**4.2. Measurement protocol**

Selection of a field of view consists of two steps, first an overview of the TEM mesh with a visible light microscope (VLM), and then selection of a specific mesh hole where a 100 μm × 100 μm mosaic image can be composed by 10 × 10 individual TXM images (see **Figure 6**). Repeating mosaics at two energies, above and below the O K‐edge can be useful for a first localization of oxygen compounds.

For the case of the O K level, XANES images are collected varying the incident photon energy with an exposure time of 2 s at each energy step. For each energy step, a flat field image (FF), i.e., an image without the sample, is recorded. An energy range of 515–580 eV ensures the most reliable spectral normalization and allows EXAFS analysis. The energy step can be adjusted between 0.1 and 0.5 eV depending on the spectral features to record. The objective zone plate lens (outermost zone width of 40 nm, 937 zones) and the CCD detector positions are automatically adjusted to maintain focus and constant magnification. The CCD motion is the slowest part of acquisition which typically requires between 1 and 2 h.

Studies of Lithium-Oxygen Battery Electrodes by Energy-Dependent Full-Field Transmission Soft X-Ray Microscopy http://dx.doi.org/10.5772/66978 105

**Figure 6.** Stepwise selection of a field of view: grid overview by VLM (a), mosaic TXM image of an individual mesh hole before (b) and after (c) the O‐K edge, and TXM field of view (d).

### **4.3. Radiation damage**

in which one wants to maximize the 2D spatial resolution and the depth of focus is not a critical parameter. More technical details on the MISTRAL beamline are reported in references

In this section, we describe the preparation, collection, and analysis of cathodes used in Li‐oxygen batteries, although most of the principles and considerations may well apply to

For images of optimal quality, we take common carbon‐coated Au TEM grids (200 mesh, 3.05 mm diameter) directly as a cathode. Gold works as a current collector that is inert in the cathode electrochemical window, while carbon provides an ideal substrate for transmission microscopy. However, also a few micrograms of super P/PVdF slurry are deposited on the TEM grids to provide a roughened surface closer to that present in practical

Electrochemical treatments are performed using homemade cell based on the Giessen battery design [13], resulting in Swagelok‐like battery arrangement. After the electrochemical treatment the cell is opened in an Ar‐filled glove box and the grid washed with DME and hexane to fully remove the electrolyte. The grid is then transferred in the microscope

tion. The samples are kept at cryogenic temperature and under high vacuum during the

Selection of a field of view consists of two steps, first an overview of the TEM mesh with a visible light microscope (VLM), and then selection of a specific mesh hole where a 100 μm × 100 μm mosaic image can be composed by 10 × 10 individual TXM images (see **Figure 6**). Repeating mosaics at two energies, above and below the O K‐edge can be useful for a first

For the case of the O K level, XANES images are collected varying the incident photon energy with an exposure time of 2 s at each energy step. For each energy step, a flat field image (FF), i.e., an image without the sample, is recorded. An energy range of 515–580 eV ensures the most reliable spectral normalization and allows EXAFS analysis. The energy step can be adjusted between 0.1 and 0.5 eV depending on the spectral features to record. The objective zone plate lens (outermost zone width of 40 nm, 937 zones) and the CCD detector positions are automatically adjusted to maintain focus and constant magnification. The CCD motion is

the slowest part of acquisition which typically requires between 1 and 2 h.

vapor to minimize atmospheric contamina-

**4. Spectromicroscopy at the O K‐edge of discharged cathodes**

104 X-ray Characterization of Nanostructured Energy Materials by Synchrotron Radiation

[18, 19].

electrodes.

measurements.

**4.2. Measurement protocol**

localization of oxygen compounds.

other systems and energy regions.

**4.1. Sample preparation and transfer**

in cryogenic condition (*T* < 110 K) under N2

A recurring issue in techniques involving high brilliance light is the sample stability to irradiation. This is often considered a major problem with organic compounds, whereas inorganic compounds are regarded as essentially stable [22]. However, Li2 O2 is also very sensitive. For instance, in our own experience, electron microscopy requires special care, as the electron beam is able to fully decompose a micrometric layer of lithium peroxide in a few seconds, as shown in **Figure 7**.

Qiao et al. [24] specifically studied the effect of irradiating lithium peroxide, oxide, and carbonate with soft X‐rays at room temperature. They found that carbonate and peroxide evolve toward oxide within several minutes of irradiation. In contrast with their respective electrochemical oxidation stability, the evolution of carbonate is faster, with oxide features appearing after 1 h for peroxide and already after 20 min for carbonate. The stability of a given compound depends on the activation process. However, in general, the high cross‐section of soft X‐rays implies a stronger interaction with matter and consequently a faster degradation than hard X‐rays, in particular when the photon energy is close above the absorption edge of elements such as C and O.

**Figure 7.** SEM image showing the effect of the electron beam on the deposits of a Li/O2 discharged cathode. The support is a carbonaceous inverse opal [23]. The deposit‐free rectangle has been obtained after scanning during about 1 min for a higher magnification image.

This problem is obviously of special relevance for biologic samples, which are always observed with soft X‐rays in cryogenic conditions. The lower nuclear thermal motion favors the reconstruction of the bonds excited by radiation, significantly improving the sample stability. For this reason, we also keep our discharged electrodes at *T* < 110 K (liquid nitrogen) during irradiation with soft X‐rays. Under these conditions, we can irradiate long enough to gain acceptable signal to noise and energy step without evident damage. As an example, **Figure 8** compares two O K‐edge XANES spectra (duration about 2 h each) of a sample recorded before (red) and after 24‐h period (blue). Spectra demonstrate very good reproducibility corresponding to no detectable radiation damage within our resolution and experimental noise.

### **4.4. Reference samples**

Reference samples are essential for correct assignment of the peaks found in the sample. They must be measured in the same conditions and preferably in the same session to allow similar instrumental peak broadening and proper spectra subtraction. This is useful to detect components that may be masked by more abundant species.

Unfortunately, commercial compounds may not always correspond exactly to the expected composition. Although also commercially available, we have chemically generated Li2 O2 from the reaction of KO2 with dicyclohexyl‐18‐crown‐6 (crown ether) in solution [25]. We have obtained images of the precipitate with different local O K‐edge XANES spectra. By inspection of the spectra at different points of the images we concluded that locally rather pure phases of both Li2 O2 and Li2 CO3 coexisted, being the first in agreement with the literature [24], and the second with a commercial sample we measured. LiO2 can only be hardly obtained in a pure

Studies of Lithium-Oxygen Battery Electrodes by Energy-Dependent Full-Field Transmission Soft X-Ray Microscopy http://dx.doi.org/10.5772/66978 107

**Figure 8.** O K‐edge XANES spectra of discharged sample before (black) and after (red) 24‐h in the same experimental condition. Reprinted with permission from Olivares‐Marín et al. [11]. Copyright 2015 American Chemical Society.

form [12, 26], and we used just a literature absorption spectrum as reference [26], mainly to confirm its energy.

#### **4.5. Data analysis**

This problem is obviously of special relevance for biologic samples, which are always observed with soft X‐rays in cryogenic conditions. The lower nuclear thermal motion favors the reconstruction of the bonds excited by radiation, significantly improving the sample stability. For this reason, we also keep our discharged electrodes at *T* < 110 K (liquid nitrogen) during irradiation with soft X‐rays. Under these conditions, we can irradiate long enough to gain acceptable signal to noise and energy step without evident damage. As an example, **Figure 8** compares two O K‐edge XANES spectra (duration about 2 h each) of a sample recorded before (red) and after 24‐h period (blue). Spectra demonstrate very good reproducibility corresponding to no detectable radiation damage within our resolution and experi-

is a carbonaceous inverse opal [23]. The deposit‐free rectangle has been obtained after scanning during about 1 min for

Reference samples are essential for correct assignment of the peaks found in the sample. They must be measured in the same conditions and preferably in the same session to allow similar instrumental peak broadening and proper spectra subtraction. This is useful to detect

Unfortunately, commercial compounds may not always correspond exactly to the expected

obtained images of the precipitate with different local O K‐edge XANES spectra. By inspection of the spectra at different points of the images we concluded that locally rather pure phases of

with dicyclohexyl‐18‐crown‐6 (crown ether) in solution [25]. We have

coexisted, being the first in agreement with the literature [24], and the

O2 from

discharged cathode. The support

can only be hardly obtained in a pure

composition. Although also commercially available, we have chemically generated Li2

components that may be masked by more abundant species.

**Figure 7.** SEM image showing the effect of the electron beam on the deposits of a Li/O2

106 X-ray Characterization of Nanostructured Energy Materials by Synchrotron Radiation

mental noise.

**4.4. Reference samples**

a higher magnification image.

the reaction of KO2

O2

and Li2

CO3

second with a commercial sample we measured. LiO2

both Li2

#### *4.5.1. Images normalization, alignment, and spectra extraction*

As a first step, each image of the energy scan is normalized to one using the corresponding FF. This normalization consists in simply dividing the image with the sample by the corresponding FF, obtaining the transmission *T* = *I*/*I* 0 as a function of the energy (see Section 3.1). In the transmission only, the measured intensity variations along the energy scan due to the sample itself are taken into account, all the contribution coming from the optical setup (beamline and microscope optics) are discarded. The normalized images (or transmission images) present drifts along the scan due to the relative movement between the ZP lens and the CCD detector. Therefore, the image sequence needs to be aligned. At Mistral for instance this is done using a in‐house software named "ctalign." Each of the images is aligned taking as reference a chosen ROI of the first image of the scan. The software uses the normalized cross‐correlation of both ROIs to detect the best matching between them. Once aligned, the absorbance images can be simply obtained from the corresponding transmission images using Eq. (10), i.e., calculating the –log of the series. The free NIH ImageJ software [27] is for instance useful to handle the aligned TXM images and doing calculations with them.

By taking the average absorbance from a given region of an image for all images, the absorption spectrum for that region is obtained. In principle, the technique gives access to a single pixel spectra analysis, practically the dimension of the selected region is limited by the lens resolution, the effectiveness of the series alignment, and single pixel spectra noise to 5–6 pixels2 .

#### *4.5.2. Spatial distribution of the discharge products*

Given that the different oxygen reduction states present in discharged metal/oxygen electrodes (superoxide, peroxide, and carbonate) are characterized by distinct absorption peak energies (around 529, 531, and 533 eV, respectively) oxygen‐state‐resolved maps can be obtained. A full quantitative approach consists in measuring a full set of pure reference samples (superoxide, peroxide, and carbonate in our case) and fit with them the obtained measured spectra. If measured reference spectra are not available some sophisticated linear algebra technique such as principle component analysis and factor analysis can also be used for the interpretation of XANES spectra. The number of principal components determined in this way was used as the basis for multivariate curve resolution‐alternating least squares (MCR‐ALS) analysis [28]. However, precise and accurate calculations of all spectral features are still difficult and not always reliable. Presently, quantitative analyses of XANES spectra using *ab initio* calculations are very rare, and a full description of absorption spectra data analysis and interpretation is beyond the scope of this paragraph. Here, we will describe a simple, qualitative/semiquantitative approach based on the construction of absorbance image differences *D* obtained from single absorbance images at specific energies. Let us consider two chemical species A and B in a thickness *t*, then the corresponding absorbance will be (from Eq. 12):

$$
\mu(E) \, t = \left( \mu\_{n, \mathbf{A}}(E) \, \rho\_A + \mu\_{n, \mathbf{B}}(E) \, \rho\_\mathbf{B} \right) \, t \tag{16}
$$

where *μ*(*E* ) *t* represents the intensity map in the absorbance image at a generic energy *E*. Now if we choose the two values of the energy position of the absorption edge peak maximum and minimum, respectively, of the species A for instance (*E*MAX,A and *E*min,A in **Figure 9**), we can write for the corresponding absorbance difference *D*<sup>A</sup>

$$\mathcal{D}\_A = \left(\mu\_{m,A}(\mathcal{E}\_{\text{MAX},A})\,\rho\_A + \mu\_{m,B}(\mathcal{E}\_{\text{MAX},A})\,\rho\_B - \mu\_{m,A}(\mathcal{E}\_{\text{min},A})\,\rho\_A - \mu\_{m,B}(\mathcal{E}\_{\text{min},A})\,\rho\_B\right)t \tag{17}$$

$$\text{and assuming } \mu\_{n,\mathbb{B}}(E\_{\text{MAX},\mathbb{A}}) \approx \mu\_{n,\mathbb{B}}(E\_{\text{min},\mathbb{A}}) \,. \tag{18}$$

$$D\_A \approx \langle \mu\_{n,A}(E\_{\text{MAX},A}) - \mu\_{n,A}(E\_{\text{min},A}) \rangle \rho\_A t \tag{19}$$

If *μ<sup>m</sup>*,*<sup>A</sup>* (*E*min,*<sup>A</sup>* ) ≪ *μ<sup>m</sup>*,*<sup>A</sup>* (*E*MAX,*<sup>A</sup>* ) ⇒ *DA* ≈ *μ<sup>m</sup>*,*<sup>A</sup>* (*E*MAX,*<sup>A</sup>* ) *ρ<sup>A</sup> t*, that is the absorbance image difference *D*A is proportional to the concentration map of the chemical specie A. Doing the same for the specie B, we have:

Studies of Lithium-Oxygen Battery Electrodes by Energy-Dependent Full-Field Transmission Soft X-Ray Microscopy http://dx.doi.org/10.5772/66978 109

**Figure 9.** Two examples of fraction estimation for different components in integrated spectra.

$$D\_{\mathfrak{g}} \approx \left. \mu\_{m,\mathfrak{g}} (E\_{\text{MAX},\mathfrak{g}}) \right\vert \rho\_{\mathfrak{g}} t \tag{20}$$

and then for the ratio between the two species: *<sup>D</sup>*\_\_\_*<sup>A</sup> DB* ∝ *ρ* \_\_*<sup>A</sup> ρB* . The corresponding estimation in integrated spectra is shown in **Figure 9**.

Under these considerations we have used absorbance images corresponding to the relative maximum absorbance of LiO2 (528.75 eV), Li2 O2 (531.25 eV), and Li2 CO3 (533.8 eV). The image at 520 eV was subtracted from that at 528.75 eV to obtain the 2D LiO2 distribution, and the image at 528.75 eV was subtracted from that at 531.25 eV to obtain the one for Li2 O2 . Also, Li2 CO3 was calculated subtracting image at 531.25 eV minus image at 533.8 eV. Three distribution images obtained were then merged with different colors.

From these maps and possibly calibrating the image intensity using reference samples with known ratio between species, it is possible to obtain profiles or images of ratios and fractions, which is useful to establish correlations or associations between species.
