Quantitative, In Situ Visualization of Metal‐Ion Dissolution and Transport Using 1H Magnetic Resonance Imaging

Abstract Quantitative mapping of metal ions freely diffusing in solution is important across a diverse range of disciplines and is particularly significant for dissolution processes in batteries, metal corrosion, and electroplating/polishing of manufactured components. However, most current techniques are invasive, requiring sample extraction, insertion of an electrode, application of an electric potential or the inclusion of a molecular sensor. Thus, there is a need for techniques to visualize the distribution of metal ions non‐invasively, in situ, quantitatively, in three dimensions (3D) and in real time. Here we have used 1H magnetic resonance imaging (MRI) to make quantitative 3D maps showing evolution of the distribution of Cu2+ ions, not directly visible by MRI, during the electrodissolution of copper, with high sensitivity and spatial resolution. The images are sensitive to the speciation of copper, the depletion of dissolved O2 in the electrolyte and show the dissolution of Cu2+ ions is not uniform across the anode.

The T1 and T2 relaxation times of water molecules are greatly affected by the presence of paramagnetic ions, such as Cu 2+ . Relaxation becomes very efficient in the presence of paramagnetic species, due to the magnetic moment of unpaired electron spins, which is 1000 times greater than nuclear magnetic moments. Hence, only micro-to millimolar concentrations are, typically, required before a reduction in the relaxation time is observed 3 .

Methods
Electrochemical cells. The corrosion cells were constructed from a polystyrene semimicro cuvette (Sarstedt, Germany) with two copper (99.98%, Aldrich) electrodes, of dimensions 0.25 × mm 51.0 mm × 3.7 mm, and a 0.50 mm diameter chloridized silver wire (Ag/AgCl) reference electrode (RE). The copper electrodes were cleaned in 10% nitric acid for 1 minute, then rinsed in nanopure water, then methanol, before left to dry for over 16 hours. Electrodes were fixed to the cuvette walls by slightly dissolving the plastic cuvette with a small drop of chloroform, leaving only the front face exposed with no air bubbles in contact with the walls. The RE was held in position using a silicone plug inside the cell, and a measurement of the RE potential pre/post experiment in 3 M NaCl relative to a conventional Ag/AgCl reference electrode (BASi, USA) showed that it drifted by -1.4 mV. The electrolyte was 0.5 M Na 2 SO 4 (> 99.0%, Sigma), prepared in ultrapure water, at pH 5.5. A 750 µL volume was injected just prior to insertion of the cell into the MRI. For the 3D mapping experiment shown in Figure 5, the electrodes in this cell were exposed to the atmosphere for 1 week to allow formation of an oxide layer, and just before introduction of the electrolyte, a small scratch was made near the centre of the anode using a needle.
MRI. 1 H MRI was performed on a Bruker DMX300 spectrometer, operating at a 1 H resonance frequency of 300.13 MHz and equipped with a 25 mm micro2.5 radiofrequency (RF) resonator, at ambient temperature of 19 ± 0.2 °C. The RF coil and corrosion cell were independently rotated in-situ to align the strip electrodes so that they were parallel with the B 0 and B 1 fields.
A series of vertical (xz) and horizontal (xy) 2D images (projections) were acquired using a RARE imaging sequence 4 to produce spin density and T 1 maps of protons in the electrolyte solution. Vertical images were acquired with a 25 mm slice thickness, a field of view (FOV) of 24 mm × 16 mm, a pixel size of 188 µm × 250 µm and RARE factor of 64. Horizontal images were acquired with a 50 mm slice thickness, a FOV of 16 mm × 8 mm, a pixel size of 250 µm × 250 µm and RARE factor of 32. Slice thicknesses were sufficiently large to ensure the entire sample was contained within each image, producing 2D images (projections) of the electrolyte. T 1 images were produced from eight 2D images with an inversion recovery delay ranging from 5 ms to 14 s, all with an echo time of 3.2 ms and a recovery time of 14.1 s, and taking 8 minutes in total to acquire. The 3D concentration maps were acquired from a single 3D image, acquired with a FOV of 24 mm × 16 mm × 8 mm, with a pixel size of 188 µm × 500 µm × 500 µm, a RARE factor of 16, a recovery time of 200 ms and resulting in an acquisition time of 25 s. In the 3D maps, the concentration was determined from image intensity, which is a function of the local T 1 relaxation time and was calibrated by imaging known copper concentrations using identical imaging parameters (see supplementary information for more details).
Potentiostat. A COMPACTSTAT.e potentiostat (Ivium), in 3-electrode mode, controlled the electrochemical cell inside the MRI, and was connected to the cell electrodes using low magnetic susceptibility RG316 coaxial wires (Radiall).
Electromagnetic noise was minimised using in-line low pass filters (DC to 48 MHz pass band; Mini Circuits, Brooklyn, NY) and an 8 mm thick copper bulkhead plate was used to cover the bore entrance for shielding. The following protocols were used: 1) the open circuit potential (OCP) was measured during MRI scanning (1120 s acquisition); 2) a galvanostatic current of 50 µA or 100 µA was applied for a fixed time (ranging between 10 and 60 s) to produce controlled dissolution of the anode.
The OCP was measured and remained in the range E = -30 mV to +150 mV. With the cell connected to the potentiostat and loaded into the MRI, the OCP was monitored while MRI setup was conducted and there was no indication of any perturbation of the potential during MRI measurements. Once the voltage had stabilised, a data set was acquired before any current was passed through the cell.
At each experimental point, the charge pulse was delivered as described above, followed immediately by the same concurrent OCP/MRI acquisition.
Optical photography. An electrochemical cell was fashioned as above using a semimicro cuvette with one plastic sidewall removed to reveal the internal, clear plastic window for unobstructed view of the fluid inside. The opposite sidewall was roughened with 1200 grit SiC paper to a smooth/translucent finish to permit diffuse but uniform backlighting. A 0.50 M Na 2 SO 4 electrolyte was mixed using 6 %-vol Universal Indicator (Fluka). The cell was clamped in place, and photography was done using a digital camera (Canon PowerShot SX600 HS) on a tripod, with a darkened room and a backlit, white paper background used for uniform, reproducible lighting. Controlled anodic dissolution was done using a series of galvanostatic pulses (0.1 mA, variable duration), each followed by a 120 s OCP measurement.

Artefacts in MRI imaging near metal electrodes
Accurate determination of ion concentrations from relaxation maps depends on acquiring accurate, distortion-free images that are used to produce those maps.
Image artefacts may arise from distortion of the primary magnetic field (B 0 ) due to the magnetic susceptibility of the metals or by interaction between the radiofrequency (RF) field and the electrodes. The RF field may induce eddy currents in bulk metals depending on the geometry of the metal and the orientation relative to the field. This problem has been characterised for various medical implants 5 , for wires, 6 Figure SI2a shows horizontal and vertical images of a single metal strip in pure water, which is aligned with the RF direction, and therefore the images are distortion-free with uniform signal intensity. With the addition of a second strip (Fig. SI2b), there is signal intensity difference inside/outside the strips, and there are distortions at the strip edges that introduce additional features at the edge of the metal strips. Neither of these effects is minimised by RF power optimisation or shimming of the B 0 field. The severity of the effect varies with fluid type. When hexane is used as the medium (Fig. SI2c), the images are not affected by the same artefacts. The RF wavelength in the medium differs from the vacuum wavelength (λ 0 ) depending on the dielectric constant, = ! / . Since the wavelength for water (ε = 81, λ = 11 cm) is considerably closer to the sample length than for hexane (ε = 1.88, λ = 73 cm), it was hypothesized that the artefact could be a near-field effect that depends on the ratio of sample height to RF wavelength and decreasing the sample height from 5 cm to 2 cm effectively minimises the artefact ( Figure SI2d).

Optical images showing change in pH using Universal Indicator
The concentration maps in Figure 3b, and concentration profiles in Figure 4a, show a reduction in the amount of Cu 2+ . Also, Figure 4b shows local Cu 2+ concentrations that appear to become negative. Both observations can be explained by cathodic reduction of oxygen with water to form hydroxide, which produces an increase in pH and a decrease in the concentration of dissolved oxygen gas. It is known 8 that pH alone does not affect the T 1 of water in the range pH 3 to 7; however, the stable species at pH > 10 are Cu 2 O and Cu(OH) 2 , both solid and therefore MRI-invisible.
The change in pH was visualised by acquiring photographs of a sample cell with 6 vol-% Universal Indicator added to the electrolyte. Figure SI4 shows  8 The alignment of the copper strips was chosen to minimise both RF absorption and distortions of the primary B 0 field, but this orientation also minimises magnetic flux changes from the field gradients used for imaging, which may induce transient voltages in the circuit when switching on/off. The magnetic field gradients of the MRI experiment produced no discernable fluctuations in the OCP.

Concentration calibration for 3D mapping
The images are normalised, so that variations in signal intensity in each image are only through variations in Cu 2+ concentration. The relationship between the normalised signal intensity (S), for the 3D, spin-echo MRI images shown in figure 5, is given by equation (S1) and is dependent on the local T 1 relaxation time, which is dependent on the concentration of Cu 2+ , the repetition time (TR) and equilibrium intensity (S 0 ). S = S 0 (1 -exp(-TR/T 1 )) (S1) S 0 is the signal intensity for the solution where TR > 5T 1 and by plotting S against (1 -exp(-TR/T 1 )), for a range of solutions with varying Cu 2+ concentrations, a straight line is produced, allowing simple mapping between S and T 1 , which leads to the determination of a [Cu 2+ ]. Figure SI5 shows (a) the images used for the calibration, and (b) the calibration data, which was fitted to equation S1 using linear leastsquares. Maps of Cu 2+ were then obtained by converting signal intensity into Cu 2+ concentration.