Current technologies for heavy metal detection include inductive coupled plasma mass spectrometry (ICP-MS), fluorescence spectroscopy, atomic absorption spectroscopy, electrochemical analysis, etc. Detection methods can vary depending on the measurement of real-time parameters.
An example of a resonator includes a body, the body having a planar surface. An aperture through the body is configured to receive a tube configured for a fluid to be tested. A gap extends into the body from the planar surface to the aperture. At least one cut extends through the body from the planar surface towards the aperture. The at least one cut extends across the gap.
In other examples, the body is cuboid. The at least one cut may be perpendicular to the gap. The at least one cut extends from the planar surface to the aperture. The at least one cut extends from the planar surface through the aperture. The resonator may include at least two cuts. The resonator is configured to produce a resonant frequency that corresponds to a water exchange rate of an ion to be detected by the resonator. The resonator may further include a dielectric material in the gap. The resonator exhibits a resonant frequency for each gap or cut in the resonator.
A system for the detection of an ion in a fluid includes a resonator with a body having a planar surface. An aperture through the body is configured to receive a tube configured for a fluid to be tested. A gap extends into the body from the planar surface to the aperture. At least one cut extends through the body from the planar surface towards the aperture. The at least one cut extends across the gap. A sample tube extends through the aperture of the resonator, the sample tube is configured to receive a sample of the fluid with the ion. A coupling loop includes a coil of wire. The coil of wire of the coupling loop is coaxially aligned with the aperture of the resonator. The sample tube further extends through the coupling loop. A vector network analyzer (VNA) is connected to the coupling loop. The VNA supplies an RF energy signal to the coupling loop. The RF energy signal is transferred to the resonator by inductive coupling. The VNA receives a reflected portion of the RF energy signal through the coupling loop and calculates a reflection coefficient. A processor receives the reflection coefficient and produces a determination if the ion is in the fluid based upon the reflection coefficient.
In examples of the system, the processor applies a support vector regressor (SVR) model to the reflection coefficient to produce the determination if the ion is in the fluid. The processor further determines a concentration of the ion in the fluid by applying the SVR model to the reflection coefficient. The SVR model is produced by a machine learning algorithm trained on datasets of reflection coefficient measurements of fluids having known ion concentrations. The SVR model is specific to a target ion and the SR model is produced by the machine learning algorithm trained on datasets of reflection coefficient measurements of the fluid having known concentrations of the target ion. The resonator is configured to produce a resonant frequency that corresponds to a water exchange rate of the target ion. The resonator is configured to produce a resonant frequency that corresponds to a water exchange rate of the target ion. In an example the target ion is lead and the resonant frequency is 7 GHz.
A method of detecting an ion in a fluid may be performed with any of the resonators or systems as described above. Examples of the method further include providing the resonator about the tube configured to receive a fluid to be tested. A magnetic field is generated with the resonator. Reflection coefficients are measured as a function of frequency across a frequency range. The measured reflection coefficients are analyzed. A detection of the ion in the fluid is determined based upon the measured reflection coefficients.
Other examples of the method include a frequency range between 10 MHz-10 GHz. Other examples of the method include a frequency range between 10 MHz-5 GHz. The measured reflection coefficients are analyzed using an ion-specific model. The measured reflection coefficients are analyzed using a model produced by training a machine learning algorithm. The model is a support vector regressor model trained using the reflection coefficients measured from a plurality of samples of known ion concentration. The support vector regressor model is trained using the reflection coefficients within a passband centered on a frequency that corresponds to a water exchange rate of the target ion. Analyzing the measured reflection coefficients includes analyzing the measured reflection coefficients within a passband centered on a frequency that corresponds to a water exchange rate of the target ion. In an example, the target ion is lead and the frequency is 7 GHz. The resonator is configured to produce a resonant frequency that corresponds to a water exchange rate of a target ion to be detected in the fluid.
Sensing systems and methods are desired to meet the sensitivity, selectivity, sensor lifetime, and real-time measurement requirements for water contamination monitoring, particularly for heavy metal detection.
A continuous, static, and non-interfering water contaminant detection system and method is disclosed herein in which RF microwave principles are used to measure specific contaminants found in water. In one example (NaCl, MgCl2, and a mixture of NaCl and MgCl2) are the representative contaminants used in reducing the detection system and method to practice. Another example described herein is the detection of lead (Pb) in a municipal water system. Other examples will be described herein and recognized from the present disclosure.
A loop gap resonator is a microwave device consisting of a hollow cylindrical loop. The hollow cylindrical loop has a consistent wall thickness. The hollow cylindrical loop includes a gap through the wall which extends for the length of the hollow cylindrical loop. Such a device is also known as a slotted tube cavity or a split ring resonator. The device can be considered as a lumped LC circuit where the loop acts as an inductor and the gap is a capacitor. In the presence of an electromagnetic signal, a magnetic field is introduced around the loop. The gap holds electric charges that create a uniform electric field. When the stored magnetic energy in the loop and stored electric energy in the gap are equal, the loop gap resonator becomes resonant.
As will be described herein, microwave measurements and machine learning algorithms are used to detect metal ions, exemplarily in drinking water. As disclosed in further detail herein, a block loop gap resonator (BLGR) is used as a sensor in a system for detecting or quantifying ion concentrations in an analyzed fluid. The BLGR is sensitive to different ions and their concentration resulting in an RF reflection spectrum (S11) amplitude and phase variations in the 10 MHz to 10 GHz range, while in other examples, narrower or larger frequency ranges may be used. A support vector regressor (SVR) algorithm is used, for example, to classify among detection of different ions and to measure a concentration of a heavy metal (e.g. Pb) ion in water. As disclosed herein, BLGRs may further exhibit multiple resonances, which have been found to enhance the sensitivities of these detections.
With reference to
A glass tube 26 extends through the aperture 14. A portion of the glass tube 26 is thus surrounded by the resonator body 12 and by the coupling loop 32. A sample to be analyzed is held within the glass tube 26. The sample may be a static sample, e.g. if the ends of the tube are stopped, or the sample may be a continuous flow through the glass tube 26. In examples herein, the sample is water, which may be municipal water, that contains one or more metal ions. As will be described in further detail herein, an external magnetic field generated by the coupling loop 32 is used to continuously measure changes in radio frequency energy in response to the magnetic field. The non-contact feature of the device allows a long sensor lifetime with high sensitivity for real-time measurements.
The output signal of the coupling loop 32 is characterized as the reflection coefficient (S11), as will be described in further detail herein. However, this output signal exhibits many peaks in the signal as a function of frequency. In examples, the output signal may exhibit 10, 20, 25, or more peaks across an operational frequency of 10 MHz to 10 GHz. These peaks are due to sensor design and the substance being tested. However, both the target ions for analysis, as well as the background chemistry of the substance being tested contribute to the peaks in the output signal. The complexity of the output signal and multiple sources contributing to the output signal present challenges in the simultaneous monitoring and analysis of the frequency and amplitude changes of all of these peaks continuously and in real-time. Therefore, multiple features of the presently disclosed BLGR system improve and facilitate this analysis.
It has been discovered that the BLGR 10 has increased sensitivity to frequencies within a band surrounding a resonant frequency of the sensor. The BLGR 10 depicted in
It will be recognized that the at least one cut 24 may extend into the resonator body 12 from the outer surface 18 of the body 12 across the gap 16 to the aperture 14. In examples, the gap 16 is aligned with the axis of the aperture 14 through the body 12, and therefore the gap 16 connects to the aperture 14 at the smallest distance from the outer surface 18 to the aperture 14, for a width (w). However, since the cuts 24 are exemplarily perpendicular to the gap 16, the cuts 24 may extend into the body 12 at various depths relative to the aperture 14. In an example the cuts 24 extend into the body 12 the same distance (w) as the gap 16. In other examples, the cuts 24 extend through the aperture 14. The cuts 24 may extend to the furthest point of the aperture 14 from the outer surface, or the cuts 24 may terminate at a distance within the aperture 14. As noted above,
The at least one cut 24 splits the capacitance between the opposing surfaces 22A, 22B into two or more capacitors. One cut 24 splits the capacitance into two, while two cuts 24 split the capacitance into three. Additional cuts provide still further capacitances in the BLGR 10. Each capacitance results in an additional resonant frequency of the BLGR 10. The lower resonance is dictated by the total capacitance and the higher additional resonant frequenc(ies) are due to the partial capacitance induced by each additional cut 24.
The resonant frequency of BLGR 10 is dependent upon the inductance (L) and the capacitance (C). The inductance and the capacitance are directly dependent on structural parameters identified above, according to the equations below.
Wherein N is the number of loops, which since the body 12 is a unitary block, a single (1) loop is provided, ε is the permittivity, and aluminum has a conductivity of 3.77×10{circumflex over ( )}7 S/m. In the BLGR 10, the resonant frequency is independent of the gap length (l). In ion sensing applications, increasing the gap length (l) allows more sample space (and thus improved sensitivity) without modifying the resonant frequency. The resonant frequency is inversely proportional to the loop radius (r). A dielectric material placed in the gap 16 changes the permittivity as well as the capacitance of the BLGR 10.
The resonant frequency is further dependent upon the quality factor (Q). The quality factor (Q) relates the amount of energy loss relative to the energy stored for a resonant system. The energy is stored by the inductor and capacitor, and the energy is dissipated by the resistor of the system. The quality factor indicates the performance of a resonator in terms of bandwidth and losses. Q is expressed as the ratio of resonant frequency to 3 dB bandwidth. A high Q value implies a low loss in the system. The higher the Q factor, the closer the system behaves as an ideal inductor. Q can be calculated as in equation (4), where ωr is the resonant frequency ωr=2πfr:
Alternatively, Q can be calculated as the resonance curve full bandwidth Δf signal (full width at half maximum power) relative to its maximum frequency fr.
As detailed herein, it has been discovered that the electrical properties of this BLGR 10 are dependent upon these physical parameters, therefore, the rest of the resonator body 12 may be designed for other considerations, for example, to provide rigid mounting surfaces, or shape to accommodate other system components or the fluid sample tube 26. This makes the BLGR 10 as described herein convenient for integration of the disclosed sensor system into existing water supply lines for contamination monitoring, as well as a durable construction.
A vector network analyzer (VNA) is an instrument that characterizes high-frequency passive and active devices and measures their effect on the amplitude and phase of swept-frequency and swept-power test signals.
A VNA generally includes operational sections: sources for stimulus, a signal-separation device, a receiver that down-converts and detects the signals, a processor for calculating results, and a graphical display connected to the processor and configured for reviewing the results. The signal source produces a stimulus for the device-under-test (e.g. the BLGR 10). The source can be used to sweep the frequency or sweep power levels of the stimulus signal. Open-loop voltage-controlled oscillators or synthesized sweepers are used to make the signal source. The signal separation block measures a portion of the incident signal using splitters or directional couplers and the incident and reflected traveling waves are separated at the input of the device under test. The measured portion of the incident signal can be used as a reference for the processor to compute the ratio-based calculations. The incident and reflected traveling waves enter the signal-detection block which may use diode detectors and tuned receivers. Conversion of RF signals to proportional DC levels is done with diode detectors. In the Receiver, local oscillators are used to translate the signal to an intermediate frequency and this signal then passes through bandpass filters. The filtered intermediate frequency signal passes through an ADC and DSP to gather magnitude and phase information before the output is transmitted, stored, and/or presented. The E8363B vector network analyzer from Agilent Technology with a frequency range of 10 MHz-40 GHz is one example of a VNA that may be used within the present disclosure.
The RF reflection spectrum (S11) is defined as the ratio of the amplitude of the reflected voltage at port one with respect to the input voltage at port one. It is a figure that quantifies the impedance discontinuity in a transmission medium. S11 is a complex quantity and the magnitude portion denoted F and the phase portion is denoted with ϕ.
Here, ZL is the load impedance and Z0 is the transmission line impedance. When line impedance is equal to the load impedance, then all the energy will be transferred to the load and nothing will be reflected back resulting Vreflected=0 and Γ=0. When ZL is an open or short circuit, then all the energy is reflected back and Γ=1. When ZL is not equal to Z0, then some of the incident wave will be reflected back and the magnitude of the reflection coefficient will be in the range between 0 and 1.
The system 30 classifies and/or measures the concentration of ions based upon analysis of the S11 data, by application of a Support Vector Regression (SVR) algorithm trained with machine learning to detect the presence of heavy metal ions and/or to quantify the concentration of heavy metal ions in the fluid sample. Several water samples were prepared with variable trace level concentrations of Pb for machine learning training as well as for detection and quantification using the BLGR. Concentrations of 1, 3, 5, 10, and 20 ppb PbCl2 were prepared as trace level contaminants in 200 mL city water by addition of 4, 12, 20, 40, and 80 uL volumes of acidified 50 ppm Pb stock solution (e.g. PbCl2 dissolved 1% v/v 16M HNO3 in DI water), respectively. Equivalent uL volumes of a bland (0 ppb Pb) stock solution (1% v/v 16M HNO3 in DI water only) were used to prepare negative controls. Both stock solutions were prepared in DI water to minimize introduction of additional monovalent (1+) or divalent (2+) metal ions. Pb concentrations were measured with different variations of BLGRs as described herein, including but not limited to single gap, double gap, and triple gap. A total of 1500 data files of each concentration were collected with each of the resonators. The SVR algorithm is trained with test S11 data sets consisting of sensor responses to multiple known samples with different ions and ion concentrations. Magnitude and phase vectors corresponding to each sample are normalized to the interval [0,1] and concatenated into a single feature vector. The SVR was trained based on error-correcting output codes comprising three binary support vector machine classifiers with 1300 data files from each solution and tested on an additional 200 files of each solution. The magnitude and phase vectors corresponding to each sample are normalized to the interval [0,1] and concatenated into a single feature vector. These normalized vectors were used to train and test the SVR algorithm. The magnitude and phase vectors correspond to each sample.
In another example of NaCl and MgCl2 ions in water, 11 different concentrations (1000 ppm-400 ppb) liquid solutions are prepared and tested with a BLGR that operates across a frequency of 10 MHz-10 GHz, or more. In other examples, the BLGR may operate across a frequency range of 10 MHz-7 GHz or 10 MHz-5 GHz. Still other ranges will be recognized from the present disclosure, in particular in relation to the water exchange rates as discussed below. As will be described in further detail herein, while the BLGR may have an operating frequency range that is larger, the analysis of the S11 measurements may be limited to a frequency band, for example, a 1 GHz frequency band. Amplitude changes and frequency shifts are noticeable among different materials and concentrations. Different test materials have different radio frequencies at which they undergo excitation and the responses are identified in S11 measurement. Therefore an algorithm can be trained with appropriate datasets to identify and/or quantify the detected ions across the range of concentrations of 1000 ppm-400 ppb. A machine learning algorithm is introduced to analyze the measured S11 data. A support vector regressor (SVR) model is trained using the measured data of various salt samples. The training data is constructed by concatenating the 20,000 amplitudes and 20,000 phase values from the measured S11 data. The hyperparameters of the SVR are optimized using 10-fold cross-Validation method. In the examples herein, the SVR model can be trained with datasets to identify a plurality of ions, as with (Pb, Na, Mg, and/or K examples described), or as will be described in further detail herein, the SVR model may be trained with datasets for a single targeted ion. Based upon the trained model, the algorithm predicts the concentration(s) of ion(s) in the liquid samples. The experimental results indicate that the device can detect concentrations as low as 400 ppb with high accuracy.
In one example, a static and non-interfering heavy metal contaminant sensor makes continuous measurements for a heavy metal contaminant (Pb) in water using an AC magnetic field generated by RF resonator. In such examples, measurements down to 1 ppb Pb concentration in water may be made. The data collected by the RF S11 measurement may be processed by an algorithm trained with machine learning.
Further a block loop gap resonator (BLGR) is disclosed to improve these measurements. Examples of the resonator sensor and measurement technique are described with respect to the detection of various ions in water including, but not limited to, Pb2+, Na+, K+′ and Mg2+. The disclosed BLGR is sensitive to different ions and their concentration resulting in S11 amplitude and phase variations in the 10 MHz to 10 GHz range. While complex mathematical models may be used to analyze the S11 data, a SVR model as described above may be used. Such analysis may result in the classification of the test ions (Na+, K+′ and Mg2+) and to measure Pb2+ concentrations in water. Information of ion concentration and classification may be more pronounced near the resonance(s) of the BLGR. In an example of a sensing system configured to detect the presence and/or concentration of a plurality of ions, an increased number of resonances in the BLGR has been found to improve detection.
In the system 30, the trained SVR 40 is stored on a computer readable medium communicatively connected to a computer 42. The computer receives the S11 data from the VNA and applies the SVR algorithm to the S11 data to produce outputs that include the identification of ions present, including for example Na+, K+′ and Mg2+ ions as well as Pb2+ ions, additionally, the concentration of the Pb2+ or other heavy metal ions is quantified.
In an example, the BLGR with more cuts 24, e.g. two cuts resulting in three resonances, was found to be the most accurate in quantifying the concentration of Pb ions in the water samples, when the results were analyzed with a general classifier SVR algorithm. Still additional cuts 24 may produce still further resonances which may help to distinguish between ions and lead to improved concentration quantification.
In another example, the BLGR may be tuned, as noted above to be sensitive to detect a particular metal ion in water. The BLGR can be tuned by the combination and size of the gap 16 and the one or more cuts 24 in the resonator body to provide a BLGR with a predetermined resonant frequency or frequencies. As will be disclosed in further detail herein, it has been discovered that a BLGR with a resonant frequency that matches the rate constant for water exchange with a particular metal ion is sensitive to detection of that metal ion. The water exchange rate is a unique property of metal ions in water. While there are various processes and techniques for measuring water exchange rate, including isotope dilution, nuclear magnetic resonance, or sound absorption, which may lead to variations in the determined exchange rate values, the phenomenon is represented as the following chemical equation:
M(H2O)xn++H2O*(M(H2O)x-1(H2O*)n++H2O (7)
Margerum, D. W., G. R. Cayley, D. C. Weatherburn and G. K. Pagenkopf, 1978. Kinetics and Mechanisms of Complex Formation and Ligand Exchange, In: A. E. Martell (ed.) Coordination Chemistry, Vol. 2 ACS Monographs 174, ACS, Washington, incorporated by reference herein in its entirety.
Table 1 below provides examples of various rate constants for water exchange with metal ions. As noted, depending upon measurement technique and background conditions, there is some variation in the values of water exchange rate for various ions. Rate constant values (k_w(s−1)) from three different sources are provided in Table 1, where n/r indicates where a value was not reported in the particular reference.
5 × 10−7
3 × 10−8
3 × 10−8
The above water exchange rates produce an ion frequency at the associated frequency (e.g. Pb2+7×109 water exchange rate=7 GHz ion frequency). The water exchange rate can be used in two ways to improve sensitivity of detection and/or quantification of the concentration of a target ion in a water sample.
First, as noted above, the BLGR can be tuned to have a resonant frequency that is the same as, or similar to, the resonant frequency of the target ion. As represented in the table above, differing conditions and quantification techniques result in some differences in the nominal water exchange rate for ions. Therefore further examples of BLGR as disclosed herein may include cuts to produce a BLGR with multiple resonant frequencies to either match the different nominal water exchange rates or to create a resonant frequency band from multiple resonant frequencies that cover the range of nominal water exchange rates. Using the example above for Pb2+ ions, the BLGR is configured to produce a resonant frequency at 7 GHz, and/or a combination of resonant frequencies to produce a resonant frequency band, for example of 1 GHz, centered on the 7 GHz frequency.
Secondly, the water exchange rates may be used to focus the training and/or implementation of the SVR model to a frequency band that is centered upon or encompasses the frequencies associated with the water exchange rate(s) for the target ion. In an example, the SVR algorithm uses a 1 GHz band centered on the frequency associated with the water exchange rate, average water exchange rate, weighted average water exchange rate, or water exchange rate observed range to focus the model training and/or focus the application of the SVR model. In an example of the use of the water exchange rate to focus the training of the SVR model to produce an ion-specific detection model, the training is focused on a frequency band including the frequency associated with the water exchange rate of the target ion. In an example of the targeted application of the SVR algorithm using the water exchange rate, the training of the SVR model is focused on a 1 GHz frequency band centered on the 7 GHz frequency. During analysis, this same 1 GHz frequency band of the S11 output signal from the BLGR is analyzed using the SVR model to detect or quantify the target ion in the water sample. In the example above, It will be recognized that these techniques that apply the water exchange rate may be used alone or in combination with one another.
It has further been recognized that resonators with low Q take into account a relatively large frequency range in sensing applications. Therefore, a BLGR 10 with a higher tube radius (r) or larger length (l) will result in a higher Q. Therefore, in examples, BLGR 10 sensors as disclosed herein may be tuned with a high Q to a narrow frequency range associated with particular ions to produce a highly sensitive sensor to the target ion, or the BLGR 10 sensors may be tuned with a low Q to provide sensing across a wide frequency range with the ability to make robust detection of the presence of a wide range of ions.
Citations to a number of references are made herein. The cited references are incorporated by reference herein in their entireties. In the event that there is an inconsistency between a definition of a term in the specification as compared to a definition of the term in a cited reference, the term should be interpreted based on the definition in the specification.
In the above description, certain terms have been used for brevity, clarity, and understanding. No unnecessary limitations are to be inferred therefrom beyond the requirement of the prior art because such terms are used for descriptive purposes and are intended to be broadly construed. The different systems and method steps described herein may be used alone or in combination with other systems and methods. It is to be expected that various equivalents, alternatives, and modifications are possible within the scope of the appended claims.
This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to make and use the invention. The patentable scope of the invention is defined by the claims and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal languages of the claims.
The present application is a US National Phase Application of PCT Patent Application No. PCT/US2021/046623, filed on Aug. 19, 2021, claiming priority of U.S. Provisional Patent Application No. 63/067,691, filed on Aug. 19, 2020, and U.S. Provisional Patent Application No. 63/074,730, filed on Sep. 4, 2020, the contents of which are incorporated by reference herein in their entireties.
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/US2021/046623 | 8/19/2021 | WO |
Number | Date | Country | |
---|---|---|---|
63067691 | Aug 2020 | US | |
63074730 | Sep 2020 | US |