Journal of Magnetic Resonance 247 (2014) 96–103
Contents lists available at ScienceDirect
Journal of Magnetic Resonance journal homepage: www.elsevier.com/locate/jmr
Cryogenic single-chip electron spin resonance detector Gabriele Gualco, Jens Anders, Andrzej Sienkiewicz, Stefano Alberti, László Forró, Giovanni Boero ⇑ Ecole Polytechnique Federale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
a r t i c l e
i n f o
Article history: Received 6 June 2014 Revised 12 August 2014 Available online 8 September 2014 Keywords: ESR Cryogenic CMOS
a b s t r a c t We report on the design and characterization of a single-chip electron spin resonance detector, operating at a frequency of about 20 GHz and in a temperature range extending at least from 300 K down to 4 K. The detector consists of an LC oscillator formed by a 200 lm diameter single turn aluminum planar coil, a metal–oxide–metal capacitor, and two metal–oxide–semiconductor field effect transistors used as negative resistance network. At 300 K, the oscillator has a frequency noise of 20 Hz/Hz1/2 at 100 kHz offset from the 20 GHz carrier. At 4 K, the frequency noise is about 1 Hz/Hz1/2 at 10 kHz offset. The spin sensitivity measured with a sample of DPPH is 108 spins/Hz1/2 at 300 K and down to 106 spins/Hz1/2 at 4 K. Ó 2014 Elsevier Inc. All rights reserved.
1. Introduction
2. Operating principle
Methods based on the electron spin resonance (ESR) phenomenon are used to investigate samples in a wide temperature range, ranging from above 1000 K [1–4] to below 1 K [5,6]. Low temperature measurements are usually performed in large microwave cavities as well as with miniaturized conductive [7,8] or superconducting [9–11] resonators. Miniaturized resonators are typically used to maximize the signal-to-noise ratio in experiments with mass-limited samples [7,12–17]. In Refs. [13,18,19] we presented single-chip integrated inductive ESR detectors, fabricated using complementary metal oxide semiconductor (CMOS) technologies, operating between 8 GHz and 28 GHz. The ESR phenomenon was detected as a variation of the frequency of an integrated LC-oscillator due to an effective variation of its coil impedance caused by the resonant complex susceptibility of the sample. Operation in the temperature range from 300 K to 77 K was demonstrated in Refs. [18,19]. Here, we report on the implementation of a 20 GHz single-chip ESR detector, based on the same operating principle, but capable of operating from 300 K down to at least 4 K. Depending on the specific sample (and, in particular on the dependence of its relaxation times on temperature), the possibility to operate down to 4 K might represent a significant advantage in terms of spin sensitivity (larger polarization, lower thermal noise) as well as in terms of information richness. The fabricated device represents also the first demonstration of a CMOS microwave oscillator operating down to 4 K.
The principle of operation of the realized single-chip ESR detector is identical to that reported in our previous work Refs. [13,18,19]. In typical experimental conditions, the oscillation frequency of an LC-oscillator coupled with an ensemble of electron spins is given by [19]
⇑ Corresponding author. Address: Ecole Polytechnique Fédérale de Lausanne (EPFL), Station 17, CH-1015 Lausanne, Switzerland. E-mail address: giovanni.boero@epfl.ch (G. Boero). http://dx.doi.org/10.1016/j.jmr.2014.08.013 1090-7807/Ó 2014 Elsevier Inc. All rights reserved.
xLC xLC v ffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ gv0
ð1Þ
where
v0 ¼
ðxLCv x0 ÞT 22 1 x0 v 0 2 1 þ T 22 ðxLCv x0 Þ2 þ c2e B21 T 1 T 2
ð2Þ
pffiffiffiffiffiffi is the real part of the magnetic susceptibility, xLC ¼ 1= LC is the unperturbed oscillator frequency, x0 = ceB0, v0 is the static magnetic susceptibility, g is the filling factor (approximately given by (Vs/Vc), where Vs is the sample volume, and Vc is the coil sensitive volume), ce is the electron gyromagnetic ratio, B1 is the microwave magnetic field, T1 and T2 are the relaxation times. As discussed in Ref. [13], the oscillator frequency variation due to the electron spin resonance phenomenon in the sample is, in first approximation, given by DxLC v ffi ð1=2ÞxLC gv0 . Consequently, the oscillator frequency variation is proportional to the real part v0 of the sample complex susceptibility v ¼ v0 jv00 . The shape of the oscillator frequency variation is thus identical to that of the dispersion signal measured in conventional continuous wave experiments. However, the dependence on the microwave field B1 is substantially different. In the conventional amplitude detection method the measured signal is linearly proportional to the precessing magnetization (i.e., to the product v00 B1 for the absorption and v0 B1 for the dispersion). In
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
97
the frequency detection case, for B1 below saturation the signal amplitude is independent of the value of B1 and decreases as 1=B21 in saturation (in the conventional method the signal is proportional to B1 below saturation and decreases as 1/B1 in saturation). Neglecting the noise contribution of the active devices in the oscillator feedback and assuming that the frequency noise spectral density (in Hz2/Hz) is only due to the thermal noise of the coil resistance and given by Sm ¼ kTRx2LC =ð2pÞ2 V 20 , the spin sensitivity (in spins/Hz1/2) is given by [18]
Nmin
pffiffiffi sffiffiffiffiffi T 3=2 R T 1 ; ffia 2 T2 B0 Bu
ð3Þ
where V0 is the oscillation amplitude, R is the coil series resistance, Bu is the coil unitary field, T is the coil (and sample) temperature, and a ffi 20 m1 kg5/2 s4 K3/2 A3. 3. Description of the single-chip detector Fig. 1 shows a photo and the block diagram of the realized chip, which consists of two LC-oscillators operating at about 21 GHz and 17 GHz, a mixer, and a frequency division module. The chip is manufactured in a commercially available 130 nm CMOS technology (IBM 8RF). The total chip surface, including the bonding pads, is about 1 mm2. The total power consumption of the chip is about 20 mW at 300 K and about 6 mW at 4 K, with the oscillators bias currents set to the minimum values which enable stable oscillations. As in our previous designs [13,19], with the exception of that reported in [18], we have integrated two oscillators to perform the first down-conversion of the oscillator frequency by a mixer. The use of two oscillators and a mixer as first step in the frequency downconversion requires a frequency divider operating at 4 GHz instead of 20 GHz, which is significantly less difficult to design. The excitation/detection octagonal coils of the oscillators have an external diameter of 200 lm, a metal width of 30 lm, a metal thickness of 7.5 lm (obtained by parallel connection of the three top metal layers available in the process, two made of aluminum and one of copper), and an inductance of about 300 pH. The capacitor is realized using interdigitated aluminum fingers. The capacitance value for the two oscillators are 180 fF and 260 fF, respectively. The microwave magnetic field B1 can be varied from 0.08 to 0.14 mT by changing the oscillator bias current Idc from 1 to 5 mA. At low temperatures, a slightly lower bias current of about 0.5 mA is sufficient to guarantee stable oscillations. The lower limit is determined by the minimum transconductance of the cross-coupled pair required for stable oscillation whereas the upper limit is given by the maximum voltage swing which can be applied across the transistors in the cross-coupled pair without damaging their thin gate oxides. The value of B1 generated by our single-chip detector is estimated by measuring the voltage at the oscillator bias node VIdc. In condition of stable oscillation the oscillation amplitude V0 ffi VIdc [20]. Hence, B1 ffi (1/2)Bu(VIdc/xLCL), where Bu ffi l0/d, d is the coil diameter, and L is the coil inductance. This B1 estimation is in agreement with saturation experiments with samples of known relaxation times. Fig. 1 shows also the schematics and block diagrams of the mixer and frequency divider. The frequency mixer needed for down-conversion of the oscillation frequency is based on a double-balanced Gilbert cell topology (see pages 368–370 of Ref. [21]). By mixing the two oscillator output voltages a signal at about 4 GHz is obtained. The sum frequency component is filtered out by the system parasitics (i.e., a low pass filter is not necessary). The frequency divider is realized by means of current model logic (CML) D-latches with resistive load (see pages 683–699 of Ref.
Fig. 1. (a) Microphotograph of the single-chip ESR detector. Idc1 and Idc2: DC power supply of the two oscillators (0.5–5 mA). VD: DC power supply for the mixer and frequency-division module (1.5 V). OUTp and OUTn: differential output signal (ffi220 MHz). The circle indicates the area where the samples are placed. (b) Block diagram of the single-chip ESR detector. (c) Schematics of the frequency mixer (Vb1 ffi 0.9 V, Vb2 ffi 0.6 V). (d) Block diagram of the frequency divider based on Dlatches. (e) Schematics of the D-latch (Vb ffi 0.6 V).
98
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
[21]). The on chip divide ratio is 16, by a cascade of four identical divide-by-two frequency dividers. The output buffer, realized as standard source follower amplifier, is capable of driving an output load of 5 pF with a voltage swing of 0.2 V at 200 MHz. The detector reported in this paper is an improved version of the one we reported in Ref. [18]. In the previous work, cooling to about 288 K was necessary to operate the detector, complicating the set-up and de facto preventing spectroscopy studies at samples temperatures above 288 K. The improved version works properly up to about 370 K. The total power consumption of the detector reported here is an order of magnitude smaller with respect to that reported in Ref. [18], allowing easier operation at low temperatures (the device reported in Ref. [18] can actually also operate down to 4 K but it requires a more efficient cooling). In order to assure operation down 4 K, i.e., well below the impurity freezeout temperature [22], we designed the integrated electronics using exclusively metal–oxide–semiconductor and degenerately doped structures.
spectrometers. The signal at the output of the frequency-to-voltage converter is demodulated by a lock-in amplifier. The lock-in internal reference signal is amplified and applied to the field modulation coils. Some experiments are performed without field modulation. In order to use the spectrometer lock-in also in these measurements, the signal at the output of the frequency-to-voltage converter is chopped using a switch controlled by the 100 kHz reference signal of the lock-in amplifier, as shown schematically in Fig. 2. The operation of CMOS devices and circuits at temperatures down to 4 K and below has been already investigated in detail [22–33]. However, no phase noise characterizations of CMOS LC-oscillators at cryogenic temperatures have been reported so far. Fig. 3 shows the square root of the measured frequency noise pffiffiffiffiffi spectral density (i.e., Sm in Hz/Hz1/2) referred to the integrated LC-oscillator output. At 300 K, the noise at 100 kHz offset from the carrier is about 20 Hz/Hz1/2. At 4.2 K, a minimum noise of about
4. Performance of the integrated detector Fig. 2 shows an illustration of the setup used to characterize the performance of the single-chip ESR detector. The chip is glued onto a printed circuit board and electrically connected by wire bonding. On the same printed circuit board a commercial amplifier is used as buffer to drive the coaxial cable carrying the chip output signal at about 200 MHz. The system operates properly in the range from 300 K down to at least 4 K (no measurements below 4 K have been performed yet). The printed circuit board is inserted in a dynamic continuous flow cryostat where the temperature is controlled by means of a temperature controller unit working with a thin film resistance temperature sensor. The signal at the output of the buffer is mixed, at room temperature, with a local oscillator signal having a frequency about 10 MHz higher. The signal at the output of the mixer (with a carrier frequency of about 10 MHz) is fed into a phase-locked loop (PLL). In most of the experiments, a magnetic field modulation at kHz frequencies is added to the static magnetic field to improve the signal-to-noise ratio as in ordinary ESR
Fig. 2. (a) Block diagram of the experimental set-up: (1) Electromagnet power supply (Bruker), (2) electromagnet (Bruker, 0–1.5 T), (3) single chip ESR detector, (4) buffer (Texas Instruments THS4304D), (5) magnetic field modulation coils, (6) power amplifier (Rohrer PA508), (7) mixer (Mini-Circuits ZAD-3), (8) signal generator (Rohde–Schwarz SMR-20), (9) phase locked loop (PLL) circuitry (see Ref. [18]), (10) RF Switch (Mini-Circuits ZYSW-2-50DR), (11) selection mode switch, (12) lock-in of the EPR spectrometer (Bruker Elexsys-II E500), (13) cryostat (Oxford Instruments CF935), (14) temperature controller (Oxford Instruments ITC503), (15) heater, and (16) temperature sensor (Lake Shore Cernox CX1050).
Fig. 3. Frequency noise spectral density of the integrated LC-oscillators at 300 K and at 4.2 K.
Fig. 4. Frequency difference between the two integrated LC-oscillators as a function of the bias current Idc2 at 300 K and at 4.2 K with Idc1 ffi 1 mA.
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
1 Hz/Hz1/2 is measured at 10 kHz offset from the carrier. Assuming R300K = 2 X, V0,300K = 1V R4.2K = 0.2 X, V0,4.2K = 1 V, the frequency noise spectral density due to the coil series resistance, which is given by Sm ¼ kTRx2LC =ð2pÞ2 V 20 , are 2 Hz/Hz1/2 at 300 K and 0.1 Hz/Hz1/2 at 4.2 K, respectively. Consequently, the measured frequency noise is about an order of magnitude larger than the coil thermal noise limit at both temperatures, probably due to the noise contribution from the cross-coupled transistor pair. The measured frequency noise at low temperatures is highly dependent on the parameters used in the temperature control loop and on the He pump settings, going from a minimum value of 1 Hz/Hz1/2 to a maximum value of 100 Hz/Hz1/2. This is due to the influence of temperature instabilities on the oscillator phase noise. Unfortunately, we are not yet able to find a strategy to reproducibly achieve the minimum frequency noise of 1 Hz/Hz1/2 in all measurements. However, this minimum frequency noise is reproducibly obtained when the chip is immersed in liquid helium.
Fig. 5. ESR spectra. The experimental ESR signal shown here (in kHz) is the amplitude of the component at the field modulation frequency of the LC-oscillator frequency. Experimental conditions notations: T is the sample temperature, B1 is the amplitude of the microwave magnetic field, Bm is the amplitude of the modulation magnetic field, mm is the frequency of the magnetic field modulation, ts is the time interval of the magnetic field sweep, Df is the equivalent noise bandwidth of the lock-in, xLC is the oscillator frequency far from the resonant magnetic field. ESR spectra of a spherical crystal of ruby sample (Cr3+:Al2O3 sample with 1% Cr3+ content) having a diameter of 122 lm placed on the 17 GHz oscillator coil and a DPPH sample having size of about (4 lm)3 placed on the 21 GHz oscillator coil at different temperatures. Experimental conditions: mm = 100 kHz, ts = 335 s, Df ffi 3 Hz, (a) B1 ffi 0.09 mT, Bm ffi 0.25 mT. (b) B1 ffi 0.11 mT, Bm ffi 0.25 mT. (c) B1 ffi 0.11 mT, Bm ffi 0.06 mT.
99
The oscillation frequency of the two integrated oscillators is dependent on temperature, increasing by about 2 GHz going from 300 K to 4 K (i.e., the oscillation frequencies are about 23 GHz and 19 GHz at 4 K). Additionally, the oscillation frequency depends on the oscillator bias current but it is independent on the applied static magnetic field (up to at least 1.5 T). Fig. 4 shows the frequency difference between the two integrated oscillators as a function of the oscillator bias current Idc2 at fixed bias current Idc1. Since the bias current of the oscillator at the lower frequency is fixed, the decrease of the frequency difference Dfosc corresponds to a decrease of the frequency of the oscillator operating at the higher frequency. This behavior is due to the larger effective gate-source capacitance at higher bias currents [34,35]. For a bias current sufficiently far from the oscillator start-up the curves at 300 and 4.2 K show similar slopes (about 100 MHz/mA) but an otherwise different behavior. At 4.2 K, the oscillator frequency show sharp transitions as a function of the oscillator bias current, with relatively flat regions between the sharp transitions. The amplitude of these transitions are in the order a few tens of kHz to a few tens of MHz. At the moment, we have no convincing explanation for this behavior. Since the ESR experiments are performed at fixed bias current, the sharp transitions observed at 4 K have no effect on the measured ESR spectra, except if we operate at a bias current very close to one of the sharp transitions. To investigate the behavior of the integrated LC-oscillator as electron spin resonance detectors we performed experiments with several different samples over the entire temperature range from 300 K to 4 K. In order to demonstrate the versatility of the realized single-chip ESR detector, we performed measurements with samples having significantly different characteristics: an exchange
Fig. 6. ESR spectra of a DPPH sample having size of about (4 lm)3 placed on the 21 GHz oscillator coil at different temperatures. Experimental conditions: Bm ffi 0.06 mT, mm = 100 kHz, ts = 42 s, Df ffi 3 Hz, (a) B1 ffi 0.09 mT, (b–d) B1 ffi 0.11 mT. See notations in Fig. 5.
100
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
narrowed standard system (DPPH), two hyperfine splitted systems having 100% and 1% concentrations (Cu2+ in TPP and in Ni(mnt)2), a zero field splitted system (Cr3+ in Al2O3), and a very broad line Fe3+ system. Fig. 5 shows measurements performed with two different samples placed on the coils of the two integrated oscillators. A 122 lm diameter spherical crystal of Cr3+:Al2O3 (Ruby G10, Saphirwerk Industrieprodukt AG, Switzerland) having a concentration of 1% of Cr3+ is placed in the center of the 17 GHz oscillator coil. A single crystal of DPPH (1,1-diphenyl-2-picryl-hydrazyl) having a volume of about (4 lm)3 is placed at the center of the 21 GHz oscillator coil. The DPPH sample is obtained by slow evaporation of a solution of DPPH powder (Aldrich D9132) in diethyl ether at room temperature in air (0.13% in volume of DPPH), as previously reported in Ref. [36]. As extensively investigated in Refs. [37,38], the ruby spectrum consists of six lines, with intensity and position strongly dependent on the orientation of the crystal with respect to the applied static magnetic field. In the inset of Fig. 5a, a background signal at B0 ffi 0.63 T, which corresponds to g ffi 2 for the 17 GHz oscillator, is clearly visible (its amplitude is 5 kHz and its linewidth is 2.3 mT). The origin of the background signal, which is no more visible at low temperatures, is unclear. Fig. 5 shows that the resonances are shifted towards higher magnetic fields at lower temperatures. This is due to the shift towards higher frequencies of the oscillators at lower temperatures discussed above. Due to the temperature dependence of the oscillators frequencies mentioned above, in Fig. 6 (as well as in Figs. 7 and 8) the obtained spectra are plotted as a function of magnetic field offset with respect to the g ffi 2 condition. Fig. 6 reports spectra of the DPPH sample obtained with narrow field sweeps about the resonance field at
different temperatures. The spectra in Fig. 5 are taken with a larger field modulation amplitude, optimized for the ruby sample (0.25 mT instead of 0.06 mT), and a faster sweep rate (3 mT/s instead of 0.07 mT/s) with respect to those in Fig. 6. Combined with the fact that the two spectra are taken with the same lock-in time constant (about 80 ms), the DPPH spectra in Fig. 5 are distorted whereas those in Fig. 6 are not. The narrow sweeps in Figs. 6 are reported to show that the obtained signal shape corresponds, although only approximately, to the derivative of a dispersion signal as expected from field modulation and Eq. (1). The opposite sign of the ruby and DPPH signals is due to the fact that the two samples are placed on two different oscillators and that the measured quantity is the difference between the two oscillators frequencies. An increase of the frequency of the oscillator operating at higher frequency corresponds to an increase of the frequency difference, whereas an increase of the frequency of the oscillator operating at the lower frequency corresponds to a decrease of the frequency difference. Since DPPH has a spin concentration of about 2 1027 spins/m3, the (4 lm)3 sample contains about 1011 spins. The ESR signal is about 20 kHz at 300 K and about 180 kHz at 30 K, as expected in the Curie-law approximation. The measured frequency noise spectral density is 30 Hz/Hz1/2 at 300 K and 20 Hz/Hz1/2 at 30 K. Consequently, the experimental spin sensitivities, as defined by Eq. (11) in Ref. [13], is about Nmin ffi 5 108 spins/Hz1/2 at 300 K and Nmin ffi 4 107 spins/Hz1/2 at 30 K. At lower temperatures the DPPH signal becomes smaller and broader. As discussed above, in the temperature range from 4 to 10 K, the frequency noise spectral density assumes values from 1 Hz/Hz1/2 to 100 Hz/Hz1/2. This corresponds to spin sensitivities in the range from Nmin ffi 106 spins/Hz1/2 to
Fig. 7. ESR spectra of single crystal of CuTPP having a volume of about (80 lm)3 at different temperatures placed on the 17 GHz oscillator coil. Experimental conditions: Bm ffi 0.25 mT, mm = 100 kHz, ts = 84 s, Df ffi 6 Hz, (a) B1 ffi 0.09 mT, (b–e) B1 ffi 0.11 mT. See notations in Fig. 5.
Fig. 8. ESR spectra of single crystal of Cu(mnt)2 in Ni(mnt)2 with a Cu concentration of 1% having a volume of about (120 lm)3 at different temperatures placed on the 21 GHz oscillator coil. Experimental conditions: Bm ffi 0.25 mT, mm = 100 kHz, ts = 84 s, Df ffi 6 Hz, (a) B1 ffi 0.09 mT, (b–e) B1 ffi 0.11 mT. See notations in Fig. 5.
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
101
amplitude does not increase following the Curie law. The 1% Cu(mnt)2 spectra have amplitudes which also do not follow the Curie law and have shapes which are highly temperature dependent. For both samples this behavior is probably due to the temperature dependence of their relaxation times. Fig. 9 shows spectra of a (200 lm)3 microcrystalline powder of synthetic b-haematin (Fe(III)-protoporphyrin-IX)2, a synthetic analogue of haemozoin, a product of haemoglobin responsible for malaria [40,41]. Due to the very broad resonance lines (about 0.2 T) with respect to the maximum magnetic field modulation achievable with our set-up (about 1 mT), a better signal-to-noise ratio is obtained by measuring the spectra without magnetic field modulation. The sample is placed at the center of the 17 GHz oscillator. Fig. 9 shows measurements performed at different temperatures. No signal is measurable at room temperature. As reported in Ref. [42] the ESR spectra of b-haematin consist of a strong signal at geff ffi 4, and a weaker signal at geff ffi 2. These measurements are taken by chopping the signal at the frequency-to-voltage converter output (see Fig. 2 and discussion above). The inset of Fig. 9d shows the geff ffi 2 component after four averages. These measurements on synthetic b-haematin show that our single-chip ESR detector has a relatively weak low frequency noise (about 300 Hz rms with a bandwidth of 13 Hz and a sweep lasting 80 s) which allows for measurements without field modulation, a desirable feature for measuring resonance lines much broader than the maximum achievable magnetic field modulation amplitude.
5. Conclusion and outlook
Fig. 9. ESR spectra acquired at different temperatures for a microcrystalline powder of synthetic b-hematin having a volume of about (200 lm)3 placed on the 17 GHz oscillator coil. Inset: blowup of the g ffi 2 region of the ESR spectrum. Experimental conditions: ts = 84 s, Df ffi 13 Hz, (a) B1 ffi 0.09 mT, (c and d) B1 ffi 0.07 mT. See notations in Fig. 5.
Nmin ffi 108 spins/Hz1/2. By means of Eq. (3), we can compute the expected spin sensitivity of the single-chip ESR detector. Assuming an Al resistivity at 4.2 K of 109 X m (typical values are in the range 1012 X m to 109 X m [39]), an Al thickness of 7.5 lm, a coil trace width of 30 lm, and a coil diameter of d = 200 lm, we obtain a coil microwave resistance R ffi 0.2 X and a unitary field Bu ffi (l0/d) ffi 6 mT/A. Assuming T1 ffi T2, B0 ffi 1 T, and T = 4.2 K, we obtain Nmin ffi 104 spins/Hz1/2. The relatively large B1 produced by the integrated oscillator saturates partially the DPPH sample, decreasing the frequency variation by approximately an order of magnitude with respect to the optimal non saturated conditions. As a consequence of the larger noise and the reduced signal amplitude, the experimentally achieved spin sensitivity is two orders of magnitude worse than the one achievable under the optimal conditions considered above (i.e., for c2 B21 T 1 T 2 < 1 and with a frequency noise only due to the coil resistance thermal noise). Fig. 7 shows spectra of a single crystal of non-diluted Cu(II)tetraphenylporphine (CuTPP, Aldrich 25182) having a volume of about (80 lm)3. Fig. 8 shows spectra of a single crystal of Cu2+-doped tetramethylammonium-bis(maleonitriledithiolato) nickel with a Cu concentration of 1% (1% Cu(mnt)2) in Ni(mnt)2) and a volume of about (120 lm)3. All spectra show the hyperfine splitting produced by the Cu nuclei (63Cu and 65Cu), having spin I = 3/2. The CuTPP spectra are similar in shape but the signal
In this work we have experimentally demonstrated that CMOS single-chip microwave LC-oscillators are a valid alternative to miniaturized resonators [7,8,17] for high spin sensitivity ESR spectroscopy on mass limited samples in the entire temperature range from 300 K down to at least 4 K. The spin sensitivity measured with a sample of DPPH is 108 spins/Hz1/2 at 300 K and down to 106 spins/Hz1/2 at 4 K. These values are more than an order of magnitude better that recent results obtained with a sample of DPPH using a miniaturized resonator [8], and similar to those obtained with a sample of E0 centers in SiO2 at 300 K and with a sample of phosphorous doped silicon (28Si:P) at 10 K also measured with a miniaturized resonator [17]. Due to the dependence of the spin sensitivity on the relaxation times, the comparison with the remarkable results reported in [17] obtained with different samples is only indicative. In the following, we describe the main advantages and disadvantages of single-chip microwave oscillators with respect to miniaturized resonators. The small size of each chip and the on-chip downconversion of the ESR signal into a robust frequency-encoded signal might allow one to create dense arrays of independent detectors that can be placed in the same magnet for simultaneous measurements of different samples. The integration of all components responsible for the spin sensitivity within a distance of 100 lm from the detection coil reduce the signal losses to a minimum. This might be particularly important at frequencies exceeding 100 GHz where conventional approaches requires expensive technologies to limit the losses and the degradation of the signal-to-noise ratio. CMOS LC-oscillators operating up to 300 GHz have been already reported [43,44] and operation at THz frequencies might be possible in the near future [45,46]. This means that the single-chip microwave oscillator approach is suitable up to the largest magnetic fields currently available. The local conversion of DC power into a near-field non-radiating microwave magnetic field obtained with the single-chip microwave oscillator might be interesting also for dynamic nuclear polarization (DNP) experiments.
102
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
The main disadvantage of our single-chip approach is the difficulty in producing low B1 fields. This problem is particularly relevant for samples easily saturated (i.e., having a large relaxation times product T1T2). Currently, with a coil having a diameter of the order of 100 lm, we can hardly produce B1 fields significantly smaller than 0.1 mT. As mentioned before, this is due to the fact that the microwave current in the integrated coil cannot be made arbitrarily small: a minimum value is required to sustain stable oscillations. We are currently studying oscillator topologies capable to produce lower B1 fields. Alternatively, we are also investigating the applicability of non steady-state techniques to our detection approach, such as the rapid scan of the magnetic field [47] or the oscillator frequency [48]. A rapid frequency scan can be easily implemented by means of a voltage controlled capacitor placed in parallel to the integrated oscillator coil. As discussed in Section 4, temperature instabilities in the cryostat affect significantly the oscillator frequency noise. This problem is caused by the fact that, contrary to conventional approaches, the microwave source is located in the cryostat. Although we believe that this problem can be solved by a careful optimization of the temperature control system or by a less temperature sensitive oscillator design, the effectiveness of these solutions is not demonstrated yet. In the near future we plan to investigate the possibility to operate the CMOS single-chip detectors below 4 K. In particular, we aim to study the behavior of the oscillator at a temperature below the superconducting transition temperature of the miniaturized coil. Due to the low critical magnetic field of Al [49], we will consider the possibility to replace the aluminum integrated coils with a post-processing integration of coils made of materials having a larger critical field, such as thin films of Nb [9,50,51]. Additionally, operation at temperatures below 1 K with a frequency above 100 GHz will allow us to investigate the behavior of the oscillator in the condition kT < hxLC. This condition combined with the dependence of the MOSFET gate-source capacitance on the oscillation amplitude, which introduces an anharmonicity in the oscillator behavior, might allow us to observe a non-trivial quantum behavior of the LC-oscillator. Acknowledgment Financial support from the Swiss National Science Foundation is gratefully acknowledged. References [1] M.T. Causa, M. Tovar, A. Caneiro, F. Prado, G. Ibanez, C.A. Ramos, A. Butera, B. Alascio, X. Obradors, S. Pinol, F. Rivadulla, C. Vazquez-Vazquez, M.A. LopezQuintela, J. Rivas, Y. Tokura, S.B. Oseroff, High-temperature spin dynamics in CMR manganites: ESR and magnetization, Phys. Rev. B 58 (1998) 3233–3239. [2] M. Bakr, M. Akiyama, Y. Sanada, In situ high-temperature ESR measurements for kerogen maturation, Org. Geochem. 17 (1991) 321–328. [3] E.M. Decastro, V. Pereira, High temperature probe for EPR measurements, Rev. Sci. Instrum. 40 (1969) 949. [4] E. Dormann, D. Hone, V. Jaccarino, High-temperature EPR in solid and molten paramagnets, Phys. Rev. B 14 (1976) 2715–2739. [5] S. Probst, H. Rotzinger, S. Wunsch, P. Jung, M. Jerger, M. Siegel, A.V. Ustinov, P.A. Bushev, Anisotropic rare-earth spin ensemble strongly coupled to a superconducting resonator, Phys. Rev. Lett. 110 (2013) 157001. [6] V. Ranjan, G. de Lange, R. Schutjens, T. Debelhoir, J.P. Groen, D. Szombati, D.J. Thoen, T.M. Klapwijk, R. Hanson, L. DiCarlo, Probing dynamics of an electronspin ensemble via a superconducting resonator, Phys. Rev. Lett. 110 (2013) 067004. [7] Y. Twig, E. Dikarov, A. Blank, Cryogenic electron spin resonance microimaging probe, J. Magn. Reson. 218 (2012) 22–29. [8] R. Narkowicz, H. Ogata, E. Reijerse, D. Suter, A cryogenic receiver for EPR, J. Magn. Reson. 237 (2013) 79–84. [9] O.W.B. Benningshof, H.R. Mohebbi, I.A.J. Taminiau, G.X. Miao, D.G. Cory, Superconducting microstrip resonator for pulsed ESR of thin films, J. Magn. Reson. 230 (2013) 84–87. [10] H. Malissa, D.I. Schuster, A.M. Tyryshkin, A.A. Houck, S.A. Lyon, Superconducting coplanar waveguide resonators for low temperature pulsed electron spin resonance spectroscopy, Rev. Sci. Instrum. 84 (2013) 025116.
[11] A.J. Sigillito, H. Malissa, A.M. Tyryshkin, H. Riemann, N.V. Abrosimov, P. Becker, H.-J. Pohl, M.L. Thewalt, K.M. Itoh, J.J. Morton, Fast, low-power manipulation of spin ensembles in superconducting microresonators, 2014 (arXiv preprint arXiv:1403.0018). [12] Y. Twig, E. Dikarov, A. Blank, Ultra miniature resonators for electron spin resonance. Sensitivity analysis, design and construction methods, and potential applications, Mol. Phys. 111 (2013) 2674–2682. [13] T. Yalcin, G. Boero, Single-chip detector for electron spin resonance spectroscopy, Rev. Sci. Instrum. 79 (2008) 094105. [14] G. Boero, M. Bouterfas, C. Massin, F. Vincent, P.A. Besse, R.S. Popovic, A. Schweiger, Electron-spin resonance probe based on a 100 lm planar microcoil, Rev. Sci. Instrum. 74 (2003) 4794–4798. [15] R. Narkowicz, D. Suter, I. Niemeyer, Scaling of sensitivity and efficiency in planar microresonators for electron spin resonance, Rev. Sci. Instrum. 79 (2008) 084702. [16] R. Narkowicz, D. Suter, R. Stonies, Planar microresonators for EPR experiments, J. Magn. Reson. 175 (2005) 275–284. [17] Y. Twig, E. Dikarov, W.D. Hutchison, A. Blank, Note: high sensitivity pulsed electron spin resonance spectroscopy with induction detection, Rev. Sci. Instrum. 82 (2011) 076105. [18] J. Anders, A. Angerhofer, G. Boero, K-band single-chip electron spin resonance detector, J. Magn. Reson. 217 (2012) 19–26. [19] G. Boero, G. Gualco, R. Lisowski, J. Anders, D. Suter, J. Brugger, Room temperature strong coupling between a microwave oscillator and an ensemble of electron spins, J. Magn. Reson. 231 (2013) 133–140. [20] P. Kinget, Amplitude detection inside CMOS LC oscillators, In: 2006 IEEE International Symposium on Circuits and Systems, vols 1–11, Proceedings, 2006, pp. 5147–5150. [21] B. Razavi, RF Microelectronics, second international ed., Pearson Education International, Upper Saddle River, NJ, 2012. [22] G. Ghibaudo, F. Balestra, Low temperature characterization of silicon CMOS devices, Microelectron. Reliab. 37 (1997) 1353–1366. [23] G. Ghibaudo, F. Balestra, A. Emrani, A survey of MOS device physics for lowtemperature electronics, Microelectron. Eng. 19 (1992) 833–840. [24] E.A. Gutierrez, L. Deferm, G. Declerck, DC characteristics of submicrometer CMOS inverters operating over the whole temperature-range of 4.2–300 K, IEEE Trans. Electron Dev. 39 (1992) 2182–2184. [25] E.A. Gutierrez, L. Deferm, G. Declerck, Experimental-determination of selfheating in submicrometer MOS-transistors operated in a liquid-helium ambient, IEEE Electron Dev. Lett. 14 (1993) 152–154. [26] J.T. Hastings, K.W. Ng, Characterization of a complementary metal–oxide– semiconductor operational-amplifier from 300 to 4.2 K, Rev. Sci. Instrum. 66 (1995) 3691–3696. [27] J. Jomaah, G. Ghibaudo, S. Cristoloveanu, A. Vandooren, F. Dieudonne, J. Pretet, F. Lime, K. Oshima, B. Guillaumot, F. Balestra, Low-temperature performance of ultimate Si-based MOSFETs, Low Temperature Electronics and Low Temperature Cofired Ceramic Based Electronic Devices, vol. 2003, 2004, pp. 118–128. [28] U. Kleine, J. Bieger, H. Seifert, A low-noise CMOS preamplifier operating at 4.2 K, IEEE J. Solid-State Circ. 29 (1994) 921–926. [29] H. Nagata, H. Shibai, T. Hirao, T. Watabe, M. Noda, Y. Hibi, M. Kawada, I. Nakagawa, Cryogenic capacitive transimpedance amplifier for astronomical infrared detectors, IEEE Trans. Electron Dev. 51 (2004) 270–278. [30] B. Okcan, G. Gielen, C. Van Hoof, A third-order complementary metal–oxide– semiconductor sigma-delta modulator operating between 4.2 K and 300 K, Rev. Sci. Instrum. 83 (2012) 024708. [31] N. Yoshikawa, T. Tomida, A. Tokuda, Q. Liu, X. Meng, S.R. Whiteley, T. Van Duzer, Characterization of 4 K CMOS devices and circuits for hybrid JosephsonCMOS systems, IEEE Trans. Appl. Supercond. 15 (2005) 267–271. [32] S.R. Ekanayake, T. Lehmann, A.S. Dzurak, R.G. Clark, A. Brawley, Characterization of SOS–CMOS FETs at low temperatures for the design of integrated circuits for quantum bit control and readout, IEEE Trans. Electron Dev. 57 (2010) 539–547. [33] F. Balestra, G. Ghibaudo, Brief review of the MOS device physics for lowtemperature electronics, Solid-State Electron. 37 (1994) 1967–1975. [34] J.E. Meyer, MOS models and circuit simulation, RCA Rev. 32 (1971). 42-&. [35] T. Ytterdal, Y. Cheng, T.A. Fjeldly, Device Modeling for Analog and RF CMOS Circuit Design, Wiley, Chichester, 2003. [36] D. Zilic, D. Pajic, M. Juric, K. Molcanov, B. Rakvin, P. Planinic, K. Zadro, Single crystals of DPPH grown from diethyl ether and carbon disulfide solutions – crystal structures, EPR and magnetization studies, J. Magn. Reson. 207 (2010) 34–41. [37] R.F. Wenzel, Y.W. Kim, Linewidth of electron paramagnetic resonance of (Al2O3)1X(Cr2O3)X, Phys. Rev. 140 (1965). 1592-&. [38] T.T. Chang, D. Foster, A.H. Kahn, Intensity standard for electron-paramagnetic resonance using chromium-doped corundum (Al2O3–Cr3+), J. Res. Natl. Bur. Stand. 83 (1978) 133–164. [39] P.D. Desai, H.M. James, C.Y. Ho, Electrical-resistivity of aluminum and manganese, J. Phys. Chem. Ref. Data 13 (1984) 1131–1172. [40] A. Dorn, R. Stoffel, H. Matile, A. Bubendorf, R.G. Ridley, Malarial haemozoin beta-hematin supports heme polymerization in the absence of protein, Nature 374 (1995) 269–271. [41] S. Pagola, P.W. Stephens, D.S. Bohle, A.D. Kosar, S.K. Madsen, The structure of malaria pigment beta-haematin, Nature 404 (2000) 307–310. [42] A. Sienkiewicz, J. Krzystek, B. Vileno, G. Chatain, A.J. Kosar, D.S. Bohle, L. Forro, Multi-frequency high-field EPR study of iron centers in malarial pigments, J. Am. Chem. Soc. 128 (2006) 4534–4535.
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103 [43] B. Razavi, A 300-GHz fundamental oscillator in 65-nm CMOS technology, IEEE J. Solid-State Circ. 46 (2011) 894–903. [44] N. Landsberg, E. Socher, 240 GHz and 272 GHz Fundamental VCOs using 32 nm CMOS technology, IEEE Trans. Microwave Theory Tech. 61 (2013) 4461–4471. [45] E. Seok, D. Shim, C. Mao, R. Han, S. Sankaran, C. Cao, W. Knap, K. Kenneth, Progress and challenges towards terahertz CMOS integrated circuits, IEEE J. Solid-State Circ. 45 (2010) 1554–1564. [46] K.K. O, S. Sankaran, C. Cao, E.-Y. Seok, D. Shim, C. Mao, R. Han, Millimeter wave to terahertz in CMOS, Int. J. Hi. Spe. Ele. Syst. 19 (2009) 55–67. [47] J.W. Stoner, D. Szymanski, S.S. Eaton, R.W. Quine, G.A. Rinard, G.R. Eaton, Direct-detected rapid-scan EPR at 250 MHz, J. Magn. Reson. 170 (2004) 127– 135.
103
[48] M. Tseitlin, G.A. Rinard, R.W. Quine, S.S. Eaton, G.R. Eaton, Rapid frequency scan EPR, J. Magn. Reson. 211 (2011) 156–161. [49] E.P. Harris, D.E. Mapother, Critical field of superconducting aluminum as a function of pressure and temperature above 0.3° K, Phys. Rev. 165 (1968) 522– 532. [50] W.J. Wallace, R.H. Silsbee, Microstrip resonators for electron-spin-resonance, Rev. Sci. Instrum. 62 (1991) 1754–1766. [51] C. Clauss, D. Bothner, D. Koelle, R. Kleiner, L. Bogani, M. Scheffler, M. Dressel, Broadband electron spin resonance from 500 MHz to 40 GHz using superconducting coplanar waveguides, Appl. Phys. Lett. 102 (2013) 162601.
Contents lists available at ScienceDirect
Journal of Magnetic Resonance journal homepage: www.elsevier.com/locate/jmr
Cryogenic single-chip electron spin resonance detector Gabriele Gualco, Jens Anders, Andrzej Sienkiewicz, Stefano Alberti, László Forró, Giovanni Boero ⇑ Ecole Polytechnique Federale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
a r t i c l e
i n f o
Article history: Received 6 June 2014 Revised 12 August 2014 Available online 8 September 2014 Keywords: ESR Cryogenic CMOS
a b s t r a c t We report on the design and characterization of a single-chip electron spin resonance detector, operating at a frequency of about 20 GHz and in a temperature range extending at least from 300 K down to 4 K. The detector consists of an LC oscillator formed by a 200 lm diameter single turn aluminum planar coil, a metal–oxide–metal capacitor, and two metal–oxide–semiconductor field effect transistors used as negative resistance network. At 300 K, the oscillator has a frequency noise of 20 Hz/Hz1/2 at 100 kHz offset from the 20 GHz carrier. At 4 K, the frequency noise is about 1 Hz/Hz1/2 at 10 kHz offset. The spin sensitivity measured with a sample of DPPH is 108 spins/Hz1/2 at 300 K and down to 106 spins/Hz1/2 at 4 K. Ó 2014 Elsevier Inc. All rights reserved.
1. Introduction
2. Operating principle
Methods based on the electron spin resonance (ESR) phenomenon are used to investigate samples in a wide temperature range, ranging from above 1000 K [1–4] to below 1 K [5,6]. Low temperature measurements are usually performed in large microwave cavities as well as with miniaturized conductive [7,8] or superconducting [9–11] resonators. Miniaturized resonators are typically used to maximize the signal-to-noise ratio in experiments with mass-limited samples [7,12–17]. In Refs. [13,18,19] we presented single-chip integrated inductive ESR detectors, fabricated using complementary metal oxide semiconductor (CMOS) technologies, operating between 8 GHz and 28 GHz. The ESR phenomenon was detected as a variation of the frequency of an integrated LC-oscillator due to an effective variation of its coil impedance caused by the resonant complex susceptibility of the sample. Operation in the temperature range from 300 K to 77 K was demonstrated in Refs. [18,19]. Here, we report on the implementation of a 20 GHz single-chip ESR detector, based on the same operating principle, but capable of operating from 300 K down to at least 4 K. Depending on the specific sample (and, in particular on the dependence of its relaxation times on temperature), the possibility to operate down to 4 K might represent a significant advantage in terms of spin sensitivity (larger polarization, lower thermal noise) as well as in terms of information richness. The fabricated device represents also the first demonstration of a CMOS microwave oscillator operating down to 4 K.
The principle of operation of the realized single-chip ESR detector is identical to that reported in our previous work Refs. [13,18,19]. In typical experimental conditions, the oscillation frequency of an LC-oscillator coupled with an ensemble of electron spins is given by [19]
⇑ Corresponding author. Address: Ecole Polytechnique Fédérale de Lausanne (EPFL), Station 17, CH-1015 Lausanne, Switzerland. E-mail address: giovanni.boero@epfl.ch (G. Boero). http://dx.doi.org/10.1016/j.jmr.2014.08.013 1090-7807/Ó 2014 Elsevier Inc. All rights reserved.
xLC xLC v ffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ gv0
ð1Þ
where
v0 ¼
ðxLCv x0 ÞT 22 1 x0 v 0 2 1 þ T 22 ðxLCv x0 Þ2 þ c2e B21 T 1 T 2
ð2Þ
pffiffiffiffiffiffi is the real part of the magnetic susceptibility, xLC ¼ 1= LC is the unperturbed oscillator frequency, x0 = ceB0, v0 is the static magnetic susceptibility, g is the filling factor (approximately given by (Vs/Vc), where Vs is the sample volume, and Vc is the coil sensitive volume), ce is the electron gyromagnetic ratio, B1 is the microwave magnetic field, T1 and T2 are the relaxation times. As discussed in Ref. [13], the oscillator frequency variation due to the electron spin resonance phenomenon in the sample is, in first approximation, given by DxLC v ffi ð1=2ÞxLC gv0 . Consequently, the oscillator frequency variation is proportional to the real part v0 of the sample complex susceptibility v ¼ v0 jv00 . The shape of the oscillator frequency variation is thus identical to that of the dispersion signal measured in conventional continuous wave experiments. However, the dependence on the microwave field B1 is substantially different. In the conventional amplitude detection method the measured signal is linearly proportional to the precessing magnetization (i.e., to the product v00 B1 for the absorption and v0 B1 for the dispersion). In
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
97
the frequency detection case, for B1 below saturation the signal amplitude is independent of the value of B1 and decreases as 1=B21 in saturation (in the conventional method the signal is proportional to B1 below saturation and decreases as 1/B1 in saturation). Neglecting the noise contribution of the active devices in the oscillator feedback and assuming that the frequency noise spectral density (in Hz2/Hz) is only due to the thermal noise of the coil resistance and given by Sm ¼ kTRx2LC =ð2pÞ2 V 20 , the spin sensitivity (in spins/Hz1/2) is given by [18]
Nmin
pffiffiffi sffiffiffiffiffi T 3=2 R T 1 ; ffia 2 T2 B0 Bu
ð3Þ
where V0 is the oscillation amplitude, R is the coil series resistance, Bu is the coil unitary field, T is the coil (and sample) temperature, and a ffi 20 m1 kg5/2 s4 K3/2 A3. 3. Description of the single-chip detector Fig. 1 shows a photo and the block diagram of the realized chip, which consists of two LC-oscillators operating at about 21 GHz and 17 GHz, a mixer, and a frequency division module. The chip is manufactured in a commercially available 130 nm CMOS technology (IBM 8RF). The total chip surface, including the bonding pads, is about 1 mm2. The total power consumption of the chip is about 20 mW at 300 K and about 6 mW at 4 K, with the oscillators bias currents set to the minimum values which enable stable oscillations. As in our previous designs [13,19], with the exception of that reported in [18], we have integrated two oscillators to perform the first down-conversion of the oscillator frequency by a mixer. The use of two oscillators and a mixer as first step in the frequency downconversion requires a frequency divider operating at 4 GHz instead of 20 GHz, which is significantly less difficult to design. The excitation/detection octagonal coils of the oscillators have an external diameter of 200 lm, a metal width of 30 lm, a metal thickness of 7.5 lm (obtained by parallel connection of the three top metal layers available in the process, two made of aluminum and one of copper), and an inductance of about 300 pH. The capacitor is realized using interdigitated aluminum fingers. The capacitance value for the two oscillators are 180 fF and 260 fF, respectively. The microwave magnetic field B1 can be varied from 0.08 to 0.14 mT by changing the oscillator bias current Idc from 1 to 5 mA. At low temperatures, a slightly lower bias current of about 0.5 mA is sufficient to guarantee stable oscillations. The lower limit is determined by the minimum transconductance of the cross-coupled pair required for stable oscillation whereas the upper limit is given by the maximum voltage swing which can be applied across the transistors in the cross-coupled pair without damaging their thin gate oxides. The value of B1 generated by our single-chip detector is estimated by measuring the voltage at the oscillator bias node VIdc. In condition of stable oscillation the oscillation amplitude V0 ffi VIdc [20]. Hence, B1 ffi (1/2)Bu(VIdc/xLCL), where Bu ffi l0/d, d is the coil diameter, and L is the coil inductance. This B1 estimation is in agreement with saturation experiments with samples of known relaxation times. Fig. 1 shows also the schematics and block diagrams of the mixer and frequency divider. The frequency mixer needed for down-conversion of the oscillation frequency is based on a double-balanced Gilbert cell topology (see pages 368–370 of Ref. [21]). By mixing the two oscillator output voltages a signal at about 4 GHz is obtained. The sum frequency component is filtered out by the system parasitics (i.e., a low pass filter is not necessary). The frequency divider is realized by means of current model logic (CML) D-latches with resistive load (see pages 683–699 of Ref.
Fig. 1. (a) Microphotograph of the single-chip ESR detector. Idc1 and Idc2: DC power supply of the two oscillators (0.5–5 mA). VD: DC power supply for the mixer and frequency-division module (1.5 V). OUTp and OUTn: differential output signal (ffi220 MHz). The circle indicates the area where the samples are placed. (b) Block diagram of the single-chip ESR detector. (c) Schematics of the frequency mixer (Vb1 ffi 0.9 V, Vb2 ffi 0.6 V). (d) Block diagram of the frequency divider based on Dlatches. (e) Schematics of the D-latch (Vb ffi 0.6 V).
98
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
[21]). The on chip divide ratio is 16, by a cascade of four identical divide-by-two frequency dividers. The output buffer, realized as standard source follower amplifier, is capable of driving an output load of 5 pF with a voltage swing of 0.2 V at 200 MHz. The detector reported in this paper is an improved version of the one we reported in Ref. [18]. In the previous work, cooling to about 288 K was necessary to operate the detector, complicating the set-up and de facto preventing spectroscopy studies at samples temperatures above 288 K. The improved version works properly up to about 370 K. The total power consumption of the detector reported here is an order of magnitude smaller with respect to that reported in Ref. [18], allowing easier operation at low temperatures (the device reported in Ref. [18] can actually also operate down to 4 K but it requires a more efficient cooling). In order to assure operation down 4 K, i.e., well below the impurity freezeout temperature [22], we designed the integrated electronics using exclusively metal–oxide–semiconductor and degenerately doped structures.
spectrometers. The signal at the output of the frequency-to-voltage converter is demodulated by a lock-in amplifier. The lock-in internal reference signal is amplified and applied to the field modulation coils. Some experiments are performed without field modulation. In order to use the spectrometer lock-in also in these measurements, the signal at the output of the frequency-to-voltage converter is chopped using a switch controlled by the 100 kHz reference signal of the lock-in amplifier, as shown schematically in Fig. 2. The operation of CMOS devices and circuits at temperatures down to 4 K and below has been already investigated in detail [22–33]. However, no phase noise characterizations of CMOS LC-oscillators at cryogenic temperatures have been reported so far. Fig. 3 shows the square root of the measured frequency noise pffiffiffiffiffi spectral density (i.e., Sm in Hz/Hz1/2) referred to the integrated LC-oscillator output. At 300 K, the noise at 100 kHz offset from the carrier is about 20 Hz/Hz1/2. At 4.2 K, a minimum noise of about
4. Performance of the integrated detector Fig. 2 shows an illustration of the setup used to characterize the performance of the single-chip ESR detector. The chip is glued onto a printed circuit board and electrically connected by wire bonding. On the same printed circuit board a commercial amplifier is used as buffer to drive the coaxial cable carrying the chip output signal at about 200 MHz. The system operates properly in the range from 300 K down to at least 4 K (no measurements below 4 K have been performed yet). The printed circuit board is inserted in a dynamic continuous flow cryostat where the temperature is controlled by means of a temperature controller unit working with a thin film resistance temperature sensor. The signal at the output of the buffer is mixed, at room temperature, with a local oscillator signal having a frequency about 10 MHz higher. The signal at the output of the mixer (with a carrier frequency of about 10 MHz) is fed into a phase-locked loop (PLL). In most of the experiments, a magnetic field modulation at kHz frequencies is added to the static magnetic field to improve the signal-to-noise ratio as in ordinary ESR
Fig. 2. (a) Block diagram of the experimental set-up: (1) Electromagnet power supply (Bruker), (2) electromagnet (Bruker, 0–1.5 T), (3) single chip ESR detector, (4) buffer (Texas Instruments THS4304D), (5) magnetic field modulation coils, (6) power amplifier (Rohrer PA508), (7) mixer (Mini-Circuits ZAD-3), (8) signal generator (Rohde–Schwarz SMR-20), (9) phase locked loop (PLL) circuitry (see Ref. [18]), (10) RF Switch (Mini-Circuits ZYSW-2-50DR), (11) selection mode switch, (12) lock-in of the EPR spectrometer (Bruker Elexsys-II E500), (13) cryostat (Oxford Instruments CF935), (14) temperature controller (Oxford Instruments ITC503), (15) heater, and (16) temperature sensor (Lake Shore Cernox CX1050).
Fig. 3. Frequency noise spectral density of the integrated LC-oscillators at 300 K and at 4.2 K.
Fig. 4. Frequency difference between the two integrated LC-oscillators as a function of the bias current Idc2 at 300 K and at 4.2 K with Idc1 ffi 1 mA.
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
1 Hz/Hz1/2 is measured at 10 kHz offset from the carrier. Assuming R300K = 2 X, V0,300K = 1V R4.2K = 0.2 X, V0,4.2K = 1 V, the frequency noise spectral density due to the coil series resistance, which is given by Sm ¼ kTRx2LC =ð2pÞ2 V 20 , are 2 Hz/Hz1/2 at 300 K and 0.1 Hz/Hz1/2 at 4.2 K, respectively. Consequently, the measured frequency noise is about an order of magnitude larger than the coil thermal noise limit at both temperatures, probably due to the noise contribution from the cross-coupled transistor pair. The measured frequency noise at low temperatures is highly dependent on the parameters used in the temperature control loop and on the He pump settings, going from a minimum value of 1 Hz/Hz1/2 to a maximum value of 100 Hz/Hz1/2. This is due to the influence of temperature instabilities on the oscillator phase noise. Unfortunately, we are not yet able to find a strategy to reproducibly achieve the minimum frequency noise of 1 Hz/Hz1/2 in all measurements. However, this minimum frequency noise is reproducibly obtained when the chip is immersed in liquid helium.
Fig. 5. ESR spectra. The experimental ESR signal shown here (in kHz) is the amplitude of the component at the field modulation frequency of the LC-oscillator frequency. Experimental conditions notations: T is the sample temperature, B1 is the amplitude of the microwave magnetic field, Bm is the amplitude of the modulation magnetic field, mm is the frequency of the magnetic field modulation, ts is the time interval of the magnetic field sweep, Df is the equivalent noise bandwidth of the lock-in, xLC is the oscillator frequency far from the resonant magnetic field. ESR spectra of a spherical crystal of ruby sample (Cr3+:Al2O3 sample with 1% Cr3+ content) having a diameter of 122 lm placed on the 17 GHz oscillator coil and a DPPH sample having size of about (4 lm)3 placed on the 21 GHz oscillator coil at different temperatures. Experimental conditions: mm = 100 kHz, ts = 335 s, Df ffi 3 Hz, (a) B1 ffi 0.09 mT, Bm ffi 0.25 mT. (b) B1 ffi 0.11 mT, Bm ffi 0.25 mT. (c) B1 ffi 0.11 mT, Bm ffi 0.06 mT.
99
The oscillation frequency of the two integrated oscillators is dependent on temperature, increasing by about 2 GHz going from 300 K to 4 K (i.e., the oscillation frequencies are about 23 GHz and 19 GHz at 4 K). Additionally, the oscillation frequency depends on the oscillator bias current but it is independent on the applied static magnetic field (up to at least 1.5 T). Fig. 4 shows the frequency difference between the two integrated oscillators as a function of the oscillator bias current Idc2 at fixed bias current Idc1. Since the bias current of the oscillator at the lower frequency is fixed, the decrease of the frequency difference Dfosc corresponds to a decrease of the frequency of the oscillator operating at the higher frequency. This behavior is due to the larger effective gate-source capacitance at higher bias currents [34,35]. For a bias current sufficiently far from the oscillator start-up the curves at 300 and 4.2 K show similar slopes (about 100 MHz/mA) but an otherwise different behavior. At 4.2 K, the oscillator frequency show sharp transitions as a function of the oscillator bias current, with relatively flat regions between the sharp transitions. The amplitude of these transitions are in the order a few tens of kHz to a few tens of MHz. At the moment, we have no convincing explanation for this behavior. Since the ESR experiments are performed at fixed bias current, the sharp transitions observed at 4 K have no effect on the measured ESR spectra, except if we operate at a bias current very close to one of the sharp transitions. To investigate the behavior of the integrated LC-oscillator as electron spin resonance detectors we performed experiments with several different samples over the entire temperature range from 300 K to 4 K. In order to demonstrate the versatility of the realized single-chip ESR detector, we performed measurements with samples having significantly different characteristics: an exchange
Fig. 6. ESR spectra of a DPPH sample having size of about (4 lm)3 placed on the 21 GHz oscillator coil at different temperatures. Experimental conditions: Bm ffi 0.06 mT, mm = 100 kHz, ts = 42 s, Df ffi 3 Hz, (a) B1 ffi 0.09 mT, (b–d) B1 ffi 0.11 mT. See notations in Fig. 5.
100
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
narrowed standard system (DPPH), two hyperfine splitted systems having 100% and 1% concentrations (Cu2+ in TPP and in Ni(mnt)2), a zero field splitted system (Cr3+ in Al2O3), and a very broad line Fe3+ system. Fig. 5 shows measurements performed with two different samples placed on the coils of the two integrated oscillators. A 122 lm diameter spherical crystal of Cr3+:Al2O3 (Ruby G10, Saphirwerk Industrieprodukt AG, Switzerland) having a concentration of 1% of Cr3+ is placed in the center of the 17 GHz oscillator coil. A single crystal of DPPH (1,1-diphenyl-2-picryl-hydrazyl) having a volume of about (4 lm)3 is placed at the center of the 21 GHz oscillator coil. The DPPH sample is obtained by slow evaporation of a solution of DPPH powder (Aldrich D9132) in diethyl ether at room temperature in air (0.13% in volume of DPPH), as previously reported in Ref. [36]. As extensively investigated in Refs. [37,38], the ruby spectrum consists of six lines, with intensity and position strongly dependent on the orientation of the crystal with respect to the applied static magnetic field. In the inset of Fig. 5a, a background signal at B0 ffi 0.63 T, which corresponds to g ffi 2 for the 17 GHz oscillator, is clearly visible (its amplitude is 5 kHz and its linewidth is 2.3 mT). The origin of the background signal, which is no more visible at low temperatures, is unclear. Fig. 5 shows that the resonances are shifted towards higher magnetic fields at lower temperatures. This is due to the shift towards higher frequencies of the oscillators at lower temperatures discussed above. Due to the temperature dependence of the oscillators frequencies mentioned above, in Fig. 6 (as well as in Figs. 7 and 8) the obtained spectra are plotted as a function of magnetic field offset with respect to the g ffi 2 condition. Fig. 6 reports spectra of the DPPH sample obtained with narrow field sweeps about the resonance field at
different temperatures. The spectra in Fig. 5 are taken with a larger field modulation amplitude, optimized for the ruby sample (0.25 mT instead of 0.06 mT), and a faster sweep rate (3 mT/s instead of 0.07 mT/s) with respect to those in Fig. 6. Combined with the fact that the two spectra are taken with the same lock-in time constant (about 80 ms), the DPPH spectra in Fig. 5 are distorted whereas those in Fig. 6 are not. The narrow sweeps in Figs. 6 are reported to show that the obtained signal shape corresponds, although only approximately, to the derivative of a dispersion signal as expected from field modulation and Eq. (1). The opposite sign of the ruby and DPPH signals is due to the fact that the two samples are placed on two different oscillators and that the measured quantity is the difference between the two oscillators frequencies. An increase of the frequency of the oscillator operating at higher frequency corresponds to an increase of the frequency difference, whereas an increase of the frequency of the oscillator operating at the lower frequency corresponds to a decrease of the frequency difference. Since DPPH has a spin concentration of about 2 1027 spins/m3, the (4 lm)3 sample contains about 1011 spins. The ESR signal is about 20 kHz at 300 K and about 180 kHz at 30 K, as expected in the Curie-law approximation. The measured frequency noise spectral density is 30 Hz/Hz1/2 at 300 K and 20 Hz/Hz1/2 at 30 K. Consequently, the experimental spin sensitivities, as defined by Eq. (11) in Ref. [13], is about Nmin ffi 5 108 spins/Hz1/2 at 300 K and Nmin ffi 4 107 spins/Hz1/2 at 30 K. At lower temperatures the DPPH signal becomes smaller and broader. As discussed above, in the temperature range from 4 to 10 K, the frequency noise spectral density assumes values from 1 Hz/Hz1/2 to 100 Hz/Hz1/2. This corresponds to spin sensitivities in the range from Nmin ffi 106 spins/Hz1/2 to
Fig. 7. ESR spectra of single crystal of CuTPP having a volume of about (80 lm)3 at different temperatures placed on the 17 GHz oscillator coil. Experimental conditions: Bm ffi 0.25 mT, mm = 100 kHz, ts = 84 s, Df ffi 6 Hz, (a) B1 ffi 0.09 mT, (b–e) B1 ffi 0.11 mT. See notations in Fig. 5.
Fig. 8. ESR spectra of single crystal of Cu(mnt)2 in Ni(mnt)2 with a Cu concentration of 1% having a volume of about (120 lm)3 at different temperatures placed on the 21 GHz oscillator coil. Experimental conditions: Bm ffi 0.25 mT, mm = 100 kHz, ts = 84 s, Df ffi 6 Hz, (a) B1 ffi 0.09 mT, (b–e) B1 ffi 0.11 mT. See notations in Fig. 5.
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
101
amplitude does not increase following the Curie law. The 1% Cu(mnt)2 spectra have amplitudes which also do not follow the Curie law and have shapes which are highly temperature dependent. For both samples this behavior is probably due to the temperature dependence of their relaxation times. Fig. 9 shows spectra of a (200 lm)3 microcrystalline powder of synthetic b-haematin (Fe(III)-protoporphyrin-IX)2, a synthetic analogue of haemozoin, a product of haemoglobin responsible for malaria [40,41]. Due to the very broad resonance lines (about 0.2 T) with respect to the maximum magnetic field modulation achievable with our set-up (about 1 mT), a better signal-to-noise ratio is obtained by measuring the spectra without magnetic field modulation. The sample is placed at the center of the 17 GHz oscillator. Fig. 9 shows measurements performed at different temperatures. No signal is measurable at room temperature. As reported in Ref. [42] the ESR spectra of b-haematin consist of a strong signal at geff ffi 4, and a weaker signal at geff ffi 2. These measurements are taken by chopping the signal at the frequency-to-voltage converter output (see Fig. 2 and discussion above). The inset of Fig. 9d shows the geff ffi 2 component after four averages. These measurements on synthetic b-haematin show that our single-chip ESR detector has a relatively weak low frequency noise (about 300 Hz rms with a bandwidth of 13 Hz and a sweep lasting 80 s) which allows for measurements without field modulation, a desirable feature for measuring resonance lines much broader than the maximum achievable magnetic field modulation amplitude.
5. Conclusion and outlook
Fig. 9. ESR spectra acquired at different temperatures for a microcrystalline powder of synthetic b-hematin having a volume of about (200 lm)3 placed on the 17 GHz oscillator coil. Inset: blowup of the g ffi 2 region of the ESR spectrum. Experimental conditions: ts = 84 s, Df ffi 13 Hz, (a) B1 ffi 0.09 mT, (c and d) B1 ffi 0.07 mT. See notations in Fig. 5.
Nmin ffi 108 spins/Hz1/2. By means of Eq. (3), we can compute the expected spin sensitivity of the single-chip ESR detector. Assuming an Al resistivity at 4.2 K of 109 X m (typical values are in the range 1012 X m to 109 X m [39]), an Al thickness of 7.5 lm, a coil trace width of 30 lm, and a coil diameter of d = 200 lm, we obtain a coil microwave resistance R ffi 0.2 X and a unitary field Bu ffi (l0/d) ffi 6 mT/A. Assuming T1 ffi T2, B0 ffi 1 T, and T = 4.2 K, we obtain Nmin ffi 104 spins/Hz1/2. The relatively large B1 produced by the integrated oscillator saturates partially the DPPH sample, decreasing the frequency variation by approximately an order of magnitude with respect to the optimal non saturated conditions. As a consequence of the larger noise and the reduced signal amplitude, the experimentally achieved spin sensitivity is two orders of magnitude worse than the one achievable under the optimal conditions considered above (i.e., for c2 B21 T 1 T 2 < 1 and with a frequency noise only due to the coil resistance thermal noise). Fig. 7 shows spectra of a single crystal of non-diluted Cu(II)tetraphenylporphine (CuTPP, Aldrich 25182) having a volume of about (80 lm)3. Fig. 8 shows spectra of a single crystal of Cu2+-doped tetramethylammonium-bis(maleonitriledithiolato) nickel with a Cu concentration of 1% (1% Cu(mnt)2) in Ni(mnt)2) and a volume of about (120 lm)3. All spectra show the hyperfine splitting produced by the Cu nuclei (63Cu and 65Cu), having spin I = 3/2. The CuTPP spectra are similar in shape but the signal
In this work we have experimentally demonstrated that CMOS single-chip microwave LC-oscillators are a valid alternative to miniaturized resonators [7,8,17] for high spin sensitivity ESR spectroscopy on mass limited samples in the entire temperature range from 300 K down to at least 4 K. The spin sensitivity measured with a sample of DPPH is 108 spins/Hz1/2 at 300 K and down to 106 spins/Hz1/2 at 4 K. These values are more than an order of magnitude better that recent results obtained with a sample of DPPH using a miniaturized resonator [8], and similar to those obtained with a sample of E0 centers in SiO2 at 300 K and with a sample of phosphorous doped silicon (28Si:P) at 10 K also measured with a miniaturized resonator [17]. Due to the dependence of the spin sensitivity on the relaxation times, the comparison with the remarkable results reported in [17] obtained with different samples is only indicative. In the following, we describe the main advantages and disadvantages of single-chip microwave oscillators with respect to miniaturized resonators. The small size of each chip and the on-chip downconversion of the ESR signal into a robust frequency-encoded signal might allow one to create dense arrays of independent detectors that can be placed in the same magnet for simultaneous measurements of different samples. The integration of all components responsible for the spin sensitivity within a distance of 100 lm from the detection coil reduce the signal losses to a minimum. This might be particularly important at frequencies exceeding 100 GHz where conventional approaches requires expensive technologies to limit the losses and the degradation of the signal-to-noise ratio. CMOS LC-oscillators operating up to 300 GHz have been already reported [43,44] and operation at THz frequencies might be possible in the near future [45,46]. This means that the single-chip microwave oscillator approach is suitable up to the largest magnetic fields currently available. The local conversion of DC power into a near-field non-radiating microwave magnetic field obtained with the single-chip microwave oscillator might be interesting also for dynamic nuclear polarization (DNP) experiments.
102
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103
The main disadvantage of our single-chip approach is the difficulty in producing low B1 fields. This problem is particularly relevant for samples easily saturated (i.e., having a large relaxation times product T1T2). Currently, with a coil having a diameter of the order of 100 lm, we can hardly produce B1 fields significantly smaller than 0.1 mT. As mentioned before, this is due to the fact that the microwave current in the integrated coil cannot be made arbitrarily small: a minimum value is required to sustain stable oscillations. We are currently studying oscillator topologies capable to produce lower B1 fields. Alternatively, we are also investigating the applicability of non steady-state techniques to our detection approach, such as the rapid scan of the magnetic field [47] or the oscillator frequency [48]. A rapid frequency scan can be easily implemented by means of a voltage controlled capacitor placed in parallel to the integrated oscillator coil. As discussed in Section 4, temperature instabilities in the cryostat affect significantly the oscillator frequency noise. This problem is caused by the fact that, contrary to conventional approaches, the microwave source is located in the cryostat. Although we believe that this problem can be solved by a careful optimization of the temperature control system or by a less temperature sensitive oscillator design, the effectiveness of these solutions is not demonstrated yet. In the near future we plan to investigate the possibility to operate the CMOS single-chip detectors below 4 K. In particular, we aim to study the behavior of the oscillator at a temperature below the superconducting transition temperature of the miniaturized coil. Due to the low critical magnetic field of Al [49], we will consider the possibility to replace the aluminum integrated coils with a post-processing integration of coils made of materials having a larger critical field, such as thin films of Nb [9,50,51]. Additionally, operation at temperatures below 1 K with a frequency above 100 GHz will allow us to investigate the behavior of the oscillator in the condition kT < hxLC. This condition combined with the dependence of the MOSFET gate-source capacitance on the oscillation amplitude, which introduces an anharmonicity in the oscillator behavior, might allow us to observe a non-trivial quantum behavior of the LC-oscillator. Acknowledgment Financial support from the Swiss National Science Foundation is gratefully acknowledged. References [1] M.T. Causa, M. Tovar, A. Caneiro, F. Prado, G. Ibanez, C.A. Ramos, A. Butera, B. Alascio, X. Obradors, S. Pinol, F. Rivadulla, C. Vazquez-Vazquez, M.A. LopezQuintela, J. Rivas, Y. Tokura, S.B. Oseroff, High-temperature spin dynamics in CMR manganites: ESR and magnetization, Phys. Rev. B 58 (1998) 3233–3239. [2] M. Bakr, M. Akiyama, Y. Sanada, In situ high-temperature ESR measurements for kerogen maturation, Org. Geochem. 17 (1991) 321–328. [3] E.M. Decastro, V. Pereira, High temperature probe for EPR measurements, Rev. Sci. Instrum. 40 (1969) 949. [4] E. Dormann, D. Hone, V. Jaccarino, High-temperature EPR in solid and molten paramagnets, Phys. Rev. B 14 (1976) 2715–2739. [5] S. Probst, H. Rotzinger, S. Wunsch, P. Jung, M. Jerger, M. Siegel, A.V. Ustinov, P.A. Bushev, Anisotropic rare-earth spin ensemble strongly coupled to a superconducting resonator, Phys. Rev. Lett. 110 (2013) 157001. [6] V. Ranjan, G. de Lange, R. Schutjens, T. Debelhoir, J.P. Groen, D. Szombati, D.J. Thoen, T.M. Klapwijk, R. Hanson, L. DiCarlo, Probing dynamics of an electronspin ensemble via a superconducting resonator, Phys. Rev. Lett. 110 (2013) 067004. [7] Y. Twig, E. Dikarov, A. Blank, Cryogenic electron spin resonance microimaging probe, J. Magn. Reson. 218 (2012) 22–29. [8] R. Narkowicz, H. Ogata, E. Reijerse, D. Suter, A cryogenic receiver for EPR, J. Magn. Reson. 237 (2013) 79–84. [9] O.W.B. Benningshof, H.R. Mohebbi, I.A.J. Taminiau, G.X. Miao, D.G. Cory, Superconducting microstrip resonator for pulsed ESR of thin films, J. Magn. Reson. 230 (2013) 84–87. [10] H. Malissa, D.I. Schuster, A.M. Tyryshkin, A.A. Houck, S.A. Lyon, Superconducting coplanar waveguide resonators for low temperature pulsed electron spin resonance spectroscopy, Rev. Sci. Instrum. 84 (2013) 025116.
[11] A.J. Sigillito, H. Malissa, A.M. Tyryshkin, H. Riemann, N.V. Abrosimov, P. Becker, H.-J. Pohl, M.L. Thewalt, K.M. Itoh, J.J. Morton, Fast, low-power manipulation of spin ensembles in superconducting microresonators, 2014 (arXiv preprint arXiv:1403.0018). [12] Y. Twig, E. Dikarov, A. Blank, Ultra miniature resonators for electron spin resonance. Sensitivity analysis, design and construction methods, and potential applications, Mol. Phys. 111 (2013) 2674–2682. [13] T. Yalcin, G. Boero, Single-chip detector for electron spin resonance spectroscopy, Rev. Sci. Instrum. 79 (2008) 094105. [14] G. Boero, M. Bouterfas, C. Massin, F. Vincent, P.A. Besse, R.S. Popovic, A. Schweiger, Electron-spin resonance probe based on a 100 lm planar microcoil, Rev. Sci. Instrum. 74 (2003) 4794–4798. [15] R. Narkowicz, D. Suter, I. Niemeyer, Scaling of sensitivity and efficiency in planar microresonators for electron spin resonance, Rev. Sci. Instrum. 79 (2008) 084702. [16] R. Narkowicz, D. Suter, R. Stonies, Planar microresonators for EPR experiments, J. Magn. Reson. 175 (2005) 275–284. [17] Y. Twig, E. Dikarov, W.D. Hutchison, A. Blank, Note: high sensitivity pulsed electron spin resonance spectroscopy with induction detection, Rev. Sci. Instrum. 82 (2011) 076105. [18] J. Anders, A. Angerhofer, G. Boero, K-band single-chip electron spin resonance detector, J. Magn. Reson. 217 (2012) 19–26. [19] G. Boero, G. Gualco, R. Lisowski, J. Anders, D. Suter, J. Brugger, Room temperature strong coupling between a microwave oscillator and an ensemble of electron spins, J. Magn. Reson. 231 (2013) 133–140. [20] P. Kinget, Amplitude detection inside CMOS LC oscillators, In: 2006 IEEE International Symposium on Circuits and Systems, vols 1–11, Proceedings, 2006, pp. 5147–5150. [21] B. Razavi, RF Microelectronics, second international ed., Pearson Education International, Upper Saddle River, NJ, 2012. [22] G. Ghibaudo, F. Balestra, Low temperature characterization of silicon CMOS devices, Microelectron. Reliab. 37 (1997) 1353–1366. [23] G. Ghibaudo, F. Balestra, A. Emrani, A survey of MOS device physics for lowtemperature electronics, Microelectron. Eng. 19 (1992) 833–840. [24] E.A. Gutierrez, L. Deferm, G. Declerck, DC characteristics of submicrometer CMOS inverters operating over the whole temperature-range of 4.2–300 K, IEEE Trans. Electron Dev. 39 (1992) 2182–2184. [25] E.A. Gutierrez, L. Deferm, G. Declerck, Experimental-determination of selfheating in submicrometer MOS-transistors operated in a liquid-helium ambient, IEEE Electron Dev. Lett. 14 (1993) 152–154. [26] J.T. Hastings, K.W. Ng, Characterization of a complementary metal–oxide– semiconductor operational-amplifier from 300 to 4.2 K, Rev. Sci. Instrum. 66 (1995) 3691–3696. [27] J. Jomaah, G. Ghibaudo, S. Cristoloveanu, A. Vandooren, F. Dieudonne, J. Pretet, F. Lime, K. Oshima, B. Guillaumot, F. Balestra, Low-temperature performance of ultimate Si-based MOSFETs, Low Temperature Electronics and Low Temperature Cofired Ceramic Based Electronic Devices, vol. 2003, 2004, pp. 118–128. [28] U. Kleine, J. Bieger, H. Seifert, A low-noise CMOS preamplifier operating at 4.2 K, IEEE J. Solid-State Circ. 29 (1994) 921–926. [29] H. Nagata, H. Shibai, T. Hirao, T. Watabe, M. Noda, Y. Hibi, M. Kawada, I. Nakagawa, Cryogenic capacitive transimpedance amplifier for astronomical infrared detectors, IEEE Trans. Electron Dev. 51 (2004) 270–278. [30] B. Okcan, G. Gielen, C. Van Hoof, A third-order complementary metal–oxide– semiconductor sigma-delta modulator operating between 4.2 K and 300 K, Rev. Sci. Instrum. 83 (2012) 024708. [31] N. Yoshikawa, T. Tomida, A. Tokuda, Q. Liu, X. Meng, S.R. Whiteley, T. Van Duzer, Characterization of 4 K CMOS devices and circuits for hybrid JosephsonCMOS systems, IEEE Trans. Appl. Supercond. 15 (2005) 267–271. [32] S.R. Ekanayake, T. Lehmann, A.S. Dzurak, R.G. Clark, A. Brawley, Characterization of SOS–CMOS FETs at low temperatures for the design of integrated circuits for quantum bit control and readout, IEEE Trans. Electron Dev. 57 (2010) 539–547. [33] F. Balestra, G. Ghibaudo, Brief review of the MOS device physics for lowtemperature electronics, Solid-State Electron. 37 (1994) 1967–1975. [34] J.E. Meyer, MOS models and circuit simulation, RCA Rev. 32 (1971). 42-&. [35] T. Ytterdal, Y. Cheng, T.A. Fjeldly, Device Modeling for Analog and RF CMOS Circuit Design, Wiley, Chichester, 2003. [36] D. Zilic, D. Pajic, M. Juric, K. Molcanov, B. Rakvin, P. Planinic, K. Zadro, Single crystals of DPPH grown from diethyl ether and carbon disulfide solutions – crystal structures, EPR and magnetization studies, J. Magn. Reson. 207 (2010) 34–41. [37] R.F. Wenzel, Y.W. Kim, Linewidth of electron paramagnetic resonance of (Al2O3)1X(Cr2O3)X, Phys. Rev. 140 (1965). 1592-&. [38] T.T. Chang, D. Foster, A.H. Kahn, Intensity standard for electron-paramagnetic resonance using chromium-doped corundum (Al2O3–Cr3+), J. Res. Natl. Bur. Stand. 83 (1978) 133–164. [39] P.D. Desai, H.M. James, C.Y. Ho, Electrical-resistivity of aluminum and manganese, J. Phys. Chem. Ref. Data 13 (1984) 1131–1172. [40] A. Dorn, R. Stoffel, H. Matile, A. Bubendorf, R.G. Ridley, Malarial haemozoin beta-hematin supports heme polymerization in the absence of protein, Nature 374 (1995) 269–271. [41] S. Pagola, P.W. Stephens, D.S. Bohle, A.D. Kosar, S.K. Madsen, The structure of malaria pigment beta-haematin, Nature 404 (2000) 307–310. [42] A. Sienkiewicz, J. Krzystek, B. Vileno, G. Chatain, A.J. Kosar, D.S. Bohle, L. Forro, Multi-frequency high-field EPR study of iron centers in malarial pigments, J. Am. Chem. Soc. 128 (2006) 4534–4535.
G. Gualco et al. / Journal of Magnetic Resonance 247 (2014) 96–103 [43] B. Razavi, A 300-GHz fundamental oscillator in 65-nm CMOS technology, IEEE J. Solid-State Circ. 46 (2011) 894–903. [44] N. Landsberg, E. Socher, 240 GHz and 272 GHz Fundamental VCOs using 32 nm CMOS technology, IEEE Trans. Microwave Theory Tech. 61 (2013) 4461–4471. [45] E. Seok, D. Shim, C. Mao, R. Han, S. Sankaran, C. Cao, W. Knap, K. Kenneth, Progress and challenges towards terahertz CMOS integrated circuits, IEEE J. Solid-State Circ. 45 (2010) 1554–1564. [46] K.K. O, S. Sankaran, C. Cao, E.-Y. Seok, D. Shim, C. Mao, R. Han, Millimeter wave to terahertz in CMOS, Int. J. Hi. Spe. Ele. Syst. 19 (2009) 55–67. [47] J.W. Stoner, D. Szymanski, S.S. Eaton, R.W. Quine, G.A. Rinard, G.R. Eaton, Direct-detected rapid-scan EPR at 250 MHz, J. Magn. Reson. 170 (2004) 127– 135.
103
[48] M. Tseitlin, G.A. Rinard, R.W. Quine, S.S. Eaton, G.R. Eaton, Rapid frequency scan EPR, J. Magn. Reson. 211 (2011) 156–161. [49] E.P. Harris, D.E. Mapother, Critical field of superconducting aluminum as a function of pressure and temperature above 0.3° K, Phys. Rev. 165 (1968) 522– 532. [50] W.J. Wallace, R.H. Silsbee, Microstrip resonators for electron-spin-resonance, Rev. Sci. Instrum. 62 (1991) 1754–1766. [51] C. Clauss, D. Bothner, D. Koelle, R. Kleiner, L. Bogani, M. Scheffler, M. Dressel, Broadband electron spin resonance from 500 MHz to 40 GHz using superconducting coplanar waveguides, Appl. Phys. Lett. 102 (2013) 162601.
