# Fileset

[JCP23-AR-03268_MDR.pdf](https://mdr.nims.go.jp/filesets/9468cbbf-a954-4c0c-9a7d-1268a17ece89/download)

## Creator

[Jun Kikkawa](https://orcid.org/0000-0003-0659-1844), Aoi Nii, Yoshiaki Sakaniwa, Naochika Kon, Marina Sakamaki, Touyou Ohashi, Nobuyasu Nita, [Koji Harano](https://orcid.org/0000-0001-6800-8023), [Koji Kimoto](https://orcid.org/0000-0002-3927-0492)

## Rights

This article may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing. This article appeared in The Journal of Chemical Physics Volume 159, Issue 17,174708  and may be found at https://doi.org/10.1063/5.0177019.[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Fast electron damage mechanism of epoxy resin studied by electron energy loss spectroscopy and electron diffraction](https://mdr.nims.go.jp/datasets/74226aae-edcf-4ee1-bddc-de891c5ed893)

## Fulltext

Fast electron damage mechanism of epoxy resin studied by electron energy loss spectroscopy and electron diffractionAccepted to J. Chem. Phys. 10.1063/5.01770191  Fast electron damage mechanism of epoxy resin studied by 1 electron energy loss spectroscopy and electron diffraction 2  3 Jun Kikkawa,1, a) Aoi Nii,2, Yoshiaki Sakaniwa3, Naochika Kon2, Marina Sakamaki3, 4 Touyou Ohashi3, Nobuyasu Nita2, Koji Harano1 and Koji Kimoto1 5 1National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan 6 2Mitsubishi Materials Corporation, 1002-14 Mukohyama, Naka 311-0102, Japan  7 3Mitsubishi Materials Corporation, 1-600 Kitabukuro, Omiya, Saitama 330-8508, Japan 8 a)Electronic address: KIKKAWA.Jun@nims.go.jp 9  10 ABSTRACT 11    The damage mechanism and exposure tolerance of epoxy resins to fast electrons 12 remain unclear. We quantitatively investigated the effects of electron irradiation on a 13 common epoxy resin by dose-dependent electron energy loss spectroscopy (EELS). The 14 results show that sp3 states of nitrogen, oxygen, and their adjacent carbon atoms were 15 converted to sp2 states, forming imine (C=N) and carbonyl (C=O) as the total electron 16 dose increased. The sp3 to sp2 conversion mechanism was proposed. The epoxy resin was 17 very sensitive to fast electrons and the original electronic states were maintained up to a 18        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770192  total dose of ~103 e− nm−2 at a low temperature of 103 K. Dose-dependent electron 19 diffraction (ED) revealed that the intra- and intermolecular geometries changed below 20 and around the total dose of ~103 e− nm−2. 21  22 I. INTRODUCTION 23 Epoxy resins are widely used in the current industry for materials such as adhesives, 24 coating ingredients, and encapsulants owing to their characteristics such as excellent 25 thermal and mechanical properties, chemical resistance, and moldability.1 They are 26 applied not only in our daily environment but also in the space environment where 27 durability to cosmic rays is required.2 Epoxy resins are generally made through chemical 28 reactions of epoxy compounds (or base resins) with curing agents (or hardeners).3 29 Although their chemical reactions and polymerized network structures are important to 30 understand because they affect the physical properties of epoxy resins, they are not fully 31 understood and are the subject of intense research.3, 4 From a functional viewpoint, in 32 addition to their macroscopic physical properties, it is also important to understand the 33 microscopic properties of epoxy resins, such as chemical bonding states at 34 heterointerfaces in adhesion.5 In space, it is also important to understand the resistance of 35        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770193  epoxy resins to cosmic rays (protons, helium nuclei, electrons, etc., with near-light speed) 36 and the damage mechanisms.2  37 Electron microscopy, a powerful microscopic characterization method, has recently 38 enabled the analysis of polymetric materials sensitive to electron beams, owing to 39 advances in measurement techniques and detector performance.6, 7 We expect that 40 measuring the chemical bonding states of epoxy resins at the atomic to nanometer scale 41 in specific areas, such as heterointerfaces in adhesion, by electron energy loss 42 spectroscopy (EELS) in electron microscopes will open the way to gaining a new 43 scientific understanding of epoxy resins. However, the resistance of epoxy resin to fast 44 electrons of 30–300 keV (i.e., common energies used in electron microscopy) and the 45 mechanism of how and which part of the epoxy resin is damaged remain unclear, although 46 there have been some previous studies.8-11  47 In this study, we use 300 keV electrons (~78 % of the speed of light) in an electron 48 microscope and quantitatively investigate the effects of electron irradiation on a cured 49 epoxy resin by EELS with a high detective quantum efficiency (DQE) camera. From the 50 results for carbon, nitrogen, and oxygen K edges and the 1–35 eV region in EELS, we 51 identify the essential nature of the damage and propose a mechanism of its formation in 52 the epoxy resin. Durability, i.e., the maximum total electron dose that maintains the 53        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770194  electronic states, for the epoxy resin is also elucidated. By electron diffraction (ED), we 54 also found that molecular geometry changes occur before electronic state changes at the 55 initial stage of irradiation. 56  57 II. EXPERIMENTAL METHOD 58 We used N,N,N',N'-tetraglycidyl-4,4'-methylenedianiline (TGMDA) (Tokyo 59 Chemical Industry Co.) as a base resin and 2-ethyl-4-methylimidazole (2E4MZ) (Shikoku 60 Chemicals Co.) as a curing agent, and mixed them in a molar ratio of 22.5:4.5 [Fig. 1(a)]. 61 The mixture was applied to surface-treated aluminum and heated at 180°C for 1 h in the 62 atmosphere to obtain an epoxy resin. The possible chemical reactions of TGMDA with 63 2E4MZ are shown in Fig. 1(b). 2E4MZ acts as a ring-opening initiator on the oxirane of 64 TGMDA to form an alkoxide, which reacts with an oxirane of another TGMDA monomer 65 to open the ring. Repeated ring-opening and chain growth through reactions between the 66 alkoxide and oxirane of TGMDA initiate polymerization.3 The detachment of the 2E4MZ 67 moiety and rebonding by heating ultimately result in a cured epoxy resin having a 68 backbone structure composed of benzenoid and side-chain parts, as shown in Fig. 1(b). 69 A microtome (Leica EM UC7 Ultramicrotome, Leica Microsystems GmbH) was used 70 to prepare thin cross-sectional specimens of the cured epoxy resin coated on aluminum 71        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770195  for EELS. We used a monochromated electron microscope (Themis Z, Thermo Fisher 72 Scientific, Inc.) with a spectrometer (Quantum 970, Gatan Inc.) and a high DQE 73 complementary metal-oxide-semiconductor (CMOS)-based direct electron detection 74 camera (K2 IS, Gatan Inc.) for electron dose-dependent EELS. All the data presented in 75 this paper were obtained using 300 keV electrons. The specimen was set on a cooling 76 holder (Elsa, Gatan Inc.). The specimen temperatures with and without liquid nitrogen in 77 the microscope were 103 K and 303 K, respectively. For electron dose-dependent EELS, 78 we used a parallel electron beam with a diameter of 216 nm, electron flux j (×104 e− nm−2 79 s−1) values of 0.84 and 3.19, energy dispersions of 0.025 eV/ch (in Sec. III-A and III-D) 80 and 0.1 eV/ch (in Sec. III-C), and convergence and collection semi-angles of ~0 mrad and 81 6 mrad. Prior to the dose-dependent EELS, we determined the measurement positions and 82 stored their coordinates with a very small j. Then, after blanking the electron beam, we 83 set j to either 0.84 or 3.19 by adjusting a condenser lens to a premeasured value. Moving 84 to the measurement position, we started the repeated acquisition of a single EELS 85 spectrum with each exposure time Δt. After starting the repetitive EELS, the specimen 86 was exposed to the electron beam at time t0. Denoting the start time of the i-th EELS as ti 87 and the number of acquisitions of a single spectrum as nf, we define the total electron dose 88 De for the i-th EELS as De = 𝑗(𝑡𝑖 + 𝑛𝑓∆𝑡 2 − 𝑡0⁄ ). Δt and nf were 0.2 s and 2–8 for a high 89        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770196  loss region (i.e., 280 eV–580 eV in Sec. III-A–III-C), and 4.9 ms and 10 for a low loss 90 region (i.e., less than 35 eV in Sec. III-D). In Sec. III-D, we collected 4×104 EELS spectra 91 from unirradiated positions in a 1 μm×1 μm area with Δt=4.9 ms using a beam scan mode 92 to obtain a single spectrum with a high signal-to-noise ratio by integrating those spectra. 93 The flux j was estimated to be ~0.01. Both the convergence and collection semiangles 94 were set to 16 mrad. The thickness of the measurement areas was calculated to be 80–90 95 nm, using the standard log-ratio method.12 In Sec. III-E, we also describe dose-dependent 96 ED at 103 K using 300 keV electrons, an electron flux of 8.07 e− nm−2 s−1 (j =8.07×10−4), 97 a selected area of 4.45 μm diameter, an exposure time Δt of 1.2768 s for single ED and 98 another CMOS-based camera (One View, Gatan Inc.) mounted on the microscope. 99 Denoting the start time of the i-th ED as ti and the exposure time of the single ED as Δt, 100 we can write the total electron dose De for the i-th ED as De = 𝑗(𝑡𝑖 + ∆𝑡 2 − 𝑡0⁄ ), similar 101 to the case of EELS. 102  103 III. RESULTS AND DISCUSSION 104 A. Dose-dependent EELS for carbon K edge at 303 K and 103 K  105 Figure 2 shows the dose dependence of the carbon K edge at 303 K and 103 K, where 106 the electron flux j (×104 e− nm−2 s−1) was set to 0.84 for the dose De (×104 e− nm−2 s−1) of 107        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770197  0.16–13.6 and to 3.19 for De of 21.0–953. The combination of spectra obtained with two 108 different j in Fig. 2 is appropriate because the flux dependence of the electron beam 109 damage in the epoxy resin was small and the dose is an important parameter in describing 110 the degree of damage. In Figs. 2(a) and 2(b), the series of spectra obtained with the same 111 j have the same y-axis scale. There are clear differences in the carbon K edge profiles 112 between 303 K and 103 K, as shown in Fig. 2. Because carbon is the main element in the 113 epoxy resin, the damage mechanism and postdamage structure are significantly different 114 between 303 K and 103 K. When De =0.16, the carbon K edge has a sharp peak at 285.3 115 eV (peak A), characterized by 1s→π* excitations at both 303 K and 103 K. The position 116 of peak A shifts ~0.5 eV from 285.3 eV at De =0.16 to ~284.8 eV at De =953 for both 117 temperatures. For 303 K in Fig. 2(a), an additional intensity at ~286.7 eV (labeled B) is 118 observed at De >21, making peak A appear to broaden toward the higher energy loss side 119 with increasing De. For 103 K in Fig. 2(b), two additional peaks appear at 287.4–287.6 120 eV (labeled C) when De ≥1.5 and at 286.6–286.7 eV (labeled B) when De ≥21, where peak 121 B is more clearly observed than at 303 K. Peaks B and C also shift ~0.1 eV to the lower 122 energy loss side in the range 1.5≤De ≤953, and ~0.2 eV in the range 21≤De ≤953 123 respectively, as De increases in Fig. 2(b). In the energy loss region of 288–293 eV, a 124 steplike profile is observed only when De =0.16 at 103 K [Fig. 2(b)], and the variations in 125        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770198  the EELS profile for 0.83≤De ≤953 at 103 K are similar to those for 0.16≤De ≤953 at 303 126 K. Peaks B and C, i.e., the presence of two energetically localized unoccupied states for 127 103 K, suggest that the carbon atoms in the epoxy resin reconstruct two stable chemical 128 bonds under electron irradiation, while they reconstruct chemical bonds in a disordered 129 manner at 303 K. It is known that the primary damage process, i.e., inelastic scattering 130 with molecular excitation, ionization, and collective molecular excitation, is dominant at 131 low temperatures and causes molecular vibrations and the breaking of bonds such as C-132 H and C-C.13 As the specimen temperature increases, secondary processes, i.e., the 133 thermal diffusion of bond-broken atoms such as H atoms, the cross-linking of molecular 134 chains, and the evaporation of atoms and molecules, affect the polymer structure.13 The 135 difference in the carbon K edge variation in Fig. 2 is attributed to the greater and lesser 136 effects of the secondary processes at 303 K and 103 K, respectively. In the following, we 137 focus on the irradiation effects at 103 K, because the carbon K edge for De =0.16 has fine 138 structures that probably reflect the original electronic structure, and there is 139 approximately only one primary damage process to be considered. 140     141 B. Carbon K edge at low total dose 142        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.01770199  Figure 3 shows the same EELS spectrum for De =0.16 as in Fig. 2(b) with the 143 reference EELS spectra of aniline and diethyl ether in the vapor phase14, 15 and their linear 144 combination for comparison. The carbon K edge of the cured epoxy resin reflects the 145 linear combination of the EELS for the benzenoid part and the EELS for the side-chain 146 part, as shown in Fig. 1(b). To ensure that the spectrum for De =0.16 in Fig. 2(b) reflects 147 the original state, we roughly assume that the carbon K edges of the benzenoid and side-148 chain parts are those for aniline and diethyl ether, respectively, and compare the carbon 149 K edge for the epoxy resin obtained with De =0.16 at 103 K and a linear combination of 150 the carbon K edges for aniline and diethyl ether. The spectrum for aniline shows four π* 151 peaks at 285.2 eV, 286.7 eV, 289.0 eV, and 290.5 eV as a result of the dominant 152 excitations of 1s (Cbz-H) → π* (e2u), 1s (Cbz -Nsp3) → π* (e2u), 1s (Cbz-H) → π* (b2g), and 153 1s (Cbz -Nsp3) → π* (b2g) and two broad σ* peaks at 294.5 eV and 302 eV, respectively, 154 in Fig. 3.14, 16, 17 The subscripts ‘bz’ and ‘sp3’ denote the atom in the benzenoid ring and 155 electronic state, respectively. The carbon K edge for the benzenoid part in the epoxy resin 156 is considered to be similar to that for aniline with a minor difference due to the presence 157 of Cbz-CH2. The spectrum for diethyl ether shows peaks as a result of excitations from 1s 158 to Rydberg 3s, 3p states, and two σ* states.15 The carbon K edge for the side-chain part 159 in the epoxy resin is considered to be similar to that for diethyl ether with partial 160        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701910  modifications. The spectrum for the linear combination of aniline and diethyl ether was 161 obtained by normalizing their intensities in the range between 282 eV and 310 eV and 162 combining the normalized spectra in a 1:1 ratio. The experimental spectrum for De =0.16 163 has peaks at 287 eV (labeled α), 289–291 eV (labeled β), 293.5 eV (labeled γ), and 301 164 eV (labeled δ) in addition to peak A. The profile for De =0.16 is similar to the linear 165 combination profile, although the intensities of peaks A and α are relatively lower than 166 those of the combination spectrum. Peak A is assigned to the 1s (Cbz-H) → π* (e2u) 167 excitations in the benzenoid part; the excitation energy for 1s (Cbz-CH2) → π* (e2u) is 168 expected to be similar to 1s (Cbz-H) → π* (e2u) as in the case of aniline.14, 18 The peak α 169 can be attributed to the excitations of 1s (Cbz -Nsp3) → π* (e2u) in the benzenoid part and 170 1s → Rydberg 3s in the side-chain part. The peak β is mainly contributed by the 171 excitations of 1s (Cbz -H) → π* (b2g) and 1s (Cbz -Nsp3) → π* (b2g) in the benzenoid part 172 and 1s → Rydberg 3p in the side-chain part. The peaks γ and δ are related to the excitation 173 to the lower and higher σ* states in both the benzenoid and side-chain parts. According 174 to the above considerations from Fig. 3, we find that the spectrum for De =0.16 at 103 K 175 mainly reflects the chemical bonding state for the original epoxy resin before irradiation. 176 In other words, the original structure, i.e., the chemical bonding network of carbon atoms, 177 is maintained up to the dose of ~103 e− nm−2 when the carbon K edge is probed. 178        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701911   179 C. Carbon K edge variation with nitrogen and oxygen K edges  180 To interpret the origin of peaks B and C that appear in the carbon K edge during 181 irradiation [Fig. 2(b)], we investigated the correlations between the carbon, nitrogen, and 182 oxygen K edges. Figure 4 shows the dose dependence of the carbon, nitrogen, and oxygen 183 K edges obtained simultaneously from the same specimen area with j = 0.83 at 103 K. 184 Figures 4(a) and 4(b) show exactly the same data, and the spectral colors in Fig. 4(b) 185 correspond to those in Fig. 4(a). Figure 4(c) shows plots of the intensity for peaks B, C, 186 D, E, and F in Figs. 4(a) and 4(c) as a function of De, using energy ranges of 0.7 eV for 187 B and C, 3.5 eV for D, and 2.2 eV for E and F to obtain the integrated intensity. 188 For the carbon K edge, peaks B (at 286.6 eV) and C (at 287.4 eV) appear with 189 increasing electron dose De, as shown in Fig. 2(b). Peak C begins to increase in intensity 190 with De, reaches a maximum at De~10, and then decreases in the range 14≤De ≤60 in Fig. 191 4(c). Peak B monotonically increases with De up to De~60 in Fig. 4(c). Peak A shifts ~0.3 192 eV to the lower energy loss side and broadens as De increases from 0.5 to 60.8 without a 193 critical decrease in intensity, as seen in Fig. 2(b). Peaks B and C can be attributed to the 194 excitation of two π* states newly formed under electron irradiation. 195        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701912  For the nitrogen K edge, the intensity in the range of 404–410 eV is dominant at De 196 =0.5. This intensity is attributed to the nitrogen 1s → σ* excitation in the original 197 benzenoid part, demonstrating that nitrogen atoms have sp3 states, as shown in Fig. 1(b). 198 Then, the intensity between 397 and 400 eV with a maximum at 399.2 eV (peak D) 199 increases rapidly up to De =34 followed by little change for 34 ≤ De ≤ 60 in Fig. 4(c). 200 Peak D and its shoulder intensities on the lower energy loss side are attributed to the 201 excitation from 1s to the newly formed π* states (i.e., two or more π* states with close 202 energy levels) under irradiation until De reaches 34. The result suggests that chemical 203 bonds of imine (C=N) are formed, and peak B in the carbon K edge and peak D in the 204 nitrogen K edge originate from the excitations to π* states in their sp2 bonds.19, 20  205 For the oxygen K edge at De =0.5, there is a broadened intensity (labeled G) in the 206 region 536–544. This intensity is attributed to the oxygen 1s → σ* excitation in the 207 original side-chain part, demonstrating that its oxygen atoms have sp3 states, as shown in 208 Fig. 1(b). The intensity of G decreases up to De ~20 and does not change in the range 209 20≤De≤60, as shown in Fig. 4(b). The intensity of peak F at 534.2 eV increases up to De 210 ~20 and decreases in the range 20≤ De≤ 60, as shown in Fig. 4(c). The intensity variation 211 of peak F in the oxygen K edge is similar to that of peak C in the carbon K edge. Because 212 peaks C and F are attributed to carbon 1s → π* and oxygen 1s → π* excitations, 213        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701913  respectively, their correlation in intensity variation suggests that carbonyl (C=O) bonds 214 are formed during irradiation.19, 21, 22 This implies that sp3-to-sp2 conversion has occurred 215 for oxygen atoms and their bonding carbon atoms. The intensity of peak E at 531.3 eV, 216 which is 3 eV lower in energy than peak F, increases up to De ~60 in Fig. 4(c). The origin 217 of minor peak E is not assigned in this paper. 218 The formation of new chemical bonds such as C=O and C=N is acceptable in terms 219 of bond breakage and subsequent hydrogen atom loss under electron irradiation. Figure 220 4(b) shows that there is no critical mass loss for carbon and nitrogen atoms. For oxygen 221 atoms, mass loss may occur only up to De~10 (i.e., no loss for De>10). Among the 222 constituents, hydrogen atoms are most likely to cause mass loss. Because C-H bonds 223 break more frequently in aliphatic chains than in aromatic compounds,13 it is likely that 224 C-H bonds in the side-chain part will break preferentially compared with Cbz-H in the 225 benzenoid part for the epoxy resin. The sputtering hydrogen atoms from C-H bonds in 226 the side-chain part can trigger the formation of new bonds such as C=O and C=N. 227  228 D. Dose-dependent EELS in the range of 1–35 eV 229 Figure 5 shows the variation of the EELS spectrum in the range of 1–35 eV for the 230 epoxy resin during irradiation at 103 K. Because of the large cross section of inelastic 231        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701914  scattering in the low-loss region, we used a shorter exposure time (Δt=0.049) with j=0.84 232 to obtain the spectra at early stages of damage in the range of De ≤ 6.8. The spectra for 233 16.3 ≤ De ≤ 399 were obtained using Δt=0.39 and j=3.2 to observe the final profile of 234 damaged states. At the initial stage, De=0.02, there is a characteristic profile in the range 235 of 3–10 eV in addition to the plasmon energy at 22.0 eV in Fig. 5. As De increases, the 236 intensities at around 4.8 eV and 6.3 eV gradually disappear and alternative broad intensity 237 maxima appear at around 5.2 eV, while the plasmon energy gradually shifts from 22.0 eV 238 at De=5.0 to 23.1 eV at De=399. The variation of the EELS profile in the range of 3.0–7.5 239 eV suggests that the primary inelastic scattering changes from π→π* single electron 240 transitions for De<5 to π-plasmon excitation for De>30. The intensities at ~5.2 eV and 241 23.1 eV at De=399 can be attributed to π and π+σ plasmons whose energies are similar to 242 those in amorphous carbon.23 A more accurate spectrum with a higher signal-to-noise 243 ratio was obtained at the initial stage by scanning the electron beam (i.e., shifting the 244 beam to fresh areas on the epoxy resin), acquiring low-dose EELS at 103 K, and 245 integrating each spectrum, as shown in Fig. 6(a). There are four peaks A, B, C, and D for 246 the epoxy resin in the range of 3.4–9.2 eV in Fig. 6(a). The local maxima for peaks A, B, 247 C, and D are located at 3.9, 4.7, 6.0, and 6.6 eV, respectively, and the onset of intensity 248 A is 3.5 eV. From the results of the Kramers–Kronig (K–K) analysis using the EELS 249        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701915  spectrum with the higher signal-to-noise ratio [Fig. 6(a)], we elucidated the real and 250 imaginary parts, i.e., ε1 and ε2 of the dielectric function as shown in Fig. 6(b), and the 251 optical absorption profile shown in Fig. 6(c). The four peaks A–D in optical absorption 252 in Fig. 6(c) correspond to peaks A–D in EELS [Fig. 6(a)]. Figure 6(c) also shows 253 reference optical absorption spectra of 4,4′-methylenedianiline (MDA),24 aniline,25 254 polyethylene glycol (PEG),26 and diethyl ether.27 The benzenoid and side-chain parts for 255 the epoxy resin have similar structures to MDA and PEG, respectively. The characteristic 256 peaks α, β, γ, and δ for aniline are contributed by electron excitations from π to four 257 degeneracy-resolved π* states,25, 28 with additional charge transfer transitions also being 258 important components of the β and γ peaks.28 Chemical shifts of these intensities occur 259 depending on the type and number of functional groups in the benzenoid ring, and the 260 shift for α and β is larger than that for γ and δ.25, 28 Peaks α and β for MDA are located at 261 4.17 eV and 5.06 eV, which are lower than those for aniline.24 For diethyl ether, the first 262 peak ζ, at 6.35 eV originates from the excitation to the 3s Rydberg state. The peak ζ also 263 shifts to the lower energy loss side as the ether compounds become more complex29. For 264 PEG, the intensity ζ is located at 5.82 eV.26 We consider that peaks A, B, and D are 265 classified as α, β, and γ+δ, respectively, and originate from the benzenoid part of the 266 epoxy resin, while peak C is classified as ζ and originates from the side-chain part. In any 267        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701916  case, we have successfully obtained molecular electronic transitions of the original epoxy 268 resin by low-dose EELS at 103 K. 269  270 E. Dose-dependent ED at initial stage of irradiation 271 From the above-described EELS experimental results, we concluded that chemical 272 bonding states (C-C, C-O, C-N) and molecular electronic states (π, π*) are stable up to 273 De~0.1. To investigate the effects of the initial electron irradiation around De=0.1, we 274 conducted dose-dependent ED with j =8.07×10−4 at 103 K. Figures 7(a) and 7(b) show 275 ED patterns with De=0.0005 (i.e., first acquisition) and De=1.211 (i.e., after irradiation 276 for 1500 s), where the scale bars indicate the magnitude of the scattering vector, q. 277 Although three halo rings labeled h1, h2, and h3 are visible in both Figs. 7(a) and 7(b), 278 the intensity distribution for h1 broadens with increasing De. By radially rotating the ED 279 patterns, we obtained the radially averaged ED profile with different De values between 280 0.0005 and 1.211, as in the logarithmic plots in Fig. 8(a). Figures 8(b)–8(d) show the 281 profiles for h1, h2, and h3 after subtracting power-law fit functions from the radially 282 averaged ED profile. The first fit ranges were 2.77–5.54 nm−1, 19.7–22.4 nm−1, and 36.8–283 39.6 nm−1 for h1, h2, and h3, respectively, and the second fit range was 76.1–94.6 nm−1 284 for all halo rings. Figure 8(a) shows that the peak for h1 decreases and broadens with 285        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701917  increasing De, with small shifts of the peak center from 12.9 nm−1 (i.e., r=0.599 nm) for 286 De=0.0005 to 13.3 nm−1 (i.e., r=0.581 nm) for De=1.211, where r is the corresponding 287 real-space distance. More specifically, r ( ≈ 2.459π q⁄ ) is either the interatomic, 288 intramolecular, or intermolecular distance calculated by assigning the peaks in Figs. 8(b)–289 8(d) to the first maximum in the Debye scattering formula.11, 30 The peak centers for h2 290 and h3 are 29.9 nm−1 (i.e., r= 0.258 nm) and 52.4 nm−1 (i.e., r= 0.147 nm), respectively, 291 for all De values in Figs. 8(b) and 8(c). Figure 8(e) shows plots of normalized intensity 292 for h1, h2, and h3 as a function of De, where intensities defined with integrated ranges of 293 8.31–17.5 nm−1, 25.0–34.3 nm−1, and 44.0–60.7 nm−1 are normalized by intensities for 294 De=0.0005. It is found that h2 has a high rate of intensity decrease, while h3 has a low 295 rate of intensity decrease.  296 The third halo ring (h3) with r=0.147 nm is attributed to the dominant interatomic 297 distance (i.e., chemical bond length). The bond lengths constituting the original cured 298 epoxy resin are expected to be approximately 0.138 nm (Cbz-Cbz), 0.142 nm (Cbz -Nsp3), 299 0.147 nm (Csp3-Nsp3), 0.143 nm (Csp3-Osp3), 0.108 nm (Cbz -H), 0.109 nm (Csp3-H), and 300 0.97 nm (O-H).31 Thus, r=0.147 nm for h3 reflects the bond lengths of C-C, C-N, and C-301 O. During sp3-to-sp2 conversion, which becomes dominant for De≳20 (Fig. 4), bond 302 lengths can be reduced from 0.147 nm (Csp3-Nsp3) to 0.128 nm (Csp2=Nsp2) and from 303        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701918  0.143 nm (Csp3-Osp3) to 0.121 nm (Csp2=Osp2). Because there is no clear peak shift or 304 profile change for h3 in Fig. 8(d), the sp3 state is dominant in nitrogen and oxygen atoms 305 at least up to De ~1.2, which is consistent with the EELS results in Fig. 4: the formation 306 of C=O, C=N starts in the range 0.5≤De≤7.2. The second halo ring (h2) with r= 0.258 nm 307 is attributed to the dominant intramolecular distance, i.e., the second nearest neighbor 308 distance between carbon, nitrogen, and oxygen atoms. The large decrease in intensity for 309 h2 in Fig. 8(e) suggests that some bond angles, probably in the side-chain part, have 310 changed considerably. The first halo ring (h1) with r~0.60 nm is attributed to dominant 311 intermolecular distances characterized by benzenoid and side-chain parts in the cured 312 epoxy resin. The decrease in peak intensity for h1 suggests that the molecular geometry 313 of the side-chain or benzenoid parts is partially changed by deformation or rotation. The 314 broadening of the peak for h1 suggests that the intermolecular distances have been 315 randomized. We propose that these changes for h2 and h1, which occur simultaneously, 316 are triggered by C-H scission,13 which can be followed by Coulomb interaction between 317 fast electrons and radicals at low temperatures. The results of dose-dependent ED indicate 318 that changes in molecular geometry dominate at intramolecular and intermolecular length 319 scales (i.e., ~0.26 nm and ~0.60 nm). 320  321        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701919  F. Damage mechanism 322 Figure 9 illustrates a possible damage mechanism of the cured epoxy resin by the fast 323 electrons at low temperatures derived from the above experimental results, in terms of 324 molecular structure (upper figure) and π-electron distribution (lower figure). Before 325 electron irradiation, nitrogen and oxygen atoms form sp3 states for their chemical bonding, 326 and the π-electron exists only on the benzenoid ring. At the initial stage of irradiation (De327 ≲0.1), C-H scissions and subsequent Coulomb interactions between radicals and fast 328 electrons can change the intramolecular geometry, such as the bond angles in the side-329 chain part, and also the intermolecular geometry (geometry between two benzenoid parts, 330 between two side-chain parts, and between benzenoid and side-chain parts). In the range 331 0.5≲De≲10, nitrogen and oxygen atoms change from sp3 to sp2 states, and the sp2 state of 332 carbon also increases, implying the formation of C=O and C=N bonds in the epoxy resin, 333 as illustrated in Fig. 9. Sputtering hydrogen atoms of -OH and its adjacent -CH can form 334 the C=O bond, while breaking a C-N bond and a sputtering hydrogen atom of -CH 335 adjacent to nitrogen atom can form the C=N bond. The mechanism in Fig. 9 implies that 336 π-electrons, initially present only at the benzenoid ring, also become present in the side-337 chain part, resulting in an increase in π-electron density throughout the epoxy resin. 338 Furthermore, as a result of the sputtering hydrogen atom of -CH2 between two benzenoid 339        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701920  rings (Fig. 1), the benzenoid rings can be connected and the π-electrons can become 340 spatially continuous. The increase in π-electron density with De is consistent with the 341 result in Fig. 5, where a predominant excitation changes from π→π* transitions (i.e., 342 single-electron excitation) to π-plasmon excitation (i.e., collective oscillation of the 343 charge distribution spread in real space by excited π-electrons) in the range of 3.0–7.5 eV 344 with increasing De. At the final stage of damage (De≳100), the profiles for 1s →σ* in the 345 carbon K edge in the range of 290–305 eV in Fig. 5 are analogous to those of amorphous 346 carbon,32 and the π-plasmon is relatively intense in Fig. 5. This suggests that the fully 347 damaged epoxy resin (De≳100) has a disordered structure mainly consisting of carbon 348 atoms with small numbers of nitrogen and oxygen atoms.23 Although the disordered 349 structure can partially have graphitic structures as a result of the joining of the benzenoid 350 rings, the presence of carbon and oxygen atoms prevents the fully damaged epoxy resin 351 from having a complete graphite-like honeycomb structure.33-35 352  353 IV. CONCLUSION 354    The damage mechanism and exposure tolerance of epoxy resin (i.e., TGMDA as the 355 base resin) to fast electrons were quantitatively investigated by dose-dependent EELS and 356 ED. It was found that the sp3 states of nitrogen, oxygen, and their adjacent carbon atoms 357        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701921  changed to sp2 states, forming C=N and C=O, as the total electron dose increased at 103 358 K. The sputtering hydrogen atoms and the breaking of C-H bonds by the fast electrons 359 can cause the sp3-to-sp2 conversion. It was proposed that the fully damaged epoxy resin 360 has a disordered structure with partially graphitized structures. The epoxy resin was very 361 sensitive to fast electrons, and the original electronic states were maintained at 103 K up 362 to a total dose of ~103 e− nm−2. Below this dose limit, we were able to measure electronic 363 states in the ultraviolet region by EELS with the high DQE camera. Intra- and 364 intermolecular geometry changes below and around the total dose of ~103 e− nm−2 were 365 also observed by ED. The results obtained in this study provide important information for 366 the electron microscopic analysis of various epoxy resins and other polymeric materials. 367 The results are also important from the viewpoint of durability and mechanisms of the 368 degradation of epoxy resins by fast charged particles in special environments such as 369 space. 370  371 ACKNOWLEDGMENTS 372 J.K. thanks O. Cretu and N. Kawamoto (NIMS) for their insightful comments. 373  374 AUTHOR DECLARATIONS 375        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701922  Conflict of Interest 376 The authors have no conflicts to disclose. 377 Author Contributions 378 A.N. and Y.S. prepared the specimens. A.N. and J.K. conducted EELS and ED 379 experiments. All authors discuss the experimental results. J.K. wrote the manuscript with 380 the support of all the authors. 381  382 DATA AVAILABILITY 383 The data that support the findings of this study are available within this article 384  385 REFERENCES 386 1 J. C. Capricho, B. Fox, and N. Hameed, Polymer Reviews 60, 1–41 (2020).. 387 2 A. Paillous, and C. Pailler, Composite 25, 287–295 (1994). 388 3 A. Shundo, S. Yamamoto, and K. Tanaka, JACS Au 2, 1522–1542 (2022). 389 4 T. Hoshino, Y. Okamoto, A. Yamamoto, and H. Masunaga, Sci. Rep. 11, 9767 (2021). 390 5 S. Horiuchi, Y. Liu, T. Hanada, and H. Akiyama, Appl. Surf. Sci. 599, 153964 (2022). 391 6 G. Haberfehlner, S. F. Hoefler, T. Rath, G. Trimmel, G. Kothleitner, and F. Hofer, 392 Micron 140, 102981 (2021). 393        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701923  7 R. Colby, R. E. A. Williams, D. L. Carpenter, N. Bagués, B. R. Ford, and D. W. 394 McComb, Ultramicroscopy 246, 113688 (2023). 395 8 T. Sasuga, and A. Udagawa, Polymer 32, 402–408 (1991). 396 9 G. Yu, D. Shangli, Y. Dezhuang, H. Shiyu, and L. Zhijun, J. Polym. Sci., Part B: 397 Polym. Phys. 44, 177–184 (2006). 398 10 W. Zhang, L. G. d. A. Melo, A. P. Hitchcock, and N. Bassim, Micron 120, 74–79 399 (2019). 400 11 K. Yoshida, H.-H. Huang, T. Miyata, Y. K. Sato, and H. Jinnai, Microscopy 72, 361–401 367 (2022). 402 12 R. F. Egerton, Electron energy-loss spectroscopy in the electron microscope, Third 403 ed. (Springer, New York, 2011). 404 13 L. Reimer, and H. Kohl, Transmission Electron Microscopy, Fifth ed. (Springer, New 405 York, 2007). 406 14 C. C. Turci, S. G. Urquhart, and A. P. Hitchcock, Can. J. Chem. 74, 851–869 (1996). 407 15 S. G. Urquhart, A. P. Hitchcock, R. D. Priester, and E. G. Rightor, J. Polym. Sci., Part 408 B: Polym. Phys. 33, 1603–1620 (1995). 409 16 Y. Joseph, M. Wühn, A. Niklewski, W. Ranke, W. Weiss, C. Wöll, and R. Schlögl, 410 Phys. Chem. Chem. Phys. 2, 5314–5319 (2000). 411        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701924  17 D. Duflot, J. P. Flament, A. Giuliani, J. Heinesch, M. Grogna, and M. J. Hubin-412 Franskin, Phys. Rev. A 75, 052719 (2007). 413 18 S. G. Urquhart, A. P. Hitchcock, A. P. Smith, H. W. Ade, W. Lidy, E. G. Rightor, and 414 G. E. Mitchell, J. Electron. Spectrosc. Relat. Phenom. 100, 119–135 (1999). 415 19 J. Alleon, S. Bernard, C. L. Guillou, O. Beyssac, K. Sugitani, and F. Robert, 416 Geochem. Perspect. Lett. 7, 37–42 (2018). 417 20 H. Yabuta, M. Uesugi, H. Naraoka, M. Ito, A. L. D. Kilcoyne, S. A. Sandford, F. 418 Kitajima, H. Mita, Y. Takano, T. Yada, Y. Karouji, Y. Ishibashi, T. Okada, and M. Abe, 419 Earth, Planets and Space 66, 156 (2014). 420 21 K. Varlot, J. M. Martin, C. Quet, and Y. Kihn, Ultmicroscopy 68, 123–133 (1997). 421 22 A. P. Hitchcock, S. G. Urquhart, and E. G. Rightor, J. Phys. Chem 96, 8736–8750 422 (1992). 423 23 S. Bhattacharyya, C. Vallée, C. Cardinaud, O. Chauvet, and G. Turban, J. Appl. Phys. 424 85, 2162–2169 (1999). 425 24 E. Mohseni, M. R. Yaftian, H. Shayani-jam, A. Zamani, and F. Piri, Iran. Polym. J. 426 29, 403–409 (2020). 427 25 K. Kimura, H. Tsubomura, and S. Nagakura, Bull. Chem. Soc. Jpn. 37, 1336–1346 428 (1964). 429        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701925  26 M. Todica, O. Stan, C. V. Pop, Ş. Răzvan, and C. Niculăescu, Rom. J. Phys. 65, 702 430 (2020). 431 27 R. K. Burdick, J. P. Villabona-Monsalve, G. A. Mashour, and T. Goodson, Sci. Rep. 432 9, 11351 (2019). 433 28 T. Ari, H. Güven, and N. Ecevit, J. Electron. Spectrosc. Relat. Phenom. 73, 13–23 434 (1995). 435 29 Q. Y. Shang, P. O. Moreno, R. Disselkamp, and E. R. Bernstein, J. Chem. Phys. 98, 436 3703–3712 (1993). 437 30 P. Debye, Ann. Phys. 351, 809–823 (1915). 438 31 A. G. Orpen, L. Brammer, F. H. Allen, O. Kennard, D. G. Watson, and R. Taylor, 439 Appendix A in Structure Correlation (VCH Verlagsgesellschaft mbH, Weinheim, 440 1994), pp. 752–858. 441 32 F. Langenhorst, and V. L. Solozhenko, Phys. Chem. Chem. Phys. 4, 5183–5188 442 (2002). 443 33 D. Ugarte, Nature 359, 707–709 (1992). 444 34 M. Chhowalla, H. Wang, N. Sano, K. B. K. Teo, S. B. Lee, and G. A. J. Amaratunga, 445 Phys. Rev. Lett. 90, 155504 (2003). 446 35 1 J. C. Capricho, B. Fox, and N. Hameed, Polymer Reviews 60, 1–41 (2020). 447        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701926  2 A. Paillous, and C. Pailler, Composite 25, 287–295 (1994). 448 3 A. Shundo, S. Yamamoto, and K. Tanaka, JACS Au 2, 1522–1542 (2022). 449 4 T. Hoshino, Y. Okamoto, A. Yamamoto, and H. Masunaga, Sci. Rep. 11, 9767 (2021). 450 5 S. Horiuchi, Y. Liu, T. Hanada, and H. Akiyama, Appl. Surf. Sci. 599, 153964 (2022). 451 6 G. Haberfehlner, S. F. Hoefler, T. Rath, G. Trimmel, G. Kothleitner, and F. Hofer, 452 Micron 140, 102981 (2021). 453 7 R. Colby, R. E. A. Williams, D. L. Carpenter, N. Bagués, B. R. Ford, and D. W. 454 McComb, Ultramicroscopy 246, 113688 (2023). 455 8 T. Sasuga, and A. Udagawa, Polymer 32, 402–408 (1991). 456 9 G. Yu, D. Shangli, Y. Dezhuang, H. Shiyu, and L. Zhijun, J. Polym. Sci., Part B: 457 Polym. Phys. 44, 177–184 (2006). 458 10 W. Zhang, L. G. d. A. Melo, A. P. Hitchcock, and N. Bassim, Micron 120, 74–79 459 (2019). 460 11 K. Yoshida, H.-H. Huang, T. Miyata, Y. K. Sato, and H. Jinnai, Microscopy 72, 361–461 367 (2022). 462 12 L. Reimer, and H. Kohl, Transmission Electron Microscopy, Fifth ed. (Springer 463 Science+Business Media, LLC, New York, 2007). 464 13 C. C. Turci, S. G. Urquhart, and A. P. Hitchcock, Can. J. Chem. 74, 851–869 (1996). 465        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701927  14 S. G. Urquhart, A. P. Hitchcock, R. D. Priester, and E. G. Rightor, J. Polym. Sci., Part 466 B: Polym. Phys. 33, 1603–1620 (1995). 467 15 Y. Joseph, M. Wühn, A. Niklewski, W. Ranke, W. Weiss, C. Wöll, and R. Schlögl, 468 Phys. Chem. Chem. Phys. 2, 5314–5319 (2000). 469 16 D. Duflot, J. P. Flament, A. Giuliani, J. Heinesch, M. Grogna, and M. J. Hubin-470 Franskin, Phys. Rev. A 75, 052719 (2007). 471 17 S. G. Urquhart, A. P. Hitchcock, A. P. Smith, H. W. Ade, W. Lidy, E. G. Rightor, and 472 G. E. Mitchell, J. Electron. Spectrosc. Relat. Phenom. 100, 119–135 (1999). 473 18 J. Alleon, S. Bernard, C. L. Guillou, O. Beyssac, K. Sugitani, and F. Robert, 474 Geochem. Perspect. Lett. 7, 37–42 (2018). 475 19 H. Yabuta, M. Uesugi, H. Naraoka, M. Ito, A. L. D. Kilcoyne, S. A. Sandford, F. 476 Kitajima, H. Mita, Y. Takano, T. Yada, Y. Karouji, Y. Ishibashi, T. Okada, and M. Abe, 477 Earth, Planets and Space 66, 156 (2014). 478 20 K. Varlot, J. M. Martin, C. Quet, and Y. Kihn, Ultmicroscopy 68, 123–133 (1997). 479 21 A. P. Hitchcock, S. G. Urquhart, and E. G. Rightor, J. Phys. Chem 96, 8736–8750 480 (1992). 481 22 S. Bhattacharyya, C. Vallée, C. Cardinaud, O. Chauvet, and G. Turban, J. Appl. Phys. 482 85, 2162–2169 (1999). 483        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701928  23 E. Mohseni, M. R. Yaftian, H. Shayani-jam, A. Zamani, and F. Piri, Iran. Polym. J. 484 29, 403–409 (2020). 485 24 K. Kimura, H. Tsubomura, and S. Nagakura, Bull. Chem. Soc. Jpn. 37, 1336–1346 486 (1964). 487 25 M. Todica, O. Stan, C. V. Pop, Ş. Răzvan, and C. Niculăescu, Rom. J. Phys. 65, 702 488 (2020). 489 26 R. K. Burdick, J. P. Villabona-Monsalve, G. A. Mashour, and T. Goodson, Sci. Rep. 490 9, 11351 (2019). 491 27 T. Ari, H. Güven, and N. Ecevit, J. Electron. Spectrosc. Relat. Phenom. 73, 13–23 492 (1995). 493 28 Q. Y. Shang, P. O. Moreno, R. Disselkamp, and E. R. Bernstein, J. Chem. Phys. 98, 494 3703–3712 (1993). 495 29 P. Debye, Ann. Phys. 351, 809–823 (1915). 496 30 A. G. Orpen, L. Brammer, F. H. Allen, O. Kennard, D. G. Watson, and R. Taylor, 497 Appendix A in Structure Correlation (VCH Verlagsgesellschaft mbH, Weinheim, 498 1994), pp. 752–858. 499 31 F. Langenhorst, and V. L. Solozhenko, Phys. Chem. Chem. Phys. 4, 5183–5188 500 (2002). 501        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701929  32 D. Ugarte, Nature 359, 707–709 (1992). 502 33 M. Chhowalla, H. Wang, N. Sano, K. B. K. Teo, S. B. Lee, and G. A. J. Amaratunga, 503 Phys. Rev. Lett. 90, 155504 (2003). 504 34 D. Lungerich, H. Hoelzel, K. Harano, N. Jux, K. Y. Amsharov, and E. Nakamura, 505 ACS Nano 15, 12804–12814 (2021). 506   507        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701930  Figures 508  509  510  511 FIG. 1. (a) Monomers of TGMDA and 2E4MZ as base resin and curing agent, 512 respectively. (b) Possible curing reaction sequence. Polymerization by alternative ring-513 opening and chain-growth processes followed by thermal detachment of 2E4MZ yields 514 the final cured epoxy resin composed of benzenoid and side-chain parts. 515   516        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701931   517  518 FIG. 2. Changes in carbon K edges for the epoxy resin with total electron dose De at (a) 519 303 K and (b) 103 K.  520   521        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701932   522  523  524 FIG. 3. Carbon K edges of the epoxy resin for De=0.16 at 103 K in Fig. 2(b), with 525 aniline,14 diethyl ether,15 and the linear combination profile of aniline and diethyl ether. 526   527        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701933   528  529  530 FIG. 4. Simultaneously acquired carbon, nitrogen, and oxygen K edges for the epoxy 531 resin as De increases, shown in (a) array and (b) overlay. (c) Intensity changes for B–F 532 with De. 533   534        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701934   535  536  537 FIG. 5. Change in EELS spectrum for the epoxy resin at 103 K in the range of 1–35 eV 538 with total electron dose De.  539   540        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701935   541  542 FIG. 6. (a) High signal-to-noise ratio EELS spectrum for the epoxy resin at 103 K with a 543 dose equivalent to De~0.01 with reference EELS. (b) Real (ε1) and imaginary (ε2) parts 544 of the dielectric function derived by K–K analysis for the EELS spectrum of De~0.01 in 545 (a). (c) Optical absorption profile of the epoxy resin obtained by K–K analysis using the 546 EELS spectrum in (a), with optical absorption spectra of 4,4′-methylenedianiline 547 (MDA),24 aniline,25 polyethylene glycol (PEG),26 and diethyl ether.27  548   549        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701936   550  551 FIG. 7. ED pattern of the epoxy resin at 103 K with total electron dose De values of (a) 552 0.0005 and (b) 1.211. The scale is in the magnitude of the scattering vector. 553   554        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701937   555 FIG. 8. (a) Logarithmic plot of the radially averaged ED patterns [Figs. 7(a), 7(b), and 556 others (not shown)] of the epoxy resin at 103 K with different total electron dose De values. 557 Net intensity profiles after subtraction of background intensity for (b) h1, (c) h2, and (d) 558 h3 with different total electron dose De values as indicated in (a). (e) Normalized 559 intensities of h1, h2, and h3 as a function of De. 560        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019Accepted to J. Chem. Phys. 10.1063/5.017701938   561  562 FIG. 9. Damage mechanism of the epoxy resin by fast electrons from the viewpoint of 563 molecular structure (upper figure) and π-electron distribution (lower figure). The sp3-to-564 sp2 conversion for nitrogen, oxygen, and their adjacent carbon atoms follows the C-H 565 scission and changes in intra- and intermolecular geometries (upper figure). The initially 566 isolated π-electrons become dense (lower figure), allowing π-plasmon excitation. 567        This is the author’s peer reviewed, accepted manuscript. However, the online version of record will be different from this version once it has been copyedited and typeset. PLEASE CITE THIS ARTICLE AS DOI:10.1063/5.0177019