Probing the beam-induced heating effect inside a transmission electron microscope by nanoparticle labels

Beam-induced heating effect on nanoscale samples is a crucial question, as it strongly inﬂu-ences the interpretation of observed unusual behaviours. This question is currently under debate without a convincing conclusion. Here, using silver nitride (Ag 3 N) nanoparticles as temperature labels, we perform an investigation on this heating effect inside a transmission electron microscope (TEM) under normal imaging conditions. Combined with experimental measurements and semi-quantitative calculations, a temperature increase of more than 100 K is estimated and conﬁrmed in the graphite carbon nitride (g-C 3 N 4 ) ﬁlms. Strong temperature gradients are found to exist in the single-end ﬁxed g-C 3 N 4 ﬁlms. The inﬂuencing factors of heat accumulation are also investigated and discussed. Findings in this study may shed some light on the understanding of the abnormal behaviours of nano-objects observed inside TEM.


INTRODUCTION
Transmission electron microscope (TEM) is one of the most powerful equipments used to characterise materials for their structures, phases, compositions, and properties. The elastic or inelastic interactions between high-energy electrons and specimen atoms enable precise determination of atom arrangements and elemental distributions, providing deep insight into the material structures and related properties. In terms of these interactions, significant side-effects usually co-occur during TEM observation and imaging, including sputtering [1], amorphisation [2], ionisation [3], carbonisation [4], local heating [5], and so forth. Among all these side-effects, the electron-beaminduced local heating is quite elusive, as it is currently neither able to directly measure the precise temperature increase inside the local nanometre-sized specimen nor able to quantitatively determine how much energy has been transformed into heat and kept inside the specimen. Although there are ways to minimise the beam-heating effect, for example, using a quite low-dose beam to image together with a substrate with very high thermal conductivity, beam-induced abnormal phenomena are still frequently observed on nanometre-sized objects during imaging [5][6][7]. It reflects that the local heating effect may hardly be avoided in many cases. Thereby, a temperature increase could exist and bring considerable interference in understanding the intrinsic properties of the samples from the observed unusual behaviours, especially for metastable and low-dimension functional materials [8,9]. As early as the 1980s, Iijima et al. showed that Au nanoparticles deposited on a carbon substrate could be excited during TEM observation [10]. The geometric shapes of the Au nanoparticles fluctuated frequently when they were irradiated by the electron beam. This phenomenon was identified as quasi-melting and further similar experiments demonstrated that the electron-beam-induced temperature effect was a key reason accounting for the phenomenon. Aside from such notable shape changes, the temperature effect is also very often doubted to play a role in many other behaviours observed in TEM, for example, the abnormal deformation behaviours of nanoscale metals, semiconductors and oxides [11], the phase transitions in nanowires and nano-belts [12], the reconstruction and recrystallisations in amorphous nanoparticles [13,14]. Generally, the electron-beam-induced temperature increase mainly depends on two aspects, that is, the used beam intensity and voltage that govern the input energy to the specimen, and the supporting substrate that determines the heat dissipation. Li et al. estimated that an electron beam having an energy between 500 and 2 keV could cause a temperature rise up to a few hundred degrees on an organic specimen [15]. Egerton et al. indicated that materials with low thermal conductivity (e.g. 0.2-2 W/mK) could be melted or thermally degraded even under a relatively low beam density [16]. Yokota et al. found that AlSi nanoalloys could be easily melted during TEM imaging due to electron thermal spikes in the particles and poor thermal conduction away from the particles [17]. In a recent study, Asoro et al. reported that even for metal nanoparticles supported on carbon films, a temperature increase of 58 K was estimated to be induced by the electron beam during normal imaging [18]. All these investigations imply that the temperature effect during TEM imaging is somehow inevitable. Unfortunately, from current studies, it is hard to make a conclusion on how severe the temperature effect is and what influence will appear subsequently. As a result, the temperature effect is often argued in the interpretations of new phenomena observed in TEM but usually without a convincing conclusion.
To probe into this question, we here perform an in-situ investigation using silver nitride (Ag 3 N)-decorated graphite carbon nitride (g-C 3 N 4 ) samples as a model. The decomposition of the Ag 3 N nanoparticles can serve as temperature labels, providing a direct profile of the temperature distribution in the g-C 3 N 4 film induced by electron irradiation.

Materials preparation
Ag 3 N nanoparticles were synthesised by dissolving silver nitrate (AgNO 3 ) in concentrated ammonia solution [19,20]. Typically, 0.25 g AgNO 3 was added into 50 ml deionised water and magnetically stirred for 30 min to form the AgNO 3 solution. Then, a 50 ml sodium hydroxide (NaOH) solution (0.1 M) was added to the AgNO 3 solution dropwise. The amount of NaOH was superfluous and thus silver oxide (Ag 2 O) nanoparticles precipitated during the reaction. After the reaction, the turbid liquid was centrifuged and washed to obtain Ag 2 O particles. Ag 3 N nanoparticles were acquired by adding the Ag 2 O particles into superfluous ammonia water. The g-C 3 N 4 films were bought from Aladdin Company. Ag 3 N-decorated g-C 3 N 4 samples were obtained by mixing the Ag 3 N nanoparticles with g-C 3 N 4 films in an ammonia solution.

TEM characterisation and in-situ heating
A 300 kV TEM (FEI, Titan) was used to conduct the experiment with a beam intensity of 1.6 × 10 4 A/m 2 . The temperatures at which Ag 3 N nanoparticles decomposed were calibrated using a heating holder (Protochips Aduro™) with electrical chip (E-chip). The E-chip contained hole-arrays as observation windows, and holey carbon films were overlaid on the windows to support the nano-specimens. By controlling the applied current, the heating temperature of the E-chip can be precisely programmed. As the nominal ramp rate of the E-chip is 1000 K/ms, the retardation of the temperature increase during heating can thus be neglected [21]. Figure 1(a) shows the topography of a typical Ag 3 N-decorated g-C 3 N 4 sample. Figure 1(b) shows the energy dispersive X-ray spectrum of the film. The strong peaks of carbon (C) and nitrogen (N) elements demonstrate the integrality of the g-C 3 N 4 film. Notably, C signal is influenced by other sources such as the supporting substrate, so the signal ratio of C to N elements varies and cannot represent the component proportion of the g-C 3 N 4 film. Figure 1(c) shows a high-resolution TEM image of an Ag 3 N nanoparticle. From the fast Fourier transform pattern shown in Figure 1(d), the crystalline structure of Ag 3 N can be identified and confirmed by the (0 -2 2), (2 -2 2) and (2 0 0) planes [22] with zone axis [0 1 1]. It is observed that the Ag 3 N nanoparticles on the g-C 3 N 4 film cannot sustain for a long time during TEM imaging. More Ag 3 N-decorated g-C 3 N 4 film samples are checked during different electron-beam conditions, and it is found that under a beam intensity of 1.6 × 10 4 A/m 2 , decomposition of the Ag 3 N nanoparticles can gradually occur. Figures 2(a)-(j) show a typical decomposition process of the Ag 3 N nanoparticles with diverse sizes. As can be seen, several Ag 3 N nanoparticles start to decompose much earlier and faster than the others. However, all the Ag 3 N nanoparticles vanish eventually under continuous irradiation, leaving only the g-C 3 N 4 frames. These asynchronous decomposition processes imply that the electronbeam-induced heat accumulation is uneven in the g-C 3 N 4 film. Thereby, the sequence of the vanishing process of the Ag 3 N nanoparticles gives direct information of the distribution of the beam-induced heating effect.

RESULTS AND DISCUSSION
To facilitate checking the heat distribution, eight Ag 3 N nanoparticles are statistically selected and their size evolutions are recorded by measuring their projection areas from the TEM image sequence. Since most of the nanoparticles keep spherical during size reduction, the change on the projection area of the nanoparticles can thus be used as a direct indicator for monitoring the evolution of the particle sizes. Figure 3 collects the measured projection areas as a function of irradiation time. Two stages can easily be identified in the evolution trends of particles 2, 3, 4, 6, 7, and 8, that is, a first stage showing a flat trend with nearly unchanged particle sizes, and a second stage showing drastic size decrease. In the first stage, the particle sizes change a little during the initial irradiation. This is possibly ascribed to the shape/structure relaxation of the pristine nanoparticles that are staying at metastable state. In the second stage, continuous size reduction appears, indicating the start of structural decomposition induced by strong heat accumulation. Notably, nanoparticles marked by 1 and 5, which are located at the edge, present a much earlier start of decomposition than the others. Their absence of the first stage implies that the beam-induced heating effect on nanoparticles located at the film edge is much more severe due to the weak heat dissipation into surroundings. These findings prove that the decomposition of Ag 3 N nanoparticles is dominated by the beam-induced heating effect but not the beam-sputtering effect.
Bulk Ag 3 N is known to decompose at a temperature around 400 K [23]. However, the decomposition temperature of Ag 3 N nanoparticles may change due to the size effect [24]. To verify the decomposition temperature of the Ag 3 N nanoparticles, a heating holder was used to carry out a calibration. Holey carbon films are applied as supporting substrate instead of g-C 3 N 4 films, as they have a much higher thermal conductivity, and this can be helpful to relieve thermal retardation. Using the same beam intensity of 1.6 × 10 4 A/m 2 , it is found that the Ag 3 N nanoparticles can sustain a heating temperature below 380 K. When the heating temperature is increased to above 380 K, the Ag 3 N nanoparticles start to decompose as shown in Figure 4(a). To double-check the accuracy of the heating E-chip, bismuth (Bi) nanoparticles are used to make a comparison. By heating the E-chip to 515 K, sublimation of Bi nanoparticles is achieved and shown in Figure 4(b). Based on our previous work [21] dealing with the sublimation speed, the change rate of Bi nanoparticle radius dr/dt can be calculated by R is the gas constant (8.314 J•mol −1 •K −1 ), T is the thermodynamic temperature, k is the Boltzmann constant (1.38 × 10 −23 J•K −1 ) and V m and M r are the molecular volume and weight of Bi nanocrystal, respectively. P ∞ is the equilibrium vapour pressure over a flat surface that, for bulk Bi, follows the empirical vapour pressure equation from experiments.
A, B, C, D and E are the fitting constants [25]. For a certain temperature, the equilibrium vapour pressure over a flat surface is a constant. Nanoparticles sublimation is supposed to follow Kelvin's law with an exponential form, and the equilibrium vapour pressure over the curved surface of Bi nanocrystal P r can be written as γ is the surface energy (490 mJ•m −2 for Bi), ρ is the density. Figure 5 shows the calculated sublimation rate with the temperature for Bi nanoparticles having an initial radius of 2, 5, 10, 20 and 50 nm at various temperatures. The sublimation process is size-dependent and with a certain temperature, the smaller the particle radius is, the easier or faster it sublimates. In the experiment, Bi nanoparticle is not perfectly spherical. The different curvatures over the particle surface may give rise to the anisotropic shape reduction. Moreover, the defects of the particle may also produce some influence on its sublimation. Although it is quite difficult to exactly characterise the precise decreasing rate of each Bi nanoparticle, the rough estimation result that a particle having a radius of around 10 nm disappears within 70 s (the red marked particle in Figure 4) is apparently in good agreement with theoretical calculations where the sublimation rate is around 0.2 nm/s at 515 K. Hence, the temperature from the heating E-chip is credible. Thereby, the decomposition temperature of Ag 3 N nanoparticles is believed to be 380 K.
A method developed by Liu and Risbud is employed to semiquantitatively calculate the heat accumulation during electron-beam irradiation by treating the system as spherical particles embedded in thin films [26]. The input power is generated from electron collisions considered as cylindrical thermal spikes [27,28], and the heat dissipation is thermal conduction from the surrounding film. When the high energy electrons interact with the sample atoms, beam-induced temperature increase can be modelled as where J is the electron beam current density; e is the electron charge; is the thermal conductivity of surrounding film; R e is the radius of the area that electrons bombard on; t is the irradiation time; f and ρ is the heat capacity and mass density of particles, respectively. Q is the total energy loss of the electron and can be briefly divided into two parts, that is, the electron excitation loss Q e and the Coulomb encounter loss Q c , which can be described as where n is the number density of atoms in the particle, Z is the atomic number, r e is the classical electron radius, β is the ratio of accelerated electron and light speed, and IZ ( = 13.54 eV) is the average ionisation potential of the electrons in the atom. T m and T a can be described, respectively, as where E is the energy of electrons and R h is Rydberg energy for hydrogen.
To make a direct comparison, Bi nanoparticles and holey carbon film are also considered in the calculation. It is important to note that, in the experiment, Bi nanoparticles located on the g-C 3 N 4 film or holey carbon film as well as Ag 3 N nanoparticles located on holey carbon film can remain well unchanged under the same irradiation condition described above. Figure 6 shows the calculated temperature increase ∆T during TEM imaging with a beam intensity of 1.6 × 10 4 A/m 2 . Obviously, a heating up of more than 100 K can be achieved when the substrate is g-C 3 N 4 films (3.5 ± 0.3 W/mK [29]). In contrast, when holey carbon films are used as substrates (129 W/mK [30]), the temperature increase is only several kelvins. These results confirm that both the electron beam condition (acceleration voltage, beam intensity, irradiation time) and the substrate material play significant effects on the heat accumulation. As a result, Ag 3 N nanoparticles located in the g-C 3 N 4 films can decompose easily if heat cannot be dissipated timely. According to the FIGURE 6 Semi-quantitative calculation results temperature increment with irradiation time for Ag 3 N and Bi nanoparticle in the g-C 3 N 4 and holey carbon film network, respectively calculation results, nanoparticles show a rapid temperature increase at first, and then the temperature increase becomes very sluggish. The rapidly upward region with several tens of seconds can well explain the existence of nearly stable stage (first stage) for particles inside the g-C 3 N 4 network shown in Figure 3. At this stage, the temperature increment is not high enough for nanoparticles to decompose. With further irradiation, nanoparticles finally get enough heat energy and decompose dramatically. The model is for the perfectly embedded nanoparticles those are in good contact with surroundings. In actual experiments, the various contact conditions between nanoparticles and surrounding films lead to the inconsistent start-point of decomposition. As for particles, those that are supported at the edge of the film, the electron-induced heat cannot be conducted through the surrounding film efficiently and the thermal conductivity used in the above equation cannot be simply treated as the one of film network. Figure 7 gives the temperature increment of Ag 3 N and Bi nanoparticles in a network having different thermal conductivity after 180 s beam irradiation. The temperature increment is small enough and can be neglected when the thermal conductivity of the film network is larger than 30 W/mK.
To estimate the heat dissipation from the supporting grid, a single-end fixed Ag 3 N-decorated g-C 3 N 4 film was carefully examined. Figure 8 shows the illustration of the film sample with two monitored regions (red boxes). After 3-min irradiation with a beam intensity of 1.6 × 10 4 A/m 2 , the Ag 3 N nanoparticles in the two regions show significant discrepancies. In the region nearby the supporting copper (Cu) grid, all the Ag 3 N nanoparticles can well retain, even for the very small ones. In sharp contrast, the Ag 3 N nanoparticles in the region that is far away from the supporting Cu grid largely vanish after beam irradiation. This evidence points out that a large temperature gradient should exist in the g-C 3 N 4 film. The key factor causing this temperature gradient is the partially supported Cu grid. Therefore, during TEM imaging, the beam-induced temperature effect should be considered comprehensively, as the ther-

FIGURE 8
Temperature gradients in g-C 3 N 4 films. The discrepant decomposition behaviours of Ag 3 N nanoparticles in a single-end fixed Ag 3 Ndecorated g-C 3 N 4 sample. The conservation and disappearance of Ag 3 N nanoparticles in the regions nearby and far away from the copper grid, respectively, suggest a strong temperature gradient mal conduction between nano-samples and supporting films as well as the thermal conductivity of the supporting film itself indeed plays crucial roles. Moreover, as for those large-scale nanowires and membranes, significant thermal gradients may be induced during TEM imaging. This may cause usual behaviours that do not originate from their intrinsic properties.

CONCLUSION
Using Ag 3 N nanoparticles as temperature labels, we have explored the temperature effect inside a TEM during normal imaging. The decomposition temperature of the Ag 3 N nanoparticles is carefully calibrated to be 380 K by heating experiments. During normal imaging, a strong beam-heating effect can be caused, and a semi-quantitative calculation indicates a temperature increase of more than 100 K in the g-C 3 N 4 films due to heat accumulation, which well agrees with the experimental observation of Ag 3 N nanoparticle decomposition during imaging. Significant temperature gradients are experimentally verified and demonstrated, possibly providing valuable information in understanding the unusual behaviours in nanomaterials.