AUSTRALIA Patents Act 1990 COMMONWEALTH SCIENTIFIC AND INDUSTRIAL RESEARCH ORGANISATION COMPLETE SPECIFICATION STANDARD PATENT Invention Title: Mineral particle material resolving X-ray imaging The following statement is a full description of this invention including the best method of performing it known to us:- 2 A system to determine the material composition of an imaged object. Field Embodiments generally concern a method and a system to determine the material 5 composition of an imaged object. Embodiments have particular application in the field of mineral liberation analysis, elemental analysis as a function of particle size, and other applications where material recognition in a volume of sample is required. Background Art 10 X-ray micro-imaging provides the possibility to study the internal structure of samples at micron scales. However, material composition of the sample is not obtained. One possible method for material identification is the dual energy X-ray imaging. Dual energy X-ray provides simultaneous, but separate signals from transmitted energy of different characteristics. The dual-energy X-ray imaging is frequently used in cargo and 15 luggage inspection systems and in specific minerals applications. Examples of specific mineral applications include the measurement of ash in coal, or the detection of diamonds in gangue ore. However, these applications rely on an approximate measurement of materials with widely different compositions. 20 Considering dual-energy X-ray imaging, two images of the investigated sample are taken at different beam energies. The attenuation of X-rays in the sample obeys Beer Lambert's law: where IE is the intensity of beam with energy E, after passing through the sample, IOEi is 25 the intensity of beam E, from the source, lEi is X-ray mass attenuation coefficient for energy Ei, p is density of the sample and x is thickness of the sample. From the two images obtained with different beam energies, we can calculate: In(,,. /IO/"2)- ,PE' 30 which reflects the material type independent of thickness and therefore can be used for the material identification. For example, the energies can be selected such that inelastic or Compton X-ray scattering dominates the total X-ray cross-section at the higher energy and photoelectric absorption dominates at the lower energy, allowing high and 3 low atomic number elements to be distinguished. This approach requires two monochromatic X-ray beams or a detector sensitive in two narrow energy windows. Isotopic sources are often used for this purpose; however they are not suitable for micro-imaging as the large volume of active material from where the X-rays or gamma 5 rays originate is too large, blurring the image obtained. Moreover, the optimal photon energies needed for good elemental discrimination using this technique are too high to image objects having a thickness in the order of micrometers, with sufficient contrast. Micro-focus X-ray tubes are more suitable for imaging of such small objects. The 10 lowest energy of photons produced and transmitted out of the tube can be as low as a few keV. However, the X-rays are generated through Bremsstrahlung and therefore their energy spectrum is not monochromatic. As the X-rays are generated from a small spot, typically with a diameter from less then one micrometer to few micrometers, their power and the subsequent X-ray flux is limited. Usage of narrow energy windows in 15 the detector system would lead to too long an exposure time being required to achieve reasonable statistical accuracy in a transmission image. A technique for material recognition which can use a low-power polychromatic beam from a standard micro focus X-ray tube is needed. 20 Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present invention as it existed before the priority date of each claim of this application. 25 Throughout this specification the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps. 30 Summary Some embodiments relate to a method to determine the material composition of an imaged object, the method comprising: separately obtaining, in each pixel of a two dimensional pixel array of X-ray 35 detectors, two or more measures of X-ray transmission through an object to be imaged, each measure corresponding to transmitted X-rays with a broad range of energies; 4 performing a signal-to-thickness calibration to determine an equivalent thickness of said object in respect of each pixel by mapping said two or more measures of X-ray transmission onto an equivalent thickness of a calibration material; in respect of each pixel, comparing said determined equivalent thickness for 5 each measure against predetermined calibration curves to identify the material composition of the object, each calibration curve representative of material thicknesses for a selected range of elements or multi-element materials; and forming an image from the identified material compositions in each pixel. 10 Respective measures corresponding to transmitted X-rays may have distinct, broad range of energies. Optionally, respective measures corresponding to transmitted X-rays may have an overlapping range of energies. The method may comprise generating a polychromatic X-ray transmission image 15 having energies less than about 100 keV. The method may comprise enabling an X-ray source to operate at a peak voltage of between 20kVp and 150kVp, preferably between 30kVp and I 00kVp, or more preferably 30kVp and 60kVp.In order that each measure corresponds to a distinct broad range of 20 X-ray energies, the method may comprise biasing the X-ray source between a high kilovolt peak (kVp) and a low kVp. Optionally, or in addition, the method may comprise controlling a dynamic filter such that the spectrum of the X-ray beam attenuated by the object is varied from one containing a broad range of energies to one biased towards higher energies. Optionally, or in addition, the X-ray detectors may be 25 energy-resolving detectors and the method may comprise detecting radiation within a plurality of energy ranges or above preset energy thresholds. Subsequently, each distinct transmission image may be obtained using a unique setting of one or more of X-ray source voltage, X-ray filter and X-ray detector energy threshold. 30 The method may comprise preparing the object for examination. In the instance where the object comprises particles, the step of preparing the object for examination may comprise spreading a layer of the particles over a supporting sheet. The layer may be a single layer. The layer may comprise two, three or more than three layers. The step of preparing the object for examination may further comprise applying an adhesive to said 35 supporting sheet to which said particles will adhere. In the instance where heavy, high atomic number particles are to be identified in a lower atomic number matrix, for 5 example the detection of gold particles in a mineral process slurry, the method may comprise injecting a sample of said slurry into a shallow container and allowing the slurry particles to settle and become stationary. 5 Preferably the method comprises obtaining more than two measures of X-ray transmission. In one embodiment three measures of X-ray transmission are obtained. In a further embodiment four measures of X-ray transmission are obtained. In other embodiments five or more measures of X-ray transmission are obtained. An advantage of such embodiments is that the additional information facilitates the identification of 10 more material species. The method may further comprise separately obtaining, in each of said pixels, two or more measures of X-ray transmission through the object to be imaged, each measure corresponding to transmitted and detected X-rays over said broad energy ranges and at 15 different angles. Obtaining said measures at different angles may comprise rotating the sample about an axis that is substantially not parallel to the X-ray beam direction. Optionally, the method may comprise translating the sample and making use of a natural divergence of the X-ray beam. A small number of unique angles, for instance from 2 to 20, or any number in between, may be used for each distinct broad energy 20 range. The method may further comprise computing a tomographic reconstruction of the particles using said obtained measurements at different angles. A thickness may be derivable from said tomographic reconstruction. The derived thickness may be used to 25 resolve overlapping materials within a single particle. The method may further comprise displaying the image formed from the identified material compositions in each pixel, said image coloured according to material type. 30 The predetermined calibration curves may be obtained by measurement or calculation over each distinct broad energy range. The method may comprise calculating said predetermined calibration curves for a range of elements or compounds and their thicknesses at each distinct energy range and storing the calculated calibration curves. Optionally, the method may comprise measuring a transmitted signal for a range of 35 elements or compounds and their thicknesses for each energy range, and storing the measured calibration curves.
6 The method may further comprise an initial step of calibrating said array of X-ray detectors. The step of calibrating said array of X-ray detectors may comprise measuring, in each pixel of said array of X-ray detectors, X-ray transmission through a 5 series of thicknesses of said calibration material. The step of calibrating said array of X-ray detectors may further comprise, for each thickness of said calibration material, separately measuring transmission of X-rays for N distinct broad X-ray energy ranges. The calibration material may comprise aluminium foils or sheets, silicon wafers, glass 10 slips or other material of thickness ranging from 1 Opm to 2cm. Some embodiments relate to a system to determine the material composition of an imaged object, the system comprising: an X-ray source adapted to generate an X-ray beam having maximum energies 15 less than about 100 keV and a focal spot size comparable to the desired image resolution; a two dimensional pixel array of X-ray detectors positioned with respect to the X-ray source and adapted to detect, in respect of each pixel in the array, two or more measures of X-ray transmission through an object to be imaged, each measure 20 corresponding to a distinct broad X-ray energy range; a processor to: determine an equivalent thickness of said object in respect of each pixel by mapping the two or more measures of X-ray transmission onto an equivalent thicknesses of a calibration material from each X-ray energy range; 25 in respect of each pixel, compare said determined equivalent thicknesses against predetermined calibration curves to identify the material composition of the object, each calibration curve representative of one material/element for a range of thicknesses; and form an image from the identified material compositions in each pixel. 30 The X-ray source may operate at a peak voltage of between 20kVp and 150kVP, preferably between 30kVp and 100kVp, or more preferably 30kV, and 60kVp. The two dimensional pixel array of X-ray detectors may be in the form of a charge coupled device (CCD), or a pixellated solid state detector respectively fitted with a 35 scintillator screen, or a single X-ray quantum counting detector system capable of discriminating the energy of individual X-ray photons. The pixel size of the detector 7 arrays may be in the range of a few microns to hundreds of microns. By placing the sample between the source and detector, the image of the sample on the detector is magnified. The magnification can be adjusted by altering the source/sample and source/detector distances to obtain the desired resolution. 5 In order that each measure of the X-ray transmission corresponds to a distinct broad range of X-ray energies, the voltage of the X-ray source may be varied. Optionally, or in addition, a series of filters may be alternately interposed such that the spectrum of the X-ray beam incident on the object is modified. The X-ray source may be 10 configured with a series of filters to enable selective modification of X-ray spectrum. Optionally, or in addition, the X-ray detectors may be energy-resolving detectors that can be operated to simultaneously or sequentially detect radiation within a plurality of energy ranges. 15 The X-ray source may be adapted such that the electron beam focal spot size, i.e. the area of the surface upon which the electron beam is impinged has a diameter (or full width-half-maximum in case of a spot with Gaussian profile) of less than about 5 microns or less than about 1 micron or less than the desired resolution. 20 The system may further comprise a stage to support a sample frame between the X-ray source and the 2-dimensional pixel array of detectors, where the object to be imaged is position-able upon the sample frame. The stage may be moveable relative to the X-ray source and the array of detectors in order to alter the magnification and the resolution of the image. Preferably the stage is at least transversely translatable. Such an 25 embodiment allows a larger area of a sample to be viewed frame by frame. In addition, or alternatively, the stage may be adapted to rotate. Such an embodiment would enable views of the sample to be obtained from different angles to facilitate 3-dimensional CT reconstruction. 30 The system may further comprise a sample frame. The sample frame may comprise a substantially rigid outer frame supporting a polymer sheet. A side of the sheet may comprise an adhesive coating to facilitate application of a layer of particles to be held in position on said sheet. The sample frame may be from a few cm square up to more than 20 cm square. 35 8 Alternatively, for the analysis of bulk materials such as a slurry, the sample frame may comprise a shallow dish or chamber adapted to hold a layer of the sample material. The upper and lower surfaces of said shallow dish or chamber is preferably substantially transparent to the X-ray used to obtain the images. For example, these could be made 5 of a suitable plastic. Provision may be made for automatically injecting a fresh slurry sample at regular intervals, allowing for automated, on-stream analysis in a processing facility. Embodiments of the invention advantageously allow for a relatively simpler sample 10 preparation, with a thin layer of particles distributed over the surface of an adhesive film. Moreover, the methodology enables significantly faster imaging (1-2 orders of magnitude) compared the today's industrial systems for mineral particle analysis. It has the direct effect of allowing large numbers of routine measurements to be made, for example for liberation analysis of trace elements such as gold. The measurement of the 15 thickness of particles allows for proper 3D mineralogical analysis. Moreover, more accurate material identification relative to existing X-ray imaging techniques is able to be obtained. Brief Description of the Drawings 20 An example of the invention will now be described with reference to the accompanying drawings, in which: Fig. 1 shows a schematic illustration of a system in accordance with the invention and upon which embodiments of the invention can be performed. Fig. 2 shows a graph of the signal calculated for a 10 ptm sample of Nickel. 25 Fig. 3 shows the projection of the calculated transmission signals for Ni converted to the equivalent thickness for the reference signal. Fig. 4 shows a graph of the calculated curves stored in the database for the material recognition. Fig. 5 shows the Monte-Carlo simulated spectrum of X-rays for an X-ray tube running 30 at 60 kVp. Fig. 6 illustrates the simulated sample used to evaluate the material recognition technique; each square contains a different element. From top to bottom, the labels show the symbol for the element, its density in g/cm 3 and the maximum thickness (microns). The colour scale shows the variation in thickness across each square 35 (microns). Fig. 7 shows a map of the recognized atomic numbers.
9 Fig. 8a shows a map of the probability of wrong element identification in Fig. 7. Fig. 8b shows a plot of the probability of wrong element identification in Fig. 7. The gray scale indicates thickness of the material in mg/cm 2 . Fig. 9a shows a map of the standard deviation of the recognized atomic number in 5 Fig. 7. Fig. 9b shows a plot of the standard deviation of the recognized atomic number in Fig. 7. The gray scale indicates thickness of the material in mg/cm2 Fig. 10 shows X-ray attenuation properties of a 50 mg/cm 2 thick Al layer and a 40 mg/cm 2 thick layer of Si. 10 Fig. 11 shows a map of the recognized atomic number with sample thickness knowledge utilized. Fig. l2a shows a map of the probability of wrong element identification in Fig. 11. Fig. 12b shows a plot of the probability of wrong element identification in Fig. I1. The gray scale indicates thickness of the material in mg/cm 2 . 15 Fig. 13a shows a map of the standard deviation of the recognized atomic number in Fig. 11. Fig. 13b shows a plot of the standard deviation of in the recognized atomic number in Fig. 11. The gray scale indicates thickness of the material in mg/cm2 Fig. 14 shows the spectrum of the Hamamatsu X-ray tube used for the experimental 20 evaluation. The spectrum was measured using the AmpTek silicon drift detector. The tube was running at 40 kVp and 30 pA. The distance between the tube window and the detector was 40 cm. Fig. 15 shows a graph of calculated (continuous curves) and measured (dots) attenuation of X-rays in aluminium foils at different system response settings. 25 Fig. 16 shows a graph of the comparison between calculated (continuous curves) and measured (dots) equivalent thicknesses obtained at different system response settings. Each colour represents different material. Fig. 17a shows a resultant image formed from the identified material compositions in each pixel where the object being imaged comprises a length of gold wire, platinum ore 30 and sand. Fig. 17b shows a black and white representation of Fig. 17a. Fig. 18a shows a photograph of various foils. Fig. 18b shows a X-ray image of the various foils shown in Fig. 18a. Fig. 18c shows a black and white representation of Fig. 18b. 35 Detailed Description 10 Fig.1 schematically illustrates components of the system (100) for determination of material composition of mineral or other particles with sizes ranging from few micrometers up to hundreds of micrometers. The system includes an X-ray source (110) for instance a Hamamatsu micro-focus X-ray tube type L8122, exchangeable X 5 ray filter (125) and a two dimensional pixel array of X-ray detectors (120) (the imaging detector). The X-ray source (110) is in the form of a commercial micro-focus X-ray tube. The X ray tube may operate at a peak voltage of between 20kV, and 150kVp, preferably 10 between 30kV, and I00kVp, or more preferably 30kV, and 60kVp. The source (110) is equipped with a Beryllium window (not shown) selected to provide maximum transmission of the low energy part of the X-ray spectrum. The system includes a stage (not shown) to support a sample frame (130) between the 15 X-ray source (110) and the 2-dimensional pixel array of detectors (120). The object to be examined is mounted on the sample frame (130). The mineral particles (135) can, for detector calibration purposes, be replaced with aluminium foil or other material sheets to obtain the signal-to-thickness calibration. 20 A processor (140) and associated database (145) is provided. The database (145) stores data representative of predetermined calibration curves for a number of different materials and a number of different energy thresholds. The database (145) further stores data representative of X-ray transmission through a calibration material at a number of different thicknesses, and a number of different operating conditions. 25 The processor (140) is adapted to determine an equivalent thickness of the object (135) in respect of each pixel of the imaging detector (120) by mapping two or more measures of X-ray transmission onto an equivalent thickness of the calibration material. The processor (140) is further adapted, in respect of each pixel, to compare the 30 determined equivalent thickness against predetermined calibration curves to identify the material composition of the object. The processor (140) is further adapted to form an image from the identified material compositions in each pixel and to display said image on the data display (150). 35 The processor is further adapted, under the instructions of software stored to database (150), to control the X-ray source 110, the X-ray filter (125) or energy discrimination 11 settings of the detector (120) and the orientation of the stage relative to the source and detector. The processor may further process the acquired images to determine statistical 5 properties of interest, such as particle size distribution, particle shape distribution, porosity or mineral liberation. In order to determine the material composition of mineral particles the detector is initially calibrated. The calibration is based on measurement of beam intensity behind 10 a range of thicknesses of a calibration material. The attenuation of the polychromatic beam in the calibration material is measured for each broad energy range of X-rays, set by the X-ray tube operating voltage, X-ray filters and/or detector energy threshold settings. 15 The principal of obtaining transmission measurements of an unknown sample at each threshold setting and correcting the measures using the signal-to-thickness calibration is illustrated in Fig. 2 and Fig. 3. In this example, the calibration material selected is aluminium foil ranging in thickness from 50 pm to 200 Rm, stacked to obtain thicker layers. The system response is modified by varying the lower energy threshold above 20 which X-ray counts are accepted. The number of energy discrimination thresholds N is set equal to two with ETHI=6.0 keV and ETH2= 13 .5 keV. Fig. 2 shows the number of counts measured for a 10 pm thick sample of Nickel (Ni) for energy discrimination threshold settings at ETHi and ETH2. Fig. 3 shows a graph of the calibration of the signal measured in Fib. 2 onto an equivalent thickness of the aluminium. The curves were 25 calculated using a computer-simulated Bremsstrahlung X-ray spectrum from a tungsten target with an electron beam of 40 keV (Fig. 5). The signal-to-thickness calibration maps the measured counts in each pixel onto an equivalent thickness of the calibration material. The resulting equivalent thickness is 30 equal to the real thickness of the sample for all thresholds only when the sample is of the same material which was used for the calibration. For other materials the equivalent thickness will be different. The signal-to-thickness calibration curves (X-ray attenuation curves) are determined for 35 all individual pixels in the 2-dimensional array. The per-pixel character of the signal-to thickness calibration means that the calibration corrects for the pixel-to-pixel variations 12 in the detection sensitivity. In addition it also corrects for non-uniformity of the X-ray beam, variations in electronic gain or noise, variations in detector response with X-ray energy and non-linearity. The calibration also corrects for the hardening (selective removal of lower-energy radiation) of the X-ray beam. Compared to the standard flat 5 field, a single signal-to-thickness calibration also into account the beam hardening due to presence of the sample and therefore also changed effective sensitivity of the imaging sensor. All these properties of the signal-to-thickness calibration simplify calculation of the material recognition database of curves as all inaccuracies in the sensor itself or the beam are corrected. 10 The database (145) of calibration curves is then determined. The system response for each broad X-ray energy range is measured or calculated for each element or compound of interest over a range of thicknesses. The expected signal in the detector for an element with atomic number Z is calculated as: 15 Iz(R,,)= S(E)- DR (E)-e-Az(E)Pzx .( -es(E)Ps,d)dE (1) where: Emin minimum energy in the X-ray spectrum, 20 Emax maximum energy in the X-ray spectrum, S(E) energy spectrum of the X-ray tube DRn(E) average signal contribution from an X-ray photon of energy E that interacts in the detector pz mass attenuation coefficient of element with atomic number Z, 25 Pz density of element with atomic number Z, x thickness of the sample, psi mass attenuation coefficient of the silicon sensor, psi density of the silicon sensor (2.33 g/cm 3 ), d thickness of the silicon sensor (300 pm). 30 The expected signal for a compound is given by equation (1) with the exponential term e-z(")Pz replaced by eZp, (E)p(x where the sum extends over all elements Z that make up the material and p(i) is the partial density of the ith element.
13 Finally, the expected detector signals are converted to expected equivalent thicknesses by determining the thickness of the reference material that yields the same X-ray signal strength. This determination is carried out separately for each broad X-ray energy range Rn. The final database then consists of a vector of equivalent thicknesses (one 5 component for each broad X-ray energy range) for each material (element or compound) and for each thickness of interest. Fig. 4 illustrates the calculated equivalent thickness for different elements for one particular choice of three broad energy ranges. As can be seen in figure 4, the 10 equivalent thickness values for separate elements generally form smooth curves, separated in N-dimensional space where N is number of broad X-ray energy ranges used. The recognition of an unknown material in one image pixel is done by finding the 15 minimum value of: (Zt) X, (ZOt2 (2) tEN r I where: 20 Z atomic number of the examined element/material, t thickness of the examined material in the database, N number of broad energy ranges, xi the measured equivalent thickness for the ith broad energy range, ei(Z,t) the calculated equivalent thickness for the ith broad energy range and 25 element with atomic number Z and thickness t, oyi error of the measured equivalent thickness for the ith broad energy range. The minimum value of / is found for the atomic number Z and the real thickness t of the element in the image. In case that recognition of compounds is required the values 30 e,(Zt) are calculated for the compounds. The parameter Z is then identified against the compound identification number in the calculated database. The range of examined thicknesses in the database can be restricted if the real thickness of the sample is known or can be separately measured. This information reduces number of database values to be checked and also reduces possible misinterpretation of the material due to statistical 35 or systematic errors in the measured signal.
14 To demonstrate the validity of this technique to correctly identify individual elements, irrespective of their thickness, and in the presence of image noise, a computer simulation was performed. The Bremsstrahlung spectrum of an X-ray tube was 5 obtained by using Monte-Carlo simulation of a 60 keV electron beam hitting a tungsten target illustrated in Fig. 5. Fig. 6 illustrates an array of different elements and material thicknesses which was used as an input for the calculation. Each element had thickness ranging from 2 tm up to a 10 maximum thickness corresponding to an areal density of 40 mg/cm 2 . Four distinct broad energy ranges were chosen, corresponding to photon having energies above thresholds of 5.9, 11.9, 18.0 and 27.0 keV. The intensity of the X-ray beam and the exposure time were chosen such that there were approximately 106 counts recorded in each pixel with an open beam for the lowest energy threshold. For each element, 15 material thickness and X-ray energy threshold, the expected detector signal was calculated using Eq (1). Poisson-distributed noise was added to each calculated value and an equivalent aluminium thickness was determined for each image pixel X-ray energy threshold setting. The element present at each image pixel was then determined by seeking the minimum value of the chi-square function given in equation (2). A map 20 of the 'measured' atomic numbers is shown in Fig. 7. The colour range shown in Fig. 7 corresponds to atomic numbers from 13 to 80. The abbreviations indicate which element was present in the appropriate cell. Figs. 8a and 8b shows the probability that the element will be incorrectly identified 25 under these conditions. Elements start to be correctly indentified once their atomic number exceeds 25(Mn) or 26(Fe). These elements have abrupt jumps in their total X ray cross-sections at energies of 6.54 and 7.11 keV respectively, corresponding to their K-shell energies. These jumps, which are just above the 5.9 keV lowest threshold energy, provide a distinctive signature. 30 Figs. 9a and 9b show standard deviation of the recognized Z. Figs. 8a and 8b and Figs. 9a and 9b show that even though the light elements (with Z<25) are difficult to recognize exactly, it is possible to differentiate them collectively from heavier elements. Elements starting from Nd are also more difficult to be recognized accurately 35 at thicknesses from 20 mg/cm 2 . Their K-shell energies are above the highest threshold (43.56 keV for Nd), but their L-shell edges are at low energies where the beam is 15 already strongly attenuated. The energy discrimination threshold levels can be adjusted or additional energy ranges can be added if precise recognition of these elements is required. Material identification of thin layers is also difficult as the difference between the open and attenuated beam is small. Longer measurement times and/or higher X-ray 5 fluxes are needed to reduce statistical counting errors. As illustrated in the graph of Fig. 10, the attenuation coefficients of the light elements do not have any significant features in the region of X-ray energies used. On the y-axis are plotted the product of the mass attenuation coefficient, density and thickness of the 10 sample. The shape of the attenuation coefficient as a function of energy is similar for all the light elements. However, the absolute values of the attenuation coefficient are different. Therefore, the low Z elements can be reliably recognized if knowledge of the areal density, that is the product of the material density and the real thickness, is available. The material thickness can be estimated approximately from tomographic 15 principles, by obtaining X-ray images of the target material from various directions. To illustrate the improvement that results from even an approximate knowledge of the material thickness, the material identification calculations were repeated with an additional input, namely the areal density measured with an assumed error of ±20%. 20 The resulting map of recognized Z is shown in Fig. 11 with each colour (Fig. Hl) corresponding to an atomic number from 13 to 80. The probability of wrong element identification is in Figs. 12a and 12b. Figs. 13a and 13b show standard deviation of the recognized Z. The error for all elements is reduced. 25 Very thin layers of material are harder to recognize as they do not exhibit sufficient attenuation. The X-ray beam passes through almost un-attenuated and therefore provides little attenuation information. Similarly, if the layer is too thick (>100 mg/cm2 under the current conditions), the low energy photons are removed from the X-ray spectrum hitting the detector and the material identification is not possible. 30 Nevertheless, a useful dynamic range can be obtained, ranging from a few microns to a few hundred microns for the assumed choice of energy ranges. The maximum thickness of material that can be recognised can be increased operating the X-ray tube at a higher voltage at the cost of degraded sensitivity for thinner material 35 layers. The optimal combination of parameters (X-ray operating voltage, beam filters 16 and/or energy thresholds) is preferably found for each application of the presented technique. Experimental evaluation 5 The described method was tested experimentally using the setup shown in Fig. 2 and described in the previous section. The measured X-ray spectrum at a distance of 40 cm from the X-ray tube running at 40 kVp and 30 ptm is shown in Fig. 14. The spectrum was measured using the AmpTek silicon drift detector. The same detector was then used for the rest of the evaluation. A comparison of the calculated (solid line) and 10 measured (round points) X-ray beam attenuations through an aluminum foil are shown in Fig. 15. There is an excellent agreement between the calculated and measured X-ray attenuations. Foil material Foil thickness [pm] Foil atomic No. Recognized atomic No. Al 145 13 13 Al 290 13 13 Fe 20 26 26 Fe 40 26 26 Fe 60 26 26 Ni 25 28 28 Ni 50 28 28 Ni 75 28 28 Cu 30 29 29 Cu 60 29 29 Cu 90 29 29 Ag 20 47 47 Ag 40 47 47 Ag 60 47 47 Sn 25 50 50 Sn 100 50 50 15 Table 1.
17 The system response was modified by setting the energy discrimination threshold in the detector to levels of 6, 9, 13 and 26 keV. The threshold levels are indicated in Fig. 14. The average signal for the open beam and at the threshold of 6.0 keV was I xl 06 counts. 5 Table I lists metal foils used for the evaluation. All tested elements were correctly recognized using the present technique. Fig. 16 depicts calculated curves of equivalent thicknesses for the tested materials (full lines). The dots in Fig. 16 represent the measured values of equivalent thicknesses for each material. 10 Figs. 17a and b show a resultant image formed from the identified material compositions in each pixel, where the object being imaged comprised a length of gold wire, platinum ore and sand. Low Z (particles of sand) and high Z (Au, Pt) objects were discriminated using a Flat Panel detector and two measurements. The first measurement was undertaken with an unfiltered X-ray beam (voltage of 40kVp, current 15 10 pA). The second measurement was undertaken with the X-ray beam filtered using a 1 mm thick glass plate. Fig. 18a shows a photograph of various foils and Figs. 18b and 18c an X-ray image of the various foils shown in the photograph of Fig. 1 8a. Low Z (Aluminium, glass, Iron, 20 Nickel) and high Z (Gold, Tin, Tungsten) foils were discriminated using the Flat Panel detector and two separate measurements. The first measurement was undertaken with an unfiltered X-ray beam (voltage of 40kVp, current 75 pA), whilst the second measurement was undertaken with the beam filtered using 1 mm thick glass plate. 25 The experimental evaluation fully supports the conclusions of the theoretical study and proves the applicability of this method for the material resolving imaging.