The presented 13C MRSI phantom is built from readily available labware, cheap materials that are available in almost all modern laboratories, such as: NMR glass tubes, centrifuge tubes, and 3D-printed structures, which can easily be replaced or renewed. This should allow almost all laboratories to reproduce our recipes and manufacturing instructions.

We used 50 mL centrifuge tubes in our design because of several advantages they offer for preclinical applications. The radial outer diameter of 28–29 mm fits into most small animal volume RF coils. When filled with saline solution around the 13C reference NMR tubes, the total volume of ~ 50 mL mimics a coil loading with an animal at a size somewhere between a large mouse and a small rat. Their cylindrical geometry, when aligned along B0, is preferable for improved B0-field shimming, resulting in low 13C signal peak widths and good peak separation. Those tubes come with standardized geometry and dimensions, even when purchased from different manufacturers or suppliers, which simplified the design of the 3D-printed reference tube holders.

The reference tube holders were designed to trap inevitable air bubbles in the saline solution at the front or rear end of the phantom and keep them far away from the MRI FOV. Designing holders that fit various numbers of NMR tubes or can be used with smaller (15 mL) or larger (> 175 mL) centrifuge tubes is simply a matter of CAD design and printing expertise. For multi-center standardization, we recommend using a single standard container and holder design. Therefore, we share our 3D-printable structures for holding NMR tubes as CAD files, which can be downloaded and used under a common open-source license [16]. Beyond preclinical use, the 13C reference vials can support potential clinical applications. Placed next to a volunteer in a 13C torso coil, they provide a stable chemical standard for calibration, and SNR assessment. This allows simultaneous 1H anatomical imaging and 13C spectral validation, helping optimize acquisition and reconstruction methods before hyperpolarized 13C experiments.

The expensive materials, i.e., the 13C labeled metabolite solutions, are to be protected, sealed, and stored under optimal conditions for use as a quality standard for as long as possible, even if the surrounding, cheaper material parts are replaced over time. This raises several challenges. Metabolites tend to have high chemical reactivity; thus, they require a sterile environment and protection from UV radiation. The first issue was resolved by replacing the oxygen with an inert argon atmosphere and adding sodium azide (NaN3) before sealing the glass air-tightly via melting. A storage under light-protected and cooled conditions—in our study, in a fridge at 6 °C—did not show any degradation of the 13C-labeled materials over months, as would be indicated by changed peak profiles or relaxation properties in the 13C spectroscopy. Sub-PPM drifts of the peak positions may arise from temperature or variation in B0-field homogeneity. Only pyruvic acid remains too reactive and forms additional molecular structures with itself, yielding a shift of the 13C signal peak resonances. With its replacement by 1-13C-labeled ethyl acetate, which has a similar chemical shift of 171 ppm, a chemically stable replacement was used in the proposed phantom. In addition, using transparent and UV-light protective glassware, e.g., borosilicate NMR tubes, can potentially extend the stability of the molecular structures, but comes at additional costs. For the sake of completeness, we note that the NMR glass itself can potentially contain impurities at micromolar concentrations, as recently reported [17]. However, using quartz glassware comes at substantially higher cost and would remove impurities that are only noticeable at very strong magnetic fields and high spectral quality—especially 13C impurities are presumably low in concentration and signal intensity. No limitations caused by unknown impurities in the 13C reference vials were experienced, even at an 11.7 T MRI. Therefore, we suggest using the standard 5 mm outer diameter glassware NMR tubes, which are widely available.

The sealing of the NMR tubes by melting the glass under a Bunsen burner flame should be well practiced in advance, before expensive materials are to be sealed. The inner length of the 50 mL centrifuge tubes requires the NMR glasses to be shortened to 90 mm. The remaining volume of the NMR vials is therefore limited to a maximum of approximately 1.2 mL. The glass melting process, unfortunately, requires a small gas bubble at the top of the liquid as heat insulation; otherwise, the Bunsen flame would boil the 13C-labeled solution and potentially destroy the reference material. We found that using a vacuum pump improves the sealing process, but it is not strictly required for airtight sealing. We recommend always testing the freshly closed side of the NMR tube for air and liquid tightness after the glass has cooled off. We found that the molten glass tips of the NMR tubes can be well reinforced against breaking by covering them with two-component epoxy adhesive (e.g., Loctite EA 9492, Henkel, Düsseldorf, Germany). The remaining gas bubbles inside the NMR tubes can cause imperfections in the static B0 field. However, if not shaken, the bubble remains in one position within the NMR tube, preferably far outside the MRI FOV, and the gas–liquid susceptibility artifacts can be compensated for with B0-field shimming routines.

In addition to sealing, reliable labeling of the NMR tubes is essential to ensure unambiguous sample identification during preparation, storage, and repeated use. We initially used waterproof markers to label the tubes at the glass surface, but the markings wore off over time when phantoms were often inserted into or removed from the holders. A more robust strategy was to number the positions in the 3D-printed holder (e.g., A–E) and track which tube was placed in each slot, which avoids the need to label the glass itself. For long-term storage, placing the tubes in a grid rack with fixed, labeled positions further simplified organization and minimized the risk of confusion. In multi-center or extended studies, additional durable labeling methods such as using nail polish and a color code system, chemical-resistant tapes covered with transparent shrink tubing, or permanent etching/laser marking on the glass, may be beneficial to guarantee that content information and preparation dates remain identifiable throughout extended storage and transport.

Relaxation times at 1.4 T and 3 T were measured in individual 5 mm NMR tubes to provide reference values for multinuclear imaging systems at clinical field strengths. In contrast, the full phantom assembly, including multiple tubes in the centrifuge tube filled with NaCl, was designed for operation at high-field preclinical systems and therefore imaged at 11.7 T. This approach yields relaxation data relevant for multinuclear imaging platforms (1.4 T and 3 T). Additionally, despite the small contribution of the urea signal in the bicarbonate image (Fig. 4e), which is most likely related to the slight resonance peak overlapping in combination with the high concentration difference between the two solutions, it also demonstrates the feasibility of the phantom for spectroscopic imaging experiments at preclinical systems (11.7 T). Further, the results confirm that the reference solutions and overall design are applicable across the range of field strengths encountered in multinuclear MRI and anticipated hyperpolarized 13C workflows.

By doping the 13C reference solutions with gadolinium, the relaxation time constants T1 and T2 can be substantially shortened. The reason and advantage of having both, long and short T1 and T2 for each sample, is for different applications: shortened T1 yield increased signals when keeping flip angle, repetition time, and averaging fixed. This is beneficial for MRI 13C channel calibration and adjustments for an experiment. Samples with long T1 relaxation constants can be used as thermally polarized experimental setups to investigate specialized spectroscopy and imaging methods for later use in HP 13C experiments, since in HP 13C, a long T1 is essential to preserve the non-equilibrium magnetization state over extended time periods. This mimics the realistic signal evolution after dissolution and transfer in hyperpolarization workflows, allowing researchers to evaluate pulse sequences, acquisition strategies, and reconstruction methods under conditions that more closely resemble true in vivo HP experiments. In the present study, relaxation times were measured only at controlled room temperature (21 °C ± 1 °C), reflecting the intended operating conditions for routine phantom use on imaging systems. Although we did not perform a systematic assessment of temperature-dependent relaxation behavior, the expected variation of ± 1 °C in controlled MRI environments is unlikely to cause relevant changes in relaxation times or to limit the usability of the phantom for routine quality assurance.

Compared to previously reported 13C phantoms, our approach addresses a complementary but distinct need. Dynamic enzyme-based phantoms [1,2,3] are valuable for mimicking metabolic conversion processes and cross-site standardization of dynamic acquisitions, but they require complex preparation and often lack long-term stability due to enzymatic degradation. Similarly, phantoms designed for sequence-independent system validation, flip-angle calibration, and coil characterization [5, 6, 12,13,14] provide vital benchmarks for MRI performance and quality assurance; however, they do not offer a stable chemical reference that can be stored and reused over years. In this sense, our design complements dynamic and metabolite-specific phantoms by providing a durable, reproducible reference standard that can support both preclinical research and multicenter harmonization efforts at low cost.