WO1991008469A2

WO1991008469A2 - System and method for monitoring substances and reactions

Info

Publication number: WO1991008469A2
Application number: PCT/US1990/006975
Authority: WO
Inventors: Bentley N. Scott; Samuel R. Shortes
Original assignee: Phase Dynamics, Inc.
Priority date: 1989-11-27
Filing date: 1990-11-27
Publication date: 1991-06-13
Also published as: AU6896691A; WO1991008469A3

Abstract

A system and method for monitoring conditions in a fluid medium. A stream of the fluid medium is flowed through a fluid container which is electrically configured as a transmission line segment and which is electrically connected to load a UHF or microwave oscillator. The oscillator is not isolated from the load, and is operated free-running, at a starting frequency which is chosen to provide a particularly strong shift in permittivity of the fluid medium, as the chemical reaction progresses. Preferably the frequency and insertion loss of the oscillator are monitored, to gauge the progress of the reaction.

Description

SYSTEM AND METHOD

FOR MONITORING

SUBSTANCES AND REACTIONS

Cross-Reference to Related Applications The present application claims priority from US application 442,980, filed November 27, 1989, and also from PCT application PCT/US 90/03849, filed July 9 1990. Both of these applications are hereby incorporated by reference.

BACKGROUND AND SUMMARY OF THE INVENTION The present invention relates to systems and methods for monitoring chemical reactions and/or changes in the composition and/or phase of chemical substances or mixtures.

Electrical Methods of Characterization

U.S. Patent 4,862,060 described a novel way to measure the water content of crude oil as it comes out of the ground. This system included a short section of piping, mechanically connected so that the fluid stream to be characterized would pass through it, and electrically connected to function as a transmission line (in the electrical sense)¹ in the feedback path of an oscillator. The oscillator was operated as a "free-running" oscillator (for reasons which will be explained below), and this system provided a very sensitive apparatus for monitoring changes (such as changes in the fraction of salt water) in the crude oil coming out of the ground.²

The system and methods disclosed in the present application provide a generally applicable method for monitoring the characteristics of a substance which includes a material (or a process flow) of interest, and also is connected electrically as part of the feedback path of an RF oscillator. Changes in the oscillation frequency provide a sensitive indicator of changes in the substance in the container. The container, in the presently preferred embodiment, is shaped as a segment of coaxial line through which fluid materials flow lengthwise, but other shapes can be used instead.

The "Load-Pull" Effect

It is well known to electrical engineers generally (and particularly to microwave engineers) that the frequency of an RF oscillator can be "pulled" (Le. shifted from the frequency of oscillation which would be seen if the oscillator were coupled to an ideal impedance-matched pure resistance), if

^JA simple electrical circuit, at low frequencies, can be analyzed as a net¬ work of discrete lumped elements, and the propagation delays between ele¬ ments can be ignored. However, at higher RF or microwave frequencies, this model is inadequate. A different and complementary way to analyze some electrical components or circuits is to model them as a "transmission line," le. an extended structure which has a distributed resistance and reactance over a finite length. Such structures behave quite differently from discrete lumped networks. An ideal uniform transmission line is completely described (electrically) by only two parameters: the phase velocity v_p and the characteristic impedance Z_o. ²An electrical oscillator must include a gain element, and a feedback path which couples the gain element's output back to the gain element's input (at least partly). The oscillator will operate at the frequency (or frequencies) where the total phase shift (through the gain element and the feedback path) is equal to an integral multiple of 360" (2π radians). the oscillator sees an impedance which is different from the ideal matched impedance. Thus, a varying load impedance may cause the oscillator frequency to shift.³

The present application sets forth various innovative methods and systems which take advantage of this effect. In one class of embodiments, an unbuffered⁴ RF oscillator is loaded by an electromagnetic propagation structure which contains, in its interior, a material for which real-time monitoring is desired. The net complex impedance⁵ seen by the oscillator will vary as the characteristics of the material in the electromagnetic

³Any electrical oscillator can be "pulled" to some extent - that is, its frequency can be shifted - by changing the net impedance seen by the oscil¬ lator. However, in many systems which use oscillators, pulling of a resonant circuit's frequency is undesirable. An oscillator which is too easily pulled may be overly susceptible to irrelevant external circumstances, such as changes in parasitic capacitance due to human proximity or temperature change. Normal techniques to avoid oscillator pulling include using isola¬ tion buffering circuits between the oscillator and the variable load, and/or using a high-Q tuned circuit to stabilize the oscillator.

⁴ An unbuffered oscillator is a oscillator without buffer amplifiers or attenuators. Amplifiers boost the output power and provide isolation from the load impedance changes. Attenuators decrease the amplitude while providing an isolation of two times the attenuation. In the load pulled oscillator configuration the oscillator feedback path that supplies the phase shift needed for oscillation is separated from the load. ⁵A "complex" number is one which can be written as A + Bi, where A is the number's "real" part, B is the number's "imaginary" part, and i² = -1. These numbers are added according to the rule

(A + Bi) + (C + Di) = (A + C) + (B + D)i, and are multiplied according to the rule (A + Bi)(C + Di) = (AC - BD) + (AD + BC)i.

Complex numbers are used in representing many electrical parameters. For example, impedance can be represented as a complex number whose real part is the resistance, and whose imaginary part is equal to the reactance (inductance or capacitance). Similarly, permittivity can be represented as a complex number whose imaginary part represents resistive loss, and whose real part represents reactive loading, by the medium, of the propagating electromagnetic wave. propagation structure varies. As this complex impedance changes, the oscillator frequency will vary. Thus, the frequency variation (which can easily be measured) can reflect changes in density (due to bonding changes, addition of additional molecular chains, etc.), ionic content, dielectric constant, or microwave loss characteristics of the medium under study. These changes will "pull" the resonant frequency of the oscillator system. Changes in the medium's magnetic permeability will also tend to cause a frequency change, since the propagation of the RF energy is an electromagnetic process which is coupled to both electric fields and magnetic fields within the transmission line.

Properties of a Dielectric in a Transmission Line

To help explain the use of the load-pull effect in the disclosed innovations, the electromagnetics of a dielectric-loaded transmission line will first be reviewed. If a transmission line is (electrically) loaded with a dielectric material (as, for example, the measurement section of the apparatus of Figure 1 is loaded by the liquid flowing through the cavity), changes in the composition of the dielectric material may cause electrical changes in the properties of the line. In particular, the impedance of the line, and the phase velocity of wave propagation in the line, may change. This can be most readily illustrated by first considering propagation of a plane wave in free space. The propagation of a time-harmonic plane wave (of frequency f) in a uniform material will satisfy the reduced wave equation

(v² + k^E = (v² + k²)H = 0, where

E is the electric field (vector), H is the magnetic field (vector), and v² represents the sum of second partial derivatives along the three spatial axes. This equation can be solved to define the electric field vector E, at any point r and time t, as

E(r,t) = E_oexp[i(k-r - ωt)], where k is a wave propagation vector which points in the direction of propagation and has a magnitude equal to the wave number k, and ω = Angular Frequency = 2πf. In a vacuum, the wave number k has a value "k_o" which is k_o = ω/c = ωμ₀e₀)*, where μ₀ = Magnetic Permeability of vacuum (4πxl0^~7 Henrys per meter), e₀ = Electric Permittivity of vacuum ((l/36π)xl0^~* Farads per meter), and c = Speed of light = (μ₀€o)^{_ 4} = 2.998x10^s meters/second. However, in a dielectric material, the wave number k is not equal to k„_. instead k = ω/(c(μ_re_r)*) = ω(μ₀μ_re₀e_r)^w, where μ_r = Relative Permeability of the material (normalized to the permeability μ₀ of a vacuum), and e_r = Relative Permittivity of the material (normalized to the permittivity e„ of a vacuum).

Thus, if the relative permeability μ_r and/or the relative permittivity e_. vary, the wave number k and the wave propagation vector k will also vary, and this variation will typically affect the load pulled oscillator frequency.⁶

he full analysis of wave propagation in a cavity or at a boundary is much more complex, but in any case wave propagation will depend on the wave number, and the foregoing equations show how the wave number k can vary as the medium changes. See generally, e.g., R.Elliott,

(continued...) Frequency Hopping in a Load-Pulled Oscillator

In a typical free-running oscillator, the oscillator frequency is defined by a resonant feedback circuit (the "tank" circuit), and can also be pulled slightly by a reactive load,⁷ as noted above. Thus, such an oscillator can be broadly tuned by including a varactor in the tank circuit.⁸

As the oscillator's frequency is thus shifted, the phase difference between the energy incident on and reflected from the load element (which is preferably a shorted transmission line segment) will change. This phase difference will be equal to an exact multiple of 180° at any frequency where the electrical length of the transmission line segment is an exact multiple of λ/4.

As the oscillator frequency passes through such a frequency (i.e. one where the transmission hne segment's electrical length is equal to a multiple of λ/4), the load's net impedance will change from inductive to capacitive (or vice versa). As this occurs, the frequency of the oscillator may change

'(...continued) Electromagnetics (1966); J Jackson, Classical Electrodynamics (2d ed. 1975); G.Tyras, Radiation and Propagation of Electromagnetic Waves (1969): R.Mittra & S Lee, Analytical Techniques in the Theory of Guided Waves (1971); L Lewin, Theory of Waveguides (1975); all of which are hereby incorporated by reference.

^e degree by which the reactive load can change the oscillator's frequency will depend on the coupling coefficient between the load and the tank circuit. Thus, an increased coupling coefficient means that the oscillator frequency will be more sensitive to changes in the load element. However, the coupling coefficient should not be increased to the point where spectral breakup (multiple frequency operation) occurs, since this would render the desired measurement of the oscillator signal impossible.

This is one type of voltage-controlled oscillator (VCO). abruptly rather than smoothly.⁹ This jump in frequency will be referred to as a frequency "hop".¹⁰

For a transmission line of length / which contains a dielectric material of relative dielectric constant e_rt the frequency at which one full wavelength (lλ) exists in the transmission hne is equal to c (the speed of light in vacuum, which is 2.995x10° meters/second) divided by the length of the line in meters and by the square root of the relative dielectric constant of the material:

Frequency- = c / (/e_r ). For example, for a one-foot-long line filled with a material having e_r = 1, / = 12 inches (= 0.3048 meters), and

Frequency_lλ = (2.995xl0⁸) / (0.3048 x 1.0) « 980 MHz.

However, since one wavelength actually contains two excursions from inductive to capacitive reactive impedances, only one-half wavelength is required to see one frequency hop of the load pulled oscillator. If the transmission hne is terminated into a short or an open, the resulting effective length is increased to twice the actual length, since a standing wave is generated (due to the energy incident at the short or open being reflected back to the input of the transmission line). In essence, the energy travels down the line, gets reflected, and travels back to the input. With this taken into account, the first frequency with a wavelength long enough to cause a frequency "hop" of the oscillator is one fourth the length calculated above, or 245 MHz.

Multiples of this first quarter-wavelength frequency will also cause the impedance seen at the input to the transmission line to go from inductive to capacitive reactance. The longer the transmission line, the greater the

This change in frequency, as the load goes from capacitive (-180 ° ) to inductive (+180° ) or vice versa, is instantaneous if the equivalent parallel resistive part is large (e.g. greater than approximately 500 ohms in a 50 ohm system).

'The amount by which the frequency shifts during the "hop" will depend on the Q of the load element (as seen by the oscillator circuit), and on the coupling coefficient between the load element and the tank circuit. number of phase transitions that will occur. Longer line length also multiplies the phase changes that are brought about by a change in the dielectric constant. For every one-quarter wavelength change in the effective (electrical) length of the line, the complex impedance seen at the oscillator changes by 180 ^β .

For example, suppose that a given oscillator, coupled into a low loss load with an electrical length of one-quarter wavelength (λ/4), provides 50 MHz of load pulling frequency change (total excursion through all phases). If the monitored material changes enough to produce a change of only one degree of phase in the electrical length of the load, the oscillator frequency will change by 138.9 kHz. This represents an absolute resolution of 7.2x10^"* degrees of phase change for each Hertz of sensitivity." For every additional quarter wavelength of line length, this sensitivity to phase is multiplied by 1.5. This is due to the change in phase being an additive function of every additional quarter wave in the measurement section.

Figure 7 shows a typical tuning frequency versus voltage plot for a VCO loaded into a shorted transmission hne. The height of the "hop" can be measured by holding the VCO tuning voltage constant, while a transmission line terminated into a short is varied in length¹² to cause a full rotation of the impedance vector seen at the VCO's input port. The resulting data of frequency versus length of the transmission line will show a jump in frequency (a delta frequency from the bottom of the "hop" to the top of the "hop") which coincides with the delta frequency of the "hop" seen when the VCO was swept using the tuning voltage.

"Even if the resolution of frequency measurement is only ±100 Hz, this would still give an accuracy of better than one-thousandth of one degree. This is vastly better resolution than is possible with vector impedance systems (such as an HP 8510 Network Analyzer). ^uch variable transmission lines are commonly used in the microwave industry, and are referred to as "hne stretchers." Thus, if the VCO is swept across a frequency band and the number of frequency "hops" was counted, the number of "hops" reveals the number of wavelengths in the transmission line.¹³

This provides a means for determination of the range of dielectric constant change in a medium even when it rotates the phase vector multiple times (and therefore, the oscillator frequency returns to the same value multiple times). If the dielectric constant of the material in the transmission line is increased, then the above equations show that the frequency of the first full wavelength is decreased by the square root of the dielectric constant. Additionally, this means that the number of wavelengths at a fixed frequency increases with increasing dielectric constant. These facts imply that the VCO tuning curve will see more "hops" as the dielectric constant is increased due to the increasing fraction or whole wavelengths encountered. Ideally, the oscillator will not cease oscillations (or break into multiple frequency oscillation or spectral breakup) into any load regardless of the load characteristics. However, this is not a strictly necessary condition for use of the disclosed method and system innovations."

Measurement of Substances with a High Microwave Loss Factor A measure of the dielectric loss of a material is typically given as the dielectric loss tangent (a unitless number) which is defined as the tangent of the imaginary part divided by the real part of the complex dielectric

"More precisely, it will be found that the wavelengths at which hops are observed are separated from each other by one-quarter of the effective (electrical) length of the measurement section.

¹⁴ The second harmonic of the oscillator frequency is typically enhanced (becoming greater in amplitude than the fundamental frequency) just before the shift from inductive to capacitive impedance (or vice versa), due to the extreme non-linearities at this point. This does not hinder the use of load pulling as a measurement technique, since the measurement is typically made outside of this region of the impedance shift from inductive to capacitive. Alternatively, the second harmonic may be filtered out of the measurement. constant. Low loss materials are typically below a loss tangent equal to or less than 0.01. When the disclosed systems are used to measure materials with a high loss factor, the material's absorption begins to dominate the load versus frequency effects, but a measurement capability still exists due to the sensitivity of the load pulling method.

However, a potential problem with highly conductive materials in an apparatus like that of Figure 1 is that the observed loss of the system may decrease for more lossy materials, since the reflection at the interface between the material under study and the microwave transition section will increase. A solution to this problem is outlined in parent application Serial No. 376,782 filed 7/7/89. As taught therein, a very good dielectric material is added as a sheath around the coaxial center conductor. This material prevents the electric field from going to zero immediately near the center conductor of the transmission line (which could otherwise occur, due to the highly conductive medium under study shorting out the electric field to the outer conductor wall). With the dielectric material as a sheath, propagation along the center rod can occur with slight loss and a small penetration of the good conductor material of the center rod. On the outer portion of the dielectric interface, the medium under study becomes the virtual outer wall of the coaxial conductor with a skin depth of propagation which encompasses the entire medium under study and terminates on the actual outer metal wall of the coaxial line. The resultant changes in the complex dielectric constant are still reflected in a change in the complex load impedance seen at the load pull oscillator and a measurement is still viable.

Additional Information from Load Pull Measurement

The disclosed innovative system and method also permits other information to be derived, regarding the substance being monitored.

Difference in Operation Frequency Additional information can be obtained by retuning the VCO, so that the frequency is forced to change, and making another measurement at a much higher frequency. Since materials change properties versus frequency, the amount of frequency change due to load pulling will vary versus the frequency of operation.

A VCO will typically be designed to cover approximately one octave above its turn on frequency. If a VCO would not give enough frequency change to see the desired range of varying parameters versus operating frequency, an additional unbuffered oscillator, which runs at any frequency required to obtain appropriate data, may be switched into the coaxial line.

When two widely spaced frequencies are measured for a medium under study with a load pulled oscillator, the difference (delta) frequency between these two measurements will be unique for a given medium. This phenomena will aid in distinguishing constituents and the progress of mixing or reaction.

Since some substances exhibit a relatively constant magnitude of complex permittivity versus frequency but others have a very strong change versus frequency, this fact provides a means by which the amount of one substance in another might be distinguished when there exists a third substance. Figure 24A shows three substances with their corresponding behavior of the magnitude of the complex permittivity versus frequency. Substance A could be a low dielectric constant low loss material such as a ceramic. Substance B would be a higher dielectric constant material that is later partially contaminated with an ionic substance which exhibits behavior shown as Substance C. The substance C would pull the oscillator less the higher the frequency of operation due to the nature of the load pull measurement.

If either a tunable VCO or two oscillators operating at widely separated frequencies were fixed to a measurement section where the described Substances were contained, the amounts of substances A, B and C could be determined from the information contained within the frequency data. Figure 24B gives volume percent content of (Substance B)/(Substance A) vs. load pull information of two oscillators at widely spaced frequencies. Curves marked 1 and 2 are for 5% content of Substance C, while curves marked 3 and 4 are for 15% content of Substance C. Figure 24C shows a resultant curve for the difference in frequency between the high and low frequency oscillators for the two cases of Substance C versus the volume percent of Substances A and B. If the system was calibrated such that this information was known, then for a real time situation there are two equations and two unknowns which can be solved for the percent content of Substances B/A and the concentration of C.

Monitoring of Insertion Loss

If the incident power and the reflected power is measured in a system where the final load is a short, the difference in powers is the insertion loss of the medium multiplied by a factor of two (since this power difference is caused by two transits through due to the path down to the short and return is a path length twice through the medium of interest). The insertion loss measurement will aid in determination of the changing conductivity of the medium or its change in absorption of the RF energy. This information can be related to the mixing or reaction products to further distinguish unique situations where the frequency change of the load pulled oscillator is not enough information or resolution by itself.

Effect of Complex Permeability The magnetic permeability μ_r can also be dynamically measured by the disclosed techniques. Since the velocity varies with (μ_re_r)^~ changes in μ_r will change the phase shift through a given physical length of line, and thus change the frequency of the oscillator.

A sample-containing waveguide, like that of the principally preferred embodiment, will typically have locations where the electric field is strong but the magnetic field is zero; at such locations only permittivity will affect the oscillator load pull frequency. However, there will also commonly be locations in a waveguide where the magnetic fields are locally strong and electric field is zero: at these locations, only the permeability will affect the propagation characteristics of the transmission hne (and therefore contribute to the oscillator frequency). A system can be built to sample (primarily) one of these parameters.

For example, to sample the permeability, the coaxial transmission line will be terminated into a short where the medium of interest is located only in close proximity to the short. A waveguide structure supports very well defined electrical and magnetic field functions, and the sample can be suitably placed in such a structure to measure primarily the permeability.

Typical compounds and substances do not have varying magnetic permeabilities and therefore, most of the discussion will involve the changing complex permittivity. But, the effects of changing complex permeability will create similar changes in the oscillator load pulling characteristics. If a substance such as barium titanate is studied, the effect of the changing permeability must be considered along with the change in permittivity unless the system is designed specifically to measure only one of these.

The Measurement Section

The transmission line selected for the majority of the measurements was a coaxial line due to its simple transverse electromagnetic (TEM) mode of propagation. The TEM mode is the simplest mode to set up and to maintain under varying conditions. If other modes were excited by a perturbation, the energy may not be recovered and therefore the information would be lost to the measurement. The coaxial line lends itself to a easily built and modified system which can encompass vast material changes to optimize both the measurement and the environmental conditions that it must work under. The diameter of the rod and the terminations may be easily altered to improve sensitivity by matching of the impedances thereby transferring more of the RF energy into the medium under study.

Addition of a good dielectric as a sheath to the center rod will provide measurements for highly conductive substances under study. The sheath must be thick enough to provide a stable field pattern between the center conductor and the conductive medium under study. The effect of adding this sheath is to in effect make the measurement as a function of the wave propagated as a skin depth in the conductive medium under study which is theoretically equal to or greater than the actual distance between the sheath and the outer coaxial wall.

If the dielectric constant of the material in the transmission line is increased, then the above equations shows that the frequency of the first full wavelength is decreased by the square root of the dielectric constant. This implies that the VCO tuning curve will see more "hops" as the dielectric constant is increased due to the increasing fraction or whole wavelengths encountered.

Coupling the Active Device

An unusual feature of the oscillator configuration used with the present invention is the separation of the load of interest from the resonant circuit proper. The configuration used isolates the two through the active device. It is the non-linear behavior of the transistor that provides the changes in frequency as the load is changed. The loop gain of an oscillator must be unity with an appropriate phase shift to cancel the negative impedance's imaginaiy part¹⁵ around the resonant loop. The initial gain of the active device must be greater than unity before oscillations can begin in order for the oscillator to be self starting. This extra gain is reduced to unity by the saturation of the active device upon establishment of the oscillations. Saturation of a device normally also changes the phase shift through the device¹*. This requires a change in the operation frequency as the load

"In a simple resistor, an increase in the current passing through the resistor will produce an increase in the voltage across the resistor. By contrast, in microwave gain diodes (or in a transistor with feedback connections) which is operating at less than its saturated current, a small transient increase in the current across the device will produce a reduction in the voltage across the device. Thus, since a simple resistor has a positive impedance, such gain devices are referred to as having a negative impedance.

"As the gain device approaches saturation, the physics of its operation will gradually change. These changes may cause the phase shift across the

(continued...) changes due to the shift in loop gain and phase by the saturated condition change in the active device.

Spectral Purity of Oscillator

It has been discovered that, in a system using a free-running oscillator as described above, spectral purity of the oscillator is an important concern.

Many microwave oscillators exhibit "spectral breakup," wherein the spectrum of the oscillator's output actually contains multiple frequencies.

In most microwave oscillators this is not a problem, since a tuned feedback element will be used to stabilize the gain element, and/or isolation or buffering stages are used to prevent the oscillator's feedback loop from being perturbed by extraneous resonances. However, in the preferred system, since such isolation stages are not used, spectral purity turns out to be quite important. For example, a spurious resonance in the feedback loop (e.g. due to a low-quality RF choke, or due to two impedance mismatches) can permit the oscillator to hop to a frequency which is determined (at least partly) by a harmonic of the spurious resonance, in which case the degree to which the oscillator frequency has been pulled by the changing load will be obscured.

In the presently preferred embodiment, a small series resistor is interposed in the RF output of the oscillator, before the measurement section connection. This resistor adds a small amount of damping, which helps to suppress oscillation at secondary frequencies).

Also, in the presently preferred embodiment, a shunt resistor is attached to the RF output of the oscillator. This resistor also adds to stability, by fixing a maximum magnitude for the load impedance seen at the RF output hne.¹⁷

"(...continued) gain device to vary significantly. Note that, in the saturation regime, the gain device behaves as a non-linear circuit element. "At frequencies where the length of the transmission line segment is a multiple of k/4), its impedance can become very large. Previous Attempts at Electrical Characterization

Various types of apparatus have been proposed for measuring the concentration of one substance in another, particularly the concentration of a liquid or flowable substance in another liquid or flowable substance. Various devices which utilize the broad concept of determining composition of matter by measuring changes in a microwave signal are disclosed in U.S. Pat. Nos. 3,498,112 to Howard; 3,693,079 to Walker; 4,206,399 to Fitzky et al; 4,311,957 to Hewitt et al; 4,361,801 to Meyer et al; 4,240,028 to Davis Jr.; 4,352,288 to Paap et al; 4,499,418 to Helms et al; and 4,367,440 and 4,429,273, both to Mazzagatti; all of which are hereby incorporated by reference.

Although various systems utilizing microwave transmissivity or signal alteration characteristics have been proposed in the prior art, certain considerations in utilizing microwave energy to detect the presence of the concentration of one medium in another have not been met by prior art apparatus. In particular, it is desirable in certain instances to be able to accurately measure, on a continuous basis, the concentration or change in concentration of one fluid in another and particularly where the concentration of one fluid is a very low percentage of the total fluid flow rate or fluid mixture quantity. It is also desirable that the signal change caused by the presence of one substance or medium in another be easily measured and be relatively error free, again, particularly in instances where measurements of low concentrations of one substance such as a fluid in another substance such as another fluid are being taken. Moreover, it is important to be able to transmit the microwave signal through a true cross section of the composition being sampled or measured to enhance the accuracy of the measurement.

Typical systems for capacitive based measurement have a capacitive element, used for parameter determination, as part of the resonant feedback loop around an active device. This method works well with very low loss systems, but oscillation ceases with even slightly lossy measurements. As the frequency is increased into the microwave region, it becomes difficult to configure the resonant feedback loop due to the increase in loss versus frequency and the wavelength becoming comparable to the path length. In this case the frequency is changed directly by the resonance change in the feedback loop which includes the element that consists of the sample to be measured. This frequency change is limited to the characteristics and loss of the feedback path and can only be changed over a narrow frequency range with out cessation of oscillations. This limits the measurement technique to small samples of very low loss.

At higher frequencies (above approximately 100 MHz), the capacitive measurement technique fails to work, due to hne lengths and stray capacitances. At such frequencies resonant cavity techniques have been employed. (For example, a sample is placed in a resonant cavity to measure the loss and frequency shift with a external microwave frequency source that can be swept across the resonance with and without the sample in the cavity.) This method uses a highly isolated microwave frequency source which is forced by the user (rather than being pulled by the changing resonance) to change its frequency. This technique too meets substantial difficulties. For example, the use of multiple interfaces without a microwave impedance match at each interface causes extraneous reflections, which tend to hide the desired measurement data. This technique too gives errors with very lossy material, but in this case it is due to the very rounded nature of the resonance curve (which is due to the low

Q of the loaded cavity). This rounded curve makes it difficult to determine both the center frequency and the 3 dB rolloff frequency closely enough to be accurate in the measurement. Another technique which is used encompasses the use of a very sharp rise time pulse to obtain time domain data, from which frequency domain values are then derived through transformation techniques.

In U.S. Patent 4.396.062 to Iskander, entitled Apparatus and Method for Time-Domain Tracking of High-speed Chemical Reactions, the technique used is time domain reflectometry (TDR). This contains a feedback system comprising a measurement of the complex permittivity by

TDR means which then forces a change in frequency of the source which is heating the formation to optimize this operation. Additionally it covers the measurement of the complex permittivity by TDR methods.

U.S. Patent 3.965.416 to Friedman appears to teach the use of pulse drivers to excite unstable, bi-stable, or relaxation circuits, and thereby propagate a pulsed signal down a transmission hne which contains the medium of interest. The pulse delay is indicative of the dielectric constant of the medium. As in all cases, these are either square wave pulses about zero or positive or negative pulses. The circuit is a pulse delay oscillator where the frequency determining element is a shorted transmission line. The frequency generated is promoted and sustained by the return reflection of each pulse. The circuit will not sustain itself into a load that is lossy, since the re-triggering will not occur without a return signal of sufficient magnitude. In addition, the circuit requires a load which is a DC short in order to complete the DC return path that is required for re-triggering the tunnel diodes.

The frequencies of operation of any pulse system can be represented as a Fourier Series with a maximum frequency which is inversely dependent upon the rise time of the pulse. Therefore, the system covered in the Friedman patent is dependent upon the summation of the frequency response across a wide bandwidth. This causes increased distortion of the return pulse and prevents a selective identification of the dielectric constant versus frequency. This also forces a design of the transmission system to meet stringent criteria to prevent additional reflections across a large bandwidth. The low frequency limit of the TDR technique is determined by the time window which is a function of the length of the transmission line. The upper extreme is determined by the frequency content of the applied pulse. In the case of this pulse delay hne oscillator, the upper frequency is determined to a greater extent by the quality of impedance match (the lack of extra reflections) from the circuit through to the substance under study. These extra reflections would more easily upset the re-triggering at higher frequencies. In one case (Figure 1 of Friedman) the return reflection initiates a new pulse from the tunnel diode and therefore sets up a frequency (pulse repetition rate) as new pulses continue to be propagated. This is in essence a monostable multivibrator with the return reflection being the trigger. The problem implied, but not completely covered with this approach, is that due to the delay in pulses, the pulse train can overlap and cause multiple triggers to occur. These are caused by the re-reflections of the original parent pulse. An additional problem is with very lossy dielectrics, which will not provide enough feedback signal to initiate the next pulse. If the dielectric medium is of high enough dielectric constant to contain more than one wavelength, or if the dielectric constant of the samples vary greatly, multiple return reflections will alter the behavior of the circuit to render it useless due to the interfering train of return and parent pulses. Figure 3 of Friedman shows a bistable multivibrator which senses the return pulse by sampling and feeding back enough phase shifted voltage to re-set the tunnel diodes. Since this device is also dependent upon the return to trigger or re-trigger the parent pulse, it suffers problems with lossy dielectrics and high dielectric constant mediums. To overcome these problems, the relaxation oscillator of Figure 4 of

Friedman was proposed that contains a RC (resistor/capacitor timing) network which will maintain the generation of pulse trains using resistor 76 and capacitor 78 with the dielectric filled transmission hne affecting the regeneration of the pulses as the reflected parent pulse voltage is returned. Since the RC time constant is defining the basic repetition rate, some improvement is obtained in reducing second order effects. The transmission line is still an integral part of the overall relaxation oscillator and lossy dielectrics may cause irregular circuit response. The proposed inverting amplifier as the pulse generator will not function at above approximately 1 MHz in frequency due to the characteristics of such inverting amplifiers. The tunnel diode can pulse up to a 100 MHz rate.

By contrast, the innovative system embodiments disclosed in the present application and its parents differ from the known prior art in using a microwave frequency generated by a free running sine wave oscillator. The preferred oscillator has the versatile capability to work into a wide variety of transmission lines or other load impedance without generation of spurious data or cessation of oscillations. It will continue to oscillate with very lossy dielectrics. It is not a relaxation oscillator or a multivibrator. The frequency of the un-isolated oscillator is dependent upon the net complex impedance of the transmission line and will work into an open circuit as well as a short circuit. The net complex impedance at the frequency of operation of the oscillator looking at the transmission line containing the medium of interest results in stable oscillations through pulling of the unisolated oscillator. Only one frequency at any one time is involved in the disclosed system proposed (not counting harmonics which are at least 10 dB down from the fundamental). This provides for well defined information and eases the transmission design criteria. This also provides for evaluation of the dielectric constant versus frequency which can improve resolution of constituents or ionic activity.

Another important difference from prior art is the separation of the load of interest from the resonant circuit proper. The configuration used isolates the two through the transistor. It is the non-linear behavior of the transistor that provides the changes in frequency as the load is changed. The loop gain of an oscillator must be unity with 180" phase shift. The initial gain of the transistor must be greater before oscillations begin in order for the oscillator to be self starting. This extra gain is reduced to unity by the saturation of the active device upon establishment of the oscillatory frequency. Saturating a device changes the gain (and accordingly the phase since it is non-linear) to maintain oscillations as the load changes. This will continue as the load changes as long as the transistor has appropriate phase and available gain to satisfy oscillations.

On-Line Characterization of Reactions The disclosed inventions use a load-pull oscillator architecture to directly monitor the changing properties of materials in a process flow. The oscillator load pull technique provides an extremely sensitive measurement of phase changes in a dielectric or semi-conducting medium. Because of the inherent sensitivity of the load-pull oscillator system, it is possible to monitor chemical reactions dynamically. Since most chemical reactions progress through several intermediate states before reaching the final reaction product, it becomes possible to correlate their characteristics to desired properties of the final product. This allows the optimization and control of yield and of product characteristics.¹⁸

Relation of Measured Electrical Parameters to Molecular and Microstructural Changes The "load pull" technique can reveal very significant information about the chemical and physical organization of the material being studied. Some of the features of interest, and the causative relations between these features and the electrical parameters which are directly measured, will now be described.

Types of Polarization¹⁹

There are four different mechanisms which can mediate compliance of molecules with an applied electric field: these include electronic

¹⁸For example, in polymerization reactions, a small change in the reaction conditions may produce a change in product molecular weight (and/or chain length and/or degree of cross-linking) which dramatically changes the mechanical properties of the polymer product.

"Maxwell's equations in their full form (as applied to a material body) distinguish the applied electric field vector E from the induced electric field vector D. The two vectors are related as

where c. is a tensor in anisotropic materials, but can usually be treated as a scalar. The induced field D can further be written as the sum of the applied field and a polarization vector P: D = e_oE + P. As the following discussion will explain, the polarization vector P can usefully be represented as a sum of four vectors:

P ^■* = P * « ctroo -l '- P ^■*^■ ionic 4 ^- ^■P* oήenuticn 4 ^■- P •^■■ _n___r__ιαal

= Pe + ' P-^ icx. + ' P^■'^"■ or + ^τ * P ««• polarization, ionic polarization, orientational, and interfacial polarization. It is important to distinguish these four mechanisms, since they appear to different degrees in different materials, and typically have different strengths and different relaxation-time characteristics.

Electronic Polarization P.

The electronic polarization P _eα^ (or P_e) represents a shift in the electron cloud of an atom (or molecule) with respect to the nucleus (or nuclei) within the cloud. This polarization has a very short relaxation time, and remains important up through optical frequencies and beyond.

Ionic Polarization Y__.

The ionic polarization P^ (or P^) is only found in ionic crystals. It represents displacement of one charged element of the crystal's unit cell with respect to the other elements of the unit cell. This type of polarization has a slower time constant, but remains significant through microwave and submilhmeter wavelengths. This type of polarization is responsible for the huge dielectric constants seen at low frequencies in ferroelectric materials such as niobates and titanates.

Orientational Polarization P..

Orientational polarization P_oπenu∞o (or ¥„) occurs when individual molecules of a substance have separate dipole moments (on a small scale). (Many substances have such atomic dipole moments.) In such substances, an applied electric field will tend to orient the molecular dipoles.²⁰ This mode of polarization is still slower, with a relaxation time which is typically on the order of microseconds (so the cutoff frequency is typically well below 1 MHz).

^An extreme example of such orientation is the practice of "poling" electrets, in which large molecules including charged groups are frozen into a polarized condition by applying an electric field. A related phenomenon, on a much slower time scale, can occur in two- phase compositions. For example, where aspherical solid grains are dispersed in a fluid medium with a lower dielectric constant, the solid grains will tend to orient along the electric field lines.

Interfacial Polarization P...

Classically, interfacial polarization

occurs in solids when charged carriers migrate to a grain boundary (or defect site, etc.). This can be the slowest of all the polarization mechanisms described, with a time constant (determined by the rate of diffusion of carriers) of the order of seconds.

A related phenomenon can also occur in two-phase compositions. For example, where small metallic grains, or droplets of salt water, are dispersed in oil, charge separation may occur across each conductive element. Where the resistivity of the conducting domains is low, the cutoff frequency in such cases may be high enough to be of interest in fluid measurement systems.

Chemical and Microstructural Differences Conducive to Analysis

The disclosed techniques and system embodiments can accordingly be used to monitor substances and reactions in many ways, by making use of many different effects.

Increased Molecular Polarization

Increased polarization of the molecule will provide a higher dielectric constant, and thereby cause a frequency shift.

Increased Orientational Polarization

This too will tend to increase the dielectric constant. Orientational polarization will typically be quite lossy at RF frequencies.

Increased Interfacial Polarization This too will tend to increase the dielectric constant, and will typically be quite lossy at RF frequencies.

Increased Ionic Polarization

Increased ionic polarization of the molecules can happen, for example, as a result of a reaction which transfers charged functional groups. This will lead to a shift in dielectric constant and distinct change in microwave loss characteristics.

Polarization Dependence on Bond Shifts

Bond positional changes cause a shift in dielectric constant due to the change in polar moment.

Moreover, the change in the interstitial fit of a sea of molecules due to a shift in the bond locations can also cause a density change which, in turn, causes a shift in dielectric constant. This can also lead to a shift from non- ionic to ionic structures.

Relocation of Functional Groups

Group site changes will tend to have an effect which is at least as strong as bond shifts. Moreover, if a molecular resonant frequency can be sampled, a strongly detectable difference may be found.

Increased Chain Length Growth of the molecule (by adding more chains, even without an associated ionic change) will cause a shift in density and, therefore, dielectric constant. This will also shift any rotational or vibrational resonances of the molecule to a lower frequency.

Changes in Ionic Bonding Character Changes in the degree of ionicity of bonds, or changes from non-ionic to ionic bonding, are easy to detect. These changes will affect the insertion loss at microwave frequencies the greatest. 26

Grouping above the Molecular Size Range

The short-range ordering of molecules such as polywater or thixotropic substances will show different microwave properties due to the "sea" of molecules' polar moment changing relative to the state of polarization or hnkage. This will be reflected in the microwave system as a change in dielectric constant.

Phase Changes

Phase changes of materials can be seen due to the change in dielectric constant. This can include introduction of a gas phase or a phase change such as oil continuous phase (droplets of oil surrounded by water as the continuous medium).

Conductivity Changes

Changes in the conductivity of a material are hkely to produce a strong shift in the electrical characteristics, for several reasons. First, increased conductivity will typically increase the RF loss, since free carriers are subject to loss mechanisms which do not apply to orbital shifts. Second, the presence of additional free carriers can increase the contribution of interfacial polarization, in a two-phase medium, if the frequency is low enough to let this become important.

Gas-Phase Measurement

The disclosed system has also been successfully applied to mixtures which are solely composed of gas-phase components. In such cases, it is usually preferable to use an oscillator frequency which is higher than would otherwise be used.

Measurement of Multiple Parameters

The load-pull oscillator architecture permits direct measurement of a number of parameters, including:

1. Oscillator frequency shift.

2. Insertion loss. 25

Molecular Resonances (Rotational and Vibrational)

Molecular resonance will be seen at specific microwave frequencies as a dip in power and a change (not unlike injection locking an oscillator) in frequency due to the changing load effect of the resonance.

Particulate Contamination

Paniculate contamination (e.g. by metal particles) will be seen as a shift in frequency with little change in power due to the small size of most contaminants. The metal particles described are seen as an artificial dielectric, due to the small areas and the effect of the interfacial polarization. This will increase the dielectric constant.

Dilution/addition

Changing the molar content of a solution will adjust the dielectric constant and ionic strength accordingly, and will shift the microwave frequency and the power, respectively.

Molecular Recombination

Of course, the formation of different compounds will alter the characteristics of the dielectric constant and the loss.

Use of "Tag" Compounds to Track Reactions

In addition to direct monitoring of reactions, monitoring can be enhanced by adding a "tag" compound into one of the ingredients. Such tagged compounds can be used to track reactions which would not otherwise have a great enough microwave response. The "tag" compounds would attach before, during, or after the chemical process has occurred, but they would not alter the reaction product. They would increase the 'Visibility" of the process to the microwave system.

Fine Structure of Dispersions

The dielectric constant of a sol or a two-phase mixture will have a direct correlation to the degree of dispersion. 3. Both oscillator frequency shift and insertion loss will vary with frequency, in ways which vary from one substance to another.

4. Both oscillator frequency shift and insertion loss will vary with temperature, in ways which vary from one substance to another. 5. Oscillator frequency and/or insertion loss may vary usefully with pH, or with the some other concentration value, so that the oscillator's behavior can be tracked during a short titration process to gain additional information.

Although signal attenuation will be the aggregate response of all contributing components, a variety of system information may be derived directly or indirectly from the direct measurements. Some of the more important ones are:

(A) reaction kinetics (reaction rate) and extent of reaction ^• ; ratio of desired compounds in final product to that produced by a competing side reaction

(C) physical phase changes in reaction vessel.

(D) reactions reflecting the condition of the catalyst. Example: If a non-ionic solution is reacted producing an ionic solution, this will cause power loss in the medium to increase due to the increased conduction in the microwave section. The slope measured at two widely separated frequencies will be vastly changed, since this is a very frequency dependent loss. An associated frequency change will be seen in the oscillator since the frequency change will be seen in the oscillator since the frequency of oscillation is dependant upon the resultant complex impedance (real and imaginary parts of the impedance).

Example: If the reaction only contains substances that during the reaction create bond position changes or re-orientation of bond groups (no- ionic exchanges or large shifts in pH), this may require a higher frequency to discern the chemical changes. Loss will not be appreciable unless the frequency is high enough to observe structure resonances (polar resonances). Widely separated tuning voltages on a given VCO will give frequency differences which are unique (due to varying dielectric properties versus frequency). A broadband sweep and the resulting location of frequency hops will indicate relative dielectric constant as well as indications of dielectric change versus frequency. Frequency hops are caused by the oscillator seeing a phase shift going from inductive to capacitive or vice versa. This creates frequency discontinuities of approximately 20 to 60 MHz (dependant upon the magnitude of the real part and the fundamental frequency of operation) when the load traverses this point in the complex plane.

Example: If the reaction has properties of progressing through various sub-classifications of reaction types, each change (from ionic to non-ionic to bond changes for example) will have changing slopes of frequency versus time and frequency versus power loss. The differentiated functions will give indication of the progress of the reaction. Further knowledge of the constituents may be discerned through the voltage sweep of the Voltage Controlled Oscillator to reveal the frequency hop positions. Example: If more specific knowledge of the chemical constituents is required, higher frequencies may be used to look at the apparent microwave resonance caused by the molecular resonance. This apparently begins to occur at wavelengths of frequencies above 9 GHz. Using the oscillator load pull technique, the molecular resonance will appear as a type of frequency hop as the VCO is tuned though its tuning voltages. The effect will be similar to that of injection locking of an oscillator. This will occur due to the sharp loss of the medium on each side of the center frequency loading the oscillator though the frequency span of the resonance; therefore, the oscillator will stop tuning as the tuning voltage increases or decreases about the point in frequency that the molecular resonance occurs.²¹

Due to the sensitivity of the load pull technique, low frequencies will give good results even for minute changes in pH or bonding position. This provides the ability to make full stream process measurements without sampling, due to the long wavelengths at the present operation frequency

^However, it should be noted that excessive pressure-induced resonance broadening may preclude this. of 200 MHz to 1 GHz. The coaxial section which is presently used will propagate in one mode and therefore does not further complex the data. The molecular resonance quasi-injection lock phenomena will require small cross section measurements (0.5 inch diameter cross-section pipe) to prevent these mode shift problems. Once again, this is due to the small wavelengths involved that will reveal molecular resonance.

The Need for Real Time Monitoring of Organic Reactions

The apparatus and techniques used in organic chemistry differ from those used in the inorganic field. There are at least two general differences which affect the chemical engineering needed:

1) The reactions of organic compounds are characteristically much slower than inorganic reactions. Thus, elevated temperatures and long reaction periods are the rule in organic chemistry, necessitating the use of reflex condensers, autoclaves, stirring devices, and similar equipment not ordinarily required in inorganic synthesis.

2) Inorganic reactions are typically "quantitative," Le. they react completely to produce a single stoichiometric compound. In distinct contrast, such a quantitative reaction is exceptionally rare among organic reactions. Yield of 80% - 90% of the theoretical are regarded excellent, yields of 50% are often acceptable, and frequently the chemical industry must be satisfied with yields of 20% - 30%.°

There are two principal reasons for the nonquantitative nature of organic reactions. First, very few species of organic matter are capable of undergoing only one reaction under a given set of experimental conditions. Side reactions almost invariably occur. The second factor limiting the yields obtained from organic reactions is the reversibility of the reaction.²³ Such

²²The remainder is undesired compound(s), and is either tolerated or separated off by a subsequent procedure (such as solvent extraction or fractional distillation), which adds cost. ^Organic reactions are often driven by relatively small differences in thermodynamic potential, and this causes the reactions to be less irreversible. reversibility places a definite limit on the yield of a product obtainable under any given set of experimental conditions.

Many competing factors can affect yield (total amount produced) and purity (degree of contamination by side reactions). For a chemical refinery, improved yield and purity both translate directly into higher gross income.

Thus, an immense amount of effort has been devoted to optimization of chemical engineering systems to increase yield and purity. In particular, a very large amount of effort has been invested in developing automatic control strategies.²⁴ However, one constraint on control strategies in chemical engineering has been that real time data collection was quite limited (primarily to temperature, pressure, and mass flows), and analyses of chemical composition had to be done off-line.

The disclosed inventions can be used for monitoring of both organic and inorganic reactions. However, due to the extreme sensitivity provided by the invention, and due to the needs just described, it is beheved that the disclosed inventions may be particularly useful for monitoring organic reactions.

"Some patent literature which appears to show the status of process control architecture in various segments of the chemical and related industries, includes 4,844,856 ("Process for automatic regulation of the soluble boron content of the cooling water of a pressurized water nuclear reactor"); 4,744,408 (Temperature control method and apparatus"); 4,713,774 ("Alkylation reactor quality control"); 4,688,726 ("Method and apparatus for controlling a particle refining process"); 4,600,570 ("Continuous controlled process for removing sulphur oxides gases from stack gases"); 4,438,499 ("Fractional distillation process control"); and 4,399,100 ("Automatic process control system and method for curing polymeric materials"). The Problem of "Process Drift"

Throughout the chemical industry, the common practice is to make manual parameter adjustments based on sidestream samples which are analyzed somewhere on site in a laboratory. Although the variety and sophistication of laboratory analytical equipment has improved over the last decade, actual production equipment is still adjusted "on the blind fly". This results in what the industry refers to as "process drift"; that is, with all the conditions the "same", yield and purity fluctuate. There are two primary reasons for this behavior. First, the feed materials may vary in their composition, since they are produced in a similarly ill-controlled fashion. Secondly, the reaction system normally consists of a series of connected chambers or zones in which parameters such as temperature are manually set. If for any reason a chemical system imbalance occurs as a phase change, this disturbance can propagate down the system undetected and without compensation.

Process Control Architectures Within the Chemical Industry

The use of closed loop control systems, in which compositional charac¬ teristics are interactively related to process parameters, is still not widespread in the chemical industry.²⁵ This seems to be due to the following reasons:

1. The industry is mature and capital-intensive, using equipment designed and built a decade or so ago. Operating procedures (as in the petroleum industry) tend to be highly formalized in practice and philosophy. 2. The improved types of equipment which could be used for monitoring are typically expensive, intended for laboratory use, and not easily converted for use in a harsh refinery type of environment. For example, chromatographs have found some usage, but are slow (response

^aOf course, many systems use programmable controllers whose inputs (measured variables) are non-compositional variables, such as temperature, pressure, mass flows, and integrals or derivatives of these. time of 5 to 20 minutes), easily contaminated, and difficult to use with high melting point materials as polymers.

Historically, the economic benefits of process yield improvement were not a primary focus of effort in applied chemical engineering. The chemical industry was founded in the days of cheap oil and cheap energy. Most of the current manufacturing facilities date from that period. The low usage of instrumentation for real time control stems from the attitude that "what we already have does the job". Only during the last few years has attention been directed toward this area. An example of recent activity is an article describing the installation of gas chromatograph for "real time"²* control of a distillation tower. Bozenhardt, "Modern Control Tricks solve distillation problems," Hydrocarbon Processing, June 1988, at 47, which is hereby incorporated by reference. This installation used a $150,000 gas chromatograph, plus about $200,000 for instrumentation and control system. During a twelve month period, this saved the refinery operator $3,000,000 in energy consumption alone, in addition to stabilizing the yield (which otherwise would drift as low as 82%) at 95%. This 13% yield swing represents 39 million pounds of lost product revenue on an annual basis. Prior to this conversion, no economic significance had been assigned to it because this performance was previously beheved unachievable.

^This was not "real time" by electrical engineering standards, since the entire system had a "dead time" of almost 20 minutes. New Measurement and Monitoring Techniques

A further class of embodiments applies a magnetic bias field to the material under test. This hardware capability can be used in several ways. Note that some of these new techniques are applicable for materials characterization, and not only for reaction monitoring.

Several apparatus embodiments for bias field application are disclosed. For example, solenoidal coils around the probe chamber can be used to make the magnetic field parallel to the direction of fluid flow. For another example, an intense bias field can be applied shortly upstream of the probe chamber, so that any paramagnetic molecules will still be in the process of relaxing to random orientation while in the probe chamber.

Skewing the Average Molecular Orientation

One way to use a magnetic bias field is simply as an additional factor in the characterization and m oring techniques described above.

Load-Pull Monitoring Under Applied Magnetic Bias Field

A simple case is where the material being monitored or characterized includes one or more components whose molecules have a net magnetic dipole moment (Le. which do not have zero net electronic spin in the ground state).²⁷ Such components will have some tendency to align themselves with an apphed DC field, and this tendency will affect the degree to which they couple to (and load) a propagating field. In general, if the RF oscillator is operating at a frequency which couples strongly (but not resonantly) to rotational transitions of a significant component of the sample flow, changes in the bias field will affect the frequency of the load- pulled oscillator. Therefore, an apphed bias field can provide yet another technique for distinguishing such species from other species.

Any dipole aligned with a bias field will have an oscillation frequency which increases with field strength, increases with dipole magnitude, and

^or example, this is true of 0₂. This is also true of molecules which have an odd number of electrons, such as KrF, CIO* NO, or N0₂. decreases with angular moment of inertia. In a normal load-pulled oscillator, the RF field is relatively weak in relation to the bias field magnitude.

In the low-density limit (e.g. in a gas at low pressure), an increase in the apphed magnetic bias field will increase the phase velocity througth the transmission line, and therefore increase the frequency of the load pulled oscillator.

Note that the foregoing considerations will apply to any molecule (or radical) which has an asymmetric permeability, regardless of whether the molecule has a net magnetic dipole moment.

If the oscillator is being operated near the frequency of a molecular resonance (in the alternative embodiment mentioned above), then the resonance frequency can be shifted or split by the applied field.²⁸

Again, it should be noted that, in a load-pulled oscillator, the electric field interactions are not strictly separable from the magnetic field interactions.

Excited States and Free Radicals

It should also be noted that, in some cases, it can be desirable to monitor a gas-phase composition which may contain a significant fraction of species in excited states, or even of free radicals. Species in excited states often will have a net magnetic dipole moment, even where the ground state of the species does not. Thus, application of a slowly modulated bias field may be particularly advantageous in such cases, to provide sensitive monitoring of excited states or of free radicals. For example, many types of lasers use a gaseous stream, which is

"pumped" (e.g. by an electrical discharge) so that it contains metastable excited species or reactive molecules. Measurement of the density of such species downstream from the lasing chamber can be used to control the

^See generally CTownes & ASchawlow, MICROWAVE SPECTROSCOPY (1955), which is hereby incorporated by reference in its entirety. apphed pumping energy or (in a pulsed laser) the frequency at which the cavity is Q-switched.

NMR Detection

A further innovative teaching set forth herein is an improved technique for nuclear magnetic resonance characterization. Nuclear magnetic resonance characterization is performed in a load-pulled oscillator configuration like that shown herein.

When a bias magnetic field is applied to a molecule with no net electron spin, the magnetic moments of the nuclei will cause a sharply resonant coupling of nuclear oscillations to an apphed field. This is the phenomenon of nuclear magnetic resonance ("NMR"), which is very well known.²⁹

NMR has been very useful in characterization of biological structures. However, it has been less useful for materials characterization generally. The apphcation of NMR methods has been hmited by the relatively low signal-to-noise ratios detected, as well as by difficulties in applying a high frequency (Le. greater than a few hundred MHz) and difficulties in detection over a wide bandwidth (Le. more than a few kHz).

By using a load-pulled oscillator (rather than direct detection) to detect the result of the apphed bias field, the disclosed invention provides a significant advance in NMR techniques. In this innovative teaching, the techniques for applying the magnetic bias are essentially the same as those previously used for NMR characterization; but the use of a load-pulled oscillator provides a substantial advance over previous detection methods. One way to think about this innovative teaching is that the perturbation field of a conventional NMR apparatus is replaced by the RF magnetic

"See, e.g., R. Feynmann, LECTURES ON PHYSICS (1964) (which is hereby incorporated by reference in its entirety), at Volume 2, Chapter 35; and MAGNETIC RESONANCE AND RELAXATION (Proc. XlVth Colloque Ampere, ed. R. Blinc 1967) (which is hereby incorporated by reference in its entirety). field of the load-pulled oscillator. The magnetic coupling can then be observed as a frequency shift in the free-running oscillation.

The capability for use of an extremely broad frequency range permits greater resolution in spectral determination. In particular, by applying a changing near-DC magnetic bias field, the behavior of load-pulled frequency with bias field can provide yet another element in a unique signature.

BRIEF DESCRIPTION OF THE DRAWING

The present invention will be described with reference to the accompanying drawings, which are incoφorated in the specification hereof by reference, wherein:

Figure 1 is a schematic diagram of a first apparatus embodiment as disclosed in the grandparent application;

Figure 2 is a section view of a portion of the combination coaxial waveguide or transmission line and fluid measurement section of the apparatus of Figure 1;

Figures 3 through 5 are diagrams showing the frequency characteristic versus the tuning circuit voltage of the unisolated oscillator of Figure 1, for its full range of frequencies under loads corresponding to certain concentrations of one hquid such as water in another hquid such as oil; and

Figure 6 is a diagram illustrating the effect of salinity of water in an oil-water mixture when measured by the apparatus of Figure 1.

Figure 7 shows a typical tuning frequency versus voltage plot for a VCO loaded into a shorted transmission hne. Figure 8 shows the physical configuration actually used to derive the data of Figures 11—20.

Figure 9 shows the electrical configuration used, with the physical configuration of Figure 8, to derive the data of Figures 11-23.

Figure 10 shows an example of a large-scale chemical process system according to the present invention.

Figure 11A shows the reaction pathways in reacting methanol with bromine. Figure 11B schematically shows the change in oscillator frequency and insertion loss when bromine is added to a 0.05M solution of aniline in benzene. Figure 11C schematically shows the change in oscillator frequency and insertion loss when bromine is added to a 0.05M solution of aniline in methanol. Figure 11D schematically shows the change in oscillator frequency and insertion loss when bromine is added to a 0.05M solution of aniline in water.

Figure 12A shows the reaction pathways in reacting maleic anhydride with isoprene. Figure 12B schematically shows the change in oscillator frequency when maleic anhydride is added to benzene, and isoprene is added thereafter. Figure 12C schematicaUy shows the change in oscillator frequency with temperature for the reaction products of the reaction of Figure 12B. Figure 12D schematically shows the change in oscillator frequency when isoprene is added to benzene, and maleic anhydride is added thereafter. Figure 13A shows the reaction pathways in reacting maleic anhydride with styrene. Figure 13B schematically shows the change in oscillator frequency when maleic anhydride is added to styrene in a benzene solution. Figure 14A shows the reaction pathways in reacting formic acid with methanol. Figure 14B schematically shows the change in oscillator frequency when formic acid is added to methanol in a benzene solution. Figure 14C schematically shows the reflux condenser arrangement used to reflux the volatile products in the reaction of Figure 14A.

Figure 15A shows the temperature dependence of osciUator frequency and measured power, with deionized water in the system. Figure 15B repeats the measurements of Figure 15A, using salt water instead of deionized water in the system.

Figure 16A shows the temperature dependence of osciUator frequency, with the system loaded with formic acid in benzene. Figure 16B shows the temperature dependence of osciUator frequency for chlorobenzene. Figure 16C shows the temperature dependence of osciUator frequency for cyclohexane. Figure 16D shows the temperature dependence of oscillator frequency, with deionized water in the system. Figure 17A shows the reaction of bromine (Br₂) with phenol. Figure 17B shows the results of monitoring this reaction.

Figure 18A shows the reaction pathways in reacting maleic anhydride with anthracene. Figure 18B schematically shows the change in oscillator frequency when maleic anhydride is added to benzene, and anthracene is added thereafter. Figure 18C schematically shows the change in oscillator frequency when anthracene is added to benzene, and maleic anhydride is added thereafter.

Figure 19 schematically shows the change in oscillator frequency and power level when a sodium hydroxide solution is mixed with an ethyl acetate solution in two stages.

Figure 20A shows the reaction pathways in reacting bromine with isoprene. Figure 20B schematicaUy shows the change in oscillator frequency when bromine is added to isoprene in a methanol solution. Figure 20C schematically shows the change in oscOlator frequency when bromine is added to isoprene in a benzene solution.

Figure 21 schematically shows the change in oscillator frequency over time, in a solution of amoφhous sUica in methanol, when the circulating pump is turned on and off. Figure 22 schematically shows the change in oscillator frequency when the load mixture is provided by a slowly polymerizing polyurethane.

Figure 23 schematically shows the large oscillator frequency difference caused, in an apparatus according to the present invention, by the substitution of used engine lubricating oil, which is near the end of its service lifetime, for new lubricating oU.

Figures 24A-24C schematically show how the varying frequency- dependence of complex permittivity of different substances can be used in on-line analysis.

Figures 25A-25C show different experimental runs, in which the absoφtion of moisture from the air by three different zeohtes was monitored. Figures 26A-26C show two different experimental runs, in which a cycle of absoφtion and desoφtion of ambient moisture by alumina was monitored.

Figure 27A shows experimental results obtained by monitoring the absoφtion of C0₂ by N-methyldiethanolamine (MDEA) in a load-pull ostillator, and Figure 27B shows a system which provides improved

"sweetening" of natural gas, by using the monitoring ability illustrated in

Figure 27A.

Figure 28 shows nondestructive analysis of the fat/protein ratio in a food product, using a load-pulled oscillator.

Figure 29 shows a modified measuring apparatus for monitoring gas- phase mixtures which may contain fractions of low volatility, and Figure 30 shows some actual results derived from test runs on various systems, using the apparatus of Figure 29. Figures 31A, 31B, and 31C show several apparatus modifications in which a magnetic bias field can be applied to the sample being tested.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

The numerous innovative teachings of the present application will be described with particular reference to the presently preferred embodiment, wherein these innovative teachings have been demonstrated in a wide variety of reactions (primarily organic). However, it should be understood that these embodiments provides only a few examples of the many ad¬ vantageous uses of the innovative teachings herein. In general, statements made in the specification of the present application do not necessarily dehmit any of the various claimed inventions. Moreover, some statements may apply to some inventive features but not to others.

Sample System Configuration

A first sample system configuration, as set forth in the grandparent application, will now be described. This system was optimized for monitoring the characteristics of a high-volume fluid flow, namely unrefined petroleum. Other system embodiments wUl be described below. Referring to Figure 1, an apparatus for measuring the concentration of a liquid in a hquid flow stream is illustrated, and is generally designated by the numeral 10. The apparatus 10 is particularly adapted for interconnection with a fluid transmission pipeline 12 for samphng the flow stream through the pipeline or by actually becoming inteφosed as a part of the pipeline. The apparatus 10 includes a fluid flow conducting and measurement section 14 comprising a conventional outer conduit section 16, spaced apart T" sections 18, and conventional weldneck pipe flanges 20. The hquid mixture to be measured for determining the concentration of one medium in the other may be conducted through the conduit 16 on a continuous basis, and the measurement section 14 may comprise part of a fluid transmission pipeline. An elongated center conductor 22 extends through the conduit 16 between opposed support end parts 24 and 26, which will be described in further detail herein in conjunction with Figure 2. The center conductor 22 may comprise a generally cylindrical rod member or tube member and is preferably coaxially arranged in the conduit 16, including the opposed end or T" sections 18. The measurement section 14 can be configured to contain a quantity of fluid or other compositions of matter without continuous or intermittent flow through the measurement section for use of the apparatus in laboratory samphng procedures, for example.

The apparatus measurement section 14 is operably connected to a source of radio frequency or so-called microwave energy comprising an unbuffered or unisolated osciUator, generally designated by the numeral 30. The osciUator 30 includes an active circuit 32 operably connected to a tuning circuit 34 and to an impedance matching network circuit 36. (It has been discovered that a system as shown in Figure 1 can be operated without the impedance-matching network, and this is preferable. In further embodiments, if the measurement section 14 may see a very wide range of dielectric constants, a PIN-diode-switch can be used to switch in circuit elements for impedance matching as needed.) The active circuit 32 is adapted to receive a constant DC voltage, V_α from a source, not shown, by way of a filter circuit 38, and the tuning circuit 34 is adapted to receive a controUable DC voltage, V-^ in the presently preferred embodiment, from another source, not shown, by way of a second filter circuit 40. An unbuffered oscUlator such as the osciUator 30 has an appreciable load puUing characteristic. The fundamental operating frequency of the oscUlator is changed as the complex load is changed on the output circuit of the oscUlator. Depending on the coupling factor of the output circuit the load puUing characteristic can be negligible or substantial. Increasing load pulling factor increases the possibility of so-called spectral breakup (multiple frequency operation) which would render the desired measurement of the osciUator signal impossible. The oscillator 30 may be of a type commerciaUy available, such as from the Watkins-Johnson Company, Scotts VaUey, California, as their Model D-827 voltage controlled osciUator. The exemplary oscillator 30 has a maximum load puUing characteristic of about 35 MHz at a nominal 1.60 GHz operating frequency into all phases of a short circuit at the end of a 50 ohm line stretcher (approximately 0.5 dB return loss). If such a line was of constant loss versus phase, the frequency of the osciUator would return to its original frequency, at any particular phase, every time the reflection co-efficient at that phase recurred with an augmentation of n360 ^• . The osciUator 30 is operably connected to the apparatus measurement section 14 through a suitable connector 44 which is in electrically conductive engagement with the center conductor 22 at the end part 24. At the other end of the load cavity, the center conductor 22 is also electrically connected, through end part 26, second connector 44, and resistance 46, back to the outer conductor 16, as iUustrated. (In the presently preferred version of the system of Figure 1, the resistor 46 is simply replaced by a short circuit. However, various other load elements could be used instead, including real, complex, or frequency-dependent impedances.) The end part 26 is also adapted to interconnect the center conductor 22 with a ten dB directional coupler 48 which is operable to sample the energy transmitted through the coaxial measurement section 14. (Of course, the coupler 48 could also be placed elsewhere in the circuit.) Now consider the electrical behavior of the system of Figure 1 as a varying oil/water mixture flows through the conduit 16. As the percentage of water in this mixture changes, the dielectric constant of the mixture will change. Therefore, the complex impedance characteristics of the measurement section 14 change too. Therefore, the operating frequency of the osciUator 30 wUl also change. The amplitude of the signal seen at mixer 52 wiU also vary as the concentration of water varies. However, the frequency characteristic provides for more accurate measurements.

The coupler 48 is connected to a receiver system which includes a mixer 52 and an isolated oscillator 54 which is tuned to provide a differential output signal. The differential output signal is amplified by amplifier 56, and its frequency is measured by frequency counter 58. The counter 58 is operably connected to a microprocessor 60, which in turn is suitably connected to a display or readout device 62. The mixer 52 may also be of a type commerciaUy avaUable from the Watkins- Johnson Company as their Model WJ-M7B. The amplifier 56 is also available from the abovementioned company as their Model WJ-A38. The frequency counter 58 may be of a type manufactured by Hewlett-Packard as their Model 5342A and the microprocessor 60 may also be of a type manufactured by Hewlett-Packard as their Model 9836. The receiver system described above may also be modified to include a signal amplitude detector, not shown. The system iUustrated in the drawing figures preferably comprises means for compensating for the temperature of the medium being measured in the measurement section 14, including a ther- mocouple 63 inteφosed in the flow path of the medium. The thermocouple 63 is suitably connected to a conversion circuit 65 to provide a suitable digital signal to the microprocessor 60 related to the temperature of the medium being measured.

In this example, the changing dielectric constant of the fluid in measurement section 14 causes the oscUlator 30 to change its operating frequency over a relatively narrow frequency band as compared with the nominal operating frequency of the oscillator. The oscillator 30, in this example, can be pulled from its nominal operating frequency through a range of about 20 MHz by the changing dielectric constant of the medium flowing through the measurement section 14 wherein the percentage of water in oil, for example, varies over a range of approximately zero to two percent of the total fluid volume. The sensitivity of the osciUator 30 to the change in the water content of the oil/water mixture is particularly high due to the operating frequency of the oscUlator since the phase change of the relatively high frequency signal is magnified to some extent by the decreased wavelength at these frequencies and the length of the measurement section 14 is multiple wavelengths. A corresponding increase in sensitivity of the system 10 can also be obtained (for a given starting frequency of oscUlator 30) by increasing the length of the measurement section 14.

By sweeping the oscUlator operating frequency across a frequency span of approximately 400 MHz (by varying the tuning voltage V_τ which is apphed to the varactor in the resonant tuning circuit 34), the sensitivity of the operating frequency for a particular tuned frequency may be determined.

As noted, in the embodiment of Figure 1, local oscillator 54 and mixer 52 provide a differential, relatively low frequency output to frequency counter 58. (However, in the presently preferred version of the system of Figure 1 downconversion is not used, and the frequency counter 58 directly counts the frequency of oscUlation.) The frequency counted by the counter 58 may be compared with frequency data stored in the microprocessor 60 and corresponding to a range of percentages of one medium in another such as water in oU. The value thus found is then suitably converted to drive a display 62, which thus displays the amount of or concentration of one medium in the other. The frequency counter 58 may include suitable analog to digital conversion devices, not shown.

As noted, the osciUator 30 has only a limited range of steady frequency deviation. If the load characteristics steadily change enough to pull the oscillator 30 beyond its limited range, the oscUlation frequency will suddenly change discontinuously, or "hop." Accordingly, measurement can be made over a broader range, by making an additional measurement to determine which range the system is operating in. For example, in combination with a system 10 as shown in Figure 1, a crude measurement can be made to ascertain whether the system is perceiving a concentration of a medium such as water in oU in the range of say zero to two percent or in a range of two percent to four percent (wherein each two percent change corresponds to the full frequency range of operation of the oscillator).

Referring now to Figure 3, there is illustrated a diagram showing the variation in the output signal frequency of the oscillator 30 over its maximum tunable frequency range when tuned by the tuning circuit 34 when the circuit is terminated into its characteristic impedance. A voltage controUed oscUlator such as the oscillator 30, when swept across its maximum range as determined by changing the tuning voltage V_τ, will exhibit a characteristic indicated by the line 70 for a perfect or balanced load. If the dielectric constant of the composition present between the conductors 16 and 22 changes (e.g. as a result of a change in the concentration of one medium, such as water, in another medium, such as oU, over concentrations in the range of zero to two percent), the oscillator 30 wiU exhibit a frequency output signal as shown in Figure 4. A curve 71 having discontinuities 72, 74, and 76, will be exhibited as the oscillator 30 is swept across its maximum frequency range. Accordingly, as the oscillator is swept across its maximum frequency range (indicated as f_t to f₂), the number of discontinuities may be counted to determine what range of change in concentration of water in oU, for example, is being measured. For example, as shown in Figure 5, a curve 73 having discontinuities indicated by the shifts 78, 80, 82, 84, 86, 88, and so on, would indicate that the osciUator 30 was measuring a change in frequency for a concentration of water in oil of say two percent to four percent. Therefore, the number of discontinuities measured per sweep of operating frequencies from i_ to f₂ can indicate what range of variation in dielectric constant is being measured which correlates with the range of concentration of one medium such as water in the other medium such as oil. Accordingly, by using an unisolated or unbuffered voltage controlled oscUlator in a circuit such as described herein, an operating frequency at a particular control voltage may indicate the concentration of water in oil, for example, if after sweeping the oscUlator across its frequency range, the number of 360" phase shifts counted are determined to determine the particular range of change of dielectric constant being experienced.

Referring now to Figure 6, there is iUustrated a diagram indicating the relationship between the oscillator signal frequency and amplitude and the effects of the salinity of a medium being measured, such as an oil-water mixture. The measurement of signal amplitude at several frequencies and a knowledge of the effect of salinity on the intercept of the frequency characteristic as a function of amplitude can correct for salinity effects on the overaU impedance seen by the oscillator 30. For example, a salt-free fluid with a particular percentage of water in oil will exhibit a signal characteristic according to the curve 91 in Figure 6, whereas the same percentage of water in a water-oil mixture with, for example, y molar percent of sodium chloride would exhibit a characteristic according to the curve 93 in Figure 6. Thus, by sweeping the frequency of the oscillator 30 across a range of frequencies, the salinity, as well as the percentage of water, can be measured.

Referring now to Figure 2, there is illustrated a sample arrangement of supporting the center conductor 22 within the measurement section 14 and terminating the center conductor at the conventional N type RF connector 44.³⁰ (The arrangements for terminating the conductor 22 at the two end parts 24 and 26 are essentially identical. Each of the conduit 'T' sections 18 is suitably welded to a conventional weldneck flange 100, as iUustrated by way of example in Figure 2, which in turn is secured to a flange 102 by conventional bolt and nut assemblies 104. The flange 102 is secured to a somewhat conical shaped reducer section 106. The internal

'"In the invention as presently practiced, the arrangement of Figure 2 has now been considerably simplified. O-rings are now included in piece 110, and epoxy cementing is not needed. space formed within the T" section 18 and the weldneck flange 100 is occupied by a generally cylindrical block 110 formed of a suitable insulating material such as a fluorocarbon plastic.

The center conductor 22 includes a generally cylindrical rod-like section 23 which is suitably supported in the block 110 and is in conductive relationship with a somewhat frustoconical conductor section 25 supported in a second support block 112 formed of an electrical insulating material. The conductor section 25 is secured to a third conductor section 114 by a conductive pin member 115. The conductor section 114 also has a somewhat frustoconical or tapered portion for reducing the diameter of the center conductor down to a portion 116 which is secured to a pin 118. The pin 118 comprises the center conductor for the connector 44. The conical tapered conductor sections 25 and 114 also prevent unwanted reflections of the signal being transmitted through the measurement section 14. Suitable insulating bushings or spacers 120 and 122 are adapted to support the conductor sections 25, 114, 116 and 118. A suitable insulating material and sealing, such as epoxy, may be injected to fill the cavity formed between the blocks 110 and 112, as indicated at 113, to prevent leakage of fluid from the interior of the conduit section 16 to the opposite ends of the measurement section 14. Thanks to the configuration of the end parts 24 and 26, there is little or no space provided which would create a void of nonflowing fluid within the measurement section 14 which might introduce errors into the determination of the concentration of one fluid in another being pumped through the measurement section. The 'T' sections 18 might be replaced by conduit portions which would introduce flow into the conduit section 16 with a more gradual change of direction to minimize turbulence which could possibly affect the frequency readings being measured by the circuit described herein.

Acquiring Data from Chemical Reaction in Progress The sample system used for the successful experiments summarized in

Figures 11-23 wiU now be described in detail. Physical Configuration and Fluid Flows

Figure 8 shows the physical configuration actually used to derive the data of Figures 11-23.

A reaction flask 810 is fed by an addition funnel 812, and is also connected to a reflux condenser 814. (The reflux condenser 814 is water- cooled, and helps to prevent the loss of volatile fractions from the system.) The temperature of the reaction flask is stabilized by a heating mantle 811, which preferably is actively heated and has a large thermal mass. The heating mantle 811 is normally controlled to maintain a constant temperature in the reaction flask 810.

A circulation pump 820 pumps liquid out of the flask 810 (through tubing 816), into measurement section 800 (through tubing 817), and back into the reaction flask 810 (through tubing 818).

Thus, the composition of the material in the measurement section 800 will correspond to the composition of the material in the flask 810. To preserve uniform temperature, a heating tape 802 is attached to the measurement section 800, and is controlled in accordance with the output of thermocouple 819 to keep the temperature of the fluid approximately uniform throughout the system. The measurement section 800 is physically shaped as a cylindrical cavity with an insulated probe rod along the axis of the cylinder. This is electrically connected to an oscillator network 830, as will now be described.

Electrical Configuration Figure 9 shows the osciUator configuration used, with the physical configuration of Figure 8, to derive the data of Figures 11-23.

Note that this configuration has some differences from the configuration of Figure 1. The load seen at line RFOUT (presented by the measurement section 800) is connected to the collector of driver transistor 910, while the tank circuit 34 is connected into the emitter-base coupling of driver transistor 910. The directional coupler 48 is now a dual directional coupler which is connected directly to the line RFOUT, instead of being separated by the length of the measurement section 14, as in the embodiment of Figure 1.

Note that a small series resistor 912 is used in the RFOUT hne. (In the presently preferred embodiment, the value of this component is 9Ω.) This resistor helps to prevent spectral breakup (by suppressing osciUation at secondary frequencies).

A shunt resistor 914 is also attached to the RFOUT line. This resistor also adds to stability, by fixing a maximum magnitude for the impedance seen at line RFOUT. (In the presently preferred embodiment, the value of this component is 562Ω.)

These two resistors wiU reduce the magnitude of the frequency hops seen, as discussed above.

The directional coupler preferably diverts only 1% of the reflected power, so that the load is still coupled closely enough to be able to pull the oscUlator. The corresponding output from coupler 48 is connected to a frequency counter and control logic, as described above. Also, the two outputs from the directional coupler are used to measure inserted power and reflected power.

Experimental Data from Reaction Monitoring Figures 11-23 show the results of a number of experiments which have demonstrated the ability of the disclosed system to monitor the progress of a wide variety of chemical reactions. In these experimental runs, the data was gathered with a system substantially as shown in Figure 8.

In the measurement system used, the frequency was read out to a resolution of 100 Hz. When the system pump is operating, the 100-Hz digit of frequency measurement displays some rapid fluctuation, due to bubbles in the system, but the 1000-Hz digit of the frequency measurement is stable.³¹ The insertion loss measurements are read out to a resolution of 0.01 dB. Again, some fluctuation was seen in the 0.01 dB digit, but the 0.1

³¹With pump off, the frequency readout is stable down to about 10

Hertz. dB digit is quite stable. Thus, in the following results, frequency measurements are reported to a resolution of only 1000 Hz, and the insertion loss measurements are reported to a resolution of 0.1 dB.

In the system used, the volume of the measurement section was 0.5/ (of a total volume of 1.5/), and the pump flow was 4 /min. Thus, the time delay to replace the volume of the measurement cavity is 0.5/4 min = 7.5 sec. This physical time constant hmits the time-domain resolution of all measurements given (except for pressure-dependent behavior, as in run 19 below). Note that significant information can be seen on a much smaller time scale, but such information may be regarded as an average over a time window of about 7.5 seconds. Note that the electrical time-domain resolution hmits are of the order of 1/f, Le. roughly a few nanoseconds.

Table of Permittivities To assist those skUled in the art in inteφreting and extrapolating from the foUowing results, the following table gives DC permittivity values e_r for several of the substances described below. The permittivities at UHF and microwave wiU be somewhat different from the DC values, but the DC values do show the low-frequency component of permittivity. Unless- otherwise specified, the following values are for the pure substance, in liquid or solid form, at room temperature and atmospheric pressure.

Figure 11A shows the reaction pathways for the reaction of liquid bromine (Br₂) with aniline ( H_«NH₂). This is an example of an addition reaction.

The foUowing data shows the behavior of this reaction in water (which is a highly polar solvent), in methanol (which is slightly less polar), and in benzene (which is nonpolar). In aqueous solution, as shown in Figure 11D, the reaction went to completion very rapidly. Note that the frequency dropped very shaφly as the hquid bromine was added. At the end of the reaction, it was found that adding more water to the solution did not shift the frequency significantly. (This provides a further technique for detecting completion of the reaction.)

Note that the insertion loss decreased shaφly as the bromine was added (as shown by the increase of the measured power level).

A detailed listing of the data points which are summarized in the curve of Figure 11D is included in the Appendix below.

Some of the measured parameters for this run are here summarized in tabular form: Frequency: Insertion Loss:

At start: 1146.466 MHz -7.19 dB

After addition of anihne: 1146.416 MHz -7.17 dB

After addition of bromine: 1145.908 MHz -6.95 dB

After equilibration: 1145.908 MHz

2. Aniline + Bromine (in Methanol)

In methanol, as shown in Figure 11C, the reaction again proceeded rapidly to completion.

A detailed listing of the data points which are summarized in the curve of Figure 11D is included in the Appendix below. Some of the measured parameters for this run are here summarized in tabular form:

Frequency: Insertion Loss:

At start: 1141.133 MHz -8.44 dB

After addition of aniline: 1141.157 MHz -8.44 dB After addition of bromine: 1140.497 MHz -8.37 dB

After equilibration: 1140.497 MHz -8.37 dB

3. Aniline + Bromine (in Benzene)

In benzene, this reaction is much slower, and does not produce a sedimented precipitate. In the test run, this reaction was performed with very dUute concentrations, at a temperature of 75-78 " C. Some of the measured parameters for this run are here summarized in tabular form:

Frequency: Insertion Loss:

At start (Benzene): 1103.4 MHz -4.47 dB

After addition of Aniline: 1102.7 MHz -4.49 dB

After addition of Br₂: -4.54 dB

After equilibration: (hours) 1100.8 MHz -4.90 dB

4. Maleic Anhydride + Isoprene (Example 1)

This reaction is an example of the important class of Diels-Alder reactions. Such reactions are very widely used. Diels-Alder reactions are also analyticaUy convenient, since they are highly specific to diene compounds which have two double bonds separated by exactly one saturated bond.

Maleic anhydride is a 1,3 diene (formally l,3-diene-3-methyl butane). Isoprene is a commonly used feedstock for making synthetic rubber.

Two different sets of measurements were taken of this reaction system.

The first run, as shown in Figure 12B, was performed at 100 °F at an initial frequency of 410 MHz. The first measurement was taken with 1.5/ of pure benzene in the system, and the frequency dropped shaφly as 1 MW (1 molecular weight, Le. a number of grams equal to the atomic weight of the substance) of maleic anhydride was added. (Maleic anhydride has a very large dielectric constant.) One MW of pure isoprene was then added. The resulting curve shows a shaφ small rise in frequency as the isoprene is added, and then shows a shaφ large drop as the reaction takes place. (The measured frequency also showed a more gradual subsequent drop, not shown in Figure 12B. This is probably due to the depletion of volatUe components over the course of the run.)

At start (Benzene): 406 MHz -0.3 dB

After addition of Maleic Anhydride:390 MHz -1.1 dB

After addition of Isoprene: 391.5 MHz -1.05 dB After equUibration: 386.9 MHz Total Shift during reaction: Δ = 19 MHz

5. Maleic Anhydride + Isoprene (Example 2)

A second series of data runs studied whether any effect could be seen by reversing the order of mixing. In this run, as shown in Figure 12D, the starting frequency was again

410 MHz in pure benzene. The reaction temperature was set at 100 ° F.

Isoprene was added first, in quantity sufficient to make the system concentration 1 molar (1M). This produced a small rise in frequency.

Some of the measured parameters for this run are here summarized in tabular form:

Frequency: Insertion Loss:

At start (Benzene): 406.5 MHz -0.3 dB

After addition of Isoprene: 406.5 MHz -0.3 dB

After addition of Maleic Anhydride: After equUibration: 387.2 MHz -1.5 dB

6. Temperature Dependence of Methylated Phthalic Anhydride

Using the reaction product of isoprene + maleic anhydride (which is primarily 4-Methyl-l,2,3,6-tetrahydrophthalic anhydride), the temperature dependence of the oscUlator frequency was studied. Studies of single- component systems help to show how the effects of temperature- and frequency-dependence can be factored out from measurements made using the disclosed innovative teachings. In some apphcations, it may also be advantageous to perform direct measurement of the conditions in a single- component system. A number of such studies have now been done. After the isoprene/maleic anhydride reaction of Figure 12D had gone to completion, a temperature cycle was performed to observe the temperature dependence of the osciUation frequency with the reaction products in the system. As shown in Figure 12C, the relation of the frequency to temperature was fairly linear, at about 50 kilohertz of shift per degree fahrenheit, over a fairly wide range. The behavior of frequency over temperature appeared to show a tail at low temperatures, Le. the frequency became more nearly constant at the lowest temperatures, rather than foUowing the linear relation.

7. Styrene + Maleic Anhydride

Figure 13A schematically shows the reaction of styrene with maleic anhydride. The experimental results of monitoring this reaction are shown in Figure 13B.

Note that, after the shaφ jump when maleic anhydride is added, the frequency continues to change fairly rapidly over time as the reaction progresses. (That is, the total frequency change over the progress of the reaction is large, and therefore the disclosed system can track the reaction's progress with high resolution.)

This reaction was conducted at 150 ^βF, with 0.5 MW of styrene and 0.5 MW of maleic anhydride. Some of the measured parameters for this run are here summarized in tabular form: Frequency: Insertion Loss:

At start (Benzene): 398.6 MHz -0.5 dB

After addition of Styrene: 398.6 MHz -0.5 dB

After addition of Maleic Anhydridβ94.1 MHz -0.75 dB

After equUibration: 392.9 MHz -0.75 dB Total Shift during reaction: Δ = 1.2 MHz

8. Methanol + Formic Acid (Esterification)

Methanol reacts with formic acid to form methyl formate and water. (This reaction is shown schematically in Figure 14A.) This is a simple example of an esterification reaction. Results from monitoring this reaction are shown in Figure 14B. This experimental run also demonstrates several methodological alternatives. The system was initially charged with approximately a 0.5 molar concentration of methanol in benzene. After the mixture stabilized, the system was brought up to the reaction temperature (140 °F in this case). Next, formic acid was added in sufficient quantity to make up a 0.5M solution. A very shaφ frequency shift resulted. Since methyl formate is fairly volatile (boUing point 34 ' C), a reflux condenser was used, as shown in Figure 14C, to retain the product. (The methyl formate product was held in vapor/condensate system, in the condenser.)

9. Temperature Dependence of Deionized Water

The frequency dependence on temperature was also tracked for a system which includes only deionized water. In this case, very odd behavior was exhibited: the measured frequency showed shaφ and repeatable dependence on temperature, including a shaφly temperature-dependent peak. This curve is seen in Figure 15A At the peak slope of this curve, the temperature-dependence of frequency is about 500 kHz per degree Fahrenheit. Note that the insertion loss curve (the lower curve in this Figure) also shows a shaφ shift at a temperature of about 127 " F.

10. Temperature Dependence of Saline Solution For a comparison run, frequency over temperature was also observed for dUute saline (at a concentration of about 1 gram of NaCl in 1.5 liters of water).³² In this case the frequency dependence is much flatter, as may be seen in Figure 15B. At the peak slope of this curve, the temperature- dependence of frequency is only about 25 kHz per degree Fahrenheit.

11. Temperature Dependence of Cvclohexane

The temperature dependence of pure cyclohexane has also been tracked. As shown in Figure 16C, this temperature-dependence was found

³²This is a quite dilute saline, with a weight percentage of about 0.07%. For comparison, the weight concentration of salts in sea water is about 3%. to be very linear, at a frequency of about 1142 MHz, with a slope of about 44.6 kHz per degree Fahrenheit.

12. Temperature Dependence of Formic Acid/Benzene

The temperature dependence of a formic acid solution (10 ml of formic acid in 1.5/ of benzene) has also been tracked. As shown in Figure 16A, this temperature-dependence was found to be very linear, at a frequency of about 1103 MHz, with a slope of about 50 kHz per degree Fahrenheit.

13. Temperature Dependence of Chlorobenzene

The temperature dependence of chlorobenzene has also been tracked. As shown in Figure 16A, this temperature-dependence was found to be very linear, at a frequency of about 1103 MHz, with a slope of about 50 kHz per degree Fahrenheit.

14. Phenol + Bromine (Substitution)

Figure 17A shows the reaction of bromine (Br₂) with phenol. This reaction is a convenient example of a substitution reaction.

Figure 17B shows the results of monitoring this reaction. Note that the frequency rises after mixing, and then gradually declines.

This reaction iUustrates several important methodological challenges:

1) The reaction is exothermic, so the temperature must be carefuUy monitored, to avoid spurious measurement due to temperature- dependence.

2) The net physical density of the reaction mixture changes steadUy as the reaction progresses.

3) One of the reaction products (at standard temperature and pressure) is a gas, which evolves wlule the reaction is in progress. The formation of gas bubbles in the solution, and the escape of those bubbles from the solution, wiU affect the electrical measurements.

At start (Benzene): 1100.5 MHz -4.4 dB

After addition of bromine: 1100.0 MHz -5.4 dB

After addition of Phenol: After equilibration: 1094.6 MHz -6.5 dB

15. Maleic Anhydride + Anthracene

Figure 18A schematically shows the reaction of Anthracene with maleic anhydride. This is a further example of a Diels-Alder reaction. It should be noted that this reaction is mildly exothermic. The two sets of experimental data summarized in Figures 18B and 18C show results of monitoring this reaction.

Figure 18B shows a run where maleic anhydride was added first, and anthracene second, in a benzene solvent. (Both reagents were added in 0.5

MW quantity.) The reaction temperature was 150* F. After a small rise when the anthracene is mixed in, the measured frequency shows a long increase, which represents the progress of the reaction.

Frequency: Insertion Loss: At start (Benzene): 398.2 MHz -0.2 dB

After addition of maleic anhydride392.8 MHz -0.3 dB

After addition of anthracene:

After equilibration: 397.15 MHz -0.1 dB

16. Anthracene + Maleic Anhydride Figure 18C shows a run where anthracene was added first, and maleic anhydride second, in a benzene solvent. (Both reagents were added in 0.5 MW quantity.) The reaction temperature was 150 *F. After a small rise when the anthracene is mixed in, and a shaφ drop when the maleic anhydride is mixed in, the measured frequency shows a long increase, which represents the progress of the reaction. Some of the measured parameters for this run are here summarized in tabular form:

Frequency:

At start (Benzene): 398.4 MHz After addition of Anthracene: 398.6 MHz

After addition of maleic anhydride393.0 MHz

After equUibration (48 hours): 397.6 MHz

17. Saponifϊcation of Ethyl Acetate

The reaction of ethyl acetate with sodium hydroxide, in aqueous solution, yields ethanol plus sodium acetate:

+ Na⁺ + OH^" →„ C_U_;OU + Na⁺ + CH₃COO^".

In an experimental demonstration of monitoring this reaction, the starting charge was 500 mi of .02M ethyl acetate, further diluted with 250 mi of water. Next, 500 ml of .02M aqueous NaOH was added. This reaction was conducted at a temperature in the range of 25-30 °C, and produced a frequency and power shift as shown.

In a further state of reaction, another 250 ml of ethyl acetate solution and another 250m/ of NaOH were again added to the reaction mixture. This produced a stiU further frequency shift, as shown. Figure 19 shows how frequency and power shifted, when this reaction was monitored using the disclosed innovations.

This reaction is conventionally used in chemistry instruction to show the use of conductivity measurement for reaction tracking. Since the hydroxyl ion OH^" dominates the conductivity of the solution, the depletion of OH^~ wiU produce a strong swing in the conductivity.

18. Isoprene + Bromine (in Methanol)

Figure 20A schematically shows the reaction of bromine with isoprene.

Figures 20B and 20C schematicaUy show two sets of experimental data which show monitoring of this reaction, using the disclosed innovations. A significant methodological point here is that isoprene boils at only

35 "C, and thus may readily flash off from the reaction mixture. Thus, in these experimental runs, more isoprene was added after the reaction has apparently gone to completion, as a check for completion.

Figure 20B shows the reaction in a polar solvent (methanol), at 74 ^β F. Note that the reaction goes to completion rapidly. Some of the measured parameters for this run are here summarized in tabular form:

Frequency: Insertion Loss:

At start (Methanol): 1141.2 MHz -8.1 dB

After addition of Isoprene: After addition of bromine:

After equUibration: 1140.2 MHz -7.8 dB

Add excess isoprene: 1140.2 MHz -8.1 dB

19. Isoprene + Bromine (in Benzene)

Figure 20C shows the reaction in a nonpolar solvent (methanol). Note that the reaction goes to completion more slowly than the reaction of

Figure 20B. Some of the measured parameters for this run are here summarized in tabular form:

Frequency: Insertion Loss:

At start (Benzene): 1108.1 MHz -4.4 dB After addition of Isoprene: 1108.2 MHz -4.4 dB

After addition of Bromine: 1107.5 MHz -4.4 dB

After reaction: 1106.9 MHz -4.4 dB

Add further 2g Isoprene: 1105.4 MHz

Add further 5g Isoprene: 1105.6 MHz Add further 2g Isoprene: 1105.7 MHz

Add further 25g Isoprene: 1107.3 MHz 20. Viscosity Dependence of α-Silica/MethanoI Mixture

The experimental data summarized in Figure 21 shows a different use of the disclosed innovations. This experiment measured fluid viscosity in situ. In order to dynamically modify viscosity, a thixotropic liquid was used. Thus, by switching the system pump on and off, the viscosity could be changed (by changing the forces on the thixotropic hquid).

Thus, this embodiment of the invention is not hmited to thixotropic or antithixotropic compositions, but can be used to monitor viscosity in situ in a wide variety of liquid compositions. In this experiment, 50 g of finely divided amoφhous silica (having a surface area of approximately 300 m²/g) was mixed into a liter of methanol. This produces a thixotropic liquid, whose viscosity is highly strain- dependent.

As shown in Figure 21, the experimental run showed that the oscillator frequency was highly dependent on the instantaneous viscosity of the mixture. The use of a thixotropic hquid makes it particularly easy to directly measure dependence on viscosity, since viscosity can be changed, by changing physical forces apphed (at the pump), with at most minimal change to other physical and chemical parameters. The data showed a frequency shift of 700 kHz in the oscUlator, depending on whether the pump was switched on or off. Note that a characteristic relaxation time of about 100 msec was seen when the pump was switched off, but the frequency rose much more shaφly when the pump was switched on. This viscosity dependence provides another example of the ability to measure short-range organization. This experiment suggests, for example, that comparable techniques might be very useful in monitoring other types of physical/chemical reactions: for example, it may be useful to detect the adhesion/cohesion changes which would indicate that the binder in a composite material has "set up".³³

detecting setup and cure times of composite materials is a very important manufacturing need in the use of the use of composite materials (such as boron fiber plus phenolic resin) for medium- or large structural components, such as aerodynamic surfaces of aircraft. 21. Formation of Low-Density Polvurethane (Polymerization)

Figure 20A schematically shows the reaction of a diisocyanate (primarily toluene diisocyanate in this example) with a polyol (a molecule containing multiple available -OH groups). Both of the isocyanate (— N=G=0) groups provide active sites which can react with a hydroxyl (- -OH) group. The matrix of bonding from such reactions creates a macromolecule, whose mechanical properties wUl depend on the molecular weight and degree of cross-linking of the bonding matrix.

This reaction produces a polyurethane polymer, and is one very simple example of the many implementations of this important class of processes.

In the experimental run demonstrated, the components used were parts

A and B of polyurethane mold compound PMC-744 from Smooth-On, Inc.,

1000 VaUey Road, GUlette NJ 07933. This compound is specified as having a pot hfe of 15 minutes, a gel time of 30 minutes, a demold time of 16 hours, and a fuU cure time of 7 days at 25 ^• C.

The disclosed innovations can be used with a very wide variety of other polymerization reactions. Two which are contemplated as particularly advantageous and predictable apphcations are in the formation of a polysulfide polymer, and in sUicone polymerization reactions.

22. Monitoring Lifetime of Lubricating Oil

To demonstrate a further class of embodiments, the system was loaded, in two tests, with new and with used engine oU. (The oil was Exxon brand diesel motor oil, API grade CC, viscosity 15W-40.) The used oil had been removed from a diesel truck engine after 100 hours of operation, Le. at the end of the oU's useful hfe.

The measured characteristics of the two oU samples were markedly different. Both were measured over a temperature range of 50 ^β C to 80 * C. The oscUlator frequency shift, between the old and new oil, was approximately 4.7 MHz. In relation to the resolution of the disclosed system, this is a huge shift. By simple linear inteφolation, this measuring technique would be estimated to have a resolution of approximately 0.02% of the oil's useful life. Even allowing a large margin of error, this is very high accuracy.

This technique can be used to monitor the aging of lubricating oil in service. Thus, oU change intervals can be reduced if needed, when unusually harsh conditions indicate that this is required. This also permits oil management to be performed far more precisely in vehicle fleets. This technique can be particularly advantageous in aviation.

Mechanical faUures wUl very rapidly change the characteristics of the oil, and this can be detected by a monitor, constructed according to the above teachings, which is permanently instaUed in the aircraft.

23. Monitoring Hydration of Zeolites Many chemical processes involve dehydration steps. Zeolites and activated alumina (A1₂0₃) are the primary species used. Zeolites are extremely useful, because their regular crystal structure includes many small apertures whose dimensions are known exactly. Because of this structure, zeohtes are often referred to as "molecular sieves." Water absoφtion by zeolites tends to be highly energetically favorable.

Water absoφtion by alumina is not as energetically favorable, but alumina can absorb more water, and it is easier to strip the absorbed water from the alumina. However, alumina dessicants have the disadvantage that they wiU tend to crystallize after repeated thermal cycling. An unloaded zeolite wiU typically have a rather low dielectric constant

(e.g. e«2). Thus, a very large change in the electrical properties can be produced by a very smaU change in the occupation of sites in the zeolite. Thus, one way of thinking about the operation of a zeolite, in a process as described herein, is that the zeolite provides amplification for changes in dielectric constant.

For example, a zeolite can be operated as a gas-phase humidity sensor. Water vapor wiU segregate into the zeolite (with very high efficiency), and this means that humidity changes - even within a range of very low humidity — can readily be detected by tracking a zeolite-loaded osciUator. Figure 25A shows the measured frequency shift as a zeolite column was allowed to absorb moisture from the air. In this case, a column packed with zeolite grains was electricaUy configured to load the oscillator. The zeolite used was a composition referred to as "3A," which is commercially avaUable from Union Carbide. Note that, as the zeolite absorbed water, the oscillator frequency shifted from about 938 MHz to about 907 MHz. Figure 25B shows another experimental run, demonstrating the measured frequency shift as a zeolite column was allowed to absorb moisture from the air. In this case, the zeolite used was a composition referred to as "4A," which is commercially available from Union Carbide. Note that, as the zeohte absorbed water, the oscillator frequency shifted from about 305 MHz to about 293.5 MHz. Figure 25C shows yet another experimental run, demonstrating the measured frequency shift as a zeolite column was allowed to absorb moisture from the air. In this case, the zeolite used was a composition referred to as "RK29," which is commercially available from Linde. Note that, as the zeohte absorbed water, the oscillator frequency shifted from about 322.44 MHz to about 310.99 MHz.

It should be noted that zeolites have many uses beyond dehydration.

They are often used to selectively take up a variety of other small molecules (such as COS or Hβ). Thus, the procedures just described can also be applied to processes where zeolites are used to scavenge other molecules (especially polar molecules).

24. Monitoring Hvdration of Alumina

Figure 26A shows a sample experimental run, where a column packed with alumina drying peUets was electrically configured to load the oscUlator. Initially, with fresh dry pellets, the oscillator frequency was 305.78 MHz. Air was then blown through the column, and, as the pellets absorbed water from the air, the oscillator frequency dropped, over a period of about 100 minutes, to about 297.82 MHz. The flow of room air was then cut off, and the column was purged with dry nitrogen. The osciUator frequency then rose to a frequency of about 302.6 MHz (at which time the flow of dry N₂ was terminated). Note that the oscillator frequency then rose to a frequency of about 302.6 MHz (at which time the flow of dry N₂ was terminated). Note that the osciUator frequency remained stable (verifying the integrity of the experimental observations).

Figures 26B and 26C show another sample experimental run, performed at room temperature, wherein a column packed with alumina drying pellets was monitored during a cycle of absorption and desoφtion of water vapor.

InitiaUy, with dry pellets, the oscillator frequency was about 307 MHz, as shown in Figure 26B. Air was then blown through the column, and, as the peUets absorbed water from the air, the osciUator frequency dropped to about 295.0 MHz. The flow of room air was then cut off, and the column was purged with dry nitrogen. The osciUator frequency then rose, as indicated in Figure 26C, to a frequency of about 305.2 MHz (at which time the flow of dry N₂ was terminated).

25. Controlling a Gas-Sweetening Process

Figure 27B shows a system which is designed for "sweetening" natural gas. Many natural gas weUs provide an output which includes a significant fraction of Hβ (which is an inconvenient contaminant in combustion, and is also extremely toxic). Such gas is known as "sour" gas. Other undesired gaseous contaminants, such as CO^ may also be present.

NormaUy reactions with complex amines are used to "sweeten" such gas, Le. to reduce the hydrogen sulfide content. An amine solvent is used to absorb the Hβ and C0₂ from the feedstock. The solvent is recycled by running it through a stripper vessel, at elevated heat and pressure, to remove the Hβ and C0₂. A sample solvent used in such systems is N- methyldiethanolamine, but a variety of other solvents are weU known for use in such systems.

Both Hβ and C0₂ are "acid gases." The various complex amines are bases, and react exothermally with H^ or CO,. The complexed acid gases are stripped in a regeneration tower, and the "lean solvent" returned to the absorbing tower. The object of these gas treating units is to lower the acid gas content, yet achieve minimum energy consumption in pumping and regeneration. In conventional sweetening process plants of the type described, the solvent is usuaUy underloaded and overcleared.

That is, it is recognized as desirable to load the solvent to no more than a certain maximum level (e.g. 75%), to retain some reserve capacity to accommodate feedstock surges. To assure this, conventional processes will commonly maintain a typical loading level of 50%, or even 40%. This is very wasteful of process heat.

Similarly, the solvent is typically held in the stripper unit longer than needed, or under more extreme conditions than needed. This too wastes process heat. Moreover, extra time at high temperature will accelerate the degradation of the solvent.

Among the innovative teachings disclosed herein is a sweetening process in which the loading and degradation of the amine reagent is monitored on-line. Thus cycle times, and the frequency with which additional amine reagent is added, can be precisely controlled. Thus, the disclosed system and method can save process heating cost, and can save on the costs of solvent maintenance.

Figure 27A shows experimental results obtained by monitoring the absoφtion of C0₂ by N-methyldiethanolamine (MDEA) in a load-pull osciUator. Note that the time scale shifts, at about data point 600, from 1 second per data point to 15 seconds per data point. The starting frequency was 315.790 MHz, and the measured end frequency was 310.238 MHz. (This implies a net shift of 5.552 MHz.)

Other commonly used amines include MEA (monoethanolamine) and DEA (diethanolamine). Various blends are formulated to achieve Hβ and C0₂ selectivity or variations thereof.

Two such load-puU oscUlator monitoring stations can be used in the system shown in Figure 27B. At the monitoring point marked "#1", the flow is of lean solvent. Monitoring at this point can be used to detect long- term degradation of the solvent charge: when the oscUlator monitoring the solvent reaches a certain trigger value, the solvent charge is pulled and purified.

Some of the benefits of the disclosed system and method could be obtained by monitoring only at point 1, or only at point 2; but it is preferable to monitor at both points.

A somewhat similar process is used for drying natural gas, except that a glycol is normal used as the solvent. The system architecture of Figure

27B is directly applicable to such systems as well. In fact, the drying of ethylene glycol has been successfully monitored using a load-pull oscillator.

26. Monitoring Content of Food Products

Another important class of apphcations is on-line monitoring of food products in a clean process.

Figure 28 shows results of sweeping the osciUator frequency across a wide range, whUe loading the cavity with three different compositions of food product, having three different fat/protein ratios.

To obtain a high-protein sample, a lean cut of beef, bought from the grocery store, was finely ground. To obtain a high-fat sample, pure beef fat was ground. The two compositions were then mechanically mixed. It was experimentally determined that the electrical characteristics of the mixture were within 3% of that which would have been calculated by straight proportional extrapolation. This does not necessarily reflect inaccuracy in the electrical measurement; it may simply reflect factors such as the presence of air in the mixture.

27. Real-Time Monitoring of Dilute Gas-Phase Mixtures

An example of a shghtly difficult gas-phase monitoring problem is monitoring humidity. For a sample load-pulled oscillator measuring instrument, which has a measuring chamber which is 1 to 3 feet long and operates from 250 MHz to 1,000 MHz, it is not possible to directly detect H₂0 in air. (Higher frequencies and longer probes theoretically would aUow this, although it would involve a more expensive system.) However, the experimental data reported above provide a simple way to achieve this goal: a zeolite is used to provide segregation of the species which is desired to be monitored, and a load-pull osciUator monitors the changing characteristics of the zeolite. The equilibrium values of absorbed H₂0 at different water partial pressure at particular temperatures may be found, from handbook data, to be: at 32°F ambient, as H₂0 partial pressure changes from 10^"* → 10^'3, the weight of H₂0 in 100 pounds of zeolite will go from 1.1 to 10.6 pounds, or from about 1% to about 10% by weight. Note that the response time of the zeolite to gas-phase compositional changes would be longer at lower temperatures than high temperatures, because of "molecular mobility," which increases with temperature.

There are other variations on this general approach. There are various forms of sieve materials which are selective to particular compounds. From For example, Union Carbide offers absorbers for Hβ, ethylene, etc.

Beside the "absorbers," there are a host of "surface interactive" materials commonly used in chromatography (both gas and liquid). If, for example, N₂ and HCN were in a gas stream, the HCN gas component may be highly diluted and beyond the detection range of a particular instrument. If the probe were packed with a particular material whic selectively interacted with HCN as it passed through the column, the net effect would be to delay, thus concentrating this HCN gas. The level of "concentrated" gas would be controUed by the concentration of HCN in the stream. The above discussion would also hold with all liquid systems. Finally, this enhanced (equilibrium) concentration within the probe allows monitoring of what would otherwise be marginally detectable levels (concentrations).

28. Real-Time Monitoring of Gas-Phase Mixtures

A further embodiment of the present invention, which has now been successfully demonstrated, provides direct real-time monitoring of gas-phase mixtures. Figure 29 shows the modified measuring apparatus preferably used for monitoring gas-phase mixtures which may contain fractions of low volatility. Note that, in this embodiment, the measurement section 2920 is surrounded by a thermal blanket (or other thermal confinement structure) to create a level-temperature zone 2930. This zone is held at a uniform temperature which is high enough to assure that the fluid flow being monitored will remain in the vapor phase without condensation. In the experiments described, the temperature was set at 85 ^β C, but of course other temperatures could be used instead. In the apparatus shown, a source flask 2940 is heated, and the resulting vapor-phase stream passes through the measurement section 2920 and condenses into a cooler coUection flask 2950. However, of course, other fluid flow arrangements would typically be used in a fuU-scale process plant.

Figure 30 shows the actual results derived from test runs on various systems, using the apparatus of Figure 29.

Nitrogen

In runs 1 and 9, pure nitrogen was flowed through the system.

The osciUator frequency was steady at about 1153.27 MHz for each of the data measurements. In this data run, and in the following data runs, the power attentuation was 0 dB. It appears that gas phase mixtures are generally less lossy than hquid phase mixtures.

Benzene

In run 2, pure benzene was evaporated from the source flask. The osciUator frequency was steady at about 1152.97 MHz for each of the data measurements.

Ethanol

In run 3, pure ethanol was evaporated from the source flask. The osciUator frequency was steady at about 1152.52 MHz for each of the data measurements. Ethyl Acetate

In run 4, pure ethyl acetate was evaporated from the source flask. The oscillator frequency was steady at about 1152.23 MHz for each of the data measurements.

2-Propanol

In run 5, pure 2-propanol was evaporated from the source flask. The oscillator frequency was steady at about 1152.46 MHz for each of the data measurements.

Azeotrope of Ethanol/Benzene (32.4/67.6 %wt/ wt) In run 6, a mixture of ethanol and benzene was evaporated from the source flask. Under mUd boiling conditions, the vapor-phase mixture which evolved was therefore the azeotropic mixture of these two components. This particular azeotropic mixture is known to have a composition of 32.4% (by mass) of ethanol, and 67.6% by mass of benzene. The oscillator frequency was steady at about 1152.76 MHz for each of the data measurements.

Azeotrope of Ethanol/Ethyl Acetate (31/69 %wt %wt)

In run 7, a mixture of ethanol and ethyl acetate was evaporated from the source flask. Under mild boiling conditions, the vapor-phase mixture which evolved was therefore the azeotropic mixture of these two components. This particular azeotropic mixture is known to have a composition of 31% (by mass) of ethanol, and 67.6% by mass of benzene. The oscUlator frequency was steady at about 1152.34 MHz for each of the data measurements.

Azeotrope of Ethanol/Water (95.6/4.4 %wt/%wt)

In run 8, a mixture of ethanol and water was evaporated from the source flask. Under mUd boUing conditions, the vapor-phase mixture which evolved was therefore the azeotropic mixture of these two components. This particular azeotropic mixture is known to have a composition of 95.6% (by mass) of ethanol, and 4.4% by mass of benzene. The oscUlator frequency was steady at about 1152.75 MHz for each of the data measurements.

Carbon Dioxide In run 10, pure C0₂ was pumped through teh system. The oscUlator frequency was steady at about 1153.20 MHz for each of the data measurements.

Uses of this Capability

One way to use this vapor-phase measurement capability is for indirect monitoring of the changing composition of a hquid mixture: the vapor- phase stream withdrawn may be a relatively small mass flow, in relation to the hquid-phase mass contained inside the reaction vessel. (Note that the vessel may be a pressure vessel if desired.) The changing composition of the vapor-phase stream can provide a proxy for direct monitoring of the hquid-phase vessel. For example, when the vapor-phase characteristics reach a target, the source flask can be cooled, or another reagent or catalyst can be introduced, or the contents of the source flask can be transferred to another processing stage.

Another way to use this capability (and the other capabilities described in this apphcation and in the parent apphcations) is to use feedback from the vapor-phase measurement to control a light source which drives photolytic reactions in vapor-phase.

Another way to use this capabUity (and the other capabUities described in this apphcation and in the parent apphcations) is to use feedback from the vapor-phase measurement to control the temperature apphed to a cracking vessel.

29. Monitoring Mixtures in a Magnetic Field

A further innovative teaching disclosed in the present application is that nuclear magnetic resonance characterization can be performed on a gas-phase mixture in a load-puUed oscillator configuration like that shown herein.

Figures 31A, 31B, and 31C show three different apparatus embodiments in which a bias magnetic field can be applied to a measurement section. In each of these embodiments, a flow of a material to be monitored or characterized is passed through a measurement section in which a substantial DC magnetic field is applied. In the example shown in Figure 3 IB, the magnetic field is transverse to the direction of fluid flow, and in Figure 31A it is parallel, but other bias field orientations could be used if desired. Optionally, for better signal-processing gain, a slow variation may be applied onto the bias field. (Similar techniques are commonly used in NMR measurement.)

In Figure 31 A, a simple solenoidal coil 3100A is wrapped around most of the measurement section 14. Thus, the magnetic field lines are generally parallel to the direction of fluid flow, and also paraUel to the coaxial line.

The fluid flow connections of the measurement section to a fluid source and a fluid sink, and the electrical connection of the free-running RF oscUlator 30 to the measurement section, are preferably the same as in the preferred embodiment of Figure 1. Figure 3 IB is generally similar to Figure 31 A, except that the magnetic bias field is applied differently. In this embodiment, the measurement section 14 lies between two poles of a wide magnet 3100B (which can be an electromagnet or even a permanent magnet). Thus, the magnetic field lines are generally peφendicular to the direction of fluid flow, and also paraUel to the coaxial line. The fluid flow connections of the measurement section to a fluid source and a fluid sink, and the electrical connection of the free-running RF osciUator 30 to the measurement section, are preferably the same as in the preferred embodiment of Figure 1.

Figure 31C shows a significantly different embodiment. Note that, in this embodiment, the free-running oscUlator 30 is not coupled directly to the measurement section 14: instead, this connection is made through an

RF switch (or diplexer) 3130. The switch 3130 is operated to couple a high-powered RF pulse from a pulsed RF source 3132 into the measurement section 14. Thereafter, the switch 3130 is reset so that the free-running osciUator 30 is coupled to the measurement section.

The switch 3130 can be implemented, for example, as a pair of pin diode switches. However, it is more preferable to make this as a diplexer.³⁴ In this case the free-running oscillator 30 and the RF power source must not operate at the same frequency. However, the RF power source can operate continuously: the switch or diplexer 3130 provides isolation between the RF power source and the free-running oscillator.

In an alternative embodiment, using the apparatus of Figure 31C, it may be preferable to apply a relatively high-powered pulse first (from a separate RF power amplifier, and possibly at a separate frequency band), and then observe the frequency shift of a load-pullec oscillator as described above. This permits relaxation phenomena to be observed.

Use of Reaction Monitoring in Larger Systems Figure 10 shows an example of a large-scale chemical process system according to the present invention.

To better explain the operation of this system, it will be described with reference to a sample embodiment where a reaction vat 1010 is used to react isoprene with bromine in a benzene solvent system, in a batch or semi-continuous process.

The reaction vat 1010 is fed by three injection pumps: a first injection pump 1022, which supphes isoprene; a second injection pump 1024, which supplies benzene; and a third injection pump 1026, which supplies bromine.

frequency wiU be reflected from the first arm, and a second tuned stub is positioned in a second arm, so that all at a first frequency wUl be reflected from the second arm. The result of this is that a third arm is coupled to the first arm at the second frequency, and to the second arm at the first frequency. A portion of the reaction vat is electricaUy connected to provide an electrical load for a real-time compositional measurement system 1030, which contains a free-running RF oscUlator (loaded by a measurement section containing an integral portion of the contents of vat 1010) and a frequency counter, like those described above. The output of the real-time compositional measurement system is fed back to provide a measured variable input for process control system 1040 (which may be a conventional microcomputer system).

The control logic preferably implemented by the process control system 1040, in operating the system of Figure 10, is as follows. (For clarity, this flow is described as a batch process.)

Benzene is added to the reaction vat 1010. Isoprene is added to the reaction vat 1010. Bromine is added to the reaction vat 1010. • Additional isoprene is metered in, by injection pump, while the frequency of the RF osciUator in measurement system is monitored, in the following control loop:

— If the frequency drops as additional isoprene is added, then continue to add isoprene (because unreacted bromine is still present); — If the frequency rises as additional isoprene is added, then stop adding isoprene, and remove product (since all bromine has been consumed).

A further example of a contemplated control application is using a real-time measurement, from an electrical configuration as disclosed, to control the feed rates of feedstock flows in a three-component azeotropic distUlation system. Specific Implications for the Chemical Industry

The disclosed innovations will be useful in a very wide variety of apphcations in the chemical industry, and also in many other type of industries, including food, pharmaceuticals, and many others. Some systems wiU be instaUed to monitor and record data that can be correlated to something of interest and be the essential counter part of those sold to the oU industry. The primary difference wUl be that the measurement section wUl have a variety of configurations since there will be few "standard" mountings. Installations will be in pipes, internal to reaction chambers, and at multiple points within distUlation columns.

Most systems wiU not be used as simply monitors, but be a part of an interactive system controlling process parameters. Also, it is reasonable to assume that most instaUations wiU involve multiple systems since more than a single input wiU be necessary to control a reaction. For example, a typical instaUation might have separate units in feed lines, the reaction chamber, and the reaction product exit line. In some cases it may be practical to run aU measurement sections from a single electronic system by multiplexing.

This equipment can be used to monitor virtually every class and type of organic reaction. This would allow control based on chemical compositional properties in both batch or continuous reactions in either hquid or gaseous systems by direct measurement.

Because the starting base components of most organic synthesis are known and tightly controUed, more specialized higher resolution equipment is possible to monitor subtle transitions. Because the dielectric range for particular chemical systems would be known, optimized equipment with 10^-3 to 10^-4 resolution is possible. It is also possible to add trace quantities of "tag" compounds to enhance the observation of particular intermediate reaction steps of interest in complex chemical sequences.

Further Modifications and Variations

It wUl be recognized by those skilled in the art that the innovative concepts disclosed in the present application can be apphed in a wide variety of contexts. Moreover, the preferred implementation can be modified in a tremendous variety of ways. Accordingly, it should be understood that the modifications and variations suggested below and above are merely Ulustrative. These examples may help to show some of the scope of the inventive concepts, but these examples do not nearly exhaust the full scope of variations in the disclosed novel concepts.

For another example, in some applications it may be useful to use TWO load-puUed oscillators in a single flow, and use a differential measurement between the two load stages, for more precise monitoring. This permits high-resolution measurement of trends in space or in time. For another example, in some applications, where relatively wideband tuning of the oscUlator is anticipated, it may be useful to use two separate measurement sections which are in proximity but have different electrical lengths. Since the two measurement sections wUl not pass through a hop at the same frequency, such a configuration provides another way to obtain accurate frequency measurement without error due to frequency hopping. The two measurement sections can be used as loads for two separate load- pulled oscillators, or can be multiplexed onto a single oscillator.

For another example, it is possible to connect a single measurement section to two (or more) free-running oscillators running at different frequencies.

For another example, in some apphcations it may be useful to use to monitor the location of a phase boundary, in a continuous process.

For another example, in some applications it may be useful to monitor physical phase characteristics, such as the degree of emulsification or colloidization in a two-phase mixture.

For another example, the disclosed innovative concepts may be particularly advantageous in environmental monitoring, e.g. to provide real¬ time monitoring of chemical contamination. This can be useful in providing systems to monitor issues such as water supply quality or acid rain.

For another example, the disclosed innovative concepts may be particularly advantageous for the food industry, for measuring the water content of a known material, or for sample analysis, or for purity monitoring (to detect tampered solutions). For similar reasons the disclosed innovative concepts may be advantageous for the pharmaceuticals industry.

For another example, the very precise compositional monitoring provided by the disclosed innovations can be used to monitor and control the fraction of a low-percentage component of a mixture. For example, this may be used to design systems in which the catalyst circulates with the reagents (and is recovered from the product), rather than using an excess of catalyst which is confined in a fixed location (using a mesh or a fluidized bed), as is conventional. A catalyst will normally be lost or consumed at a moderate rate over time, and the compositional monitoring permitted by the present invention can avoid excessive catalyst consumption, by allowing the total fraction of catalyst present to be reduced without any risk of falling below predetermined a minimum catalyst fraction. This can be particularly advantageous where the catalyst is a finely divided metal, as is common.

For another example, the very precise compositional monitoring provided by the disclosed innovations can be used in polymer synthesis to monitor and control the characteristics of the reaction mixture. As is well known to polymer chemists, there are a large number of additives which can be used to modify polymer properties by affecting the molecular weight, degree of cross-linking, and/or formation of heterogenous domains in the product. Some of these additives are not consumed, and many of them have a large effect in very small concentrations. The sensitive measurement of composition (and of changes in molecular bonding and conformation) permits more precise control of such additives, and also of other inputs (such as thermal curing or photochemical energy input).

For example, the disclosed methods can be used to precisely monitor changes in physical density of materials, or in the smaU-scale structure of materials. For example, very small changes in the solid/gas ratio of a fluidized bed can be detected, especially in the range where the gas film thickness between adjacent solid particles changes significantly. For another example, the disclosed methods may be used to momtor the degree of solid-solid linkage in a gel or aerogel.

In further alternative embodiments, the disclosed innovations can be used for dynamic monitoring (and/or for control based on dynamic monitoring) of any of the foUowing: monitoring reaction kinetics (reaction rates); analysis of dynamic components of reaction and reactivity; analysis of isomeric components and transitional equilibrium; determination of various equilibrium constants based on the identification of particular species; determination of solubility/insolubility constants; determination of extent of completion of reaction; monitoring changes in physical phase; monitoring intermediate reaction components which may control yield or properties; monitoring/determination of reaction mechanisms; monitoring/determination of ionization dissociation constants; monitoring the state of catalysts (e.g. solid/liquid solubility or ionic states); component characteristics from polar contributing components (e.g. with ionic groups, asymmetric unsaturated bonds, nucleophilic groups, or electrophilic groups); determination of particular component concentration.

For another example, the very precise compositional monitoring provided by the disclosed innovations can be used to rigorously monitor a product for contamination by undesired intermediates. This can be particularly advantageous in the food and pharmaceuticals industries, since it broadens the range of process which can be used economically while stUl meeting product purity standards.

For another example, the very precise monitoring of short-range organization, in two-phase compositions, which is provided by the disclosed innovations can be used to momtor and control the fraction of a high-cost component of a slurry, gel, or other multiphase system.

For another example, the very precise monitoring of short-range organization which is provided by the disclosed innovations can be used to momtor and control the flow characteristics of two-phase inputs to a continuous process. Chemical processes can now be designed to use slurries as inputs, and to make use of both the liquid and the solid components of the slurry, with confidence that the flow rate and composition of the slurry can be accurately controlled.

For another example, if it is desired to operate the disclosed system at higher microwave frequencies, coupling to the variable load can be accom- plished with a probe into a cavity, rather than a simple coaxial line. Such probes can readUy be configured to couple primarily to the electric field, or primarily to magnetic field.

That is, a general teaching is that an electromagnetic propagation structure is both part of a fluid stream which connects its contents to a chemical system of interest, and also part of an electrical circuit from which real-time characterization of the fluid stream can be derived. This cavity should preferably not have multiple spurious resonance modes at the frequency of interest. (For example, if a coaxial hne has a radius which is much smaUer than a quarter-wavelength at the frequencies of interest, the frequencies where a terminated segment of that line change from inductive to capacitive wiU be determined merely by the effective electrical length of the line.) It is preferable, although not absolutely necessary, that the electromagnetic propagation structure should have only one class of modes in the frequency band of interest. The electromagnetic propagation structure is most preferably a shorted coaxial segment, but may less preferably be a resonant cavity or other structure.

For another example, the disclosed system can alternatively be operated at a frequency which corresponds to the second harmonic of the cavity. In such a system, the fuU frequency of the oscUlator is preferably fed into the load, but a filter is used to extract the second harmonic component. By measuring insertion loss at the second and higher harmonics, a profile of insertion loss over a wide range can readUy be obtained. (The onfy hardware change needed is an appropriate filter stage.) For another example, it is not strictly necessary to use a closed chamber for the measurement section. Alternatively, An electrical probe structure could simply be placed in close proximity to the material to be monitored. (With bulk solids, this may be necessary.) For another example, the disclosed innovative systems could also be used as an analytical tool, for analysis of samples off-line.

For another example, a bias electric field (instead of, or in addition to, a bias magnetic field) can be applied to change the load-pulling characterisitcs of a sample. This is particularly convenient where the measurement section includes a central coaxial conductor, as in the embodiment of Figure 1, since the bias electric field can be apphed between the central conductor and the shell.

As will be recognized by those skilled in the art, the innovative concepts described in the present application can be modified and varied over a tremendous range of apphcations, and accordingly their scope is not limited except by the aUowed claims.

Claims

CLAIMSWhat is claimed is:

1. A system for controlling a process, comprising: a voltage-controUed oscUlator, which includes a gain element capable of providing substantial gain at frequencies greater than 100 MHz; a feedback path, coupling the output of said gain element to the input thereof, said feedback path including a tunable resonant circuit; an electromagnetic propagation structure which is RF-coupled to load said oscillator and in which electromagnetic wave propagation is electricaUy loaded by a portion of a medium undergoing said process; means for monitoring the frequency and insertion loss of said osciUator to ascertain changes in the composition of the medium; and means for controlling one or more process variables in accordance with the output of said monitoring means.

2. The system of Claim 1, wherein said electromagnetic propagation structure comprises a transmission line segment which permits only one mode of propagation at the operating frequency of said osciUator.

3. The system of Claim 1, wherein said electromagnetic propagation structure comprises a shorted transmission line segment.

4. The system of Claim 1, wherein said electromagnetic propagation structure includes a hoUow portion therein, through which said medium undergoing said process can flow.

5. A method for controlling a reaction process, comprising the steps of: providing a voltage-controlled osciUator, which includes a gain element capable of providing substantial gain at frequencies greater than 100 MHz, and a feedback path, coupling the output of said gain element to the input thereof, said feedback path including a tunable resonant circuit; flowing a stream of said fluid medium, in which said process is expected to be taking place, through a fluid container which is electrically configured as a transmission line segment and which is electrically connected to load said oscillator; operating said osciUator at a frequency chosen to provide a particularly strong shift in electrical parameters in accordance with the progress of said process in said fluid medium; monitoring frequency and insertion loss of said osciUator to ascertain the progress of said process in said fluid medium.

6. The method of Claim 5, wherein said osciUator is operated at a microwave frequency which is near a molecular resonance in said fluid medium.

7. The method of Claim 5, wherein one of said controUed variables is heat flow to a particular vessel.

8. The method of Claim 5, wherein one of said controUed variables is flow of an input stream to a particular vessel.

9. The method of Claim 5, wherein one of said controUed variables is flow of a product stream from a particular vessel.

10. The method of Claim 5, wherein one of said controUed variables is flow of a bottom product stream from a particular vessel.

11. The method of Claim 5, wherein one of the products of said reaction is expected to be strongly ionic, and wherein insertion loss is measured at at least two widely separated frequencies.

12. The method of Claim 5, wherein said transmission line segment has dimensions which permit only one mode of propagation at frequencies in the neighborhood of said starting frequency of said oscUlator.

13. A method for monitoring changes in the small-scale structure of a medium of interest, comprising the steps of: providing an osciUator, which includes a gain element capable of providing substantial gain at frequencies greater than 100 MHz, and a feedback path, coupling the output of said gain element to the input thereof, said feedback path including a tunable resonant circuit; flowing a stream of said fluid medium, in which said process is expected to be taking place, through a fluid container which is electrically configured as a transmission hne segment and which is electrically connected to load said oscUlator; operating said oscUlator at a frequency chosen to provide a particularly strong shift in electrical parameters in accordance with the progress of said process in said fluid medium; monitoring frequency and insertion loss of said oscillator to ascertain the progress of said process in said fluid medium.

14. The method of Claim 13, wherein said medium of interest is a two- phase flowable composition.

15. A method for controlling an absoφtion/desoφtion process, comprising the steps of: providing a feedstock flow which contains at least one undesired impurity; flowing said feedstock flow, in a first reaction vessel, past a solvent which preferentially absorbs said impurity; monitoring the fraction of said impurity in the flow of loaded solvent from said first reaction vessel, using a load-pulled oscillator, and controlling the conditions of said first vessel, and/or the flow rates of said solvent and said feedstock, to maintain the loading of said loaded solvent at a desired level; flowing said loaded solvent into a second reaction vessel, wherein said loaded solvent is stripped, to reduce the concentration of said undesired impurity therein to a target level; monitoring the concentration of said impurity in the flow of stripped solvent from said second reaction vessel, using a load-puUed oscUlator; and controlling the conditions of said second vessel, and/or the flow rate of said loaded solvent, to maintain the loading of said stripped solvent at said target level.