Multidimensional mapping of spin-exchange optical pumping in clinical-scale batch-mode 129Xe hyperpolarizers.

We present a systematic, multiparameter study of Rb/(129)Xe spin-exchange optical pumping (SEOP) in the regimes of high xenon pressure and photon flux using a 3D-printed, clinical-scale stopped-flow hyperpolarizer. In situ NMR detection was used to study the dynamics of (129)Xe polarization as a function of SEOP-cell operating temperature, photon flux, and xenon partial pressure to maximize (129)Xe polarization (PXe). PXe values of 95 ± 9%, 73 ± 4%, 60 ± 2%, 41 ± 1%, and 31 ± 1% at 275, 515, 1000, 1500, and 2000 Torr Xe partial pressure were achieved. These PXe polarization values were separately validated by ejecting the hyperpolarized (129)Xe gas and performing low-field MRI at 47.5 mT. It is shown that PXe in this high-pressure regime can be increased beyond already record levels with higher photon flux and better SEOP thermal management, as well as optimization of the polarization dynamics, pointing the way to further improvements in hyperpolarized (129)Xe production efficiency.

Introduction

The nuclear spins of xenon and other noble gases can be hyperpolarized (HP) to order unity by the process of spin-exchange optical pumping (SEOP).[1,2] In this two-step process, the electron spins of an alkali metal vapor such as rubidium are first polarized by the absorption of angular momentum from circularly polarized light. Spin-exchange collisions between the alkali metal atoms and 129Xe then transfer the angular momentum to the 129Xe nuclear spins through Fermi hyperfine interactions, resulting in a high non-Boltzmann distribution of 129Xe spin states that increases the detection sensitivity of NMR/MRI.[3−6] The two most common approaches to hyperpolarize 129Xe via SEOP are termed continuous flow(7−14) and stopped flow(15−25) (sometimes also referred to as “batch mode”) with respect to the delivery of Xe gas to and from the polarization cell. N2 gas is typically added to the gas mixture to quench alkali metal fluorescence.[7,26,27] The batch-mode/stopped-flow systems are attractive not only because of their relative simplicity but also because they can operate in a xenon-rich regime that obviates the need to separate the polarized Xe from the N2 (or He) via cryocollection, eliminating a potential source of polarization loss[28] as well as facilitating applications using quadrupolar noble gas isotopes.[5] This production method has also been scaled up for automated production of clinically required quantities.[24,29,30]

Regardless of the polarization method, HP noble gases have seen wide application varying from fundamental physics experiments[31−34] to NMR/MRI applications including molecular biosensors,[35−37] probing structural aspects of cage molecules and proteins,[18,38−43] and studies of porous materials (to name only a few).[15,45−47] However, it has been biomedical applications that have largely driven the development of hyperpolarized MR techniques over the past decade; indeed, for gas imaging in particular, HP 129Xe can be used to assess lung function and report on functional and microstructural abnormalities.[6,48−50] A useful figure of merit for 129Xe hyperpolarizers is the total 129Xe magnetization, MXe, delivered in a clinically useful gas volume, typically ∼0.5–1 L at 760 Torr. MXe is determined by the product of nuclear spin polarization PXe and 129Xe concentration [Xe], i.e., MXe ∝ PXe[Xe]. It is therefore important to maximize MXe through both PXe and [Xe], which is challenging because PXe generally decreases as [Xe] within the SEOP-cell increases[10] (mostly because of increased alkali metal spin-destruction rates from non-spin-conserving collisions with Xe).[51,52] But fundamentally laser photons are the source of 129Xe hyperpolarization; thus, the decreasing cost of laser diodes narrowly tuned to the alkali metal rubidium D1 wavelength (794.8 nm)[53] has made economically feasible the higher photon fluxes required to improve MXe when Xe partial pressures are high.

In the present work, a 200 W laser diode array (LDA) was used in a 3D-printed, automated 129Xe polarizer[29] to study SEOP dynamics as a function of xenon density, laser power, and SEOP-cell temperature. More specifically, the SEOP polarization conditions at several partial pressures of natural abundance Xe (26.44% 129Xe isotope enrichment) were studied: (i) 275 Torr Xe and 1725 Torr N2, (ii) 515 Torr Xe and 1485 Torr N2, (iii) 1000 Torr Xe and 1000 Torr N2, (iv) 1500 Torr Xe and 500 Torr N2, and (v) 2000 Torr Xe and 200 Torr N2, where the gases were loaded with an accuracy of ±25 Torr. The reader is also directed to Supporting Information for detailed descriptions of the experimental setup, which represents the second-generation device of our HXTC consortium (Figure 1).[24,30] For each SEOP cell loading, data were obtained for a range of incident laser power levels (approximately 100, 125, 140, or 170 W) and variable SEOP-cell surface temperatures ranging from 42 to 92 °C. For each condition, measurement of 129Xe polarization dynamics allowed the rate constant for PXe accumulation (γSEOP) and the maximum attainable steady-state PXe value [PXe(t → ∞) or 129Xe Pmax] to be determined from exponential fits. The temperature of the SEOP cell was monitored by a thermistor mounted directly to its surface; temperature control allows the Rb concentration in the gas phase to be varied.[54,55]

Figure 1

(a) Spin exchange optical pumping (SEOP) 3D-printed polarizer used to perform all in situ experiments. (b) Side view of laser aligned to the SEOP oven, SEOP cell, and near-IR spectrometer. (c) Top view, with the oven lid removed to show SEOP cell.[29]

Methods

Spin-Exchange Optical Pumping (SEOP) Polarizer

The SEOP 3D-printed portable polarizer (Figure 1) consists of a 200 W frequency narrowed volume holographic grating (VHG) laser diode array (LDA), a custom 3D-printed thermoelectric cooling (TEC) optical pumping (OP) oven, a 0.5 L SEOP cell, an electromagnet providing 47 kHz 129Xe and 1H Larmor frequencies, in situ NMR polarimetry endowed by a Magritek Kea2 system, and a Magritek 88 mm bore magnet for ex situ NMR polarimetry and MRI (Magritek, Wellington, New Zealand). The components of the polarizer[29] and 47.5 mT MRI[56] have been discussed in detail previously[29] and thus are only discussed briefly here.

In Situ Low-Field NMR and IR Spectroscopy

In situ NMR polarimetry for these experiments was performed via single-shot 129Xe NMR at 47 kHz (Figure 2a) calibrated against 1H NMR at the same frequency from a sample of thermally polarized water doped with 10 mM CuSO4 inside a 0.5 L SEOP-cell phantom (200 000 scans, Figure 2b). The polarizer allows the Rb electron spin polarization, PRb, to be estimated by comparing the integrated intensities of transmitted laser spectra measured with and without the applied magnetic field (e.g., Figure 2c).[19,24] For each set of conditions, PXe was sampled every 5–20 min throughout SEOP; the process is repeated by either destroying the 129Xe polarization with a series of “crusher” pulses or allowing it to decay with the laser off. The time-course examples in Figure 2d,e show the excellent reproducibility of PXe, PRb, and γSEOP in these experiments (and those values were not sensitive to the application of the rf pulses). Once steady-state 129Xe polarization was achieved, growth curves can be extracted (e.g., Figure 2f) and fit to an exponential: PXe(t) = Pmax[1 – exp(−γSEOPt)]. In the absence of SEOP, in-cell room-temperature (rt) measurement of the spin–lattice relaxation time constant (T1) can be obtained after steady-state PXe has been achieved by quickly bringing the cell to room temperature to minimize the Rb gas-phase concentration, turning off the laser, and performing in situ NMR polarimetry while the polarization decays; for example, the data in Figure 2g were fit to an exponential decay curve: PXe(t) = PXe(0) exp[−ΓXet], where ΓXe (=1/T1) is the 129Xe spin-destruction rate, here exhibiting an ultralong in-cell 129Xe T1 of 150.5 ± 2.5 min (or 2.5 h). Particularly when optimizing SEOP under the regimes of high [Xe] and laser power, it is also important to be observant for the onset of positive feedback effects that give rise to dramatic increases in [Rb] and laser absorption over time (and ultimately poorer PXe). Examples showing the manifestation of such “Rb-runaway”[9,53,570] effects are provided in Figure 2h, which shows behavior where relatively small increases in cell surface temperature result not only in reduced peak PXe but also in reduced PXe over time. Such effects are discussed in greater detail in Results and Discussion.

Figure 2

(a) Example of an in situ low-field 129Xe NMR spectrum from a SEOP-cell during SEOP (single scan, B0 = 4.00 mT). (b) Corresponding 1H NMR spectrum from a thermally polarized water reference sample using 200 000 scans, B0 = 1.10 mT. (c) Examples of field-cycled near-IR spectra of laser light transmitted through the SEOP-cell used to estimate PRb: room temperature before SEOP (dark gray), during SEOP with B0 electromagnet on (blue), and during SEOP with B0 electromagnet turned off (red). (d, e) Examples of data sets for studying time-resolved SEOP build-up and decay kinetics using a cell containing 1000 Torr (each) of Xe and N2 gas (143 W laser power, 65 °C). (d) Plot showing reproducibility of PXe accumulation following the application of >500 rf “crusher” pulses that nearly zero-out the 129Xe polarization (time periods marked by vertical green bars); PRb (red circles) was sampled via field-cycled near-IR spectroscopy (c) before and after application of the crusher pulses. (e) Similar to (d), with 129Xe NMR signals acquired with different interpulse durations and with polarization decay observed after turning the laser off (times demarked with vertical arrows); here 129Xe decay was observed with the SEOP cell temperature maintained at 65 °C. Pulse delay (PD) refers to timing between NMR acquisitions during build-up. (f) Exponential buildup of 129Xe polarization during the SEOP process for a cell filled with 2000 Torr of Xe and 200 Torr of N2. (g) T1 decay of HP 129Xe at r.t. obtained with the laser turned off. (h) Time-course examples showing the temperature-dependent effects of nonequilibrium “Rb runaway”[9,53,570] in a 1500 Torr Xe SEOP cell using only 100 W laser power: a normal build-up curve at 72 °C (black squares), a mildly distorted build-up curve at 82 °C (red circles), and a significantly distorted build-up curve at 92 °C (orange triangles). All spectra were recorded with a surface coil using small radiofrequency (rf) excitation pulses with little to no measurable effect on 129Xe magnetization. Except for the fitting curves in (f) and (g), connecting lines are meant only to guide the eye.

Ex Situ Low-Field NMR Spectroscopy and MRI Imaging

Ex situ PXe at 47.5 mT was calculated by comparing the HP 129Xe signal with the 13C signal at 508 kHz 13C Larmor frequency from a reference sample of thermally polarized sodium 1-13C-acetate dissolved in D2O (Figure 4a). The 129Xe T1 relaxation time inside the polypropylene phantom sphere was ∼9.2 min (Figure 4c), which is sufficient for short-term storage of HP 129Xe. Moreover, this 129Xe T1 value was used in parallel experiments to precisely calibrate the rf excitation pulse for the 47.5 mT rf probe shown in Figure 5f; the image signal decay is due to both T1 decay and excitation rf pulses. A y-slice projection of fast gradient echo (GRE) imaging with millimeter-scale spatial resolution without slice selection shows the excellent 129Xe signal intensity (Figure 5a). All 20 images (Figure 5b–e, selected images shown) were acquired identically with TE = 4.0 ms, TR ≈ 80 ms (limited by the spectrometer electronics response time), 50% k-space sampling, 64 × 64 imaging matrix with 72 × 72 mm2 field of view (FOV), and a spectral width of 20 kHz. Given the relatively long T1 in the phantom (Figure 4c), the decay of the hyperpolarized signal was primarily due to rf-pulse-induced polarization loss in Figure 5f. The calibrated rf pulse width for the flip angle (2.7° ± 0.1°) was used for calculation of % PXe for the HP 129Xe post-transfer in Figure 4 (see Supporting Information for details).

Figure 4

Ex situ 47.5 mT NMR spectroscopy of HP 129Xe gas expanded into a phantom. (a) 13C NMR spectroscopy using thermal 13C polarization (% P13C = 4.1 × 10–6%) of a 17.5 mL reference sample of 5.2 g of sodium 1-13C-acetate dissolved in D2O. 256 averages were acquired at 508 kHz resonance frequency with a 90° square excitation rf pulse and a repetition time (TR) of 200 s. Acquisition time was 100 ms. (b) Ex situ 129Xe NMR spectroscopy of HP 129Xe gas ejected from the polarizer, 0.61 mmol of 129Xe spins (% PXe= 51 ± 2%); cell loading was 1000 Torr of Xe and 1000 N2. The spectrum is acquired at 558.6 kHz 129Xe resonance frequency with a single scan (2.7° excitation rf pulse) and TR = 200 ms. The rf pulse is calibrated by monitoring signal decay in the MRI images (see Figure 5) and accounting for T1 relaxation of the HP 129Xe in the phantom (c).

Figure 5

Hyperpolarized 129Xe gas MRI at 47.5 mT. (a) y-slice/projection across the center of the image shown in (b). (b–e) Selected MRI gradient echo (GRE) images from a series of 20 images. All 20 images were acquired identically with TE = 4.0 ms, TR ≈ 80 ms (limited by the spectrometer electronics response time), 50% k-space sampling, 64 × 64 imaging matrix with 72 × 72 mm2 field of view (FOV), and a spectral width of 20 kHz. (f) Decay of HP signal primarily due to rf-pulse-induced polarization loss. The temporal decay of the signal measured between individual images within 20-image series was used to calibrate rf pulse width (2.7° ± 0.1° corresponding to 20 μs at ∼80 mW) using T1 determined by the data shown in Figure 4c, because the signal decay in Figure 5f is due to both T1 decay and excitation rf-pulse-associated magnetization losses.

Results and Discussion

The dependence of 129Xe polarization and its dynamics as functions of temperature, photon flux, and xenon partial pressure was systematically studied under stopped-flow operation in the regimes of high xenon density and photon flux. Results for five Xe:N2 SEOP-cell compositions at four different LDA incident powers (approximately 100, 125, 140, and 170 W) with SEOP-cell surface temperatures ranging from 42 to 92 °C are displayed in Figure 3: Figure 3a provides example plots of % Pmax and γSEOP as functions of SEOP-cell surface temperature for a cell containing 1000 Torr of Xe and 1000 Torr of N2 and illuminated by 100 W of laser power from the LDA. Such data were used to create contour plots (“maps”) of 129Xe % Pmax and γSEOP for each Xe density as functions of laser power and SEOP-cell surface temperature (Figures 3c–l); the highest values achieved for % Pmax for each Xe:N2 mix studied are summarized in Figure 3b and Table 1 (corresponding numerical values for all data points in Figure 3 are tabulated in the Supporting Information, Table S1).

Figure 3

Mapping the conditions of Rb/129Xe SEOP for five different Xe partial pressures. (a) Example data set showing the dependence of % Pmax and γSEOP on the SEOP-cell surface temperature, here for a 1000 Torr Xe pressure cell (also filled with 1000 Torr N2) with 100 W incident laser power. (b) Dependence of % Pmax and γSEOP on Xe partial pressure for five cell Xe:N2 compositions using 170 W incident laser power and operating at the optimal cell temperature for each Xe loading. (c-l) % Pmax and γSEOP maps for five cell Xe:N2 compositions showing the interdependence on incident laser power and SEOP-cell surface temperature.

Table 1

Summary of Maximum 129Xe % Pmax, M′Xe (M′Xe = (% P)(C129Xe)), and Other 129Xe Hyperpolarizer Metrics Achieved for Five Gas Mixturesa

Xe/N2 partial pressure (Torr)	C129Xe (mM)	% PXe(max) (%)	% PXe(max,app) (%)	M′Xe(max) (mM)	production cycle time (min)	apparent production rate (L/h)
275/1725	3.9	95 ± 9	13 ± 1	3.7 ± 0.3	51	0.94
515/1485	7.4	73 ± 4	19 ± 1	5.4 ± 0.3	53	0.91
1000/1000	14.3	60 ± 2	30 ± 1	8.5 ± 0.3	98	0.49
1500/500	21.4	41 ± 1	31 ± 1	8.8 ± 0.2	84	0.57
2000/200	28.6	31 ± 1	28 ± 1	9.0 ± 0.2	98	0.49

Apparent or usable 129Xe hyperpolarization % Pmax(app) is computed according to ref (23) to reflect that 129Xe is diluted by N2 gas as follows: % PXe(max,app) = [% PXe(max)](pXe/ptot) where pXe is partial pressure of Xe and ptot is total mixture pressure. The production cycle time (for producing of ∼0.8 L of HP 129Xe/N2 gas mix) is calculated as the sum of 2/γSEOP (i.e., when the bulk (87% of Pmax) of 129Xe hyperpolarization is established, in addition to a reasonable (25 min long) interval necessary to unload/reload the OP cell (with cool-down/reheating procedure described in ref (24)) with xenon mix; γSEOP value corresponding to a maximum value of % Pmax was used for every gas composition. The production rate is calculated by dividing ∼0.8 L gas volume expanded in the Tedlar bag during each production cycle by the production cycle time.

The data in Figure 3 exhibit several trends. First, increasing cell surface temperature gives rise to an exponential increase in γSEOP (e.g., Figure 3a), consistent with the expected exponential increase in the Rb gas-phase concentration [Rb].[54,55] This dependence of γSEOP on [Rb] arises from the relation[1,2]where γSE and kSE are the Rb/129Xe spin-exchange rate and cross-section, respectively. Thus, the behavior of γSEOP mostly reflects the spin-exchange rate, since generally kSE[Rb] > ΓXe or kSE[Rb] ≫ ΓXe under our conditions. (At the highest temperatures studied, kSE[Rb] ≫ ΓXe; at the lowest temperatures kSE[Rb] can approach or become less than ΓXe, but ΓXe is expected to have a more mild dependence on surface temperature that trends in the opposite direction.[57])

However, Pmax exhibits significantly different behavior, for example, peaking at ∼72 °C for the data in Figure 3a. Pmax is given bywhere ⟨PRb⟩ is the spatial average of the local Rb electron spin polarization, PRb(r), which itself is determined bywhere γOP(r) is the local Rb optical pumping rate (the integrated product of the laser flux at position r and the Rb absorption cross section) and ΓRb is the Rb electronic spin destruction rate (which is essentially proportional to [Xe] under our conditions[51,52]). Intuitively from eq 2, 129Xe Pmax → ⟨PRb⟩ when kSE[Rb] ≫ ΓXe, which occurs at higher temperatures. However, having higher Rb densities generally translates into greater optical density, which in turn gives rise to reduced transmittance of the laser light and hence poorer illumination throughout the cell, lower γOP, and ultimately reduced ⟨PRb⟩, thereby decreasing Pmax. Thus, % Pmax initially grows with increasing temperature as more Rb is vaporized (e.g., Figure 3a), but once [Rb] becomes too high, overall ⟨PRb⟩ decreases in accordance with eq 3, resulting in lower 129Xe % Pmax at some of the highest temperatures studied.

The highest % Pmax values in the contour plots of Figure 3c–l were always achieved at the maximum LDA power of 170 W. However, as the Xe density increased, the optimal temperature decreased from 92 to 62 °C, in qualitative agreement with our previous results obtained at a much smaller scale.[21,53] This inverse relationship between Xe density and optimal cell surface temperature, an effect amplified by the use of frequency-narrowed lasers,[22] may be explained in part by the fact that as [Xe] rises, Xe-induced Rb spin-destruction becomes increasingly dominant; thus, lowering the cell temperature helps maintain a sufficient “photon-to-Rb” ratio to ensure high global ⟨PRb⟩ and hence higher % Pmax (provided that the cell 129Xe T1 is sufficiently long[24]). The effect may also be exacerbated by greater in-cell temperature gradients caused by (i) greater absorption of laser energy and (ii) the several-fold lower thermal conductivity of Xe compared to that of N2.[58,59] Indeed, the effects of differential heating are also manifested in the γSEOP maps: γSEOP (and hence γSE) is not a constant of exterior cell temperature but shows some variation. For example, the value at 100 W, 82 °C for the 2000 Torr Xe gas composition is nearly twice that for the 275 Torr Xe gas composition; overall, apparent γSEOP values tend to increase with increasing laser power and [Xe], consistent with higher-than-expected Rb vapor densities (and higher internal temperatures) under these conditions.

Next, there is a clear indication that all the studied Xe densities benefited from the increased laser power (see also Figure S1). Consequently, the use of LDA power greater than 170 W should lead to further increases in % Pmax. Furthermore, if the increased heat load could be mitigated, greater LDA power would allow for operation in the regimes with higher [Rb], thereby increasing γSEOP and HP 129Xe production rate.

Data are absent from some regions of the SEOP maps in Figures 3c–l. These regions were avoided because the build-up rate γSEOP was found to be excessively long, the % Pmax values were clearly low, and/or the conditions would render an unfavorably high [Rb], resulting in undesirable effects dubbed “Rb pre-runaway” or “Rb runaway”. The phenomenon of “Rb runaway” takes place when undissipated heat from laser absorption or cell heating rapidly compounds the amount of Rb in the vapor phase over a short time;[9,21] the increasing [Rb] results in decreasing PRb in more poorly illuminated regions of the cell and hence more laser absorption and heat dissipation from the gas into the inner surface of the cell (and Rb pools) in a self-reinforcing pattern. The effect can be characterized by its severity: In full Rb runaway, one sees a dramatic decrease in the amount of laser light transmitted through the cell over time which may be followed by deteriorating PXe and even elevated exterior cell surface temperatures. The behavior is also hysteretic, as simple temperature reduction to normal operating regimes generally fails to regain efficient SEOP. On the other hand, a more mild condition (here termed “Rb pre-runaway”) does not have as pronounced a manifestation in the transmitted laser’s near-IR spectroscopy but is readily observed during measurements of the kinetics of PXe build-up. Examples of “Rb pre-runaway” can be seen in Figure 2h (orange trace), where PXe grows, passes the maximum, and then dips. The effect is less pronounced in Figure 2h (red trace) and nonobservable in Figure 2h (black trace). While true “Rb runaway” causes % PXe to irreversibly deteriorate and requires a restart of the SEOP procedure from initial conditions to lower [Rb], “Rb pre-runaway” is not hysteretic and can be more easily controlled by reducing the cell temperature. However, it results in lower polarization (Figure 2h). In any case, these deleterious effects are more problematic for higher Xe densities because of the greater Rb spin-destruction rates (and hence greater light absorption from more poorly polarized Rb, as well as any possible contributions from reduced thermal conductivity).

Table 1 summarizes the maximum achieved 129Xe polarization (% Pmax) for every Xe:N2 mix studied. The results show not only the trend of decreased % Pmax with increasing Xe in-cell pressure but also a corresponding decrease in γSEOP measured at these optimal conditions, a finding that predominantly reflects the lower concentration of Rb vapor that must be attained to achieve maximal 129Xe polarization at higher Xe densities. Nevertheless, the optimization process allows the total magnetization (M′Xe) to continue to grow despite the decrease in % Pmax,[21] as the 129Xe density increases faster than % Pmax decreases. While the 129Xe polarization values (and amounts) are significantly higher here than those in ref (21), the improvement in M′Xe from 1000 to 2000 Torr of Xe is more marginal. Higher laser power may provide further improvements in % Pmax (and M′Xe) at high [Xe] by allowing operation with higher Rb densities and hence higher γSEOP rates. Other useful metrics describing the overall hyperpolarizer performance summarized in Table 1 include the apparent % PXe(max) due to Xe dilution by N2 gas (% PXe(max,app)), production cycle time, and apparent production rate of hyperpolarized gas (L/h). % PXe(max,app) is a useful metric[23] because it takes into account HP Xe dilution by N2 gas, which has not been eliminated because the HP Xe cryocollection step was obviated. Production cycle time corresponds to the time necessary to complete the production of ∼0.8 L of HP Xe/N2 gas composition and return the hyperpolarizer (i.e., gas reloading, etc.) to the same step in the operational cycle. Computed in this fashion production cycle time was used for estimating the apparent production rate of the hyperpolarizer in liters of hyperpolarized Xe/N2 mixture per hour. The production rate in L/h is truly the characteristic of continuous-flow hyperpolarizers, and the apparent production rate values computed in Table 1 should be used with care for direct comparison with continuous-flow hyperpolarizers, because the batch-mode method used here produces a single batch per each production cycle, and there is no produced HP 129Xe until the cycle is finished.

To validate the in situ NMR results, the polarized contents of the SEOP-cell filled with 1000 Torr of Xe and 1000 Torr of N2 was transferred into an evacuated (<10–3 Torr) 0.05 L hollow polypropylene sphere located in a rf probe of a 47.5 mT imaging system[56,60,61] (see Supporting Information for details). In-cell PXe was measured in situ as 54 ± 5% before the transfer, and a PXe value of 51 ± 2% was detected in the 47.5 mT preclinical MRI scanner (558.6 kHz 129Xe Larmor frequency), corresponding to polarization enhancement ε > 11 000 000 after the gas transfer (Figure 4b). The HP 129Xe transfer from the polarizer was performed without a cryocollection process.[24,25,30] Figure 5 also demonstrates the feasibility of millimeter-scale MRI of hyperpolarized 129Xe at very low magnetic fields using frequency optimized rf coils.[56]

Conclusions

Simultaneous optimization of various SEOP conditions (Xe density, cell surface temperature, and photon flux) combined with previously reported SEOP hardware improvements[24,29,30] yielded greatly improved % PXe. Indeed, very high values of % PXe and M′Xe were demonstrated here for dense (up to 2000 Torr of Xe in 2200 Torr total) Xe gas mixtures, in part enabled by optimized laser illumination throughout the cell, ultralong in-cell 129Xe relaxation times, and efficient thermal management that also allows for diligent avoidance of “Rb runaway” regimes. The SEOP condition maps provide guidance for the production of highly polarized 129Xe gas at different xenon densities for a wide variety of applications ranging from materials science to biomedical imaging. Furthermore, our results indicate that the PXe values at higher Xe densities are still laser-power-limited. Thus, while the benefit in total Xe magnetization was less substantial at the highest Xe densities studied, the advantage will likely be improved when more powerful LDA instrumentation is available provided that the greater thermal loads can be mitigated. Finally, the highly reproducible maps of γSEOP build-up rates, combined with automated fine control of cell conditions and real-time spectroscopic feedback, should also allow optimization of multiexponential Xe polarization dynamics, pointing the way to multifold improvements in HP 129Xe production efficiency.