WO2007128140A1 - Magnetic resonance spectrometer suitable for integration on a single chip - Google Patents

Abstract

A magnetic resonance device is disclosed which is suitable for integration on a single chip. The device integrates oscillators for frequency generation, excitation/detection coils and a detector in a manner that allows for production on a single chip by a standard CMOS process. In one embodiment, a dispersion-type spectrometer is provided. Two LC-type oscillators operating at different fre- quencies are provided, the difference frequency being monitored. The coils employed in the resonant circuits of the LC-type oscillators serve as excitation coils. On resonance, the inductance of one of the coils is affected, while the other coil remains unaffected. Another embodiment of the device is an absorption-type spectrometer which operates differentially. Two identical groups of excitation and detection coils (L11, L22, LD1; L21, L12, LD2) are arranged in a symmetric manner around an active core region, and a difference signal from the detection coils (LD1, LD2) is detected while the sample is close to only one of the detection coils.

Description

MAGNETIC RESONANCE SPECTROMETER SUITABLE FOR INTEGRATION ON A SINGLE CHIP

Field of the invention

The present invention relates to a magnetic resonance device.

Background of the invention Electron spin resonance (ESR), also known as electron paramagnetic resonance (EPR), is a spectroscopic technique that detects species having unpaired electrons. ESR is widely used in a variety of different fields, e.g., in solid-state physics, where ESR may be used for the identification and quantification of radicals (molecules with unpaired electrons); in chemistry, where ESR may be used for identifying reaction pathways; or in biology and medicine, where ESR may be used for tagging biological spin probes. Many other applications exist.

Today, spectrometers in the X-band range (8-10 GHz) of resonance frequencies are most commonly employed. Commercial or experimental spectrometers also exist in the S band (around 3 GHz), K band (around 23 GHz), Q band (around 35 GHz) and W band (around 95 GHz). Generally, sensitivity increases with increasing resonance frequency. Commercial X-band spectrometers can reach spin sensitivities around 10¹⁰ spins/Gauss (10¹¹ spins/mT) at room temperature, with lower figures attainable only with special setups.

The basic physical concepts of ESR are analogous to those of nuclear magnetic resonance (NMR), another magnetic resonance technique that has found an even wider range of applications. Because of the larger magnetic moment of the electron as compared to the magnetic moments of nuclear spins, generally weaker magnetic fields and higher frequencies are employed in ESR than in NMR. ESR or NMR spectroscopy of ultra-small scale structures (e.g., protein single crystals, single cells, monolayers of biomolecules etc.) is often not feasible with conventional setups. It has been proposed to use special techniques such as magnetic force detection for applications with small samples. Such setups are, however, very complex, expensive and difficult to operate. In parallel, there have been efforts to miniaturize the probes for conventional magnetic resonance devices for applications with small samples. In particular, microcoils on glass substrates have been developed for ESR (e.g., G. Boero et al., Rev. Sci- ent. Instr. 74, 4794-4798 (2003)) or NMR (e.g., C. Massin et al., Sensors and Actuators A 97-98, 280-288 (2002)). Such coils are normally used with spectrometer components such as oscillators, amplifiers or mixers implemented in a conventional, discrete manner, which still require considerable space. Relatively long connections between these components, usually in the range of several meters, are required, which may cause undesired losses and nonlinearities. In addition, frequency tuning and impedance matching to the impedance of the connecting cables and the input impedance of the signal preamplifier can be difficult for such microprobes. Furthermore, a spectrometer setup using micro- probes is still relatively expensive, costs being comparable or even exceeding those of a conventional spectrometer setup.

In Boero et al., Rev. Scient. Instr. 72, 2764-2768 (2001 ), it was proposed to provide an integrated NMR probe on a single chip. Planar spiral coils, a radio- frequency (rf) preamplifier, a mixer and an audio-frequency amplifier were integrated on the probe using a standard CMOS process. However, frequency gen- eration was done externally by a standard stand-alone rf generator and a standard rf power amplifier, and the rf signals for excitation had still to be transmitted to the probe through relatively long cables.

Summary of the invention It is an object of the present invention to provide a magnetic resonance device which can be manufactured in a highly integrated form and at low cost. This object is achieved by a magnetic resonance device according to claim 1 or 15.

It is another object of the present invention to provide a method of operating such a device. This object is achieved by a method according to claim 14 or 26.

Advantageous embodiments of the present invention are laid down in the dependent claims.

Thus, according to a first aspect of the present invention, a magnetic resonance device is provided which may be called a differential dispersion-type magnetic resonance device. The device comprises an inductor arrangement for generating a high-frequency magnetic field at a sample location and for receiving a response of a sample placed in the sample location to said high-frequency magnetic field; a first and a second oscillator, said first oscillator being adapted for generating a current in said inductor arrangement at a first oscillator frequency and said second oscillator being adapted for generating a current in said inductor arrangement at a second oscillator frequency different from said first oscillator frequency; and a detection circuit for detecting said response of said sample and for providing an electrical output signal that reflects said response, wherein said detection circuit is adapted for detecting a change of a difference frequency between said first and second oscillator frequencies due to a change of inductance in said inductor arrangement when magnetic resonance in said sample occurs.

In this embodiment, magnetic resonance is detected by monitoring a frequency difference. On resonance, the sample influences the inductance of at least one of the inductors of the inductor arrangement, thereby changing the eigenfre- quency of at least one of the oscillators. Preferably, the inductor arrangement, the oscillator device and the detection circuit are part of an integrated circuit on a single semiconductor substrate ("chip"), in particular, a CMOS chip. The device may be manufactured by standard semiconductor production processes, in particular, standard CMOS (com- plementary metal-oxide semiconductor) processes. In this context, a CMOS process includes variants such as BiCMOS etc. The inductor arrangement is preferably formed from a metallization layer deposited on the substrate and selectively etched to provide a continuous conductive path.

Preferably, the inductor arrangement includes separate first and second inductors or groups of inductors, the first oscillator being adapted to generate a current in said first inductor or group of inductors and the second oscillator being adapted to generate a current in said second inductor or groups of inductors. These separate inductors or groups of inductors can be considered as separate "excitation coils". It is preferably the inductance change on magnetic resonance in one of these excitation coils which leads to a frequency change of one of the oscillators. This frequency change can be monitored by directly monitoring the difference of the frequencies of the oscillators or by picking up the high- frequency field generated by the excitation coils and detecting the difference frequency.

Each of the inductors defines an active area, i.e., an area in which a high- frequency current through the inductor generates an appreciable magnetic field. If the inductor takes the form of a coil with one or more full turns, it is the inside of the coil which may be regarded as the active area. Preferably, the active areas of the two inductors overlap to form the sample location. In this manner, any disturbances at the sample location leading to changes of inductance which are unrelated to magnetic resonance affect both inductors equally and thus cancel.

For achieving the best possible stability and lowest possible susceptibility to disturbances, it is advantageous to use a symmetrized setup as far as possible. In particular, it is preferred that the first inductor and the second inductor are arranged in a substantially symmetric configuration around an axis or plane of symmetry. It is. further preferred that the inductors are identical in shape and that they are arranged in a substantially orthogonal configuration (i.e., at an an- gle of 90 degrees to each other) in order to minimize magnetic and capacitive couplings for avoiding injection-locking of the two oscillators. It is also preferred that the first and second oscillators are arranged in a substantially symmetric configuration around an axis or plane of symmetry. Such a highly symmetric setup is particularly suited for integration on a chip because of the high dimen- sional reproducibility of semiconductor production processes, which ensures that the two inductors are as much identical in their properties as possible.

Traditional magnetic resonance devices employ a frequency generation or local oscillator system which is followed by a power amplifier, and the outputs of the power amplifier are then fed to an excitation coil. In the present invention, it is proposed to preferably integrate the functions of frequency generation and amplification into one and the same unit by utilizing an LC-type oscillator, in particular, a voltage-controlled oscillator (VCO), whose coil(s) serve(s) as the excitation coil. In particular, each of the first and second oscillators is an LC-type oscillator, and each of the first and second inductors is part of a resonant circuit of one of said LC-type oscillators. This provides a particularly simple way of translating an inductance change into a frequency change, and it removes the necessity of separate power amplifiers for driving the inductors. In the context of the present invention, an LC-type oscillator is an oscillator whose output fre- quency is determined by a resonant circuit formed by at least one inductor ("coil") and at least one capacitance. In the most simple cases, one coil and one capacitor are connected in parallel or in series to form the resonating circuit; however, generally more complicated coil-capacitor configurations are employed.

The detecting circuit preferably comprises a mixer having first input terminals receiving an output signal of said first oscillator, said mixer having second input terminals receiving an output signal of said second oscillator, and said mixer having output terminals providing a signal at a difference frequency, said difference frequency being the difference of frequencies provided at said first and second input terminals. Preferably, the detecting circuit further comprises a high-frequency buffer for converting said mixer output into a binary signal.

For providing a low-frequency output which may readily be processed by standard equipment, the detection circuit preferably further comprises a frequency divider providing an output signal at a frequency corresponding to said differ- ence frequency divided by a predetermined factor. In this manner, a difference frequency which will typically be in the range of a few hundred MHz may be converted to a frequency in the lower MHz range, which may readily be processed.

The magnetic resonance device according to the present invention is suitable for both ESR and NMR. However, preferably it is adapted to detect electron spin resonance. In particular, the device may be operated a range of frequencies above 1 GHz, which is not interesting for standard NMR applications. Operation even above 8 GHz has been demonstrated experimentally. Thus, the device is suited for operation in the X-band of microwave frequencies.

The present invention thus proposes to provide all essential elements of a magnetic resonance device, including frequency generation, inductors and signal detection, in a manner that is readily integrated on a single chip (= semiconduc- tor substrate). This extremely high degree of integration enables the manufacture of low-cost magnetic resonance devices for diverse fields of application, including low-cost magnetometry, operation as a sensor, spectroscopy of very small samples or of a large number of samples in parallel, magnetic resonance surface microscopy, and many more. Already from this list it is apparent that the range of applications is not limited to benchtop spectroscopic applications, but that completely novel applications of magnetic resonance are possible which have previously not been considered due to the excessive cost of magnetic resonance devices.

A complete magnetic resonance device can thus be manufactured at a fraction of the cost of prior-art devices. The device is easily operated, as it requires not much more than a voltage supply and some standard equipment for the recording of spectra. In particular, many of the difficulties of prior-art devices which were due to the necessity of transmitting the high-frequency excitation currents and the very small resonance signals through cables to and from a probe are overcome by the present invention, as no such cabling for such cur- rents or signals is required.

A method of operation of the differential dispersion-type magnetic resonance device according to the present invention may comprise the steps of placing a sample in said sample location and placing the device in an external magnetic field; operating said first oscillator at said first oscillator frequency and said second oscillator at said second oscillator frequency; sweeping said external magnetic field through a magnetic resonance of said sample; - monitoring said difference frequency as a function of said external magnetic field to obtain a spectrum.

According to a second aspect of the present invention, a magnetic resonance device is provided which may be called a differential absorption-type spec- trometer. The device comprises an inductor arrangement for generating a high-frequency magnetic field at a sample location and for receiving a response of a sample placed in the sample location to said high-frequency magnetic field due to magnetic resonance of paramagnetic species in said sample, said inductor arrangement including a first excitation coil, a second excitation coil having substantially identical characteristics as said first excitation coil and being spatially removed from said first excitation coil, a first detection coil being magnetically coupled to said first excitation coil, and a second detection coil having substantially identical characteristics as said second detection coil and being magnetically coupled to said second excitation coil, wherein said sample location is closer to said first detection coil than to said second detection coil; an oscillator device operable to provide high-frequency currents at the same frequency to said first and second excitation coils; and a detection circuit for detecting said response of said sample, wherein said detection circuit is adapted for detecting a difference of induced voltages in said detection coils due to energy absorption by the sample on resonance..

In this context, the term "coil" is to be understood to include any inductor or arrangement of inductors capable of generating or picking up a magnetic field. Thus, two groups of excitation/detection coils are provided, which are essentially identical in their properties, but spatially separated from each other. Only one of them receives the sample, while the other one serves as a reference for differential detection. Such a device acts as a differential absorption-type spectrometer. A very simple differential detection scheme results if the detection coils are simply connected in an anti-series configuration, thus providing only the difference of any voltages induced in these coils.

Again, the device is preferably implemented on a single semiconductor chip, i.e., the inductor arrangement, the oscillator device and the detection circuit are preferably part of an integrated circuit on a single semiconductor substrate, in particular, a CMOS-type integrated circuit.

For ensuring the best possible stability of the frequency generation branch and for ensuring that both groups of excitation coils receive high-frequency currents of exactly the same frequency and amplitude, preferably two oscillators are operated in parallel in a conductor-sharing configuration. In particular, in a preferred setup, the oscillator device comprises a first and a second LC-type oscil- lator. Each LC-type oscillator has a first and a second output terminal, where the first output terminals of said first and second oscillators are connected with each other and said second output terminals of said first and second oscillators are also connected with each other. Each LC-type oscillator comprises a first and a second inductor, which are part of a resonant circuit of said LC-type oscillator. The first inductor of said first oscillator and the second inductor of said second oscillator are connected in a series configuration to form said first excitation coil, and the second inductor of said first oscillator and the first inductor of said second oscillator are connected in a series configuration to form said sec- ond excitation coil. In this manner, the LC-type oscillators "share" their resonating coils, which ensures that the two oscillators operate in perfect synchrony. In many embodiments, the connecting point between each inductor pair will additionally be connected to a current source or to a supply line.

Conceptually, an LC-type oscillator device comprises an oscillator core constituted of the semiconductor elements, in particular, the transistors and varactors, and the oscillator coil, which will generally be etched from a metal layer. In an advantageous embodiment, the coils are arranged symmetrically with respect to the oscillator core. In particular, preferably said first excitation coil and said first detection coil form a first group of coils, said second excitation coil and said second detection coil form a second group of coils, and said first and second groups of coils are arranged in a axially or mirror-symmetric configuration with respect to said oscillator core.

Preferably, the excitation and detection coils overlap each other. In particular, it is advantageous if the first detection coil is arranged within said first excitation coil, and said second detection coil is arranged within said second excitation coil.

Preferably, the magnetic resonance device provides a low-frequency output. To this end, the detection circuit preferably comprises a mixer having first input terminals receiving signals from output terminals of said oscillator device, said mixer further having second input terminals receiving signals derived from said detection coils, and said mixer having output terminals providing a signal at a difference frequency, said difference frequency being the difference of frequencies provided at said first and second input terminals. Since these frequencies are generally the same, a DC output is provided, the voltage of that output varying according to the magnetic resonance behavior of the sample during a field of frequency sweep.

In order to provide the mixer with a sufficiently large input voltage, an amplifier circuit (in particular, a low-noise amplifier) receiving a difference signal from said first and second detection coils and supplying an amplified difference signal to said mixer is suggested. The amplifier circuit is preferably matched in its input impedance to the detection coil system formed by the first and second detection coils connected in an anti-series configuration. Also the output imped- ance should be matched to the input impedance of the mixer. In this manner, the amplifier circuit serves at the same time as an impedance transformer. For achieving matching, the amplifier circuit may comprise at least one pair of matching coils for adjusting its input and/or output impedance. The matching coils are then preferably arranged in a symmetric manner with respect to said inductor arrangement.

A method of operation of the differential absorption-type magnetic resonance device according to the present invention may comprise the steps of placing a sample in said sample location and placing the device in an external magnetic field; operating said oscillator device to provide high-frequency currents to said first and second excitation coils; sweeping said external magnetic field at constant frequency of said oscillator device or sweeping said frequency of said oscillator device at constant magnetic field through a magnetic resonance of said sample; monitoring said electrical output signal of said detection circuit to ob- tain a spectrum.

Brief description of the drawings

The invention will be described in more detail in connection with an exemplary embodiment illustrated in the drawings, which show:

Fig. 1 a block diagram of a differential ESR spectrometer;

Fig. 2 a block diagram of a differential ESR spectrometer adapted for on- chip implementation;

Fig. 3 a block diagram of a differential ESR spectrometer further opti- mized for on-chip implementation;

Fig. 4 a circuit diagram of a CMOS LC-type twin voltage-controlled oscillator (VCO);

Fig. 5 a circuit diagram of a CMOS LC-type inductor-sharing twin VCO;

Fig. 6 a circuit diagram of a CMOS LC-type circular-geometry twin VCO; Fig. 7 a MOS differential low-noise amplifier (LNA);

Fig. 8 a MOS passive mixer;

Fig. 9 a block diagram of an on-chip ESR spectrometer according to a first embodiment; Fig. 10 a coil network layout; Fig. 11 a sketch of a three-layered compensation coil;

Fig. 12 an equivalent circuit for the coil network; Fig. 13 a diagram illustrating coil dimensions for a 6 GHz spectrometer; Fig. 14 a diagram illustrating coil dimensions for an 8 GHz spectrometer; Fig. 15 a micrograph of a CMOS die for a dual spectrometer according to a first embodiment of the present invention for operation around 6

GHz;

Fig. 16 a spectrum obtained with a spectrometer operating at 6 GHz; Fig. 17 a spectrum obtained with a spectrometer operating at 8 GHz; Fig. 18 a micrograph of a CMOS die for a spectrometer according to a second embodiment;

Fig. 19 a block diagram of a spectrometer according to a third embodiment, in particular, a dispersion-type spectrometer, adapted for on-chip implementation; Fig. 20 a sketch illustrating the detection principle for the spectrometer according to the third embodiment; Fig. 21 a circuit diagram of a first VCO; Fig. 22 a circuit diagram of a second VCO;

Fig. 23 a circuit diagram of a mixer;

Fig. 24 a circuit diagram of a high-frequency buffer; Fig. 25 a block diagram of a programmable frequency divider; Fig. 26 a block diagram of a high-frequency divide-by-four block; and Fig. 27 a micrograph of a CMOS die for a spectrometer according to a third embodiment of the present invention.

Detailed description of the invention In the following, preferred embodiments of the present invention are described which conceptually build on a highly symmetric, differential approach to the setup of a magnetic resonance device. In order to make this approach better understood, some of the principles of differential detection will first be described in rather general terms with reference to Figs. 1 to 3, which serve illustrate the concept of differential detection.

Fig. 1 shows a sketch of a differential ESR spectrometer built with conventional components. The outputs of two conventional, identical spectrometers, shown on the left- and right-hand sides of Fig. 1 , are fed to a differential detector 105. In each spectrometer, microwaves are generated by a microwave source 101 , 101' typically comprising a klystron. The microwaves are fed to a circulator 102, 102'. The circulator directs the microwaves to a sample cavity 103, 103'. Microwaves reflected back from the sample cavity are directed by the circulator 102, 102' to the differential detector 105, which detects the difference between the reflected amplitudes from the left cavity 103 and the right cavity 103'. Identical loads 104, 104' close the circulator. Only the left-hand cavity 103 contains an actual paramagnetic sample. This cavity 103 or both cavities 103, 103¹ are placed in an external static magnetic field Bo, whose field strength is chosen such that the electron Larmor frequency (the product of the gyromagnetic ratio of the electron and the magnetic flux density) is close to the microwave frequency. A comparably small, time- dependent, modulating magnetic field is added to the external static field to yield a slowly varying total magnetic field. As long as the total magnetic field is at a value such that the electron Larmor frequency is different from the microwave frequency, both the left cavity 103 and the right cavity 103' behave essen- tially identically, and no difference signal will be detected at the output of differential detector 105. However, when the total magnetic field is at a value where the electron Larmor frequency matches the microwave frequency (the resonance condition), the sample will absorb energy from the microwave field in the cavity 103, while no such absorption can occur in the empty cavity 103'. There- fore, on resonance a difference signal will be detected at the output of detector 105.

The setup of Fig. 1 is not directly suitable for being integrated on a chip, as it is difficult to realize circulators or cavities on a chip by known semiconductor pro- duction processes. A spectrometer setup which is better adapted for on-chip integration is shown in Fig. 2. In this setup, no circulator is required, and the microwave cavity is replaced by an arrangement of coils (or, more generally, inductors). A local oscillator (LO) implemented as a voltage-controlled oscillator (VCO) 201 provides a high-frequency AC voltage. This voltage is fed to a power amplifier (PA) 202, which feeds an amplified high-frequency current to an excitation coil 203. Two detection coils 205, 205' are magnetically coupled to the excitation coil 203. A sample is placed in the vicinity of only one detection coil 205. The detection coils 205, 205' are arranged in an anti-series configuration, i.e., they are connected in series with opposite sense of current flow. Therefore, an output voltage will be present at the terminals of the detection coil arrangement only if the induced voltages in the two coils are different. This output voltage is fed to a low-noise amplifier (LNA) 206. A mixer 207 receives the output of the LNA 206 and of the LO 201 and provides at its output a signal at the difference frequency, the sum frequency being filtered out by a low-pass filter integrated in the mixer. This low-frequency signal is further amplified by an audiofrequency (AF) amplifier 208. Both the mixer output and the AF amplifier output are made available as output signals of the spectrometer.

In operation, the magnetic field is periodically swept through the resonance condition at a fixed LO frequency (or, equivalent^, the LO frequency is swept while the magnetic field is kept constant). When magnetic resonance occurs, the sample will absorb energy, thus modifying the induced voltage in the detection coil 205 while leaving the induced voltage in the other detection coil 205' unaffected. The difference between these induced voltages is detected and provided as a low-frequency signal at the output of the spectrometer.

Fig. 3 shows a slight variant of the device of Fig. 2, which is specifically adapted for being implemented on a single chip by a CMOS process. The excitation coil 303 is now part of an LC-type VCO 300 whose frequency is determined by the inductance of the excitation coil 303 and the capacitance of a varactor (a voltage-controlled capacitance) 302. The LC tank circuit formed by parallel connec- tion of the excitation coil and the varactor is driven by a CMOS active section 301 serving both as a VCO core and as a power amplifier. Thus, the functionalities of a VCO, a power amplifier and of the excitation coil are integrated into a single block which may directly be implemented on a semiconductor chip. In the following, several different embodiments of on-chip spectrometers will be described, which are all based on a symmetric design conceptually similar to that of Fig. 3. The first two embodiments are so-called absorption-type spectrometers, while the third embodiment is a so-called dispersion-type spectrometer. In the following, components which may be used for implementing a CMOS absorption-type spectrometer will be described with reference to Figs. 4 to 8. Subsequently, it will be described how such components may be combined into a highly symmetric magnetic resonance device. For practical implementation of the frequency source, a "circular-geometry" or "balance-geometry" twin VCO structure 6 is suggested for achieving high symmetry, as discussed in Yalcin et al., Proc. IEEE Radio Frequency Integrated Circuits Symposium, pp. 687-689, Long Beach, USA, June 2005. The proposed structure is shown in Fig. 6. Its design is best understood with reference to Figs. 4 to 6. Fig. 4 shows two identical standard CMOS VCOs whose outputs are connected in parallel. The left VCO, designated as VCO1 , comprises a pair of transistors Q11 , Q12 (here, more specifically, NMOSFETs = n-type metal oxide semiconductor field effect transistors). These two transistors are connected with their source electrodes to a common current source that is formed by a transistor Q13 controlled by a gate voltage at an electrode designated as "Core". The drain electrode of each of the two transistors is connected to the gate electrode of the other transistor. Two varactors C11 , C12 in series configuration are connected between the drain electrodes of the transistors in each pair and are con- trolled by a tuning voltage at a tuning electrode T for adjusting the output frequency of the VCO. The drain electrodes of the transistors are connected to a supply voltage Vdd via two coils L11 , L12 having the same inductance. The transistor pair together with the varactor pair forms a first VCO core V1. The second VCO, designated as VCO2, is identical to the first VCO. Its VCO core is designated as V2, its two coils are designated as L21 and L22. By connecting the outputs VOut+ and VOut- of the two oscillators in parallel, it is ensured that a current having the same frequency flows through all four coils L11 , L12, L21 and L22, even if there are slight differences in inductance.

Synchronization of the currents in the coils of the twin VCO may be further improved by a slight modification of the connection scheme, as shown for the "inductor-sharing" VCO 5 in Fig. 5. Now the coils L11 and L22, which belong to different oscillators, are connected to a common Vdd electrode. Likewise, the coils L12 and L21 are now connected to a second common Vdd electrode. Clearly, this setup is electrically equivalent to the setup of Fig. 4. Conceptually, however, the two VCOs now "share" their coils. With this distribution of the coils over the two coupled VCOs, the two Vdd electrodes may be physically separated, as shown in Fig. 6. This yields the "circular- geometry" twin VCO design 6 that will be actually used in the first embodiment of the present invention. Here, coils L11 and L22 are physically separated from coils L21 and L22 by the VCO cores V1 and V2 in the center between the coils. The overall design is perfectly symmetric. In principle, it is possible to place further circuit blocks vertically between the two VCO cores V1 and V2 without destroying the perfect symmetry of the setup.

This circular-geometry twin VCO design enables a highly efficient generation of an AC magnetic field for excitation of the sample spins by the coils (or, more generally, inductors) L11 , L12, L21 and L22, while at the same time providing a reference (LO) signal for detection at its outputs VOut+ and VOut-.

Fig. 7 shows a possible CMOS implementation of a low-noise amplifier (LNA), as discussed, e.g., in Yang et al., Topical Meeting on Silicon Monolithic Integrated Circuits in RF Systems 2001 , pp 12-17. The amplifier 7 is designed in a highly symmetric manner. It comprises two identical branches fed by a common current source formed by transistor Q75. The first branch comprises two transis- tors Q71 , Q72 connected in a cascode configuration in order to achieve sufficient gain. A first input electrode AIn+ is connected to the gate electrode of the first transistor Q71. A bias voltage may be applied via a bias electrode B through resistor R71. Input impedance matching is achieved by an input capacitance C71. The drain electrode of second transistor Q72 is connected to the supply voltage Vdd via a first matching inductance LM1 for output impedance matching. The drain of second transistor Q72 is also connected to the first output electrode AOut-. The second, identical branch (transistors Q73, Q74, capacitance C72, resistor R72, matching coil LM2) receives an input voltage at electrode AIn- and provides an amplified output voltage at electrode AOut+. The two branches together achieve a symmetric amplification of the voltage difference between terminals AIn+ and AIn- and provide an amplified signal as a voltage difference between terminals AOut+ and AOut-. The transistors and ca- pacitances together form an LNA core 70.

In contrast to standard preamplifiers usually employed in NMR or ESR, which usually have input and output impedances of 50 Ohms, the input and output impedance of the LNA employed here is typically much lower. The input imped- ance is directly matched to the impedance of the detection coil, while the output impedance matches the input impedance of the subsequent mixer.

Fig. 8 shows a possible CMOS implementation of a standard differential passive mixer 8, as discussed, e.g., in Crols et al., IEEE Journal of Solid-State Circuits, Vo. 30, 1995, pp. 1483-1492. The high-frequency (radiofrequency or microwave frequency) input signal to be mixed with the local oscillator signal is fed to the two input terminals RF+ and RF- of the mixer. The RF+ terminal is connected to the source electrode of two transistors Q81 and Q83, while the RF- terminal is connected to the source electrodes of transistors Q82 and Q84. The local oscil- lator signal from the twin VCO is fed to input terminals LO+ and LO-. Terminal LO+ is connected to the gate electrodes of transistors Q82 and Q83, while terminal LO- is connected to the gate electrodes of transistors Q81 and Q84. The drain electrodes of transistors Q81 and Q82 are connected to form the first output terminal Out+, while the drain electrodes of transistors Q83 and Q84 are connected to form the second output terminal Out-. The transistors are operated in the triode region for maximum linearity at small signal inputs. They are switched by the comparably large LO voltage. Capacitances C81 , C82 serve as low-pass filters which pass only the low-frequency component of the output signal, i.e., the difference frequency signal. A bias voltage may be applied at ter- minal B via resistors R81 , R82 connected to the output terminals. Transistors Q81 , Q82, resistor R81 and capacitance C81 together form a first half mixer, transistors Q83, Q84, resistor R82 and capacitance C82 together form a second half mixer. The two half mixers may be spatially separated without destroying the symmetry of the mixer.

Fig. 9 shows a complete block diagram of a fully integrated on-chip ESR spectrometer assembled from the components of Figs. 6 to 8. In the center, the core of a low-noise amplifier (LNA) of the type of Fig. 7 is provided. On each side of the LNA core along a first direction (in Fig. 9, above and below the LNA), the upper and lower half, respectively, of a mixer of the type of Fig. 8 is arranged. On the free sides of the half mixers along the first direction (above and below the half mixers), the VCO cores V1 and V2 of oscillators VCO1 and VCO2, respectively, are provided. The LNA core, the half mixers and the VCO cores together form the active section of the spectrometer setup.

On both sides of this active section along a second direction perpendicular to the first direction (to the left and right in Fig. 9), coils for excitation, detection and impedance matching are provided. On the first (left) side of the active section, coil L11 extends from VCO core V1 to a Vdd terminal at the left side of Fig. 9. Likewise, coil L22 extends from VCO core V2 to the same Vdd terminal. Together, these two coils form a first excitation coil. A first detection coil LD1 is provided inside this excitation coil.

Output impedance matching of the LNA is achieved by a pair of identical first matching coils LM11 and LM12 arranged outside the excitation coil (above and below the excitation coil in Fig. 9). Each of these matching coils is connected between an output terminal of the LNA core (which, at the same time, is one of the input terminals of the mixer) and a Vdd terminal. Thus, in electrical terms, the matching coils act as being connected in parallel, i.e., they act like a single inductance with half the inductance value of the individual coils LM11 and LM12. Together, they form the matching inductance LM1 of Fig. 7.

On the other side of the active section along the second direction (to the right in Fig. 9), an identical arrangement of coils is provided. In other words, a second excitation coil is formed by coils L12 and L21 connecting VCO cores V1 and V2, respectively, to supply voltage Vdd. A second detection coil LD2 is provided inside the excitation coil. A pair of identical second matching coils LM21 and LM22 are provided next to the second excitation coil and connected between the second output of the LNA and the supply voltage Vdd in order to provide output impedance matching for the LNA.

The detection coils LD1 and LD2 are connected with each other in an anti- series configuration. The two remaining terminals are connected to the inputs AIn+ and AIn- of the LNA. In this way, differential detection is achieved, i.e., only the difference between the induced voltages in coils LD1 and LD2 will be applied to the input terminals of the LNA. Each of the two LNA outputs AOut+ and AOut- is fed to the input RF+ and RF-, respectively, of one half mixer. The half mixers receive the Vout+ and Vout- outputs of the twin VCO at their LO+ and LO- inputs. The low-frequency outputs MOut+ and MOut- of the half mixers serve as the output of the spectrometer. Core and tune (T) terminals are provided for controlling the oscillators.

Altogether, this setup constitutes a complete spectrometer on a single chip, comprising a frequency generation section, coils for excitation and detection, and a detector section which includes an LNA and a mixer. At its output, the spectrometer provides low-frequency (audio) output signals which can be readily processed by (possibly remote) standard recording equipment like a transient recorder. The setup is almost completely axially symmetric around an axis through the center of the sheet and perpendicular to the sheet plane. This high symmetry leads to excellent stability, small offsets and an extremely low susceptibility to external disturbances. Due to the high precision of standard CMOS processes, symmetry can be preserved on the sub-micrometer scale when the setup is implemented in CMOS technology. The setup is therefore especially suitable for on-chip implementation, even though in principle the same setup could also be used for a conventionally, discrete implementation.

In the following, some considerations for the practical design of the CMOS implementation will be made with reference to Figs. 10 to 15.

Fig. 10 shows a preferred layout of the coils. Only the left arrangement of coils in Fig. 9 is shown. This arrangement has been rotated clockwise in the paper plane by 90 degrees relative to Fig. 9. The excitation coil, having terminals E1 and E2, is formed by only one single turn of conductive material. One half turn constitutes coil L11 , the other half turn coil L22, the point where these two half turns meet being connected to the supply voltage Vdd. Each half turn consists of three straight stretches arranged in a crescent shape. At one end, the two oppositely bent crescents are joined by another straight stretch (in Fig. 10, a horizontal stretch) connected to the supply voltage Vdd. The straight stretches are arranged along seven of the eight edges of an (almost regular) octagon. At their other ends, electrodes E1 and E2 are formed.

Since the excitation coil has only one single turn, it might be more appropriate to use the more general term "inductor" instead of the term "coil". However, for the purposes of this document, these terms are considered to be synonyms.

Inside the excitation coil, the detection coil LD1 is formed. It is likewise formed by one single turn of conductive material. Like the excitation coil, it is composed of seven straight stretches along the edges of an almost regular octagon. Its terminals are designated as D1 and D2.

For both the excitation and detection coils, high quality factors (Q factors) are desired. The excitation coil is loaded by the detection coil. In order to ensure oscillation and low phase noise, the excitation coils should have as high a Q factor as possible. The detection coil is loaded by the sample, which behaves like an equivalent ohmic resistance proportional to the inductance. For the de- tection to be sensitive to this resistive effect, the ohmic series resistance of the detection coil must be as low as possible. Therefore also the detection coil should have a high Q factor. For achieving a high Q factor, several aspects are important.

Outside of the excitation coil, to its left and right in Fig. 10, two identical matching coils LM11 and LM12 with terminals M1 , V1 , M2 and V2 are provided. Each of these coils is implemented as a multi-turn, multi-layered (in particular, two- or three-layered) coil. Fig. 11 illustrates how such a coil may be implemented in three conductive layers. Instead of one or two turns per layer, as shown in Fig. 11 , more turns per layer are possible. Such multi-layered coils are advantageous because a relatively large inductance is required for matching and be- cause this requirement is even doubled as the two coils act in parallel for reasons of symmetry. For details about the design, modeling and fabrication of such coils, reference is made to Tang et al., IEEE Journal of Sold-State Circuits 37, 471-480 (2002). The arrangement of the matching coils on both sides of the excitation coil has been chosen in a manner to minimize cross couplings be- tween the matching coils and the excitation and detection coils as much as possible. Cross talk is further minimized by the fact that the two matching coils have opposite senses of current flow, which cancels cross talk to a very good approximation.

Fig. 12 shows an equivalent circuit diagram for modeling the behavior of the coil network formed by the coils L11, L12, LD, LM11 and LM12. The excitation coil formed by coils L11 and L22 is modeled by an inductance LE, a capacitance C_E and an ohmic (parasitic) resistance RE due to imperfect isolation of the coil trace. Likewise, the detection coil LD is modeled by an inductance LD, a capaci- tance CD and an ohmic (parasitic) resistance RD. The matching coil is modeled by an inductance LM, capacitance CM and ohmic parasitic resistance RM, with capacitances at the supply terminals V1 and V2 being approximated by zero and ohmic parasitic resistance at these terminals being approximated by infinity (i.e., assuming perfect electrical isolation of the coil traces). Since the coils are spatially close to each other, couplings exist. These couplings are modeled by a capacitance CED and an inductive coupling KED between excitation and detection coil, and by a capacitance C_EM and an inductive coupling KEM between the excitation coil and each matching coil. With these parameters, the electric behavior of the coil network can be modeled to a good approximation.

For on-chip implementation of the coil network, S-parameters for the target frequency range were extracted by a 2.5-D field solver ("Momentum RF", Agilent Technologies, Palo Alto, US). LSE optimization was applied to match the S- parameter values derived from the equivalent circuit of Fig. 12 to S-parameter values obtained via a field solver. The final error in S-parameter matching was below 0.1%. The obtained model could be shown to be valid within a range of a few hundred MHz. This bandwidth is rather small compared to the typical operation frequencies of the device (here in the 6 to 8 GHz range). On the other hand, the bandwidth is sufficient since the bandwidth of the device is limited already by the tuning range of the VCO frequency, which is in the 300 to 400 MHz range for the chosen implementation. Thus, the design parameters of the coil network could be modeled and optimized for a sufficient bandwidth and to sufficient accuracy.

Two spectrometers operating at different frequencies were designed with the aid of such optimization. The first spectrometer was designed for operation at 6 GHz (i.e., between the S band and the X band). The second spectrometer was designed for operation at the lower end of the X band around 8 GHz. The initial reason for first implementing a 6 GHz spectrometer was that the correctness of the simulation results could not be guaranteed over 6 GHz due to lack of verified device simulation models. Therefore, the actual performance of the first spectrometer operating at 6 GHz could be best expected to be close to the simulated performance, while the second spectrometer carried a higher risk of failure, as the performance of the second spectrometer could only be determined from on-chip measurements.

Both spectrometers were implemented in a standard mixed signal 0.35 μm CMOS process (AMS). In such a process, transistor performance is well characterized up to 6 GHz, and it is known that NMOS transistors have sufficient gain even at 9 GHz. The process also offers advanced options such as thick metal layers, metal-metal capacitors, high-Q varactors, SiGe npn bipolar tran- sistors, etc. In principle, however, other CMOS processes may be employed instead, such as the more expensive DSM (deep sub-micron) process, as long as these processes allow for the fabrication of high-Q factor coils and of transis- tors having sufficient gain in the desired frequency range.

The excitation and detection coils were etched from a "thick" metal layer (metal- 4 layer of the AMS process, 2.5 micrometers thickness) in order to obtain a high quality factor. For achieving a high Q factor, good conductivity and good substrate isolation are important in order to achieve a low series resistance and a high resistance to the substrate. Also coil geometry can be quite critical for achieving a high Q factor, and this geometry was accordingly optimized. Figs. 13 and 14 show the final geometry of the excitation and detection coils LE and LD for the 6 GHz and 8 GHz implementations, respectively. Q factors in the range of 12...15 were achieved for such coils. The trace widths b1 , b4 and inner radii b2, b3 of the excitation and detection coils, respectively, are given in Table 1 below, together with further design parameters of both implementations.

Table 1: Design parameters for two implementations of an on-chip ESR spectrometer

Fig. 15 shows a micrograph of a CMOS die containing two spectrometers, the upper one for operation at 6 GHz, the lower one for operation at 8 GHz. A prototype of this chip was manufactured and tested. Fig. 16 shows a spectrum obtained from a standard DPPH sample placed on one of the two detection coils of the 6 GHz spectrometer, Fig. 17 shows a corresponding spectrum obtained with the 8 GHz spectrometer. The spectra are not pure absorption spectra due to slight phase differences between LNA and VCO outputs. However, this does not influence the ESR detection performance. For each spectrum, the VCO frequency was kept fixed, the external magnetic field was swept, and the output signal at the output of the AF amplifier was monitored. The spectra demonstrate the functioning of both spectrometers. From such measurements, the sensitivity was determined. The sensitivity of the 6 GHz spectrometer turned out to be in the range of 2... 3 x 10¹⁰ spins/Gauss, depending on actual operating conditions, while the sensitivity of the 8 GHz spectrometer was even slightly better. These figures are comparable to the sensitivity achieved in commercial spectrometers. With further optimization, in particular, careful matching and higher microwave gain, sensitivity may be expected to improve by up to about an order of magnitude, thus even exceeding the sensitivity of today's commercial X-band spectrometers.

For both spectrometers, the frequency range can be adjusted by adjusting the core and tune voltages at terminals "Core" and "T" of the twin VCO. Table 2 shows the relationship between core voltage and frequency range as well as rf magnetic flux density Bi as determined for the two spectrometers.

Table 2: Influence of core voltage on frequency range and rf magnetic field

Fig. 18 shows a CMOS die for a second embodiment of an absorption-type spectrometer according to the present invention. The design is mirror-symmetric rather than axially symmetric. Again, a circular-geometry twin VCO as in Fig. 6 was chosen for frequency generation. The coils L11 , L12, L21 and L22 of the VCO are indicated in Fig. 18, coils L11 and L22 forming the first (left) excitation coil and coils L12 and L21 forming the second (right) excitation coil. The detection coils LD1 and LD2 are again disposed within the excitation coils. In contrast to the first embodiment, the connecting traces of each detection coil are ar- ranged diagonally, in Fig. 16 towards the center and bottom of the Figure, crossing the traces of the excitation coil. They are fed to an improved 3-stage bipolar LNA. Output matching of the LNA is achieved by matching coils LM 1 and LM2. In comparison to the first embodiment, only one matching coil instead of two matching coils connected in series is employed for each LNA branch. Therefore, the matching coils can be kept smaller. They are spatially removed from the excitation coil and therefore have decreased cross-talk to the excitation and detection coil system. The LNA output is fed to an improved, low-noise mixer having optimized transistor sizes and bias voltage and from there to a low-frequency (AF) amplifier. This setup is implemented on a chip of only 2 x 2 mm² size.

The LNA was implemented as a three-stage amplifier in SiGe BiCMOS technology, as suggested, e.g., in one of the following references: Winkler et al., Proc. SPIE Second International Symposium on Fluctuations and Noise, Maspalo- mas, Gran Canaria (Spain), May 2004, vo. 5470, Noise in Detectors and Circuits II, pp. 185-192; Wang et al., Proc. IEEE International Radar Conference 2005, pp 27-30; Plouchart et al., Proc. IEEE Custom Integrated Circuits Conference 1999, pp. 217-220. Such an LNA has higher gain and lower noise figures at GHz than a pure MOS LNA.

In the following, a dispersion-type spectrometer constructed on the basis of similar principles as the absorption-type embodiments discussed above will be discussed with reference to Figs. 19 to 26. In this device, magnetic resonance is not detected through energy absorption of the sample on resonance, as in an absorption-type spectrometer, but through a change of the real part of the (complex) AC magnetic susceptibility on resonance, which is a dispersion ef- feet. Therefore, this type of spectrometer is called a dispersion-type spectrometer.

The overall setup is illustrated schematically in Fig. 19. A first VCO designated as VCO-A is formed by a current source 401 and a VCO core 402 (including a varactor), connected by a first pair of excitation coils LA1 and LA2. Likewise, a second VCO designated as VCO-B is formed by a current source 401 ' and a VCO core 402', connected by a second pair of excitation coils LB1 and LB2. The coil pairs are arranged overlapping with each other in a manner that they enclose a common region, into which the sample 403 is placed. The outputs of both VCOs are fed to a mixer 404. The mixer output is fed to a high-frequency buffer (HFB) 405 followed by a programmable frequency divider 406 for dividing the frequency by a value M. The output of the frequency divider constitutes the spectrometer output FOut.

In operation, the spectrometer is placed into an external magnetic field. One of the oscillators, e.g., VCO-A, is tuned to a first frequency close to, but sufficiently away from the electron Larmor frequency in the given external field such that no resonance occurs yet. The other oscillator (here, VCO-B) is tuned to a second frequency, the frequency difference exceeding the spectral range to be investi- gated. Typically, the frequency difference will be set to a few MHz up to a few hundreds of MHz at an operating frequency of several GHz. This frequency difference is extracted at the output of mixer 403. The frequency-difference signal is shaped into a square-wave (binary) signal by the high-frequency buffer 404 and divided down by frequency divider 405 to a frequency which can easily be counted by commercial frequency counters.

The field is now swept through the resonance. The expected response of the output frequency to a sweep of the magnetic field Bo is illustrated in Fig. 20. Away from resonance, the output frequency Δv at terminal FOut will be constant at the initially chosen difference frequency Δvo, independent of the value of the magnetic field B₀. When the magnetic field Bo approaches the magnetic reso- nance at the frequency of VCO-A, however, the inductance of VCO-A will be changed by the resonating sample, which will lead to a change in the eigenfre- quency of the oscillator VCO-A. On the other hand, oscillator VCO-B, having a different eigenfrequency, will be unaffected by the resonance around the eigen- frequency of VCO-A. Therefore, the frequency difference between the two oscil- lators will change. This frequency difference will result in a frequency change at the output terminal FOut, as shown in the left trace of the resonance curve in Fig. 20. This frequency change may be readily recorded by a frequency counter, and thus a resonance spectrum may be obtained. When the magnetic field is further swept through the condition at which the Larmor frequency matches the frequency of VCO-B, a second resonance occurs, which will lead to a frequency change as shown as the right trace of the resonance curve in Fig. 20. These traces are mirror-symmetric, and either one of these traces represents the complete (dispersion) spectrum.

The two oscillators VCO-A and VCO-B are arranged in a geometrically orthogonal configuration. In particular, the respective inductors LA1 , LA2 and LB1 , LB2, respectively, are pairwise orthogonal to each other. This minimizes magnetic and capacitive couplings between the two oscillators despite the overlap of the inductor loops. Such couplings are undesired because they may lead to a phenomenon called injection-locking. Injection-locking is a concept of locking the signal generated by one oscillator to one of the harmonics of the other oscillator via magnetic or capacitive couplings. This phenomenon is undesired in the present case, and the oscillators should behave independently of each other.

In the following, ways of implementing such a dispersion-type spectrometer on a single chip in CMOS technology will be discussed. Fig. 21 shows a possible design of the first VCO, called VCO-A. A standard CMOS VCO setup is employed. The drain and gate electrodes of (NMOS) transistors Q25 and Q26 are mutually connected. Their source electrodes are connected to ground. The drain electrodes are fed from a common current source constituted by a (PMOS) transistor Q27 via coils LA1 and LA2. These electrodes also constitute the output terminals VAOut+ and VAOut-. Two varactors are connected in parallel between these two electrodes, each varactor being formed by a transistor pair having shorted and interconnected gate and drain electrodes. The common node of the first pair, Q21 and Q22, is connected via a resistor to ground, thus yielding a fixed first capacitance. The common node of the second pair, Q23 and Q24, may be provided with a tuning voltage at tune electrode TA. In this manner, VCO-A may be tuned within a range of a few hundred MHz. Transistors Q21 to Q26 form the VCO core, while the current source constituted by transistor Q27 is spatially separated from this core by the traces of coils LA1 and LA2.

Fig. 22 shows a possible design of the second VCO, called VCO-B. This VCO is almost identical to the first VCO, except that the common node of the first pair of transistors, Q21¹ and Q22¹, is connected via a resistor to the supply voltage Vdd rather than to ground. Thus, a capacitance results which is different from the capacitance of the corresponding varactor in the first VCO. In this manner, the frequency range to which VCO-B may be tuned is different from the frequency range of VCO-A.

Fig. 23 shows a CMOS implementation of a Gilbert mixer suitable to be employed in the present embodiment. Such a mixer design is known in the art, see, e.g., Gilbert, IEEE Journal of Solid-State Circuits, Vol. 32, 1997, pp. 1412- 1423 and Harvey et al., IEEE International Symposium on Circuits and Systems 2001 , vol. 4, pp. 786-798.

Fig. 24 shows a CMOS implementation of a standard high-frequency buffer suitable to be employed in the present embodiment. Such a design is well known in the art.

Fig. 25 shows a block diagram of a programmable frequency divider. A divide- by-four block HF%4 may be enabled or disabled by a select voltage at selection terminal S. The input frequency is provided to terminals D+ and D-. Depending on the select voltage, the input frequency is either divided by four or passed to a chain of TSPC (true single-phase clocked) divide-by-two blocks %2. A possible CMOS implementation of a ring-counter type divide-by-four block is illustrated in Fig. 26. The input signal is provided at terminals C+ and C-. The output is pro- vided at terminals Q+ and Q-. Terminals X+, X-, Y+, Y-, Z+ and Z- are connected as indicated to form a CMOS ring counter. Such a ring counters are known in the art.

Fig. 27 shows a CMOS die for a thus implemented dispersion-type spectrome- ter. The coils LA1 , LA2, LB1 and LB2 are indicated. Each such coil is formed by a single, substantially straight strip of conducting material, where the term "coil" is again to be understood in a broad manner as meaning "inductor". The conducting strips of coil pair LA1 and LA2 are parallel to each other with a distance of 100 micrometers and a width of 20 micrometers. The likewise parallel con- ducting strips of coil pair LB1 and LB2 have the same dimensions and cross the strips of coil pair LA1 and LA2 at right angles. Together, the four coils enclose a square detection region R in which the sample is to be placed.

The present design proposes that the areas enclosed by the coils of the two VCOs overlap to form a common detection region. While placing the sample exactly in the detection region leads to the best sensitivity, the exact sample location is not critical in this embodiment. By the way of example, even if the sample is larger than the square region, the spectrometer will still work, as long as there is sufficient sample material within or near that coil pair whose high- frequency current is at the resonance (Larmor) frequency of the sample, such that the sample will influence the inductance of that coil on resonance. An advantage of this design is that the device is relatively robust to a misplacement of the sample, and that it may readily be used with samples which exceed the size of the chip. A spectrometer designed in this manner may thus be used for applications such as scanning the surface of a large sample, for gas-phase detection of radicals, for detection of radicals in liquids into which the (sealed) spectrome- ter is immersed, etc. Such applications are not as easily possible with the absorption-type design of the first and second embodiments, since there an exact placement of the sample into the area of one of the excitation and detection coils is required, while no sample should be present in the area of the other detection coil.

It should be noted that the device will still work if the coil pairs LA1 , LA2 and LB1 , LB2, respectively, do not overlap. They may even be spatially separated. Resonance will be detected as long as some sample material is in the vicinity of at least one of the coil pairs such as to influence the inductance of that coil pair on resonance. However, having overlapping coils has the advantage that any disturbances in the region enclosed by one coil pair (e.g., external fluctuating magnetic fields etc.) will also influence the other coil pair in the same manner and will therefore be cancelled at the output of the spectrometer. This significantly improves the stability, reliability and noise figure in practical applications.

While the invention has been described with reference to embodiments for electron spin resonance, all embodiments of the invention may equally well be employed for nuclear magnetic resonance. In this case, however, the frequencies for excitation and detection will be generally significantly lower. In particular, with presently available magnets, these frequencies will normally be in the range below about 2 GHz (the field of the strongest superconducting magnets commercially available at the present time corresponds to a proton Larmor frequency of 900 MHz, significantly higher fields presently being obtainable only by hybrid systems). For many interesting applications, the frequencies will be in the range of only a few kHz to a few tens of MHz. For such applications, coils with larger inductances are required, and it might thus be more appropriate to use coils with more than one turn. Such modifications are, however, readily possi- ble. In addition, the concepts of symmetrizing the frequency generation and detection scheme applied above are equally well applicable to NMR as to ESR.

While the invention has been described with reference to specific embodiments, it is to be understood that the invention is by no means limited to these embodiments. Many modifications are possible. Just to name a few variants, other types of oscillators, amplifiers or mixers may be used, as they are well known in the art of analog integrated circuits. While three particularly advantageous absorption- and dispersion-mode designs have been described in detail, other de- signs are possible that integrate the functionalities of frequency generation and detection on the same chip, including designs with lower or even no symmetry. On the other hand, the symmetry concepts described herein are also useful in magnetic resonance devices that are not implemented on a single chip. While the invention has been described with reference to spectrometer embodiments, the devices are easily modified to be used as part of a magnetometer or of a sensor for paramagnetic species. In this spirit, many further modifications are possible.

List of reference signs

101 , 101" Source

102, 102' Circulator

103, 103' Cavity

104, 104' Load

105 Differential detector

Out Output terminal

B₀ Static magnetic field

201 Local oscillator

202 Power amplifier

203 Excitation coil

204 Sample

205, 205' Detection coil 206 Low-noise amplifier

207 Mixer

208 Audio-frequency amplifier

LO Local oscillator

PA Power amplifier

LNA Low-noise amplifier

AF Audio-frequency amplifier

300 CMOS LC-type voltage-controlled oscillator (VCO)

301 CMOS active section

302 Varactor

303 Excitation coil

304 Sample

305, 305" Detection coil

306 Low-noise amplifier (LNA)

307 Mixer

308 Audio-frequency amplifier

VCO1. VCO2 Voltage-controlled oscillator

V1 , V2 VCO core

L11 , L12, L21 , L22 Inductance (excitation coil)

C11 , C12 Varactor

Q11. Q12, Q13 Transistor

T Tune electrode

Core Core electrode

Vdd Supply voltage

VOut+, VOut- VCO output terminal

7 LNA

70 LNA core

AIn+, AIn- LNA input terminal

AOut+, AOut- LNA output terminal B Bias electrode

LM1 , LM2 Matching coil

Q71 , Q72, Q73, Q74, Q75 Transistor

R71 , R72 Resistor C71 , C72 Capacitance

8 Mixer

LO+, LO- Mixer LO input terminal

RF+, RF- Mixer rf input terminal

MOut+, MOut- Mixer output terminal

Q81 , Q82, Q83, Q84 Transistor

R81. R82 Resistor

C81 , C82 Capacitance

LDI , LD2 Detection coil

LM11 , LM12, LM21 , LM22 Matching coil

E1 , E2 Terminal of excitation coil

D1 , D2 Terminal of detection coil M1 , M2, V1 , V2 Terminal of matching coils

RD, RE, RM Resistance (equivalent circuit) CD, C_E, CM, C_ED, CEM Capacitance (equivalent circuit)

LD. L_E, LM Inductance (equivalent circuit) KM, KED Coupling (equivalent circuit)

LE Excitation coil

LD Detection coil b1 , b4 Trace width b2, b3 Inner radius

Signal LM1 , LM2 Matching coil

VCO-A, VCO-B Voltage-controlled oscillator LA1. LA2, LB1. LB2 Excitation coil

HFB High-frequency buffer

%M Frequency divider

FOut Frequency output terminal

401 , 401 ' Current source 402, 402' VCO core

403 Sample

404 Mixer

405 High-frequency buffer

406 Frequency divider

VAOut+, VAOut- Output terminal

VBOut+, VBOut- Output terminal

TA, TB Tune electrode Q21 , Q22, Q23, Q24, Q25, Q26, Q21 ', Q22¹ Transistor

RF+, RF- Radiofrequency input terminal

A+, A-, B+, B- Mixer input terminal

BIn+, BIn- HFB input terminal BOut+. BOut- HFB output terminal

NR2 NOR gate

HF%4 High-frequency divide-by-four block

IV Inverter %2 TSPC Divide-by-two block

D+, D- Divider input terminal

S Divider selection terminal X+, X-, Y+, Y-, C+, C- Terminal

Claims

Patent claims

1. A magnetic resonance device, comprising - an inductor arrangement (LA1 , LA2, LB1 , LB2) for generating a high-frequency magnetic field at a sample location and for receiving a response of a sample (403) placed in the sample location to said high-frequency magnetic field; a first and a second oscillator (VCO-A, VCO-B), said first oscillator (VCO-A) being adapted for generating a current in said inductor arrangement at a first oscillator frequency and said second oscillator (VCO-B) being adapted for generating a current in said inductor arrangement at a second oscillator frequency different from said first oscillator frequency; and a detection circuit (404, 405, 406) for detecting said response of said sample and for providing an electrical output signal that reflects said response, wherein said detection circuit is adapted for detecting a change of a difference frequency between said first and second oscillator frequencies due to a change of inductance in said inductor arrangement when magnetic resonance in said sample occurs.

2. The magnetic resonance device of claim 1 , wherein said inductor arrangement, said oscillator device and said detection circuit are part of an integrated circuit on a single semiconductor substrate.

3. The magnetic resonance device of claim 2, wherein said integrated circuit is a CMOS-type integrated circuit.

4. The magnetic resonance device of one of claims 1 to 3, wherein said inductor arrangement includes a first and a second inductor (LA1 , LA2; LB 1 , LB2), said first oscillator (VCO-A) being adapted for generating a current in said first inductor (LA1 , LA2) and said second oscillator (VCO-B) being adapted for generating a current in said second inductor (LB1. LB2).

5. The magnetic resonance device of claim 4, wherein each of said first and second oscillators (VCO-A, VCO-B) is an LC-type oscillator, each of said first and second inductors (LA1 , LA2; LB1 , LB2) being part of a resonant circuit of one of said LC-type oscillators.

6. The magnetic resonance device of claim 4 or 5, wherein said first inductor (LA1 , LA2) has a first active area, wherein said second inductor (LB1 , LB2) has a second active area, and wherein said first and second active area overlap to form the sample location.

7. The magnetic resonance device of one of claims 4 to 6, wherein said first inductor (LA1 , LA2) and said second inductor (LB1 , LB2) are arranged in a substantially symmetric configuration around an axis or plane of symmetry.

8. The magnetic resonance device of one of claims 4 to 7, wherein said first and second inductors (LA1 , LA2; LB1 , LB2) are identical in shape and arranged in a substantially orthogonal configuration.

9. The magnetic resonance device of one of claims 1 to 8, wherein said first and second oscillators (VCO-A, VCO-B) are arranged in a substantially symmetric configuration around an axis or plane of symmetry.

10. The magnetic resonance device of one of claims 1 to 9, wherein said detecting circuit comprises a mixer (404) having first input terminals (A+, A-) receiving an output signal of said first oscillator (VCO-A), said mixer having second input terminals (B+, B-) receiving an output signal of said second oscillator (VCO-B), and said mixer having output termi- nals (MOut+, MOut-) providing a signal at said difference frequency, said difference frequency being the difference of frequencies provided at said first and second input terminals.

11. The magnetic resonance device of claim 10, wherein said detecting circuit further comprises a high-frequency buffer (405) for converting said signal at said difference frequency into a binary signal.

12. The magnetic resonance device of claim 10 or 11 , wherein said detect- ing circuit further comprises a frequency divider (406) providing an output signal at a frequency corresponding to said difference frequency divided by a predetermined factor.

13. The magnetic resonance device of one of the preceding claims, wherein said device is adapted to detect electron spin resonance.

14. A method of operation of a magnetic resonance device according to one of claims 1 to 13, comprising the steps of placing a sample (403) in said sample location and placing the de- vice in an external magnetic field (Bo); operating said first oscillator (VCO-A) at said first oscillator frequency and said second oscillator (VCO-B) at said second oscillator frequency; sweeping said external magnetic field through a magnetic reso- nance of said sample; monitoring said difference frequency as a function of said external magnetic field (Bo) to obtain a spectrum.

15. A magnetic resonance device, comprising - an inductor arrangement (L11 , L12, L21 , L22, LD1 , LD2) for generating a high-frequency magnetic field at a sample location and for receiving a response of a sample placed in the sample location to said high-frequency magnetic field due to magnetic resonance of paramagnetic species in said sample, said inductor arrangement including a first excitation coil (L11 , L22), a second excitation coil (L12, L21) having substantially identical characteristics as said first excitation coil and being spatially removed from said first excitation coil, a first detection coil (LD1) being magnetically coupled to said first excitation coil, and a second detection coil (LD2) having substantially identical characteristics as said second detection coil and being magnetically coupled to said second excitation coil, wherein said sample location is closer to said first detection coil than to said second detection coil; an oscillator device (VCO1 , VCO2) operable to provide high- frequency currents at the same frequency to said first and second excitation coils; - a detection circuit (7, 8) for detecting said response of said sample, wherein said detection circuit is adapted for detecting a difference of induced voltages in said detection coils.

16. The magnetic resonance device of claim 15, wherein said inductor ar- rangement, said oscillator device and said detection circuit are part of an integrated circuit on a single semiconductor substrate.

17. The magnetic resonance device of claim 16, wherein said integrated circuit is a CMOS-type integrated circuit.

18. The magnetic resonance device of one of claims 15 to 17, wherein said detection coils (LD1 , LD2) are connected in an anti-series configuration.

19. The magnetic resonance device of one of claims 15 to 18, wherein said oscillator device comprises a first and a second LC-type oscillator

(VCO1 , VCO2), each LC-type oscillator having a first and a second output terminal (VOut+, VOut-), said first output terminals of said first and second oscillators being connected with each other and said second output terminals of said first and second oscillators being connected with each other, each LC-type oscillator comprising a first and a second inductor (L11 , L12; L21 , L22) being part of a resonant circuit of said LC- type oscillator, said first inductor (L11 ) of said first oscillator (VCO1 ) and said second inductor (L22) of said second oscillator (VCO2) being connected in a series configuration to form said first excitation coil, and said second inductor (L12) of said first oscillator (VCO1) and said first inductor (L21 ) of said second oscillator (VCO2) being connected in a series configuration to form said second excitation coil.

20. The magnetic resonance device of one of claims 15 to 19, wherein said oscillator device (VCO1 , VCO2) comprises an oscillator core (V1 , V2) constituted of semiconductor elements in said oscillator device, wherein said first excitation coil (L11 , L22) and said first detection coil (LD1 ) form a first group of coils, wherein said second excitation coil (L21 , L12) and said second detection coil (LD2) form a second group of coils, and wherein said first and second groups of coils are arranged in a axially or mirror-symmetric configuration with respect to said oscillator core (V1 , V2).

21. The magnetic resonance device of claim 20, wherein said first detection coil (LD1) is arranged within said first excitation coil (L11 , L22), and wherein said second detection coil (LD2) is arranged within said second excitation coil (L12, L21 ).

22. The magnetic resonance device of one of claims 15 to 21 , wherein said detection circuit comprises a mixer (8) having first input terminals (RF+, RF-) receiving signals from output terminals of said oscillator device, said mixer further having second input terminals (LO+, LO-) receiving signals derived from said detection coils, and said mixer having output terminals (MOut+, MOut-) providing a signal at a difference frequency, said difference frequency being the difference of frequencies provided at said first and second input terminals.

23. The magnetic resonance device of claim 22, further comprising an am- plifier circuit (7) receiving a difference signal from said first and second detection coils (LD1, LD2) and supplying an amplified difference signal to said mixer (8).

24. The magnetic resonance device of claim 23, wherein said first and sec- ond detection coils (LD1 , LD2) are connected in an anti-series configuration to form a detection coil system, said detection coil system having a predetermined impedance, wherein said amplifier circuit (7) has an input impedance matching said predetermined impedance, and wherein said amplifier circuit (7) has an output impedance matching an input im- pedance of said mixer (8).

25. The magnetic resonance device of claim 24, wherein said amplifier circuit comprises at least one pair of matching coils (LM1 , LM2; LM11 , LM12, LM21 , LM22) for adjusting its input and/or output impedance, said matching coils being arranged in a symmetric manner with respect to said inductor arrangement.

26. A method of operation of a magnetic resonance device according to one of claims 15 to 25, comprising the steps of - placing a sample in said sample location and placing the device in an external magnetic field (B₀); operating said oscillator device to provide high-frequency currents to said first and second excitation coils (L11 , L22; L21 , L12); sweeping said external magnetic field (B₀) at constant frequency of said oscillator device or sweeping said frequency of said oscillator device at constant external magnetic field (B₀) through a magnetic resonance of said sample; monitoring said electrical output signal of said detection circuit to obtain a spectrum.