Nuclear spin polarization of cesium ions in the salt was enhanced during optical pumping of cesium vapor at high magnetic field. Significant motional narrowing and frequency shift of NMR signals were observed by intense laser heating of the salt. When the hyperpolarized salt was cooled by blocking the heating laser, the signal width and frequency changed during cooling and presented the phase transition from liquid to solid. Hence, we find that the signal enhancement is mostly due to the molten salt and nuclear spin polarization is injected into the salt efficiently in the liquid phase. We also show that optical pumping similarly induces line narrowing in the solid phase. The use of powdered salt provided an increase in effective surface area and signal amplitude without glass wool in the glass cells.

Optically-pumped cesium (Cs) vapor enhances nuclear spin polarization of Cs ions in the hydrides,1 the halides,2 and a borosilicate glass.3 Because the coupling with electron spin is absent once the atoms are ionized, the nuclear polarization has a sufficiently-long longitudinal relaxation time T1 to be transferred to the salts. At the salt surface, magnetic dipole interaction (dipole-dipole cross relaxation)4 between the ions of the same nuclear species provides spin polarization transfer from vapor to solid,5 similar in effect to spin-exchange optical pumping of noble gases.6 When the glass matrix is softened at high temperature, the polarized atoms are ionized and injected into the glass structure,3 suggesting that the polarized ions can replace the thermally-polarized ions at the surface of hot salt (called ion-exchange). The salt clusters would be polarized by the spin transfer from the atoms, and then condensed on the salt particles. In the salt, dipolar interaction is the primary mechanism of spin diffusion near the room temperature2,7 and diffusive ion migration is effective at T ≳ 300 °C.8,9 The spin diffusion length, D T 1 , remains in the balance of D (diffusion coefficient) and T1 in the solid and the liquid phases.10 The above-described picture of spin-injection optical pumping is shown in Fig. 1.

FIG. 1.

Spin-injection optical pumping of Cs salts. The Cs atoms are optically pumped and the salt particles are heated by another laser. The spin polarization of the atoms is injected, directly or via the salt clusters, by polarized ion injection, ion exchange, and spin exchange between nuclei at the salt surface. The spin diffusion in the salt is due to dipolar interaction and ion migration. The salt is polarized by the spin diffusion length beneath the surface (penetration depth).

FIG. 1.

Spin-injection optical pumping of Cs salts. The Cs atoms are optically pumped and the salt particles are heated by another laser. The spin polarization of the atoms is injected, directly or via the salt clusters, by polarized ion injection, ion exchange, and spin exchange between nuclei at the salt surface. The spin diffusion in the salt is due to dipolar interaction and ion migration. The salt is polarized by the spin diffusion length beneath the surface (penetration depth).

Close modal

In a controlled environment, hyperpolarized noble gases have been studied about spin transfer from xenon and helium to surface nuclei,11–13 surface thin films,14,15 and spin relaxation and atomic diffusion in the frozen xenon.16–18 In a similar approach, the physics of spin injection into the salts can be studied and it opens up the possibility for NMR probing of surface state and nuclear polarization transport in condensed materials. At present, however, spin physics in the polarized salts is unclear because the system of salt and metal in the glass cell was easily damaged by laser heating for accurate measurements. Necessary for coexistence of molten salt and proper density of atomic vapor, but inhomogeneous temperature distribution gave different NMR line shapes. Although a glass wool increased local temperature and surface area of the salt in contact with the polarized vapor, the pumping light was randomly polarized in the scattering medium and became unsuited for optical pumping at low magnetic field.7 

The observations are confusing as described above, but able to be organized with respect to temperature. In this paper, we report on spin-injection optical pumping of molten Cs salts, CsCl and CsI. Because of chemical shift differences of 133Cs NMR signal, the phase changes from liquid to solid are clearly observed at 9.4 T. It shows that the signal enhancement arises mostly from the molten salt during optical pumping. From diagnosis based on many NMR signals, the signal enhancement and motional narrowing are also found in the solid phase. In addition, the signal of powdered salt is optically enhanced without glass wool until molten droplets fuse together.

Optical and NMR alignment in the bore of a superconducting magnet is shown in Fig. 2. Two types of glass cells were used: ground powder of salt was dispersed in a quartz-glass wool in the cells (ϕ 10 mm o.d.); and the salt powder was placed at the bottom of the cells (ϕ 6 mm o.d.) without glass wool. The salt particles were piled up and each diameter was a few micrometers measured with a microscope. A small amount of Cs metal was added into the glass cells such that the salt powder was not covered with too much metal. These were not chemically reacted at the room temperature before the optical experiment.19 Most of the glass cells were filled with nitrogen gas of 2 kPa and some cells had no buffer gas. The presence of the buffer gas did not show significant difference in the enhanced NMR signal by optical pumping at 9.4 T. Details of manufacturing the glass cells are described by Ref. 5. Temperature in the NMR probe box was measured by a non-magnetic thermocouple placed at approximately 10 mm from the glass cell. The cw outputs of optical pumping and heating lasers were introduced vertically from below. Therefore, Cs metal moved upward and a moderate density of Cs vapor was expected near the hot salt. The number densities of Cs atoms and Cs2 molecules were, respectively, 7 × 1013 cm−3 and 3 × 1010 cm−3 at the temperature of 120 °C.20 Despite heat transfer via the metal vaporization, the sparse structure of salt powder in glass wool resulted in small thermal conductivity and large temperature gradient. It is possible to irradiate the glass cell from above and below to decrease the gradient, however, this reduces sample life.3 The intense laser irradiation melted a part of glass wool during the measurement, meaning that local temperature reached the melting point of quartz glass and the powdered salts were distributed in a wide temperature range. The number density of the CsCl molecules in the gas phase was 2 × 1015 cm−3 on the solid at the melting point and an order of magnitude larger density was expected in equilibrium with the liquid.21 The number density distribution of the clusters (CsCl)n was noticeable,19,22 because the spin relaxation due to quadrupolar and spin-rotation interactions becomes less effective when the cluster size increases. Because of local elevation of temperature, we were able to detect the signal from the hot salt with a standard NMR coil, and yet at the same time it resulted in a lack of information on the salt temperature by an external sensor.

FIG. 2.

Optical and NMR alignment near the cylindrical glass cell. NMR probe was electrically heated at typically 100 °C. Tuning capacitors were isolated by a partition plate from the high temperature region. Optical pumping from Ti:sapphire laser (1.2 W, 852 nm) and heating from fiber laser (≤50 W, 1060 nm) were introduced to the glass cell from below. A single rf pulse was applied to detect free-induction decay signal by a saddle coil. Quartz-glass wool was added to the cell (ϕ 10  mm o.d.). The grains of Cs salt are shown by red balls and the droplets of Cs metal are not explicitly shown. Quartz-glass tube was used for fitting the narrow glass cell (ϕ 6  mm o.d.) to the coil.

FIG. 2.

Optical and NMR alignment near the cylindrical glass cell. NMR probe was electrically heated at typically 100 °C. Tuning capacitors were isolated by a partition plate from the high temperature region. Optical pumping from Ti:sapphire laser (1.2 W, 852 nm) and heating from fiber laser (≤50 W, 1060 nm) were introduced to the glass cell from below. A single rf pulse was applied to detect free-induction decay signal by a saddle coil. Quartz-glass wool was added to the cell (ϕ 10  mm o.d.). The grains of Cs salt are shown by red balls and the droplets of Cs metal are not explicitly shown. Quartz-glass tube was used for fitting the narrow glass cell (ϕ 6  mm o.d.) to the coil.

Close modal

The D2 absorption line of Cs atoms, 6s2S1/2(mS)↔6p2P3/2(mJ), was split by the electron Zeeman effect at 9.4 T, where mS and mJ are magnetic quantum numbers of the electron. The pumping laser was tuned to one of the lines and linearly polarized before the glass cell. The nuclear spin polarization of pumped atoms, 〈mI〉/I, depends on the laser frequency and polarization. The polarized atoms transport the nuclear polarization within the buffer gas. Since the polarization decays at the salt surface, the diffusion nuclear spin current flows to the surface.5 As shown in Fig. 3, each polarized light induces the nuclear spin current at the high field. There are two main processes: direct optical pumping and collision-induced pumping. The former induces the spin current via ‘forbidden’ transition, where angular momentum is conserved among nuclei, electrons, and photons with the help of hyperfine interaction. The spectral width is approximately 10 GHz with hyperfine splitting and Doppler broadening. The latter induces positive (negative) nuclear spin current when the ground-state sublevel mS = − 1/2 (+1/2) is depopulated. Therefore, the largest spin current occurs via ‘allowed’ transition, as shown in Table I. The broadening ∼50 GHz is due to a slow rate of collision-induced spin exchange between nuclei and electrons. The spin current induced by well-defined polarized light was discussed thus far. In the experiment, atomic polarization very close to the surface is most important since the nuclear polarization is injected to the salts at the surface. The light polarization is modified by light scattering at rough surfaces and by light emission and re-absorption of atoms due to the surface spin relaxation.5,23 The conversion of light polarization is more notable by using glass wool and powdered salts. Therefore, different from optical pumping in free space, the randomly-polarized light is assumed for spin-injection optical pumping. At the transitions D2(a) and D2(h), light polarization mixing has little influence on the sign of spin current because the atoms absorb only a single component of light polarization, as shown in Table I. At D2(c) and D2(f), the spin currents have the same sign for various combinations of light polarizations and pumping processes. For the other transitions D2(b), D2(d), D2(e), and D2(g), the induced current depends on the situation. At D2(b), for example, the spin current is positive by σ+ pumping and can be inverted by direct optical pumping of π light. Therefore, the sign depends on the mixing ratio of the σ+ and π lights near the surface. The observed nuclear polarization at D2(b) is described in the section V.

FIG. 3.

Diffusion nuclear spin current calculated at the side wall of cylindrical cell, induced by σ (blue), π (green), and σ+ (red) lights with uniform power density of 100 mW/cm2 and 2 kPa N2 at 100 °C. The positive (negative) sign means that the positive (negative) nuclear polarization flows to the wall. Laser detuning is measured with respect to the center of the D2 line. The Zeeman splitting is large enough to see the shape of lines (a–h) at 9.4 T. The current of narrow width is induced by direct optical pumping. Collision-induced pumping is closely related with depopulation pumping via allowed transition, and the induced current is broad because the optical transition rate becomes the same order of the hyperfine-shift collision rate ∼1 s−1 at the detuning of several tens of GHz.5 

FIG. 3.

Diffusion nuclear spin current calculated at the side wall of cylindrical cell, induced by σ (blue), π (green), and σ+ (red) lights with uniform power density of 100 mW/cm2 and 2 kPa N2 at 100 °C. The positive (negative) sign means that the positive (negative) nuclear polarization flows to the wall. Laser detuning is measured with respect to the center of the D2 line. The Zeeman splitting is large enough to see the shape of lines (a–h) at 9.4 T. The current of narrow width is induced by direct optical pumping. Collision-induced pumping is closely related with depopulation pumping via allowed transition, and the induced current is broad because the optical transition rate becomes the same order of the hyperfine-shift collision rate ∼1 s−1 at the detuning of several tens of GHz.5 

Close modal
TABLE I.

Positive (+) and negative (−) nuclear spin currents are induced by optical pumping of the light polarizations (σ, π, σ+). Optical transitions D2(n) (mS, mJ), respectively, correspond to the eight lines shown in Fig. 3. Collision-induced pumping (c) and direct optical pumping (d) are presented as c/d, and dominant component is indicated by red color. The blank means that the spin current by resonant transition is negligible.

 
 

In a state of thermodynamic equilibrium, the nuclear polarization of Cs ions in the salt is parallel to the external magnetic field because of positive nuclear dipole moment. The signals of thermal ions will be shown as positive. In the optical experiment, atomic nuclear polarization is injected into the salt by spin and ion-exchange mechanisms at the surface. The injected polarization, parallel to atomic nuclear polarization, is compared to the thermal polarization. Atomic property related with spin injection, such as atom density and characteristic time of motion at the surface, depends on the temperature and the phase, solid or liquid. Therefore, when used with spin-injection optical pumping, the NMR measurement is potentially sensitive to the surface state such as the liquid layer known as premelting24,25 and prewetting26 below the melting point. The enhancement of nuclear polarization and the longitudinal spin relaxation time of Cs ions depend less on the species of anion and the applied magnetic field than on the crystal symmetry.3 Because of small electric quadrupole moment, a single resonance peak of 133Cs NMR signal is observed in cubic symmetry of each Cs halide, but the quadrupole interaction is detectable in low symmetry crystals and the surface.27 The relaxation time T1 is long enough for spin injection, for example, 590 s in the solid CsCl at 120 °C.2 The resonance frequency is 52.5 MHz and far from the resonance in the metal 53.26 MHz. According to magic-angle-spinning NMR, the line widths in CsCl and CsI crystals are, respectively, 74 Hz and 394 Hz at 9.4 T and room temperature. The width by static NMR is 800  Hz for both halides.28 Ion migration and dipolar interaction between Cs nuclei contribute to the diffusion of spin polarization from the surface into the salt. Generally, ionic conductivity is a function of temperature such as the Boltzmann factor. In addition, the conductivity in CsCl is substantially changed by the phase transition between the CsCl (cubic) and NaCl (octahedral) types at approximately 480 °C and by melting at 645 °C. Since dipolar interaction is nearly independent of temperature, ion migration becomes effective to spin transport by heating. The diffusion coefficient of Cs ions is approximately 10−9 cm2 s−1 in the solid CsCl at the melting point.8,9 Therefore, we measured nuclear polarization of the salts in a wide range of temperature, over the melting point. Table II shows characteristic temperatures of materials in the glass cells.

TABLE II.

Characteristic temperatures for Cs metal, CsCl, CsI and borosilicate glass. Phase transition (pt) between the NaCl and CsCl structural types,9 melting point (mp), boiling point (bp),19 softening point (sp), and working point (wp, glass blowing temperature).29 

T (°C) 480 626 645 685 820 1252 1280 1297
  CsCl  CsI  CsCl  Cs  glass  glass  CsI  CsCl 
  pt  mp  mp  bp  sp  wp  bp  bp 
T (°C) 480 626 645 685 820 1252 1280 1297
  CsCl  CsI  CsCl  Cs  glass  glass  CsI  CsCl 
  pt  mp  mp  bp  sp  wp  bp  bp 

For the salt CsCl in the cell temperature of 110 °C, the thermal 133Cs NMR signal was observed at 3.6  kHz from the reference frequency of 52.5 MHz. As shown in Fig. 4, the signal was enhanced during optical pumping and heating for a few minutes. A major part of the enhanced signal was positive at approximately 2.7  kHz by pumping at D2(h) and D2(b), and negative at D2(a) and D2(c). The other major peak was near the frequency of thermal signal. Total enhancement of the double-peaked signals was measured by the spectral area and the maximum was 25, which corresponds to mean nuclear polarization of 0.02 %. The condition of optical pumping and heating was changed by beam position, salt and metal distribution, density of atomic vapor, polarization of scattered light, temperature and its gradient, and a history of the cell use. Therefore, the resonance frequency and the amplitude of the peaks varied with time and the particular glass cell, which is the problem to be solved. The fluctuation made it difficult to study what the salt temperature was best for nuclear spin injection.

FIG. 4.

Single shot measurement of optically enhanced 133Cs NMR signal of CsCl. Cs metal, CsCl powder, quartz-glass wool, and N2 gas were sealed in the glass cells (ϕ10). The signal was positively (red) enhanced by optical pumping at D2(b) and D2(h), and negatively (green) at D2(a) and D2(c). The signals in different glass cells (A, B, and C) are indicated by dashed, solid, and dotted curves, respectively. The amplitude can be compared only within each glass cell. NMR frequency is measured with respect to the synthesizer frequency 52.5 MHz. The thermal signal at 3.6 kHz and 110 °C (black) is ten times expanded in the vertical scale (gray). Wavy-form noises are shown in the expanded signal measured in a single shot. The center frequency and FWHM of Gaussian fitting are shown in the inset.

FIG. 4.

Single shot measurement of optically enhanced 133Cs NMR signal of CsCl. Cs metal, CsCl powder, quartz-glass wool, and N2 gas were sealed in the glass cells (ϕ10). The signal was positively (red) enhanced by optical pumping at D2(b) and D2(h), and negatively (green) at D2(a) and D2(c). The signals in different glass cells (A, B, and C) are indicated by dashed, solid, and dotted curves, respectively. The amplitude can be compared only within each glass cell. NMR frequency is measured with respect to the synthesizer frequency 52.5 MHz. The thermal signal at 3.6 kHz and 110 °C (black) is ten times expanded in the vertical scale (gray). Wavy-form noises are shown in the expanded signal measured in a single shot. The center frequency and FWHM of Gaussian fitting are shown in the inset.

Close modal

The resonance frequency and the width, obtained by fitting two Gauss functions, are shown in the inset of Fig. 4. The linewidth of the low-frequency peak (380  Hz) is narrower than the width (800 Hz) of the thermal signal. Since this peak is of the hot salt as described below, the narrowing is caused by motional averaging by fast ion migration. The width measured at 0.56 T was 78 Hz.10 No matter how intensively the salt was laser heated at 9.4 T, the signal was not narrowed further but the glass cell was melted. One reason is that ion motion is ineffective at high field: narrowing begins when the hopping frequency comes close to the line width in a solid, and further narrowing is due to hopping near the resonance frequency, which is simply achieved at low field.30 One other reason is that the signal is inhomogeneously broadened in a temperature gradient since the NMR frequency depends on the temperature significantly at high field.

In a single shot measurement during optical pumping, the inhomogeneously broadened signals consisted of narrow lines with respectively different frequencies, and the enhancement was mostly observed between 2.6 kHz and 3.1 kHz. The signal amplitude is proportional to the amount of salt and the nuclear polarization at the corresponding temperature. The spin injection efficiency can be extracted when the salt particles are in similar contact with the polarized vapor. By the external sensor, it is difficult to measure the local temperature of the salts in laser heating. Therefore, the NMR frequency of thermal signal was used as the temperature scale. As shown in Fig. 5, the resonance frequency was shifted from 3.6 kHz to 2.6 kHz and the linewidth was narrowed by an increase in heating laser power. The salt signal disappeared by further heating, and then another peak appeared at 1 kHz. The newly observed signal was detected during the process of melting the glass cells. For a number of measurements of the salt in the melted glass, no signals were observed in the frequency gap between 1 kHz and 2.6 kHz. Because the salt melts below the softening temperature of the borosilicate glass (see Table II), the narrow signals near 2.7 kHz should be from molten salt. Since the powdered salt was placed at the bottom of glass cell (ϕ 6) with no glass wool, temperature difference was small across the sample. The linewidth 170 Hz, which is larger than 78 Hz at 0.56 T, was an attainable minimum in residual temperature gradient at 9.4 T.

FIG. 5.

Thermal 133Cs NMR signal of CsCl averaged several times without optical pumping. There were Cs metal, CsCl powder, and N2 gas in the glass cells (ϕ 6) and no glass wool. It was almost the same condition as in the optical pumping experiment. Laser heating made the signal narrowed and moved toward lower frequency. The narrowest line (FWHM = 170 Hz) was observed at 2.7 kHz. By more intense heating, the salt signal disappeared and one other peak appeared at 1 kHz in the process of melting the glass. A peak has never been observed at 1–2.6 kHz. The signals are plotted on different vertical scales. The signal frequency and width are measured within a reasonable range of variations, as shown in Fig. 11.

FIG. 5.

Thermal 133Cs NMR signal of CsCl averaged several times without optical pumping. There were Cs metal, CsCl powder, and N2 gas in the glass cells (ϕ 6) and no glass wool. It was almost the same condition as in the optical pumping experiment. Laser heating made the signal narrowed and moved toward lower frequency. The narrowest line (FWHM = 170 Hz) was observed at 2.7 kHz. By more intense heating, the salt signal disappeared and one other peak appeared at 1 kHz in the process of melting the glass. A peak has never been observed at 1–2.6 kHz. The signals are plotted on different vertical scales. The signal frequency and width are measured within a reasonable range of variations, as shown in Fig. 11.

Close modal

Based on the above-described thermal signals, we study temperature dependence of the enhanced signal by constant power of optical pumping and heating lasers, where the position and the beam profile of both lasers were also unchanged. The temperature of the glass cell was increased by the electric heater in the NMR probe. Therefore, the salt temperature was increased mostly by the lasers but also changed a little according to the cell temperature. As shown in Fig. 6, the signal at 2.7 kHz was enhanced with an increase in the cell temperature. The frequency and the width were little changed because the salt was heated close to the maximum temperature (softening point) of borosilicate glass cell. Most of amplitude came from the polarized molten salt. The signal was also enhanced at 3.4 kHz, where the frequency corresponds to 120 °C, despite a decrease in the amount of the solid by melting. It means that the powdered salt was distributed in a wide temperature range; one was in the solid phase, the others were in the liquid phase, and someone could be melted at the surface of solid particle.

FIG. 6.

Enhanced 133Cs NMR signal of CsCl measured at the fixed power of heating and pumping lasers. There were Cs metal, CsCl powder, quartz-glass wool, and N2 gas in the glass cell (ϕ10). The cell was heated by increasing power of the electric heater and the temperature was measured by the thermocouple. In addition to the signal at 2.7 kHz, the signal was also enhanced at 3.4 kHz though solid salt was largely lost to melting. Since the thermal signals in the solid phase (dotted curves) were almost the same at different temperatures, the detection efficiency of NMR circuit was unchanged in the temperature range. The inset shows FWHM of signals at 2.7 kHz (•) and 3.4 kHz (▴).

FIG. 6.

Enhanced 133Cs NMR signal of CsCl measured at the fixed power of heating and pumping lasers. There were Cs metal, CsCl powder, quartz-glass wool, and N2 gas in the glass cell (ϕ10). The cell was heated by increasing power of the electric heater and the temperature was measured by the thermocouple. In addition to the signal at 2.7 kHz, the signal was also enhanced at 3.4 kHz though solid salt was largely lost to melting. Since the thermal signals in the solid phase (dotted curves) were almost the same at different temperatures, the detection efficiency of NMR circuit was unchanged in the temperature range. The inset shows FWHM of signals at 2.7 kHz (•) and 3.4 kHz (▴).

Close modal

The NMR frequency at the melting point of the salt gives key information about the temperature range for efficient spin injection. The hot salt was cooled by blocking the pumping and heating lasers, and the enhanced signals were recorded during the cooling process. Because of no laser irradiation during signal acquisition, the signal amplitude, frequency, and width changed stably and monotonously, as shown in Fig. 7. Generally, the sample temperature ceases to change at the freezing point due to release of latent heat. Therefore, the phase transition is detected as the stationary point in a conventional temperature-time curve.31 In our measurement, the chemical shift and the width should change significantly at the freezing point since they were, respectively, different between the liquid and the solid phases, as shown in Fig. 5. In spite of slow temperature change at the freezing point, the measured frequency rapidly increased at the ten-second time point in Fig. 7(c), and the signal width was discontinuously broadened in Fig. 7(d). Assuming no supercooling, the phase change from liquid to solid occurred at approximately 3.1 kHz. Since the signal was most enhanced in the frequency range of the liquid phase, nuclear polarization was injected to the molten salt more than the solid. A broad peak overlapped at the high frequency side for the first signal in Fig. 7(a). Based on the frequency, a small amount of salt remained solid near the melting point. Figure 7(b) shows the nuclear polarization during the cooling. The decay time 13.9 s was four orders of magnitude larger than in the Cs metal. Due to tip angle by rf pulse 35 (0.6 rad), it became smaller than longitudinal relaxation time in the salt.2 According to the phase diagram of the CsCl-Cs system,19 the freezing point depression is negligible by the metal impurity. Therefore, the observed melting point was very close to that of pure salt.

FIG. 7.

(a) 133Cs NMR signals of CsCl (solid curves) were recorded every 3 s after pumping and heating lasers were blocked. The tip angle by a single rf pulse was approximately 35 (0.6 rad). Due to nuclear polarization decay by rf pulses, a few signals became smaller than the thermal signal at 90 °C (dotted curve). Cs metal, CsCl powder, quartz-glass wool, and N2 gas were sealed in the glass cell (ϕ10). (b) Signal enhancement calculated from the spectral areas of the enhanced and the thermal signals. (c) Center frequency of Gaussian fitting. The top horizontal axis presents the frequency, 3.49 kHz, of the thermal signal. (d) FWHM of fitting function. The dotted line presents the line width of the thermal signal. Two Gauss functions are used for fitting before 10 s. The solid curves in (b), (c), and (d) are intended only as visual aids.

FIG. 7.

(a) 133Cs NMR signals of CsCl (solid curves) were recorded every 3 s after pumping and heating lasers were blocked. The tip angle by a single rf pulse was approximately 35 (0.6 rad). Due to nuclear polarization decay by rf pulses, a few signals became smaller than the thermal signal at 90 °C (dotted curve). Cs metal, CsCl powder, quartz-glass wool, and N2 gas were sealed in the glass cell (ϕ10). (b) Signal enhancement calculated from the spectral areas of the enhanced and the thermal signals. (c) Center frequency of Gaussian fitting. The top horizontal axis presents the frequency, 3.49 kHz, of the thermal signal. (d) FWHM of fitting function. The dotted line presents the line width of the thermal signal. Two Gauss functions are used for fitting before 10 s. The solid curves in (b), (c), and (d) are intended only as visual aids.

Close modal

The glass cell was warmed up by the heating laser (beam diameter ≈ cell diameter) at 50 W for sufficient time until all the salt melted. After that, the salt was slowly cooled down by decreasing the laser power to 20 W. As shown in Fig. 8(a), the signal was positively (negatively) enhanced by continuous optical pumping at D2(b) (D2(c)). The thermal signal was measured by detuning the pumping laser from atomic transitions, with a similar heating effect on the salt. A single enhanced peak was observed at the frequency of the liquid phase and no signal from the solid phase. These positive, negative, and thermal signals were recorded every five seconds in the three cooling sequences. The salt temperature was scanned enough over a wide range in the liquid phase. Since the thermal signal was constant in amplitude during the acquisition sequence, the sensitivity of the NMR circuit was reliable in spite of wide temperature range. Owing to slow cooling, the center frequency and the linewidth were reproduced for two sequences, as shown in Figs. 8(b) and 8(c). Because of small rf pulses of approximately 12 (0.21 rad), the spin polarization was sufficiently enhanced at every acquisition time. As shown in Fig. 8(d), the pumped signals were nearly as large as the thermal signal at 0 s and the enhancement was increased by cooling. The relaxation time T1 and the diffusion coefficient of Cs ions D are temperature dependent. Therefore, the spin-diffusion length should be maximized at an optimum temperature. The maximum enhancement of 40 was obtained at 2.97 kHz in that glass cell. The optimal frequency corresponds to the salt temperature approximately 700 °C, estimated by linear interpolation of the softening point of glass (820 °C, 2.65 kHz) and the melting point of salt (645 °C, 3.10 kHz). It is much higher than the temperature, 400 °C, previously estimated from motional narrowing at low magnetic field.10 As shown in the previous figures, the signals by continuous pumping and heating have the maximum amplitude below 2.97 kHz, meaning that the salt was heated beyond the optimal temperature.

FIG. 8.

133Cs NMR signals were recorded every 5 s by single rf pulses of a small tip angle ∼12 (0.21 rad). The heating laser power was decreased from 50 W to 20 W at 0 s during continuous pumping. Cs metal, CsCl powder, quartz-glass wool, and N2 gas were sealed in the glass cell (ϕ10). (a) Positive (D2(b) pumping, red curve), negative (D2(c), green), and thermal (the pumping laser was detuned from absorption lines, black) signals were measured at approximately 40 s. Time variations of (b) center frequency, (c) FWHM, and (d) enhancement were extracted from the positive (▴, red) and negative (▾, green) signals. Since the salt temperature varied similarly for different series of acquisition, time variations of frequency and line width were reproduced in both measurements. The integrated area of the thermal signal was almost constant through the other run.

FIG. 8.

133Cs NMR signals were recorded every 5 s by single rf pulses of a small tip angle ∼12 (0.21 rad). The heating laser power was decreased from 50 W to 20 W at 0 s during continuous pumping. Cs metal, CsCl powder, quartz-glass wool, and N2 gas were sealed in the glass cell (ϕ10). (a) Positive (D2(b) pumping, red curve), negative (D2(c), green), and thermal (the pumping laser was detuned from absorption lines, black) signals were measured at approximately 40 s. Time variations of (b) center frequency, (c) FWHM, and (d) enhancement were extracted from the positive (▴, red) and negative (▾, green) signals. Since the salt temperature varied similarly for different series of acquisition, time variations of frequency and line width were reproduced in both measurements. The integrated area of the thermal signal was almost constant through the other run.

Close modal

In the above experiment, the use of glass wool provided an increase in surface area of the salt in contact with the polarized vapor. A large enhancement was obtained at the expense of the simplicity of sample configuration. In this section, we demonstrate hyperpolarization of the salt without glass wool. As shown in Fig. 2, the powdered salt was placed at the bottom window of the cell (ϕ 6) and the droplets of Cs metal were on the upper walls. The salts were colored blue-purple by thin metal film.32 Figure 9 shows the signal enhanced by the factor of 15 during D2(b) pumping and laser heating. The center frequency of 2.65 kHz corresponds to the temperature near the glass softening point. It means that the salt was mostly in the liquid phase and the signal in the solid phase was small. Counterintuitively, we have often observed the solid signal by increasing the heating laser power. One possible explanation is that a part of molten salt moves from a locally hot place to a relatively cold place, and solidifies. Since nothing supports the granular salt, the liquid droplets come together by heating. The larger polycrystals were found on the bottom window after the optical experiment. When the salts were melted prior to the experiment, the signal was not significantly enhanced by optical pumping. Therefore, in-situ melting of the salt during measurement is important to produce liquid droplets with a large surface area in contact with the polarized vapor.

FIG. 9.

133Cs NMR signals enhanced by optical pumping at D2(b) and laser heating of 20 W (blue curve) and 25 W (red). Cs metal and CsCl powder were sealed in the glass cell (ϕ 6) with no glass wool and no N2 gas. The beam diameter of heating and pumping lasers was matched with the size of the cell. Compared with the thermal signal measured at 120 °C (black), the signal at 3.5 kHz (red) became narrow by laser irradiation. The signal width and enhancement are shown in the inset.

FIG. 9.

133Cs NMR signals enhanced by optical pumping at D2(b) and laser heating of 20 W (blue curve) and 25 W (red). Cs metal and CsCl powder were sealed in the glass cell (ϕ 6) with no glass wool and no N2 gas. The beam diameter of heating and pumping lasers was matched with the size of the cell. Compared with the thermal signal measured at 120 °C (black), the signal at 3.5 kHz (red) became narrow by laser irradiation. The signal width and enhancement are shown in the inset.

Close modal

A different cesium salt CsI was tested for hyperpolarization without glass wool. Because the chemical shift difference is 200 Hz at most between liquid and solid,28,33 inhomogeneous broadening should be small in a temperature gradient and we do not measure the salt temperature and the phase change by the NMR signals. Figure 10(a) shows the signal development during laser heating and optical pumping at D2(b). Compared to the thermal signal, the positive signal was first narrowed and then inverted by continuous laser irradiation. The negative sign, opposite from the signals shown in Figs. 4, 8, and 9, has been observed both in CsCl and CsI. A similar signal was reported in Fig. 9 of Ref. 5 for reasons unknown at that time. Optical pumping at D2(b) consists of several competitive optical and collision-induced processes, as shown in Table I. Among them, only direct optical pumping with π light inverts nuclear polarization. The linearly-polarized light was introduced along the magnetic field with no π component before the glass cell. A part of the σ light was converted to the π light by light scattering at the rough surface of the powdered salt. The processes induced by the σ and π lights canceled out each other, resulting in a small and inverted signal. For large enhancement, the laser was tuned to the transition D2(a) in the same glass cell. Even if the pumping light was randomly polarized, atomic vapor was pumped only by the σ+ light. Therefore, the signal was stably and negatively enhanced with the best enhancement of −16, as shown in Fig. 10(b). The line width of 394 Hz was smaller than the thermal signal, and similar to those measured in the glass-wool cells33 and in magic angle spinning.28 

FIG. 10.

Optically enhanced 133Cs NMR signals of CsI at the cell temperature 120 °C. Cs metal, CsI powder, and N2 gas were sealed in the glass cell (ϕ6) with no glass wool. (a) The thermal signal was positive (dotted curve). During D2(b) pumping and laser heating, the signal was narrowed and then inverted as a small negative dip (solid curves). (b) The signal was inverted by D2(a) pumping and laser heating. The number on each curve presents the time of laser irradiation. Note different vertical scale for positive and negative signals. Inset shows the signal width and enhancement by D2(a) pumping.

FIG. 10.

Optically enhanced 133Cs NMR signals of CsI at the cell temperature 120 °C. Cs metal, CsI powder, and N2 gas were sealed in the glass cell (ϕ6) with no glass wool. (a) The thermal signal was positive (dotted curve). During D2(b) pumping and laser heating, the signal was narrowed and then inverted as a small negative dip (solid curves). (b) The signal was inverted by D2(a) pumping and laser heating. The number on each curve presents the time of laser irradiation. Note different vertical scale for positive and negative signals. Inset shows the signal width and enhancement by D2(a) pumping.

Close modal

As shown in Fig. 9, the solid-state signal at 3.5 kHz was enhanced during optical pumping and heating, and it was narrower than the thermal signal at 120 °C. In contrast, the broadened signals were observed during cooling of the polarized salt, as shown in Fig. 7. Since the linewidth was uncorrelated with nuclear polarization, the width change was not caused by spin ordering. To see a full picture of the linewidth and the center frequency, the above figures and other data sets are summarized, as shown in Fig. 11. These were measured in various experimental conditions with respect to glass wool, cell temperature, heating laser power and focusing, and optical pumping transition. Nonetheless, a difference in the linewidth is clearly presented in the solid phase. It seems to be due to high ion-mobility in a hot solid.8–10 However, the motional averaging alone could not describe the observed narrowing since it reproduces the identical width at the same temperature (frequency) in a homogeneous isotropic material. There should be another important point to consider for the observed width in the solid phase. This issue remains unresolved.

FIG. 11.

Linewidth and center frequency of Gaussian fitting. These are extracted from Fig. 5 and the other measurements (∘) for thermal signal, Fig. 4 (▴), Fig. 6 (▾), Fig. 7 (◂) for cooling without pumping and heating, Fig. 8 and similar data sets (▸) for cooling with pumping and weak heating, Fig. 9 (■), and the others with optical pumping (•). The error bar presents the fitting error. The vertical line shows approximately the frequency at the melting point of CsCl. The solid and dotted curves indicate, respectively, the lower boundaries of scattered width of the enhanced Δ νe and thermal signal Δ νt with Δ ν e 2 = Δ ν t 2 + ( 330 Hz ) 2 . The marks of width ∼ 1 kHz and frequency ∼ 3 kHz are artifacts of Gaussian fitting to the heavily-overlapped spectra.

FIG. 11.

Linewidth and center frequency of Gaussian fitting. These are extracted from Fig. 5 and the other measurements (∘) for thermal signal, Fig. 4 (▴), Fig. 6 (▾), Fig. 7 (◂) for cooling without pumping and heating, Fig. 8 and similar data sets (▸) for cooling with pumping and weak heating, Fig. 9 (■), and the others with optical pumping (•). The error bar presents the fitting error. The vertical line shows approximately the frequency at the melting point of CsCl. The solid and dotted curves indicate, respectively, the lower boundaries of scattered width of the enhanced Δ νe and thermal signal Δ νt with Δ ν e 2 = Δ ν t 2 + ( 330 Hz ) 2 . The marks of width ∼ 1 kHz and frequency ∼ 3 kHz are artifacts of Gaussian fitting to the heavily-overlapped spectra.

Close modal

Above the melting point (frequency < 3.1 kHz), the enhanced signal is broader than the thermal signal. Because of various temperature gradient, the width measured in continuous heating (▴, ▾, ■, •) is more scattered than in cooling (◂, ▸). The gradient is minimum at the lower boundary of the scattered data (solid curve). By deconvolution to separate the linewidth of the thermal signal (170 Hz) from the enhanced signal (370 Hz), residual broadening is Δ ν = 37 0 2 17 0 2 = 330  Hz at the center frequency of 2.7 kHz. Assuming Δ ν as constant, the linewidth of thermal signal (dotted curve) is obtained from the solid curve. The thermal signal originates from whole droplets and the enhanced signal is from their surface layer. Therefore, we estimate the broadening originating from the difference of the magnetic susceptibility between the surface film and the inner bulk Δ χ, as follows,15 

Δ ν ν Δ χ ,
(1)

for negligible thickness of film and spherical shape of salt particles. Since Δ ν is proportional to the resonance frequency ν, the broadening is detectable at high magnetic field. The susceptibility difference of solid and liquid (Δ χ ∼ 0.5/52 500 ∼ 1 × 10−5) and the observed broadening (Δ ν/ν ∼ 0.33/52 500 ∼ 0.6 × 10−5) are consistent with Eq. (1). The spectral shape was smoothed and broadened by averaging with respect to the size and the shape of droplets in the glass cell. In view of the linewidth (78 Hz) at 0.56 T,10 the measured broadening is a function of magnetic field. A systematic frequency difference has not been observed for the positively and negatively enhanced signals, even though it is expected for the quadrupolar broadening. Therefore, it is assumed reasonably that the salt was melted at the surface. The quadrupolar coupling at the surface can be averaged out by ionic motion in the liquid phase. For further study on spectral broadening, the higher magnetic fields are necessary for clear measurement of chemical shifts and susceptibility effects at the surface.

The nuclear spin polarization of molten cesium salts, CsCl and CsI, was investigated by optical pumping of cesium vapor in the glass cells. Natural cooling of hot spin-polarized salts enabled the measuring of the phase change from a liquid to a solid. Based on the NMR frequency at the melting point, we found that the liquid-state signal was enhanced more than the solid state. Despite a small spin relaxation time in the liquid phase, total efficiency of spin polarization injection reached a maximum above the melting point. The signals were narrowed by motional averaging of local field, but broader than expected from the thermal signals. The residual broadening can be explained by assuming that the surface layers of the salt were polarized by the pumped vapor. It shows the potential of spin-injection induced nuclear polarization to probe a surface state. However, it is not clear why the solid-state signal was also narrowed during optical pumping. The use of powdered salt provided an increase in effective surface area, obviating the need for glass wool in the cells. Since the salt droplets fused with each other and consequently became the larger ones, the capability of spin polarization was lost after a few hours of laser heating. In the glass cells used in this work, the metal and the salt moved upward from the bottom by intense laser heating. The salt was inevitably subject to a temperature gradient. The signal was enhanced by optical pumping as long as the metal and the salt droplets were left at high temperature. It was approximately a half-day experiment for a single glass cell. It is important to develop methods for increasing surface area of the salts in a long-lasting way.

This work was supported in part by JSPS KAKENHI Grant No. 25610115.

1.
K.
Ishikawa
,
B.
Patton
,
Y.-Y.
Jau
, and
W.
Happer
,
Phys. Rev. Lett.
98
,
183004
(
2007
).
2.
K.
Ishikawa
,
Phys. Rev. A
84
,
033404
(
2011
).
5.
K.
Ishikawa
,
B.
Patton
,
B. A.
Olsen
,
Y.-Y.
Jau
, and
W.
Happer
,
Phys. Rev. A
83
,
063410
(
2011
).
6.
T. G.
Walker
and
W.
Happer
,
Rev. Mod. Phys.
69
,
629
(
1997
).
7.
K.
Ishikawa
,
Phys. Rev. A
84
,
013403
(
2011
).
8.
I. M.
Hoodless
and
R. G.
Turner
,
J. Phys. Chem. Solids
33
,
1915
(
1972
).
9.
J.
Arends
and
H.
Nijboer
,
Solid State Commun.
5
,
163
(
1967
).
10.
K.
Ishikawa
,
Phys. Rev. A
84
,
061405(R)
(
2011
).
11.
T.
Rõõm
,
S.
Appelt
,
R.
Seydoux
,
E. L.
Hahn
, and
A.
Pines
,
Phys. Rev. B
55
,
11604
(
1997
).
12.
B.
Driehuys
,
G. D.
Cates
, and
W.
Happer
,
Phys. Rev. Lett.
74
,
4943
(
1995
).
13.
K. L.
Sauer
,
R. J.
Fitzgerald
, and
W.
Happer
,
Phys. Rev. A
59
,
R1746
(
1999
).
14.
G.
Tastevin
,
P.
Nacher
,
L.
Wiesenfeld
,
M.
Leduc
, and
F.
Laloë
,
J. Phys. (Paris)
49
,
1
(
1988
).
15.
D.
Raftery
,
H.
Long
,
L.
Reven
,
P.
Tang
, and
A.
Pines
,
Chem. Phys. Lett.
191
,
385
(
1992
).
16.
W. M.
Yen
and
R. E.
Norberg
,
Phys. Rev.
131
,
269
(
1963
).
17.
R. J.
Fitzgerald
,
M.
Gatzke
,
D. C.
Fox
,
G. D.
Cates
, and
W.
Happer
,
Phys. Rev. B
59
,
8795
(
1999
).
18.
N. N.
Kuzma
,
B.
Patton
,
K.
Raman
, and
W.
Happer
,
Phys. Rev. Lett.
88
,
147602
(
2002
).
19.
M. A.
Bredig
,
Molten Salt Chemistry
,
V. Mixture of Metals with Molten Salts
(
Interscience
,
New York
,
1964
).
20.
A. N.
Nesmeyanov
,
Vapor Pressure of the Elements
(
Academic Press
,
1963
).
21.
G. E.
Cogin
and
G. E.
Kimball
,
J. Chem. Phys.
16
,
1035
(
1948
).
22.
Y. J.
Twu
,
C. W. S.
Conover
,
Y. A.
Yang
, and
L. A.
Bloomfield
,
Phys. Rev. B
42
,
5306
(
1990
).
23.
B. A.
Olsen
,
B.
Patton
,
Y.-Y.
Jau
, and
W.
Happer
,
Phys. Rev. A
84
,
063410
(
2011
).
24.
T.
Tanaka
,
T.
Mitsui
,
K.
Sugiyama
,
M.
Kitano
, and
T.
Yabuzaki
,
Phys. Rev. Lett.
63
,
1390
(
1989
).
25.
26.
S.
Staroske
,
D.
Nattland
, and
W.
Freyland
,
Phys. Rev. Lett.
84
,
1736
(
2000
).
27.
A. R.
Haase
,
M. A.
Kerber
,
D.
Kessler
,
J.
Kronenbitter
,
H.
Krüger
,
O.
Lutz
,
M.
Müller
, and
A.
Nolle
,
Z. Naturforsch. A
32
,
952
(
1977
).
28.
S.
Hayashi
and
K.
Hayamizu
,
Bull. Chem. Soc. Jpn.
63
,
913
(
1990
).
29.
W. H.
Kohl
, in
Handbook of Materials and Techniques for Vacuum Devices
, edited by
I.
Glass
(
Springer
,
New York
,
1995
).
30.
A.
Abragam
,
Principles of Nuclear Magnetism
(
Oxford University Press
,
1961
).
31.
V. A.
Simon
,
Z. Anorg. Allg. Chem.
395
,
301
(
1973
).
32.
S.
Villalba
,
A.
Laliotis
,
L.
Lenci
,
D.
Bloch
,
A.
Lezama
, and
H.
Failache
,
Phys. Rev. A
89
,
023422
(
2014
).
33.
K.
Ishikawa
,
Micropor. Mesopor. Mater.
178
,
123
(
2013
).