The role of contaminations in ion beam spectroscopy with freestanding 2D materials: A study on thermal treatment

In the past decade, two-dimensional (2D) materials evolved to a major field of research in physics, chemistry, and materials science. Since the discovery of stable monolayer graphene in 2004,¹ the family of 2D materials has grown, covering an interesting range of optical, mechanical, and electronic properties. The prime candidates are semi-metal graphene, semi-conducting transition metal dichalcogenides (TMDs), and insulating hexagonal boron nitride.² This variety makes 2D materials promising candidates for future applications, e.g., in electronics,^3–5 optoelectronics,⁶ or molecular sieving.^7,8

However, often, the properties of 2D materials are significantly worse in real applications than they should be based on their intrinsic material characteristics. The latter are sometimes even hard to determine in a controlled laboratory environment.^9,10 The main reason is that any surface and, especially, surface-only 2D materials are densely covered with contaminations made of mostly hydrocarbons, which are adsorbed as adventitious carbon from ambient air exposure during sample handling,¹¹ and possibly residuals from transfer and/or production processes.^12–14 Adsorbates may change material properties, e.g., due to electronic doping or by unintentional chemical functionalization^1,15,16 and can also facilitate particular processes such as self-healing of pores observed for graphene.¹⁷ Thus, working with 2D materials always requires dealing with adsorbates, which makes appropriate cleaning techniques inevitable.¹⁸ Leaving the sample itself unaffected while removing contaminants is a challenging task: frequently used surface cleaning methods such as ion sputtering would severely damage the single monolayer of interest.

Thermal annealing is therefore the most common way to achieve cleaning of atomically thin materials. There are various approaches to thermal treatment both in air and in vacuo, e.g., resistive heating via a heating wire or radiative heating. Recently, a laser annealing method was introduced to locally increase temperature in freestanding 2D materials.¹⁹ In general, this results in cleaned areas up to several 100 nm² but still leaves contaminated spots. Cleanliness of a sample is usually determined via (scanning) transmission electron microscopy [(S)TEM],^19,20 atomic force microscopy,²¹ or Raman spectroscopy^20–22 and describes the ratio of clean vs contaminated areas. However, it is crucial to consider analysis techniques used to determine cleanliness and material properties of a sample. A clean state may mean different things in different methods. The key is to find a way to discriminate results obtained from clean areas from those obtained from contaminated ones. In atomically resolved microscopy, e.g., (S)TEM, clean areas can be directly identified, and an experiment can subsequently be performed on these spots. In Raman spectroscopy,^23–25 for example, where typically sample areas of several 10 μm² to mm² are probed, this is not as simple.

We study the transmission of (highly charged) ions through freestanding 2D materials. Here, we present a technique to evaluate cleanliness of single layer graphene after thermal annealing based on data filtering according to ion time of flight (TOF) or energy loss, respectively. Therefore, the energy loss of charged particles is considered as a criterion for the cleanliness of a sample. Our studies of exit charge states and TOFs for transmitted ions are accompanied by a STEM and scanning electron microscopy/energy dispersive x-ray spectroscopy (SEM–EDX) analysis. Our results show that large clean areas are developing, surrounded by heavily contaminated parts. We discuss what cleanliness means in the case of ion beam spectroscopy.

We perform ion spectroscopy with the ion beam spectrometer at TU Wien to study the interaction of highly charged ions (HCIs) with two-dimensional materials. A commercially available Dresden EBIS-A electron beam ion source²⁶ is employed to produce Xe ions with charge states of Xe¹⁺ to Xe⁴⁴⁺ with kinetic energies of 1 keV–400 keV. Selection of charge states is performed by means of a Wien filter.²⁷ Our ion beam diameter is 0.5 mm–1 mm. Further details on the facility are given in a recent publication.²⁸ 

Presented results were gained using freestanding single layer graphene commercially available from Graphenea.²⁹ Samples are grown by chemical vapor deposition and transferred onto TEM grids with the help of a spin-coated sacrificial poly(methyl methacrylate) (PMMA) layer, whose residues make up the major part of sample contaminations. An additional 10 nm–20 nm thick Quantifoil carbon support structure containing holes with a diameter of 2 μm holds the suspended single layer of graphene.

Similar to the methods discussed in the studies of Tripathi et al.,¹⁹ two techniques are used to prepare cleaned samples: a 1Ncl15 heating wire integrated into the target holder and a Lasertack³⁰ laser diode (6 W, 445 nm) focused to cover the target position (∼10 mm²) serve as tools to clean the samples in situ. Temperatures up to 400 °C can be achieved on the target holder by Ohmic heating. The temperature induced by the laser diode cannot directly be measured, but it stays below the melting/sublimation temperature of the grid even under continuous irradiation. STEM was measured in a Nion UltraSTEM 100 (60 kV electron acceleration voltage) acquiring high angle annular dark field (HAADF) data (80 mrad–200 mrad). SEM employed a FEI Quanta 250 FEG SEM with a EDAX SDD Octane Elite 55 EDX system.

Our experiments in transmission geometry yield information on energy loss and exit charge states of the projectiles after transmission through 2D materials and ion-induced electron emission from the materials.^31–33 The latter serves as a start trigger for TOF measurements, which in turn allows the analysis of the projectiles’ kinetic energy loss. The exit charge state q_out is assessed via deflection of transmitted particles in a pair of deflection plates, which is detected angle-resolved on a position sensitive RoentDek delay line microchannel plate (MCP) detector [Fig. 1(b)]. The MCP detector also provides the stop signal for the projectiles’ time of flight. All measurements are recorded in coincidence in a list-mode, which allows restriction of parameters for analysis in retrospect.³⁴ 

A TOF spectrum of 170 keV Xe³⁸⁺ transmitted through a single layer of graphene is given in Fig. 1(a). The sample was cleaned using Ohmic heating. Details on the effect of the cleaning method are given in Sec. III. We find two distributions: ions transmitted through graphene with shorter (red) and ions transmitted through support material with larger TOFs (blue). Coincident detection of both, TOF and exit charge state q_out of the projectile, makes it possible to analyze these two TOF regimes separately: exit charge state 2D maps taking into account only ions transmitted through graphene or support material, respectively, are displayed in Fig. 1(b). A projection of both spectra on the exit charge state axis, integrating the information over scattering angles, is given in Fig. 1(c), allowing a direct comparison. Note that the width of each distribution is mostly determined by statistics of charge exchange and only to a minor extent from contaminations.

Ohmic sample heating was also applied at HZDR (Dresden, Germany) using a standard Prevac PTS 1000 IR RES/C-K sample holder to establish clean sample conditions for (highly charged) ion spectroscopy using an electrostatic analyzer (ESTAT) to study charge exchange.³⁵ The impact of these cleaning techniques on our studies is discussed in the following sections.

We heated a single layer graphene sample stepwise to 360 °C over the course of one day using the 1Ncl15 heating wire. Note that due to low count rates for highly charged ions, the acquisition of one spectrum takes about 2 h. Transmission measurements with 146 keV Xe³⁰⁺ were done after each step. The target temperature dependent behavior of the TOF is given as well as the number of captured electrons of the projectile in Fig. 2(a). A direct comparison of the exit charge state distribution before (T = 25 °C, orange) and after (T = 360 °C, red) the cleaning procedure is shown in Fig. 2(b). The mean value shifts by 12, and the low charge state tail vanished in the annealed case. Note that there is some count rate at charge states <10 (and somewhat more at q_out = 0), which may be attributed to ions transmitted through still contaminated areas. We detect ions with their incident charge state even after transmission through single layer graphene. This is due to transmission through larger defective areas in the target, which cannot be fully avoided during growth and transfer. Because of scattering at the material boundaries, there might also occur a second charge state distribution with a high exit charge state (as seen here at Xe²⁸⁺, red).

For ions transmitted through thin materials, the measured TOF is linked to energy loss: less energy loss corresponds to shorter time of flight of a projectile. While heating a single layer graphene sample, a decrease in TOFs is observed, indicating a decrease in material thickness and thus a removal of surface contamination [Fig. 2(a)]. Similar behavior is observed for exit charge states, where again less captured electrons indicate cleaner samples. Slight changes are already observed between room temperature and ∼100 °C. The cleanest state is, however, only achieved between 250 °C and 360 °C. An exact determination of the critical temperature for cleaning of graphene is not possible since for T > 250 °C, simultaneous heating and evaluation of TOF is not applicable. In this region, slight pressure increase and background signals due to emitted light from the heating wires produce noise in our electron detector.³² In order to study the TOF after annealing at higher temperatures (>250 °C), the heater is turned off and a measurement is performed. It should be noted that the sample temperature may drop during data acquisition in this case. Laser-assisted cleaning yields comparable results, which are achieved within a few seconds only. Because of low count rates for highly charged ions, no time-dependence of the cleaning process can thus be measured.

After ending thermal treatment, the TOF and charge state spectra do not change for the first 5 h–6 h until the sample is sufficiently cooled down (<100 °C) for contaminants to be adsorbed again. Once temperatures of <100 °C are reached, but while keeping the samples at a pressure of ∼5 × 10⁻⁹ mbar, re-contamination occurs within 1 h–2 h. This leads again to lower exit charge states by ∼8 and longer TOFs by ∼10 ns for 146 keV Xe³⁰⁺ [Fig. 2(a)].

Re-cleaning the sample afterward following the procedure described above shows a different behavior: the cleanest state (marked as red area) is now achieved already at temperatures <200 °C [(Fig. 2(a), violet circles].

Figure 3 shows TOF spectra of 170 keV Xe³⁸⁺ transmitted through a single layer of graphene. Measurements before and after cleaning of a previously virgin sample are pictured in orange and red/blue, respectively. Here, cleaning was performed using the laser diode for a few seconds and additionally heating the target to 180 °C to prevent re-contamination during measurements. For the contaminated sample before cleaning, we observe a very broad TOF distribution (note counts at >100 ns) and a large angular spread of the transmitted ions. This is due to multiple scattering, and consequently, we also observe a diminishing overall count rate on our detector in forward direction. In addition, projectiles transmitted through Quantifoil are mostly neutral. This may also contribute to the diminishing count rate as there is a lower detection efficiency of the MCP for neutral particles.^36,37 After cleaning, the TOF distribution gets narrower and the angular spread gets smaller, leading to better statistics on our detector. The TOF spectrum separates into two distributions: cleaned single layer graphene and support material (Quantifoil).

SEM and STEM images of a cleaned single layer graphene sample are presented in Fig. 4. Note that this sample has been transported in ambient air between the HCI/annealing setup and the SEM/STEM. The images of the sample therefore reflect not only the cleaned state (PMMA and adsorbed hydrocarbon removal during in situ UHV annealing as probed by the HCI experiments) but also the effect of a subsequent additional exposure to ambient air and therefore additional post-anneal hydrocarbon adsorption.¹¹ In the SEM image in Fig. 4(a), we show one hole in the Quantifoil support. The sample is found to be homogeneously contaminated on the whole suspended membrane. We use STEM in HAADF mode to estimate the spatial contamination level distribution in equivalent numbers of graphene layers based on the roughly linear intensity scaling of the HAADF STEM with specimen thickness and assuming as an upper limit the same density for the hydrocarbon adsorbates as for graphene.^38,39 In Fig. 4(b), a STEM image is intensity color coded to show the thickness of the imaged material in equivalent number of graphene layers (cf. top right of the image). We find four typical regions labeled A–D in Fig. 4(c): A identifies atomically clean single layer graphene membrane (i.e., no contamination at all) and B and C mark an increase in contamination levels with 4–5 (B) and ∼10 (C) equivalent graphene layer thickness attached to the monolayer graphene membrane. The particle-like deposits at D identify heavy/thick contamination of between 20 and 70 (D) equivalent graphene monolayers, which is roughly as thick in equivalent graphene layer thickness as the Quantifoil support layer.

We additionally performed a SEM–EDX analysis to check the elemental composition of the thick contaminants for possible heavy impurities from the TEM grid material. The contaminations appear in SEM–EDX as (hydro-)carbon, consistent with naturally occurring adventitious hydrocarbon adsorbates from atmospheric exposure and even under ultra-high vacuum conditions and PMMA residues. This fact further justifies the calibration of the STEM HAADF signal to graphene layer equivalents in Fig. 4.

In prior work of Gruber et al.,³⁸ we have done similar HAADF STEM contamination analysis of graphene monolayer membranes that had been transferred without PMMA but not additionally cleaned. For these, we found adventitious hydrocarbon adsorption from ambient air exposure to a maximum equivalent graphene layer thickness of ∼6. Based on this prior analysis and our current sample preparation history including PMMA transfer, we identify regions C and D with their significantly higher equivalent graphene layer thicknesses (C: ∼10 and D: 20–70) in Fig. 4(a) as redistributed and de-wetted residual PMMA from UHV annealing (i.e., also present during HCI experiments),^19,20 while region B (4–5) is attributed to readsorbed adventitious hydrocarbon from air exposure during transport to STEM (i.e., after HCI experiments).^11,38 Thus, regions A + B are suggested to have been atomically clean during HCI experiments (∼40% of sample area), while C + D remained contaminated also during HCI experiments (heavily contaminated region D ∼25% of sample area).

Our thermal treatment does not remove all contaminants from a sample, but rather yields patches of clean parts well distinguishable from even heavier contaminated parts. This sample condition will further on be denoted as ‘cleaned.’. The following discussion will justify this denotation.

Contaminants of virgin samples consist mainly of residuals from production and transfer processes (e.g., PMMA) and adsorbed adventitious hydrocarbons from ambient air exposure.^12,13,20 Presumably, the slight changes in the beginning of the first annealing process reflect desorption of the latter kind and evaporation of water molecules. The first kind, PMMA, is cleaned only at higher temperatures (consistent with Ref. 20), explaining the sudden change in both spectra between 250 °C and 360 °C. This also concurs well with Refs. 40 and 41, where critical temperatures of ∼300 °C and ∼400 °C were reported, respectively, to remove residuals from single layer graphene. Not surprisingly, as only hydrocarbons are re-adsorbed again afterward, the clean state is then already reached at a lower temperature of ∼200 °C (Fig. 2). Keeping the sample at this temperature at a pressure of ∼5 × 10⁻⁹ mbar conserves cleaned sample conditions as it is crucial for long-term measurements.

We showed that virgin single layer graphene samples from a PMMA based transfer method behave as a homogeneous (contaminated) sample: only one broad TOF distribution is seen in Fig. 3. After cleaning, shorter TOFs and less captured electrons of transmitted HCIs are measured. At the same time, a distribution with longer TOFs and lower exit charge states also appears (Fig. 1), i.e., the distribution splits into two. This indicates the existence of two distinct thickness regimes on the sample after cleaning. The homogeneous surface coverage decomposes into uncovered and strand-like thick parts.

As observed in STEM, thermal treatment of 2D samples leads to clustering of existing contamination leaving clean (∼40%) and heavily contaminated (∼25%) areas (Fig. 4). TEM images of pristine Graphenea samples are presented in Ref. 19 and show homogeneous coverage conditions. Our findings are in line with the studies of Lin et al.²⁰ who also revealed formation of strip-like patterns and wrapping nanoparticles from PMMA-leftovers after cleansing. Tripathi et al.¹⁹ also present TEM images of graphene annealed at 600 ^○C with clean areas of ∼10 nm² in-between contaminated regions, even though in this case samples were transferred to TEM in vacuo. The heavily contaminated spots (D) in Fig. 4(a) have thicknesses comparable to the 10 nm–20 nm carbon support, which also results in similar energy loss and TOFs of transmitted HCIs. We cannot, however, distinguish between Quantifoil and heavily contaminated regions in our measurements because due to their similar nature, ions transmitted through either part of the sample match both in charge exchange and TOF. We suggest that we see both influences, Quantifoil and contaminations, in the TOF distribution with higher TOFs. The effect of the support on the TOF and charge exchange was checked in a separate control experiment without graphene and is in fair agreement with the data from the contaminated areas here.

A comparison of data achieved using cleaned samples to a recent study on this field³⁸ is presented in Fig. 5. The graph shows the number of captured electrons—the difference of incident and exit charge state—of Xe projectiles for three different incident charge states and various ion energies. Data of cleaned samples measured in Vienna using a MCP detector and in Dresden using an ESTAT, respectively, show good agreement, which confirms our findings. In contrast to our Graphenea samples we used for the present work, the samples studied by Gruber et al. were transferred onto a TEM grid without the use of PMMA and did not experience any cleaning prior to measuring. This translated to a clear difference in observed contamination levels in STEM measurements, as outlined above. Thus, a difference in the charge exchange behavior does not come unexpected and accounts for only ∼35% in relation to our new results of clean samples here. In case of virgin samples based on PMMA transfer processes, we find a much larger difference of ∼85% [Fig. 2(b)]. However, following the analysis of Gruber et al., we still find neutralization times for highly charged ions transmitted through graphene in the order of ∼10 fs. This leaves the conclusions of previous studies unaffected.^38,42

It should also be noted that the decrease in the number of captured electrons for clean graphene in comparison to pristine samples is accompanied by an increase in the yield of emitted electrons. This observation is currently a subject of further investigations.

A study on the effect of thermal treatment of 2D materials in ion beam spectroscopy was performed. We found that, in particular, PMMA-related adsorbates tend to form clustered, heavily contaminated spots besides clean areas on freestanding single layer graphene due to thermal treatment. Both Ohmic heating and laser irradiation yield good results, whereas the latter is faster. Keeping the sample at an elevated temperature under UHV conditions preserves the clean state throughout extended measurement times. The energy loss of transmitted highly charged ions varies significantly between either sample conditions. Coincident detection of exit charge state and time of flight of projectiles allows us to exclude ions with long time of flights in analysis. Hence, we devised and presented a procedure to study clean single layer graphene probed with highly charged ions.

Thermal treatment will, most likely, always result in a decomposition of clean and contaminated areas, and as a consequence, any measurement technique applied must be able to discriminate both surface parts from one another. We use TOF and charge exchange determination to do so, and in STEM, (atomic) resolution imaging can do the same.¹⁶ Special care must be taken with other methods where this is not straightforward.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

This research was made possible by the Austrian Science Fund (FWF) (Grant No. Y 1174-N36) and the Deutsche Forschungsgemeinschaft (DFG) (Grant Nos. WI 4691/1-1 and 322051344). A.N., B.C.B., and R.A.W. gratefully acknowledge funding from TU Wien’s competitive Innovative Projects program. J.S. thanks the Doctoral College TU-D for financial support. We acknowledge use of the facilities at the University Service Centre for Transmission Electron Microscopy (USTEM), TU Wien, for parts of this work.