time series image group I m (t) is the process of converting the known sample S k The plasma PL is intermittently introduced into the measurement area MA set as the plasma PL, and the detection unit time ΔT is set in advance. u ) is obtained by photographing the light emitted in the measurement area MA at the photographing timing.

[0159] Although the embodiments of the present invention have been described above, the above embodiments merely represent a part of application examples of the present invention and are not intended to limit the technical scope of the present invention to the specific configurations of the above embodiments.

[0160] For example, the sample characteristic Ch is not limited to the composition, particle size, particle structure, and particle size distribution of the particles p contained in the liquid sample S described in the above embodiments, as long as it can be obtained from the above time / space spectrum I d (t, X, Y), then any other characteristics of the particle p or the characteristics of the liquid sample S itself can be applied. For example, it is also possible to improve the time resolution (shorten the detection unit time ΔT u ) and the time / space spectrum I d In (t, X, Y), by analyzing the boiling point of the liquid component of the liquid sample S or the melting point of the solid component, the chemical coupling structure (for example, carbon structure, etc.) of the elements constituting the liquid sample S is determined as the sample characteristic Ch.

[0161] In addition, although Figure 1 , an example is shown in which the analyzer 39 is provided in the sample analysis system 10 separately from the high-speed camera 38, but the function of the analyzer 39 may be incorporated into the high-speed camera 38. In addition, the analyzer 39 may be configured in an external system that can communicate with the sample analysis system 10 of this embodiment.

[0162] For example, in the above-mentioned embodiment, the time series image I is generated by the image generation unit (high-speed camera 38) from the two-dimensional image detected by the detection device (image intensifier 36). m (t), and obtain the optical time / space spectrum I through the analyzer 39 d However, in the above-mentioned sample analysis system 10, the high-speed camera 38 may be omitted, and a method of not generating a time series image I may be adopted. m Directly measure the temporal / spatial spectrum from a 2D imaged (t, X, Y) device.

[0163] In the above embodiment, an example is described in which the image intensifier 36 that detects the spatial distribution of light from the spectrometer 34 as a two-dimensional image is used as the detection device. However, a detection device other than the image intensifier 36 may be used. For example, a configuration may be adopted in which a device that can detect the spatial distribution of light from the spectrometer 34 as a three-dimensional image is used, and the three-dimensional image is measured by the measurement device.

[0164] In addition, by the desired opening and closing cycle △T f The structure of the opening and closing mechanism 20a of the dripper 20 is opened and closed to intermittently introduce the sample droplets S d The example of is described. However, it is not limited to this, and a structure for supplying a solid or colloidal sample or the like may also be adopted. In this case, even if the opening and closing mechanism 20a is not used, the sample can be introduced in the form of particles. Therefore, a small dropper 20 including an arbitrary mechanism for introducing the sample particles one by one may be used instead of the opening and closing mechanism 20a. In addition, a storage medium storing a program for realizing the functions of the image analysis unit 40 and the characteristic determination unit 41 in the first embodiment, and a storage medium storing the learned model M in the second embodiment are included in the scope of matters recorded in the specification at the time of application in this application.

[0165] [Example]

[0166] Hereinafter, the present invention will be described in further detail by way of examples. In addition, the present invention is not limited to the above-mentioned examples.

[0167] (Example 1)

[0168] use Figure 1 The sample analysis system 10 shown in the figure creates a plasma PL time series image group I by using the following devices and measurement conditions: m (t), and the time / space spectrum I d (t, X, Y) were analyzed.

[0169] [Device]

[0170] Droplet controller 30: MD-K-130 70 μm (manufactured by Microdrop Technologies)

[0171] Spectrometer 34: Shamrock SR-750 (manufactured by Andor Technology Co., Ltd.)

[0172] Image intensifier 36: Image intensifier unit C10880-13F (manufactured by Hamamatsu Photonics Co., Ltd.)

[0173] High-speed camera 38: FASTCAM Mini AX100 540K-M-16GB (manufactured by Photron)

[0174] [Measurement conditions]

[0175] Sample: Two dispersions containing suspected microplastics (polystyrene beads) with a particle size of 3 μm and 10 μm respectively

[0176] Power supplied to coil 22 (ICP power): 400w

[0177] ·Frame rate of high-speed camera 38: 10000fps (detection unit time △T u :100μs)

[0178] ·Opening and closing cycle of the opening and closing mechanism 20a (droplet frequency): 100 Hz

[0179] Droplet diameter (diameter of the opening of the opening and closing mechanism 20a): 50 μm

[0180] ·Measurement area MA: 10mm

[0181] Time series image group I including the effect of excitation light from microplastics with particle sizes of 3 μm and 10 μm m (t) Sample droplet S d Arrived at the luminous starting point P O1 The following four time series images i m_t0 ~i m_t3 .exist Fig.10 (a) and Fig.10 (b) shows the time series images i extracted for microplastics with a particle size of 3 μm and 10 μm, respectively. m_t0 ~i m_t3 .

[0182] In addition, the time series images i of the extracted microplastics with a particle size of 3 μm and 10 μm are analyzed by the analyzer 39. m_t0 ~i m_t3 Analyze and find the time / space spectra I corresponding to the image coordinates (X, Y) and the detection time t d (t, X, Y). In addition, each time / space spectrum I is calculated in a range corresponding to the image coordinate X in the wavelength domain of the measurement target element. d (t, X, Y) is integrated to obtain the spectrum I at each time / position d2(t, Y). Fig.11 (a) and Fig.11 (b) shows the time / position spectra I of microplastics with a particle size of 3 μm and 10 μm, respectively. d2 (t, Y). In addition, Fig.11 (a) or Fig.11 The multiple curves shown in (b) show the time / position spectrum I corresponding to the value of the image coordinate Y d2 (t, Y).

[0183] [Results and Investigation]

[0184] In microplastics with a particle size of 3 μm and 10 μm, in the time series images i m_t0 ~i m_t3 and time / position spectrum I d2 In particular, in the time series image group I m In (t), it can be seen that the excitation of microplastics with a particle size of 3 μm occurs earlier than that of microplastics with a particle size of 10 μm. d2 In (t, Y), the intensity extreme value (extreme value from carbon) appears earlier than that of the particle size 10 μm, and the extreme value intensity is smaller than that of the particle size 10 μm. This shows that even if the particles have the same composition, if the particle size is different, the time / space spectrum I d (t, X, Y) also changes. That is, it can be considered that the time / space spectrum I d (t, X, Y) Analyze the difference in particle size among particles of the same composition.

[0185] (Example 2)

[0186] use Figure 1 The sample analysis system 10 shown analyzes the particle size distribution of microplastics contained in the liquid sample S using the following apparatus and measurement conditions.

[0187] [Device]

[0188] Same as Example 1

[0189] [Measurement conditions]

[0190] · Sample: A dispersion containing two types of microplastics (polystyrene beads) with particle sizes of 3μm and 10μm. The concentration of 3μm particles was adjusted to 25% and the concentration of 10μm particles was adjusted to 5% regarding the probability of particles existing in one droplet.

[0191] ·Measurement time: 2s

[0192] The other conditions are the same as those in the first embodiment.

[0193] By the same method as in Example 1, the time / position spectrum I of each droplet is d2 (t, Y) is accumulated to obtain the time spectrum I d3 (t). Fig.12 (a) shows the time spectrum I per droplet (per shot) d3 (t).

[0194] Right now, Fig.12 Each scale on the horizontal axis in (a) represents each small droplet Dp (about 1 ms interval) in the total number of drops (200 times) in the measurement time 2s. On the other hand, the vertical axis represents the time spectrum I in each small droplet D d3 (t) is the cumulative value. In addition, Fig.12 (b) shows the Fig.12 (a) is the particle size distribution obtained based on the results shown.

[0195] [Results and Investigation]

[0196] In the sample with two particle sizes of 3 μm and 10 μm mixed, each extreme value can be confirmed as the particle size distribution. That is, it can be considered that the time / space spectrum I d (t, X, Y) to obtain the particle size distribution.

[0197] (Example 3)

[0198] use Figure 1 The sample analysis system 10 shown in the figure creates a plasma PL time series image group I by using the following devices and measurement conditions: m (t), and the time / space spectrum I d (t, X, Y) were analyzed.

[0199] [Device]

[0200] Same as Example 1

[0201] [Measurement conditions]

[0202] Sample 1: A mixture of 100 ppm of Ag nanoparticles (average particle size: 90 nm) and 10 ppm of Au nanoparticles (average particle size: 20 nm)

[0203] Sample 2: A mixed solution obtained by mixing core-shell particles (average particle size: 90 nm, core diameter: 20 nm) with an outer shell of Ag and a core of Au at 100 ppm

[0204] Power supplied to coil 22 (ICP power): 400w

[0205] ·Frame rate of high-speed camera 38: 50000fps (detection unit time △T u :20μs)

[0206] ·Opening and closing cycle of the opening and closing mechanism 20a (droplet frequency): 100 Hz

[0207] Droplet diameter (diameter of the opening of the opening and closing mechanism 20a): 50 μm

[0208] ·Measurement area MA: 5mm

[0209] The time series image group I obtained for the mixed solution involving sample 1 and sample 2 is extracted. m (t), from the sample droplet S d Arrived at the luminous starting point P O1 A measurement unit time △T u From the previous measurement unit time △T u Eleven time series images of m_t0 ~i m_t10 .exist Fig.13 In FIG. 1 , the time series images i of sample 1 are shown. m_t0 ~i m_t10 In addition, Fig.14 In FIG. 1 , the time series images i of sample 2 are shown. m_t0 ~i m_t10 .

[0210] In addition, the time series images i extracted from the sample 1 and the sample 2 are analyzed by the analyzer 39. m_t0 ~i m_t10 Analyze and find the time / space spectra I corresponding to the image coordinates (X, Y) and the detection time t d (t, X, Y). In addition, each time / space spectrum I is calculated based on the wavelength domain of the measurement target element (the sum of the wavelength domains of Ag and Au), the image coordinate X, and the image coordinate Y. d (t, X, Y) is integrated to obtain the spectrum I at each time d3 (t). Fig.15 (a) shows the time spectrum I of sample 1. d3 (t). In addition, Fig.15 (b) shows the time spectrum I of sample 2. d3 (t).

[0211] [Results and Investigation]

[0212] In Example 3, between the mixed solution obtained by mixing Ag nanoparticles and Au nanoparticles, namely sample 1, and the mixed solution obtained by mixing Ag / Au core-shell particles, namely sample 2, there is a difference in the time series images i.e., m_t0 ~i m_t10 And time spectrum I d3 (t) These two sides are different. In particular, it can be seen from the above measurement results that, relative to the sample 1 containing Ag nanoparticles and Au nanoparticles in a separated state, in the sample 2 in which the Ag nanoparticles and the Au nanoparticles are coupled and a core-shell is used, the extreme value of the Au nanoparticles constituting the core is offset in the direction of the delay of the excitation. By this, it can be seen that even if the combination of the elements of the particles contained in the sample is the same, if the coupling form (structure) of each element is different, the time / space spectrum I d (t, X, Y) also changes. That is, it can be considered that the time / space spectrum I d (t, X, Y) analyzes the differences in particle structure among particles of the same composition.

[0213] This application claims priority based on Japanese Patent Application No. 2020-028438 filed with the Japan Patent Office on February 21, 2020, and all the contents of that application are incorporated into this specification by reference.

Claims

1. A sample analysis system, wherein: have: A droplet device that intermittently introduces a sample into a measurement area set as a plasma; a luminescence detection device for detecting luminescence of the measurement area at a detection timing of a predetermined period set in advance; an analysis device for analyzing the sample based on the detected luminescence, The analysis device comprises: a distribution calculation unit that calculates, based on the detected light emission, a distribution of light intensity corresponding to a detection timing, a position in the measurement area, and a wavelength component of the light emission, that is, a time / space light intensity distribution; A characteristic specifying unit calculates a feature quantity related to a sample characteristic representing a property of the sample from the temporal / spatial light intensity distribution, and specifies the sample characteristic based on the feature quantity.

2. The sample analysis system according to claim 1, wherein: The sample characteristics include at least one of an element, a particle size, and a particle structure of particles constituting the sample.

3. The sample analysis system according to claim 2, wherein: The sample is a liquid sample obtained by mixing one or more particles with a predetermined liquid. The droplet device includes an opening and closing mechanism for opening and closing a discharge port so as to introduce the liquid sample into the plasma in the form of droplets of a desired diameter. The analyzing device adjusts an opening and closing cycle of the opening and closing mechanism according to a concentration of particles included in the liquid sample.

4. The sample analysis system according to claim 2 or 3, wherein: The characteristic determination unit calculates a first extreme value intensity, a detection timing when the first extreme value intensity is used, and a position on the measurement area when the first extreme value intensity is used as the feature quantity, wherein the first extreme value intensity is the extreme value intensity when the time / space light intensity distribution is integrated with the wavelength component, The characteristic determination unit determines a particle size based on the characteristic amount.

5. The sample analysis system according to claim 4, wherein: The sample characteristics include the coupling form of one or more elements constituting the particles, i.e., the particle structure. The characteristic determination unit further determines a particle structure based on the feature amount.

6. The sample analysis system according to claim 4, wherein: The sample characteristics include the existence ratio of each particle when the sample contains particles of different particle sizes. The characteristic determination unit calculates a second extreme value intensity and a detection timing when the second extreme value intensity is used as the feature quantity, wherein the second extreme value intensity is the extreme value intensity when the time / space light intensity distribution is integrated with the wavelength component and the position on the measurement area, The characteristic determination unit determines the existence ratio based on the feature amount.

7. The sample analysis system according to claim 5, wherein: The sample characteristics include the existence ratio of each particle when the sample contains particles of different particle sizes. The characteristic determination unit calculates a second extreme value intensity and a detection timing when the second extreme value intensity is used as the feature quantity, wherein the second extreme value intensity is the extreme value intensity when the time / space light intensity distribution is integrated with the wavelength component and the position on the measurement area, The characteristic determination unit determines the existence ratio based on the feature amount.

8. The sample analysis system according to any one of claims 1 to 3, wherein: The characteristic determination unit is composed of a learned model that takes the temporal / spatial light intensity distribution as input and the sample characteristic as output, The learned model is obtained by executing machine learning in which the temporal / spatial light intensity distribution associated with a known sample is set as input and a known characteristic representing the property of the known sample is set as output, The learned model operates the analysis device by taking the temporal / spatial light intensity distribution obtained for an unknown sample as input and outputting an unknown characteristic representing a property of the unknown sample.

9. A method for generating a learned model, the method comprising generating a learned model that takes a temporal / spatial light intensity distribution associated with a sample as input and a sample characteristic representing a property of the sample as output, wherein: The method comprises performing a machine learning process by taking the temporal / spatial light intensity distribution associated with a known sample as input and taking a known characteristic representing the property of the known sample as output, The known sample is intermittently introduced into a measuring area set as plasma, the luminescence generated in the measuring area is detected with a detection timing of a predetermined period set in advance, and the time / space light intensity distribution is obtained by calculating the distribution of the light intensity corresponding to the detection timing, the position on the measuring area, and the wavelength component based on the detected luminescence.

10. A sample analysis method, performed using the learned model generated by the learned model generation method according to claim 9, wherein: The temporal / spatial light intensity distribution obtained for the unknown sample is used as an input to the learned model, and the output data is determined as an unknown characteristic representing the property of the unknown sample.

11. A method for generating a learned model, the method comprising generating a learned model that takes a time series image group consisting of a plurality of time series images related to a sample as input and takes a sample characteristic representing the properties of the sample as output, wherein: The method includes performing machine learning by taking the time series image group related to a known sample as input and taking a known characteristic representing the property of the known sample as output, The known sample is intermittently introduced into a measurement region set as plasma, and light emission generated in the measurement region is photographed a plurality of times at a photographing timing of a predetermined period set in advance, thereby obtaining a time series image group consisting of the plurality of time series images.

12. A sample analysis method, performed using the learned model generated by the learned model generation method according to claim 11, wherein: The time series image group obtained for the unknown sample is used as an input of the learned model, and the output data is determined as an unknown characteristic representing the property of the unknown sample.

13. A method for analyzing a sample, wherein: The sample is intermittently introduced into the measurement area set as plasma. The light emission generated in the measurement area is detected at a detection timing of a predetermined period set in advance, Based on the detected light emission, a distribution of light intensity corresponding to the detection timing, the position in the measurement area, and the wavelength component of the light emission, that is, a temporal / spatial light intensity distribution is calculated; calculating a characteristic quantity related to sample characteristics representing the properties of the sample from the temporal / spatial light intensity distribution, The sample characteristic is determined based on the feature amount.

Images