Desorption characteristics of selenium and tellurium thin films

Interest in the growth of chalcogenide-based thin films, though dating back to the mid-20th century, has seen a renewed interest over the past 10 years due to the unique electronic states found in chalcogenide-based compounds that cover a broad range of novel physical phenomena, such as Dirac or Weyl semimetals,^1–5 topological insulators,^6–10 high mobility semiconductors, superconductors, ferroelectrics, thermoelectrics, photovoltaics, and building blocks for diode lasers and detectors.^11–21 Furthermore, the relatively lower melting temperature compared to other functional materials, such as oxides, and accompanied by this a much-reduced ideal growth temperature range to balance sufficient surface diffusivity while keeping bulk diffusion to a minimum²² makes this material class ideally suited for the potential back end of line integration into existing semiconductor manufacturing processes. The high vapor pressure of the chalcogen elements makes it relatively simple to generate fluxes in physical vapor deposition systems by supplying chalcogens using low-temperature thermal effusion cells.²³ This situation renders the growth process of molecular beam epitaxy (MBE) ideally suited because a growth window for chalcogenide compounds can be easily accessed by oversupplying chalcogen.^24–27 However, favorably high vapor pressures are accompanied with a large volatility—a challenge if the crystallization temperature is comparable to or higher than the growth temperature and the grown film evaporates congruently, thus limiting growth rates and causing unintentional re-evaporation,²⁸ or incongruently thus leaving chalcogenide vacancies behind,^29–33 making it challenging to avoid the formation of native point defects. Furthermore, different species of chalcogen molecules have been reported to exist in the vapor phase over elemental chalcogen.³⁴ Due to the many stable polytypes, thermal evaporation was found to produce a chemical composition of chalcogen vapors ranging from atomic species up to 12-count multiatomic chains and rings, which showed significant variation with the vapor temperature.^34–40 The tendency to form chains rather than atoms or dimers in the gas phase adds additional ambiguity to rationalize the growth process better and to quickly improve film quality grown by MBE. Specifically, the presence of less reactive polyatomic species in the supplied chalcogen flux that is not dissociated into more reactive atomic species at the growth front to allow for an efficient incorporation into the growing film requires a much larger oversupply to compensate for the existing volatility and lack of reactivity.^24,26,41,42 Therefore, understanding the chemical state of the supplied species and their surface adsorption and desorption kinetics is advantageous.

Previously, we have investigated the temperature-dependent adsorption behavior of selenium and tellurium and reported a temperature-dependent sticking coefficient. A sharp drop in the sticking coefficients was found for both selenium and tellurium at temperatures as low as 80 and 200 °C, respectively.⁴³ Remarkably, while the sticking coefficient of tellurium dropped to zero, selenium maintained a sticking coefficient of about 0.2 at a temperature of 80 °C. Measuring the sticking coefficient for selenium at higher temperatures was not possible due to the design of the experiment and remained an open question.

In this work, we studied the desorption kinetics of Se and Te and determined the temperature dependence in a more direct way using a heated quartz crystal microbalance (QCM). Se and Te films with varying film thickness were first deposited at room temperature on the pristine quartz crystal surface used as a QCM sensor from thermal effusion cells in an MBE system. The QCM temperature was then increased and the desorption rate, i.e., mass loss from the QCM surface was detected. A significant increase in Te mass loss was found to be around 190 °C, with the entire film being evaporated within 20 °C. In contrast, the desorption kinetics of the Se films occurred via a two-stage process with a smaller percentage of the film desorbing at (65 ± 5) °C and the main portion evaporating around (90 ± 10) °C. From this two-stage behavior, it is concluded that either the evaporated Se was present as at least two chemically distinct species or occupied adsorption sites with different binding energies in the film.^44,45 The direct and quantitative determination of the chalcogen desorption process demonstrated here provides important insights into the kinetics of chalcogenide-based film growth in MBE and points toward the need for a more detailed understanding of the chemical composition state of Se and Te supplied from thermal evaporation sources during growth and their binding process into a chalcogenide film.

A schematic of the used R450 MBE reactor from DCA Instruments is shown in Figs. 1(a) and 1(b). The growth chamber dedicated to the growth of chalcogenides was equipped with a heated QCM (Colnatec) mounted on a linear motion arm that could be moved into the sample position. Pure silicon dioxide monocrystals (RC-cut) were used as oscillating quartz crystals with “Inficon” patterned gold electrodes deposited on the hromium underlayer (PhillipTech). Electrode surfaces had a nominal rms roughness of 7 μm. Quartz crystals were operated at 5.989 MHz.

An air-cooling system integrated into the QCM head allowed stabilization of the quartz crystal at low temperatures. The surface temperature of the crystal was monitored through an Xi80 infrared (IR) camera (Optris). The thermal IR camera image picturing the inside of the growth chamber with the inserted QCM is shown in Figs. 1(c) and 1(d) at room temperature and at an elevated temperature of about 200 °C, respectively. Either Se or Te was then evaporated at room temperature onto a pristine QCM crystal from single-filament Knudsen effusion cells at typical source temperatures of 165 and 225 °C, respectively, resulting in a flux of 1.79 × 10¹⁴ Se atoms/(cm² × s) and 8.44 × 10¹³ Te atoms/(cm²× s), which translated into effective deposition rates of 0.49 and 0.29 Å/s, respectively. A schematic of the rate- and thickness-time dependence recorded by the QCM during room temperature film deposition on the quartz crystal surface is shown in Fig. 1(e), where the yellow and green panels indicate the times that the chalcogenide shutter was closed (shutter state = zero) and open (shutter state = 1), respectively. The background pressure in the MBE reactor during measurements was kept at around 5 × 10⁻⁹ Torr. A thermal load onto the QCM from the hot evaporation sources during film deposition at room temperature, which would manifest as spikes in the QCM rate measurement upon shutter opening and closing was not observed as the Se/Te cell operating temperatures were low and their orifice positioned at a 35-cm distance from the QCM head. The tooling factors to extract effective deposition rates for the QCM measurements for both fluxes were determined by physical thickness measurements of Se and Te films. X-ray reflectivity was used to extract the thickness of Se capping layers grown at room temperature on Bi₂Se₃ films. For Te, the physical thickness of a Te film on a sapphire single crystal sample was obtained using secondary electron microscopy images of the sample's cross section.

After Se/Te deposition at room temperature [Fig. 1(e)], the QCM crystal was heated using a fixed constant output power. During the heating process, the mass loss of the precedently deposited chalcogenide film from the QCM crystal in terms of loss in film thickness and desorbing material rate was monitored as sketched in Fig. 1(f), while the chalcogenide shutter was kept closed (shutter state = 0). After each deposition and desorption measurement, the QCM crystal was cycled to its maximum temperature to recover a pristine QCM surface.

Various temperature-time curves of quartz crystals heated with constant internal heater power setting are shown in Figs. 2 and 3. The QCM heater output power level was chosen to be well above the threshold heater power for a vanishing sticking coefficient for Te and Se respectively, thus ensuring that eventually the entire film of Te and Se layer can be desorbed from the quartz crystal surface during the heating cycle while keeping the heating power constant.⁴³ The quartz crystal temperature was measured by averaging the IR camera signal over the quartz crystal area available for deposition. After completion of the heating cycle with the pristine quartz crystal, Se and Te films of different thicknesses ranging from 100 nm to 1200 nm were deposited at room temperature onto the pristine crystal and the temperature-time dependence of the now Se/Te-coated crystal was measured again. No thickness and/or history dependent differences in the heating behavior of the QCM crystal were found that could potentially lead to unintentional artifacts and misinterpretation of the subsequent desorption rate measurements. For the desorption experiments of Te films, a heating power level of 30% was used, resulting in a linear temperature increase for the first 30 min up until a quartz crystal temperature of about 150 °C after which the temperature rate was reduced and entered a plateau region after about 75 min (see Fig. 2). For desorption experiments of Se films in Fig. 3, a lower heating power of 20% was used. Here, a linear temperature rise for about half an hour was found up to a quartz crystal temperature of 70 °C, which was considerably reduced after that, but did not saturate even after 100 min. In both cases, the time dependent heating characteristics were high reproducible irrespective of the history and the amount of chalcogenide film deposited, demonstrated by the temperature taken after 100 min shown in the insets of Figs. 2 and 3. In the case of the higher heater power levels, the temperatures reached for different Te film thicknesses were in random order and differed by less than 6 K from the heating cycle of the pristine quartz crystal. Using a smaller output power of 20%, the quartz crystal temperature difference after 100 min of heating varied by less than 1 K and in a random fashion from the Se thickness and heating cycle number. This high reproducibility of the quartz crystal temperature-time characteristics that were found to be independent of history and film thickness deposited enabled simple temperature-dependent desorption experiments that could be performed sequentially and with little experimental effort.

Changes in temperature or thermal gradients endured by the quartz crystal affect its mechanical and elastic properties inducing shifts of the resonant frequency of the crystal and, therefore, manifesting as artifacts in the mass changes on the quartz crystal surface that need to be accounted for in the analysis of the temperature-dependent desorption behavior.^43,46,47 The temperature dependence of the QCM rate signal for the two cases of tellurium and selenium deposition are shown in Figs. 4 and 5, respectively. Figure 4(a) shows the temperature-dependent QCM rate signal with tellurium films of varying thickness. This rate was measured as a function of time like the quartz crystal surface temperature and is represented as the QCM rate as a function of temperature. The rate-temperature dependence for the pristine crystal with no tellurium film was found to exhibit a distinct behavior different from a constant line with 0 Å/s desorption rate due to the aforementioned thermal artifacts. At low temperatures up to about 60 °C, the rate showed a damped oscillation and then declined linearly with temperature up to about 160 °C where it then increased and finally reached the nominal 0 Å/s desorption rate at 200 °C. Similar low-temperature artifacts were recorded for all Te-coated crystals. At higher temperatures, the rate signal for Te films of different thicknesses diverted significantly from the pristine crystal curve, which is highlighted in the inset of Fig. 4(a). A minimum of the rate signal was smaller and occurred at higher temperatures with increasing Te film thickness, attributed to the desorption kinetics of the Te film additionally contributing to the rate signal.

The net desorption rate signal is the difference of the rate-temperature curve for a Te film with given thickness subtracted from the temperature induced rate signal of the pristine quartz crystal and is shown in Fig. 4(b). To eliminate the underlying temperature induced rate shift of the QCM in the nondesorption regime between 100 and 150 °C, a linear correction regression was applied in addition to all curves to level the data at a net desorption rate of zero at around 125 °C in Fig. 4(b). Aside from the temperature induced low artifacts below 60 °C, a noticeable Te desorption occurred around 160 °C. Highlighted by arrows in Fig. 4(b), the maximum Te film desorption rate (R_max) increased and reached a higher temperature (T_max) for thicker Te films until the upper temperature bound of the measurements was reached, where the thickest Te film of 1200 nm was fully desorbed. The Arrhenius plot of the desorption rate in the regime of interest from about 2.35 × 10⁻³ K⁻¹ (152 °C) to 2.10 × 10⁻³ K (203 °C) is shown in the inset of Fig. 4(b). The slope in the linear regime corresponds to $Ea/kB$ with the thermal activation energy for Te evaporation $Ea$⁠. Fitting results [light-yellow lines in the inset of Fig. 4(b)] for $Ea$ are summarized in Table I together with the temperature-dependent desorption characteristics for Te and ranged from 1.725 to 2.037 eV.

Despite the large spread of activation energies extracted from our experiments, values are corresponding to previously reported energy barriers for Te desorption. Specifically, activation energies of about 1.7, 2.0, and 1.8 eV were found for Te₂ and Te₅ evaporation from solid bulk Te and liquid Te.^48,49 Dissociation energies for Te are known to differ substantially depending on the predominant Te species and were additionally found to depend on the physical samples size used for evaporation. Thermal dissociation of bulk Te results in a vapor of short spiral Te chains of less than 10 atoms, with predominantly Te₂ molecules and 10% of Te₅ clusters among much smaller contributions of other Te species.^48–50 Other reports state that the contribution of Te₃ to Te₇ species constitutes less than 5% in the vapor besides Te₂.⁴⁰ As the physical sample size for Te evaporation is reduced to large clusters in the range of tens of atoms, Te₅ was found to dominate over Te₂ and other species in the vapor with activation potentials around 1.0 eV.⁴⁹ Smaller Te clusters dissolved in predominantly Te₅, Te₆, and Te₇ species, and an energy barrier of 3.4 eV was reported for the dissociation of a Te₂ molecule into atomic Te.⁴⁹ The experimentally determined values indicate that the predominantly desorbing Te species are Te₂ and Te₅.

We have extracted the x values—temperature of maximum desorption rate (T_max) in °C and y values—maximum desorption rate (R_max) in Å/s for each sample thickness and desorption extrema shown and highlighted by arrows in Fig. 4(b) and plotted the numbers (Table I) versus the initial deposited film thickness for Te in Fig. 4(c). Dotted lines are a guide to the eye only, not fits. The trend of T_max over Te film thickness in Fig. 4(c) follows an inverse exponential behavior, and the same can be stated about R_max. However, more data points and/or theoretical modeling are necessary to pinpoint the exact mathematical relationship between T_max and R_max and the film thickness for Te desorption.

Integrating the net desorption rate provides a sanity check for our measurements as it allows us to confirm that all the previously room temperature deposited Te was indeed completely desorbed. Figure 4(d) shows the normalized Te film thickness as a function of quartz crystal temperatures calculated from the desorption rate shown in Fig. 4(b). Starting at 165 °C, noticeable film thickness loss is recorded for the thinnest sample, which systematically shifted to higher temperatures up to above 190 °C for the thickest films. Regardless of the film thickness, desorption proceeded rapidly with increasing temperature until about 20%–10% of the initial film was left on the surface at 180 and 190 °C for the two thinnest films with 100 and 200 nm thickness, respectively. From the reduction of the desorption rate toward evaporating the complete Te film, it is concluded that either initial Te is more tightly bonded to the quartz crystal surface than to a Te surface or that different Te species desorb off the surface of tellurium and the quartz crystal with potentially more volatile Te species predominantly desorbing initially at lower temperatures, while others only desorb at progressively higher temperatures. For 800 nm and thicker films, the desorption loss over temperature is maximized and remains unchanged until the entire film is lost at 198 °C. The last 10% of thinner films desorb with significantly reduced rates over a larger temperature range of 10–20 K. Marked by circles in Fig. 4(d) are the normalized thickness values at T_max as derived from Fig. 4(b), which consistently shift to smaller portions of the initial film and higher temperature with increasing film thickness.

The measured total desorbed film thickness for Te—t_des—is plotted in Fig. 4(e) over the initially deposited film thickness—t_dep. This ratio consistently ranges above 1.0 and diverges no more than 20% of the starting film thickness. Our desorption measurements, thus, reflect the entire deposited amount of material.

The desorption characteristics of Se are presented in Fig. 5. Figure 5(a) shows the temperature-dependent QCM rate signal with selenium films of varying thickness using the same methodology as described for Te above. Again, initial temperature change induced artifacts are present in the rate measurements, albeit at a much-reduced scale compared to Te since a smaller heater power was used, resulting in smaller temperature changes. In contrast to Te, the Se desorption was found to evolve in a two-step process with a significant desorption peak around 65 °C—desorption regime 1—and the second desorption peak at about 30 °C higher temperature—desorption regime 2.

In the net desorption rate plotted in Fig. 5(b), both desorption peaks reached higher peak values and shifted to higher temperatures with increasing film thickness, which is highlighted by arrows for desorption regime 2. Remarkably, these trends were different for both peaks, which can be seen more clearly by following the behavior of T_max and R_max of both desorption regimes over film thickness in Fig. 5(c). T_max shows an inverse exponential behavior in both desorption regimes with saturation temperatures of 65 and 98 °C. While R_max for desorption regime 1 saturated to about 0.2 Å/s in the thick film limit [see Table II and data in red in Fig. 5(c)], R_max scaled with film thickness up to a max. value exceeding 0.5 Å/s in regime 2. It is also noted that while for thinner films R_max was higher in desorption regime 1, resulting in a higher mass loss at lower temperatures for film thicknesses up to 400 nm, R_max was found higher in desorption regime 2 for thicker films.

Arrhenius plots for the two desorption regimes between 3.2 × 10⁻³ K⁻¹ (39 °C) and 2.65 × 10⁻³ K⁻¹ (104 °C) are shown in the inset of Fig. 5(b). Contrary to Te evaporation, the slopes were found to be independent of the Se film thickness and both desorption regimes exhibited similar slopes. (Fitting was not possible for regime 2 in the 100-nm sample.) Fitting results of the Arrhenius plot for Se evaporation are shown as light-yellow lines in the inset of Fig. 5(b) and yielded activation energies of (1.3 ± 0.1) eV, and (1.3 ± 0.2) eV for desorption regimes 1 and 2, respectively. The details of the desorption characteristics for Se films including all extracted activation energies are summarized in Table II. The existence of two distinct desorption regimes implies that there are at least two physisorbed states of Se in the film with different desorption kinetics. Se was previously reported to evaporate predominantly as Se₆, followed by Se₅, as well as Se₇ and Se₉, but also other, smaller molecules like Se₂ to Se₄, taking up about 28% of the vapor.^34,40 Furthermore, it was found that with increasing temperature, Se progressively dissociated into smaller molecules, where Se₂ had a dissociation energy of about 3 eV.^34,51 Beyond this, it is unclear if and how those energy barriers vary among the different Se molecules. The thermal desorption energy barrier extracted here for Se is in good agreement with values found in the literature.^52,53 A similar energy barrier for both desorption regimes implied either that the Se species composition in the desorbing vapor remained unchanged in both regimes, or that desorption energy barriers for different Se species varied much less compared to Te. Assuming similar Se species in the desorbing vapor for regimes 1 and 2 is supported by the identical activation energies. In addition, if specific Se species would dominate desorption regime 1, which were supplied in identical composition to samples with different thicknesses, the total amount of desorbing Se molecules would have scaled with the Se film thickness, which is not observed. The saturation behavior of desorption regime 1, on the other hand, points rather to a physisorbed state in the film, which possesses a maximum number of occupied sites like a surface or interface layer.

The limited capacity of desorption regime 1 is demonstrated further in Figs. 5(d) and 5(e). The normalized film thickness in Fig. 5(d) shows a maximum decrease in regime 1 to 0.5, 0.65, 0.8, 0.9, and 0.93 for the 100-nm to 1200-nm-thick films. The same trend can be seen in the lower branch curve (filled squares) of total desorbed film thickness—t_des—over nominally deposited film thickness—t_dep—for regime 1 in Fig. 5(e). Converted into an absolute loss in film thickness, this amounts to 50, 70, 80, 80, and 80 nm in regime 1, i.e., it saturates to an equivalent of 80 nm as soon as desorption in regime 2 dominates over regime 1 (at a thickness of 400 nm or more). This means that Se preferably adsorbs into the state from which desorption at low temperatures is possible. The maximum number of occupation sites for this physisorbed state, which amounts to an equivalent thickness of 80 nm, are filled in films with a thickness of 400 nm or more. Similar to surface or interface effects, which dominate in thin films and are less pronounced with increased film thickness, desorption out of those physisorbed states dominates in regime 1 for thin films. However, the maximum capacity expected for a surface or interface layer would range in the single digit nm regime, i.e., much lower than the observed 80 nm. Unless our physical thickness measurement calibrations are plagued with error, the high capacity of the physisorbed state, thus, points to another mechanism that has yet to be revealed.

As a sanity check, the cumulative amount of desorbed material from desorption regimes 1 and 2 in Fig. 5(e) ranges above the total initially deposited amount and diverges from the latter less than 10%, confirming that also for Se our desorption measurements reflect the entire deposited Se material.

In summary, we presented a direct and quantitative determination of the chalcogen desorption process. The reported temperature ranges and rates at which full desorption is archived provide a framework for the commonly used procedure of Se/Te capping and decapping of air sensitive materials. It was found that desorption rates for Te had a single sharp peak with a maximum shifting from 0.2 to 0.8 Å/s and occurring at temperatures from 175 to 200 °C with increasing film thickness. An energy barrier for desorption of about 1.8 eV was observed. Residual amounts of Te were desorbed over a much larger temperature window and remained on the quartz crystal surface for longer times. In contrast to the single desorption peak of Te, Se desorbed in a two-stage process with a first desorption peak of around 65 °C, while remaining Se desorbed around 95 °C. For films thinner than 400 nm, a larger amount of Se was found to be desorbed at lower temperatures, with a larger rate at low temperatures compared to the high temperature desorption peaks, but not larger than 0.2 Å/s. For thicker films, the desorption rate of the low-temperature peak was barely exceeding, 0.2 Å/s, while the maximum rate of the higher temperature desorption peak increased to 0.5 Å/s. Both desorption regimes are characterized by identical energy barriers of about 1.3 eV independent of the Se film thickness. However, due to the observed saturation of Se desorption at low temperatures, which does not scale with the film thickness, we hypothesize the presence of a physisorbed state such as a specific site or bonding state in the Se film likely at the surface or interface, which causes Se desorption at low temperatures. These insights into the kinetics of chalcogen film growth point toward the need for a more detailed understanding of the chemical composition states of thermally generated Se and Te vapors from evaporation sources and/or directly at the growth front to further engineer chalcogenide-based film growth.

The instrument and experimental work were funded by The Pennsylvania State University Two-Dimensional Crystal Consortium—Materials Innovation Platform (2DCC-MIP). 2DCC-MIP is supported by NSF Cooperative Agreement Nos. DMR-1539916 and DMR-2039351. Funds for data compiling and analysis work came from the center for 3D Ferroelectric Microelectronics (3DFeM), an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences Energy Frontier Research Centers program under Award No. DE-SC0021118.

The authors have no conflicts to disclose.

Derrick S. H. Liu: Data curation (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – review & editing (equal). Maria Hilse: Conceptualization (equal); Formal analysis (lead); Methodology (equal); Supervision (equal); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Roman Engel-Herbert: Conceptualization (equal); Formal analysis (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.