Introduction

The development of high entropy materials originates as metallic alloys. For centuries, during the Copper, Bronze, and Iron DOI: 10.1002/adma.202402442 Ages, alloy systems have been predominantly focused on introducing additive elements in a matrix composed of a single principal element. In the modern age, the development of binary alloy systems only began in the late 20th century. High entropy alloy (HEA) research has been exponentially gaining attention since 2004 [1,2] due to their potential for improved mechanical properties, such as elevated temperature mechanical strength [3,4] wear resistance, [5] creep resistance, [6] and ductility. [7] Novel functionalities, such as magnetism, electronic structure tuning, thermoelectric properties, superconductivity, hydrogen storage, catalytic performance, and oxidation resistance, have also been studied since the first interest in HEAs. [8][9][10][11][12][13][14] Growth in the fundamental knowledge about HEAs has led to the concept of stabilizing multiple principal elements into a single phase material through the use of high entropy for oxides, [15] carbides, [16,17] borides, [18] nitrides, [19] sulfides, [20] fluorides, [21] and aluminosilicides. [22] These high entropy materials exhibit high Li-ion conductivities, [23] high Li storage capacity, [24] narrow and tailored band gaps, [25] colossal dielectric constants, [26] complex magnetic phases, [27] and high catalytic activities. [28,29] Despite the rapid development of high entropy materials, one experimental challenge remains elusive-atomic scale characterization to identify elements. Such high-specificity measurements will provide invaluable insight into the correlations among the local elemental distribution, the resulting microstructures, and the properties and functionalities of high entropy materials.

Element sensitive measurements rely on elemental fingerprints, such as the core level electron binding energies of elements. Therefore, the elemental micrographs for HEAs are typically recorded using tools such as scanning electron microscopy (SEM) with energy dispersive X-ray spectroscopy (EDX), [30] or scanning transmission electron microscope (STEM) with corresponding EDX mapping. [31] However, SEM does not have the resolution down to the atomic scale. Though STEM can provide images at the atomic scale, it measures columns of atoms that involve all elements in an HEA, limiting the ability of revealing local elemental distributions. On the other hand, scanning tunneling microscopy (STM) offers high spatial resolution of the surface layer, down to the atomic scale. [32,33] However, STM detects the electronic density of states (DOS) near the Fermi energy, which does not contain elemental fingerprints. Only in limited examples, elemental STM images have been reported with the following possible mechanisms: 1) differences in the work function, such as Cu atoms on Mo(110); [34] 2) tunneling through adsorbates on the STM tip for a Pt 25 Ni 75 (111) surface; [35] 3) differences in the surface states on the Fe alloying with Cr(100) top surfaces; [36] and 4) differences in the DOS near the Fermi energy, such as Cu on W(110), [37] and PdAg(100) and (111) [38] surfaces. It is also worth noting that recently, the subtle variations in the differential conductance (dI/dV) spectra near the Fermi energy in mixed binary-element materials have been distinguished with the assistance of a machine learning model. [39] Nevertheless, STM is not typically considered as an element sensitive instrument.

Synchrotron X-ray STM (SX-STM) can enable elementsensitive imaging at the nanometer scale. Previous work has demonstrated a 2-nm spatial resolution in resolving Ni islands grown on a Cu(111) substrate. [40] Likewise, far field (≈5 nm tip-sample distance) X-ray absorption spectroscopy (XAS) of a Eu 3+ ion in a [Eu(pcam) 3 X] 2+ complex, corroborated the potential of high resolution XAS measurements. [41] Recently, single atom chemical fingerprinting has been demonstrated with near field (STM tip in tunneling regime) XAS for Fe and Tb in supramolecular assemblies. [42] The molecular complexes in these previous reports can isolate target ions and have been major steps in advancing the capabilities of SX-STM. However, it has yet to be demonstrated that this capability can be applied to distinguish single atoms within a lattice, where atoms cannot be spatially separated. The random distribution of elements in an HEA lattice provides an ideal platform to examine and quantitatively determine these experimental limits. In this work, through the use of SX-STM, the elemental mapping at the subnm scale is demonstrated with a nearly equimolar five-element CrMnFeCoNi HEA. The results indicate that SX-STM can comprehensively reveal elemental mappings at the sub-nm scale on the surfaces of high entropy and compositionally complex materials.

Experimental Section

A nearly equimolar CrMnFeCoNi alloy was synthesized using arc melting under 1 atm of inert Ar gas, using equimolar ratios of high purity powders (>99.95%) of each constituent element. To ensure compositional homogeneity, each sample was overturned and re-melted five times at ≈3500 K. The sample stayed on a water-cooled copper plate during melting and solidification to cool it to room temperature. The HEA was cut into a 1 mm thick slice, where one side was polished with 1200 grit sandpaper and 0.5 to 0.05 μm DIAMAT polycrystalline diamond paste, in preparation for STM measurements. Prior to the SX-STM measurements, the sample was cleaned in an ultra-high vacuum environment with a base pressure of 5 × 10 -10 torr through multiple cycles of hot Ar + ion sputtering (460 C, 3 kV, 7 mA, for 20 min) and annealing (500 C for 30 min), followed by slow cooling at a rate of ≈2.5 °C min -1 down to 300 °C and then cooled naturally afterward. The surface quality was checked with STM, using a Pt/Ir tip, and the resultant topography showed high density terrace steps, with step heights varying from 0.5 to 2 nm across the region measured (Figure 1f).

Characterization

The bulk elemental composition of the HEA was determined through SEM-EDX to be Cr 20 Mn 26 Fe 16 Co 18 Ni 20 . The configurational entropy of this sample was calculated to be 1.596R ≈ 13.269 J K -1 mol -1 with the equation

Dendritic structures are observed in the elemental images with EDX, as shown in Figure 1a-e. These images show that Cr, Co, and Fe aggregate to form dendritic arms in a branch-like structure, while Mn and Ni aggregate to form interdendritic regions. The formation of the dendritic microstructure is due to the uncontrolled cooling rate of arc melting. [43,44] Despite the formation of the dendritic structure, the elements are not fully segregated. Mn and Ni may still be found in the CrCoFe-rich region with lower relative concentrations. The crystal structure was determined by X-ray diffraction (XRD), as shown in Figure 1g, to be a predominantly face-centered cubic (FCC) structure, which is the expected phase of an equiatomic CrMnFeCoNi alloy. [2] The degree of crystallinity was calculated using the XRD data in Figure 1g with the equation

SX-STM

The SX-STM measurements were carried out at the Advanced Photon Source's XTIP beamline (4-ID-E). Throughout the experiment, the sample and tip are kept at a 90 K and 8 × 10 -10 torr environment. A monochromatic X-ray beam is sent through an optical chopper operating at 650 Hz and is aligned with the STM tipsample junction under grazing incidence geometry as shown in Figure 2a. A specially fabricated STM tip [45,46] collects a combination of conventional STM tunneling current (I STM tunnel in Figure 2), X-ray-assisted tunneling current (I X-ray tunnel in Figure 2), and X-ray induced photo-ejected electrons (I sample pass in Figure 2). [47][48][49] It has been determined that the X-ray-assisted tunneling dominates the tip current channel, hence enhancing the ability to collect elemental information with high spatial resolution. [42] During the experiment, the tip-sample junction (≈1 nm separation) is illuminated by an X-ray beam of ≈10 × 10 μm 2 . Photo-ejected elec-trons from both the tip and samples are expected. However, due to the nm scale tip movement, the photo-ejected electrons from the tip form a constant background and are subtracted during the data analysis.

For SX-STM measurements, a specialized coaxial SX-STM tip is employed, consisting of a chemically etched W tip which is coated with an insulated layer (250 nm thick SiO 2 ) with only The I tip and I sample currents in the SX-STM system are collected and processed independently through lock-in amplifiers, which reference the chopper frequency to separate conventional (green) and X-ray excited currents (purple and red). A lock-in amplifier separates sample current into X-ray excited photo-ejected current with the X-ray unblocked (red). Another lock-in amplifier separates tip current into conventional (green) and X-ray-assisted (purple) tunneling currents. The conventional tunneling current (green) is used for the feedback loop to control STM scanning. A further description of the signal processing can be found in ref. [40]. The inset depicts the tunneling junction with photo-ejected electrons that are far from the tip apex and not collected by the tip (red), nearby photo-ejected electrons that are collected by the SX-STM tip (blue), and a combination of X-ray-assisted and conventional STM tunneling 20-100 nm of the tip apex exposed. This design enhances the spatial resolution by limiting the detection of the photo-ejected electrons from the sample only to the tip apex. Furthermore, the currents measured by the tip are dominated by the X-rayassisted tunneling current, which is spatially localized to the subnanometer and atomic scales. [40,50,51] As illustrated in Figure 2b, the tip channel contains three routes of electron paths: 1) conventional tunneling (no elemental sensitivity, surface sensitive to the topmost atomic layer with atomic scale lateral resolution); 2) X-ray photo-ejected electrons from samples collected by the exposed part of the SX-STM tip (elementally sensitive, probing depth around 1 nm based on the electron mean free path of the escaping electrons, and 20-100 nm lateral resolution of the exposed part of the SX-STM tip); and 3) X-ray-assisted tunneling (elementally sensitive, surface sensitive to the top most atomic layer and atomic scale lateral resolution due to the tunneling mechanism). The tip channel data is dominated by the third mechanism as will be discussed later and demonstrated by ref. [42], hence the observed atomic scale elementally sensitive images. When the sample and tip currents are monitored separately while sweeping the X-ray energy, the global XAS signals (similar to conventional total electron yield (TEY) measurements) can be obtained from the sample current channel, which probes a deeper (≈1 nm deep) and larger surface area (larger than 10 × 10 μm 2 X-ray beam spot size under the grazing incident geometry). The spatially resolved local XAS signals can be detected from the tip channel, which is dominated by X-ray-assisted tunneling, and probes mainly the surface layer with element sensitivity and atomic scale lateral resolution due to the tunneling mechanism.

The chopper frequency in the SX-STM controls the rate of the X-ray beam to be turned "on" and "off" at the tip-sample junction. A description of the setup details can be found in ref. [52].

The tip-sample junction distance is maintained through a feedback loop using the tip currents measured in the "off" periods. In other words, the feedback loop is not impacted by the measured current during the "on" periods. During the "on" period, the tip current contains conventional STM tunneling current, the photoejected electrons, and the X-ray-assisted tunneling current. The local XAS information was collected through the tip channel during the "on" periods using a lock-in amplifier referenced with the chopper frequency. It is expected that the global XAS signal from the sample channel remains the same while the tip scans at a nm scale. This provides a good reference of the elemental composition of the surface under the X-ray-illuminated region, which is discussed in the following section.

Results and Discussion

For the given elements of interest (Cr, Mn, Fe, Co, and Ni), the Labsorption edge for all five elements can be covered by the X-ray energy range from 500 eV to 900 eV. The sample channel XAS spectra for all five elements are shown in Figure 3a-e. All five principal elements are detected, as expected. On the other hand, for a particular tip location, the tip channel XAS spectra of the five elements are shown in Figure 3f-j. Cr is not detected at this tip location as the L 3 peak is small comparable to the noise level, while other elements are still present. Note that due to the local measurements by the tip, the XAS signal of the tip channel is smaller than that from the sample channel, on the order of 1/10 the signal. The photo-ejected electrons collected by the exposed STM tip apex are expected to be the tip channel signal. Since the sample channel signal is expected to be unchanged when the tip is scanning in nm distance, the intensity of the XAS from the sample channel can be used to normalize the tip channel signal to reveal the relative intensity of each element across different tip locations. This is rather important since the intensity of the synchrotron X-ray may fluctuate between different measurements. The tip channel amplitudes (A tip ) and the sample channel amplitudes (A sample ) are extracted via Gaussian function fittings from the characteristic L 3 peaks for each element. Then, the tip channel amplitudes (A tip ), are normalized with the sample channel amplitude (A sample ), namely

, at each tip location. Propagated error values of normalized tip channel amplitudes are calculated and reported in Table S1, Supporting Information. In a 3 × 5 grid of tip locations, the relative elemental XAS signals are shown in Figure 4a-e. For easy analysis, they are further normalized into values between 0 and 1, where 1 represents the maximum value of ( The local XAS of the five elements measured in the 3 × 5 grid of tip locations with neighboring pixel separation of 5 Å can be analyzed to extract the elemental distribution with sub-nm precision. The XAS of the L 3 peaks in the tip channel measured in the 3 × 5 grid are shown in Figure S6, Supporting Information and summarized in Figure 4a-e. These pixel plots resemble the elemental distribution mappings in a similar way as the SEM-EDX images in Figure 1a-e, but at a significantly smaller scale, down to the sub-nm scale. To reveal the elemental distribution near the scanning area, a model with a 2D square lattice of the surface is simulated, where each atomic site is assigned with one of the five principal elements. In the model, the lattice constant of the FCC structure is set to be 3.586 Å, the value obtained from the XRD measurements, making the nearest neighbor interatomic distance d = 2.536 Å on the surface structure. The local XAS signal is simulated by the convolution of a spatial resolution function (Gaussian function), f (x, y

), where 𝜎 represents the spatial resolution, as illustrated by the Gaussian function in Figure 4k; x i and y i are the positions of the i th atom of a particular element. N represents the number of atoms of that particular element. The value f(x, y) is calculated for each element and normalized to 1 to be compared with the experimental data shown in Figure 4a-e. Three major factors impact the simulated XAS mappings: 1) spatial resolution, 2) tip position inside of the surface square lattice unit cell, and 3) element distributions across the surface of the scanning area. The optimal simulated XAS mappings are shown in Figure 4f-j to R up = 0.734) value as can be seen in Figure S2, Supporting Information. The discrepancy in the simulated mappings is also obvious as shown in Figure S3, Supporting Information. Finally, the element distribution is the most impactful factor in simulated mapping. Finally, by changing a single Fe atom into Cr atom, as shown in Figure S4, Supporting Information the correlation coefficients of both Fe and Cr drop ≈10% compared to the optimized one (Figure 4l). This shows that the local XAS mapping is extremely sensitive to the elemental distributions in sub-nm scale.

It is important to note that the model presented here does not include contributions from the second-layer atoms. The contributions from the second layer atoms are neglected with the rationale that the X-ray-assisted tunneling limits the probing depth by the tunneling effect (to the topmost atomic layer). Even though the mean free path of the photo-ejected electrons (with the kinetic energy of few hundreds of eV) may be on the order of two-three atomic layers (≈1 nm), the dominating signal in the tip channel is the X-ray-assisted tunneling. Thus, the small deviation between the simulated and experimental XAS mappings might originate from the contributions from the second atomic layer. It is possible to include the second layer elemental distribution into the model, however, the lower contribution from the second layer atoms makes it more arbitrary to build the elemental distributions and is thus ignored here.

In summary, the sub-nm scale detailed elemental distributions of an HEA surface are extracted utilizing the SX-STM technique. The X-ray-assisted tunneling makes the local XAS mappings feasible to reveal complex elemental configurations on the surface. This method can be potentially applied to other high entropy ma-terials or less entropic materials that require distinguished elements with high spatial resolution. The results shown here enables future research in correlating sub-nm scale elemental distributions to the physical properties of high entropy materials, such as the cocktail effects, lattice distortions, compositional freedom, etc. Combining the analysis of this work with existing XRD, EXAFS, and pair distribution functions, may provide greater insights in understanding the HEAs. Also, combining the results provided here with the molecular adsorption sites with atomic scale STM images may shed light on the understanding of catalytic selectivity in various catalytic processes.