Crystallize It before It Diffuses: Kinetic Stabilization of Thin-Film Phosphorus-Rich Semiconductor CuP2.

Numerous phosphorus-rich metal phosphides containing both P-P bonds and metal-P bonds are known from the solid-state chemistry literature. A method to grow these materials in thin-film form would be desirable, as thin films are required in many applications and they are an ideal platform for high-throughput studies. In addition, the high density and smooth surfaces achievable in thin films are a significant advantage for characterization of transport and optical properties. Despite these benefits, there is hardly any published work on even the simplest binary phosphorus-rich phosphide films. Here, we demonstrate growth of single-phase CuP2 films by a two-step process involving reactive sputtering of amorphous CuP2+x and rapid annealing in an inert atmosphere. At the crystallization temperature, CuP2 is thermodynamically unstable with respect to Cu3P and P4. However, CuP2 can be stabilized if the amorphous precursors are mixed on the atomic scale and are sufficiently close to the desired composition (neither too P poor nor too P rich). Fast formation of polycrystalline CuP2, combined with a short annealing time, makes it possible to bypass the diffusion processes responsible for decomposition. We find that thin-film CuP2 is a 1.5 eV band gap semiconductor with interesting properties, such as a high optical absorption coefficient (above 105 cm-1), low thermal conductivity (1.1 W/(K m)), and composition-insensitive electrical conductivity (around 1 S/cm). We anticipate that our processing route can be extended to other phosphorus-rich phosphides that are still awaiting thin-film synthesis and will lead to a more complete understanding of these materials and of their potential applications.

Introduction

Phosphorus readily forms homoelement bonds in the solid state. Accordingly, over 100 phosphorus-rich binary metal phosphides containing both P–P bonds and metal–P bonds have been synthesized in bulk form.[1,2] Often, these compounds have semiconducting properties and decompose into elemental phosphorus and a metal-rich phosphide (with only metal–P bonds) at high temperatures.[2] Thin-film synthesis of P-rich materials would help determine their technological potential and their compatibility with established materials and processes. In addition, growing these materials in thin-film form would be desirable for high-throughput characterization of their properties as a function of composition and process conditions. However, reports of polycrystalline P-rich phosphides as thin films are very scarce and seem to be limited to basic characterization of ZnP2 and CdP2 deposited by evaporation of powders of the presynthesized compounds.[3,4] Thin-film growth from elemental or gaseous sources would significantly simplify the synthesis process. However, the high P partial pressure required to stabilize these P-rich compounds poses additional challenges for thin-film growth with respect to bulk synthesis. The classic method of heating the elements in powder form in a sealed ampule cannot easily be extended to phosphorization of metal thin films. Because the volume of a thin film is very small, it is difficult to achieve a sufficiently high P partial pressure without excessive P recondensation on the film. On the other hand, open-system processes with fixed gas flow rates are more controllable, but the combination of a high P partial pressure, high temperature, and a continuously flowing P source requires careful safety measures.

Like many other P-rich phosphides, bulk synthesis of CuP2 as a single-crystal or powder is well established,[5−10] but there are no reports of thin-film growth. CuP2 is a semiconductor that has been proposed as a solar absorber,[11] thermoelectric material,[12,13] electrocatalyst for hydrogen and oxygen evolution,[14] and a component in composite anode materials for Li-[15−17] and Na-based batteries.[18,19] Although CuP2 has been incorporated in electrochemical devices, its optoelectronic and thermoelectric characterization is incomplete. For example, the optical absorption coefficient of CuP2 crystals has only been measured in the weak absorption region just above its 1.4–1.5 eV band gap,[8,10] so it is impossible to evaluate its performance as a light absorber in the visible region. For thermoelectric applications, the properties needed to calculate the quality factor zT have only been measured separately on different CuP2 specimens in single-crystal or powder form. A potential method for growing phosphorus-rich phosphide thin films is reactive sputtering. We have recently shown the feasibility of this deposition technique for various metal-rich phosphide compounds.[20−22]

In this work, we present a relatively simple two-step process route to grow polycrystalline CuP2 thin films as semiconductors of potential technological interest. First, we deposit amorphous CuP2+ by reactive sputtering in a PH3-containing atmosphere. The advantage of this process step is that sufficient P can be incorporated in the films at room temperature at a relatively low PH3 partial pressure (0.1 Pa). In the second step, we crystallize the CuP2+ films by rapid thermal annealing (RTA) at atmospheric pressure under an inert gas flow. With this two-step process, high phosphorus partial pressures at high temperatures are avoided. We find that the crystallization step must be kinetically facilitated by employing amorphous precursors of sufficiently similar composition to the desired CuP2 stoichiometry. We investigate the optical properties of CuP2 over a broad spectral range and conduct comprehensive temperature-dependent thermoelectric characterization (including the zT value) up to room temperature. We find a remarkably high optical absorption coefficient (above 105 cm–1 in the visible region), low thermal conductivity (1.1 W/(K m)), composition-insensitive electrical conductivity (1 S/cm), and a moderate native doping density (1015–1017 cm–3) potentially suitable for photovoltaic applications.

Results and Discussion

Structure and Bonding

Bonding in CuP2 has some interesting features that are worth a brief analysis. Bulk CuP2 crystallizes in the monoclinic structure shown in Figure , with space group P21/c.[7] The structure consists of alternating sheets of CuP4 tetrahedra and of homoelement-bonded P atoms in planes parallel to (100) (Figure ). Each CuP4 tetrahedron shares an edge and three corners with other analogous tetrahedra. The existence of anion–anion bonding is a key qualitative difference between P-rich compounds like CuP2 and most optoelectronic compounds such as III–V and II–VI semiconductors. The generalized 8 – N rule[24] can then be used to interpret bonding. In this framework, one may assign the −1 oxidation state to one-half of the P atoms, since they are bonded to two other P atoms and three Cu atoms. The remaining P atoms have three P–P bonds and one Cu–P bond and are formally neutral, as the three homoelement bonds complete their octet. To achieve charge neutrality, Cu should then be in the +1 oxidation state. While explicit calculations[12] indicate that only about 30% of this charge is actually transferred to P due to significant covalency, they also confirm that the charge is only accepted by the P atoms that are in the −1 oxidation state. Thus, CuP2 is a relatively rare example of a compound with mixed anion valence.

Figure 1

Monoclinic P21/c structure of CuP2. (a) View emphasizing the sheets of CuP4 tetrahedra and of single-bonded P atoms. (b) View emphasizing the short Cu–Cu distance (dashed line) between two edge-sharing CuP4 tetrahedra.[23]

Another peculiar feature of the P21/c structure of CuP2 is that pairs of Cu atoms are quite close to each other (2.48 Å).[7] Comparing this distance to the bond length of metallic Cu (2.55 Å) and the metallic radius of single-bonded Cu (2.49 Å)[25] suggests that some metallic Cu–Cu bonding is to be expected. This is confirmed by calculation of a nonzero electron localization function between the two Cu atoms and by experimental analysis of phonon modes in CuP2.[26] These Cu–Cu dimers were recently shown to vibrate anharmonically as a rattling mode and strongly scatter acoustic phonons.[26] This is the key feature enabling low lattice thermal conductivity in CuP2 in spite of its relatively high acoustic conductivity, thus making it interesting for thermoelectrics.

Synthesizability

CuP2+ thin films with a broad range of x (positive and negative) could be deposited by reactive sputtering in a PH3/Ar atmosphere at room temperature by using either a Cu target, a Cu3P target, or both at the same time (see the Experimental Details and the x-axis in Figure a). The main available parameters to tune x are the RF power on the targets and the PH3 partial pressure (see the Supporting Information). Decreasing the power led to higher P contents due to a more P-enriched target surface and/or to a lower flux of Cu at the substrate, promoting phosphorization at the substrate. Higher PH3 partial pressures can be achieved by increasing the total pressure or the PH3 concentration in Ar. Because the PH3 concentration was limited to below 5% in our setup, we had to employ a relatively high sputter pressure (2 Pa ≃ 15 mTorr) to obtain films of CuP2 stoichiometry. The wide tunability of the P content in Cu–P films was also observed in our recently reported amorphous B–P films by reactive sputtering.[21] This compositional flexibility is likely related to the ability of P to form homoelement bonds in the films and segregate as an elemental impurity. Hence, we assume that the excess P in CuP2+ films with x > 0 is mainly bonded to other P atoms.

Figure 2

Effect of postannealing on amorphous CuP2+ films. (a) Change in P/Cu ratio after annealing for films of different initial compositions and thicknesses under constant annealing conditions (400 °C for 5 min). Vertical and horizontal lines indicate the CuP2 and Cu3P stoichiometries. The gray diagonal line corresponds to P/Cu ratios that are not modified by annealing. The dashed line is a guide to the eye, indicating that the films with initial composition closest to CuP2 are the ones with least severe P losses. (b) Change in P/Cu ratio after annealing for films of CuP2.4–CuP2.7 initial composition and similar thicknesses under different annealing conditions. (c) Calorimetry experiment on an initially amorphous 90 nm thick CuP2.5 film deposited on a Si3N4 membrane. If we assume a baseline for the heat capacity as shown in the figure, the energy released in the 300–400 °C region (area under the baseline) is estimated as 13 meV/atom.

As-deposited CuP2+ films did not exhibit any X-ray diffraction (XRD) peaks (Figure S2), so we crystallized them in an RTA furnace at atmospheric pressure under a N2 flow. Loss of phosphorus at moderate temperatures is a well-known phenomenon in many P-rich phosphides.[2] We also observed P losses in all our postannealed CuP2 films (Figure a,b). However, the dependence of these P losses on the initial composition of the as-deposited films is not trivial.

In Figure a, we compare the P/Cu ratio before and after annealing at 400 °C for 5 min for various initial compositions between CuP1.3 and CuP4.5. The P/Cu ratio is measured by X-ray fluorescence (XRF) so it represents an average through the depth of the film. Several interesting trends can be identified. First, thicker films generally experience milder P losses because P located deeper in the film requires a longer time to diffuse out. Second, sufficiently thick films with initial composition in the CuP2.2–CuP2.7 range can be “locked” into the desired CuP2 stoichiometry by annealing (Figure b). Third, films with severe P losses tend to approach the Cu3P composition after annealing. Cu3P is the most commonly reported binary stoichiometry in the Cu–P system.[5,6,27,28]

Last, and most surprisingly, we find that highly P-rich initial compositions do not help achieve a higher P content in the postannealed films. In fact, the opposite is true. When the initial composition is in the (P-rich) CuP4.1–CuP4.5 range, the postannealed composition is around CuP0.5 (Figure a). When the initial composition is much poorer in P (CuP1.3–CuP1.4 range), the postannealed composition is similar, around CuP0.4 (Figure a). On the other hand, when the initial composition is in an intermediate CuP2.2–CuP2.7 range closer to the desired CuP2 stoichiometry, P losses upon annealing are much slower in films of comparable thickness.

Using these atomically dispersed precursors with moderate P excess with respect to the target CuP2 stoichiometry, the necessary species for forming crystalline CuP2 are readily available within a subnanometer distance of their ideal crystallographic site. This enables fast crystallization of monoclinic CuP2. On the other hand, solid-state diffusion processes responsible for P losses have longer characteristic lengths, on the order of the film thickness (in our case, hundreds of nanometers). Thus, there is an optimal annealing time that is sufficient for CuP2 to crystallize but insufficient for substantial P losses to occur. Presumably, the lower total energy achieved by crystallizing the originally amorphous CuP2 film (Figure c) also helps delay P evaporation.

Conversely, precursor films that are too P-rich require solid-state diffusion to form a crystalline CuP2 phase because Cu atoms are too far apart in the initial amorphous phase. This Cu diffusion process now competes with the unwanted P diffusion leading to P evaporation. As a result, P losses are much faster. These findings are summarized in the qualitative diagram shown in Figure .

Figure 3

Qualitative convex hull of the Cu–P system at two different temperatures. Among competing phases, only Cu3P and the elements are considered. At 0 K, the Gibbs free energy only consists of enthalpy H, so the convex hull is drawn following DFT enthalpy calculations as available on the Materials Project database.[27] CuP2 is found to be a stable phase (on the convex hull), and Cu3P is slightly metastable (above the convex hull). At 700 K, our experiments indicate that CuP2 is destabilized. Part of the reason may be a large entropic term TS for elemental P, which is in gaseous form at this temperature. When higher energy amorphous precursors (diamond data points) are rapidly heated to 700 K, decomposition into Cu3P and gaseous P is thermodynamically favored. However, CuP2 formation is kinetically facilitated when the initial composition of the precursors is sufficiently close to the CuP2 stoichiometry (red diamond).

To visualize the P loss process, we image a film with final composition CuP1.3 by scanning electron microscopy (SEM, Figure ). Two phases can be clearly distinguished on the micrometer scale: a porous polycrystalline matrix with grain size around 30 nm and islands of more compact morphology. The intensity ratio between the Cu and the P peaks in energy-dispersive X-ray spectroscopy (EDX) increases by a factor ∼5.5 when moving from the matrix to the islands (Figure S1). Thus, we conclude that the matrix consists of CuP2 and the islands consist of Cu3P. The mechanism of conversion from CuP2 to Cu3P appears to be diffusion of Cu in the plane of the substrate, contributing to the enlargement of seed Cu3P islands. At the same time, P gradually evaporates elsewhere. Because the most stable gaseous form of phosphorus[29] at our annealing temperatures is P4 and no intermediate solid phases between CuP2 and Cu3P are observed, the CuP2 decomposition reaction can be written as 12CuP2(s) → 4Cu3P(s) + 5P4(g).

Figure 4

Morphology of a postannealed film with overall CuP1.3 composition. (a) Low-magnification SEM image showing a dual-phase morphology with a CuP2 matrix and Cu3P islands. (b) High-magnification SEM image showing porosity in the CuP2 matrix. Phase identification was performed on the basis of spatially resolved EDX spectra (Figure S1).

Figure b shows the effect of annealing temperature and time on the final composition. As expected, increasing the annealing time at fixed temperature results in more severe P losses (compare the data from 30 s versus 5 min annealing time at 400 °C). In general, longer annealing times can be tolerated at lower annealing temperatures. For example, annealing at 350 °C for 15 min yields about as many P-poor samples as the case of annealing at 400 °C for 5 min. Note that the spread of final P/Cu ratios sometimes obtained for films of otherwise similar initial P/Cu ratios, and the thicknesses is mainly caused by the different positions of the samples inside the furnace. The samples located further downstream with respect to the gas flow tend to lose less P, possibly because they are exposed to a finite P partial pressure due to P evaporating from the samples further upstream.

We observe a decrease in the apparent heat capacity of an as-deposited CuP2.5 film at around 300 °C by nanocalorimetry (Figure c). This indicates an exothermic signal, which may be related to the transition from the amorphous to the (more stable) polycrystalline state. The heat capacity baseline needed to calculate the heat of crystallization is not straightforward to define. If we assume the baseline shown in Figure c, we estimate the crystallization energy of CuP2 as 13 meV/atom. Although the uncertainty on this value is substantial, some qualitative conclusions can still be drawn. The calculated formation enthalpy of CuP2 from the elements in their standard state is 112 meV/atom.[27] Because this value is much larger than the estimated crystallization energy, most of the formation energy has probably already been released during formation of the amorphous compound. The thermal energy of a solid at 400 °C is approximately 3kT = 174 meV/atom by using the Dulong–Petit law. Thus, the extra stabilization achieved by crystallizing CuP2 is only a small fraction of the thermal energy available at that temperature.

It is also interesting to consider typical values for the calculated energy difference between the most stable amorphous configuration and most stable crystalline polymorph for a given material. This quantity has been calculated in a previous study for 41 material systems (mainly oxides) at 0 K.[30] The energy difference varies between ∼50 and ∼500 meV/atom depending on the material. The significantly lower crystallization energy measured in CuP2 could indicate that the entropic contribution to the total energy is substantially higher in the amorphous state than in the crystalline state at finite temperatures. Higher entropy is indeed expected in the amorphous state due to higher disorder, and it would contribute to reducing the energy difference between the amorphous and crystalline state of CuP2 at ∼700 K with respect to 0 K. Although this explanation is plausible, it is also possible that CuP2 and other non-oxide compounds simply exhibit different energetic trends than the computationally investigated selection of compounds. Computational analysis of the energetics of a more diverse range of amorphous material systems would certainly be useful.

Stability

Previous work on CuP2 single crystals does not comment on their stability under ambient conditions. On the basis of simple observations on our thin-film samples, we suggest that the air stability of CuP2 should be further investigated. A change in color is consistently observed in as-deposited CuP2+ after few hours of exposure to ambient air, signaling a reaction that is not limited to a surface layer of a few nanometers thickness. For this reason, the films characterized in this work were annealed immediately after deposition. After annealing, the bulk properties of the films appear to be stable for a longer time (at least a few days) under ambient conditions, as judged by their visual appearance and electrical conductivity. The higher reactivity of amorphous CuP2+ may be due to the extra P present before annealing and to the higher energy associated with the amorphous state (Figure c). Both the amorphous and the polycrystalline films appear to be stable in a N2 atmosphere.

After either type of film has been exposed to air for a sufficiently long time, the reaction front has reached the back surface of the film, as evident by visual inspection through the glass substrate. The exact details of the CuP2+–air reaction are currently unknown. However, XRF measurements reveal a large decrease in P/Cu ratio after prolonged exposure to air, indicating that the reaction involves P losses. We suspect that the high sputter pressure (2 Pa) necessary to obtain a P/Cu ratio above 2 in our growth setup may explain why the reaction of CuP2 films with air is not limited to a surface layer. Films sputtered at high pressure are generally more porous and more air sensitive due to their higher surface area.[31] Thus, we cannot conclude that CuP2 films are intrinsically unstable in air. The stability of CuP2 films sputtered at lower pressures or deposited by other techniques should be investigated to clarify this issue.

Structural and Vibrational Properties

In agreement with nanocalorimetry results, the originally amorphous CuP2+ films only begin to show crystalline XRD peaks above 300 °C annealing temperature (Figure S2). Beyond this lower limit, it is possible to obtain polycrystalline CuP2+ films in the P21/c structure under various annealing conditions. As long as the final composition is close to the nominal CuP2 stoichiometry, XRD patterns of films processed under different annealing conditions are rather similar (Figure S2). As an example, the XRD pattern of a CuP2 film annealed at 450 °C for 30 s (Figure a) contains all the peaks expected for the P21/c structure, without major preferential orientation effects and without clear peaks from secondary phases above the noise level. The XRD peak positions closely match the positions of the reference bulk CuP2 sample,[7] indicating that structural parameters (including the short Cu–Cu distance) are about the same in thin-film and bulk CuP2. XRD peaks from Cu3P in the hexagonal P63cm structure are observed in CuP2+ films when x < 0 (Figure S2). However, the threshold value of x at which Cu3P peaks begin to appear strongly varies with annealing conditions. When x > 0 (P-rich films), no XRD peaks associated with secondary phases like Cu2P7 or elemental phosphorus are observed up to the most P-rich composition reached in this study (CuP2.2).

Figure 5

Structural and vibrational characterization of a film with CuP2 stoichiometry after postannealing at 450 °C for 30 s. (a) Experimental XRD pattern together with the reflections expected for randomly oriented CuP2 in the monoclinic P21/c structure.[7] XRD patterns under other annealing conditions are shown in Figure S2. (b) Experimental Raman spectrum with labels for the identified peak positions. The total phonon density of states of CuP2 in the P21/c structure, as calculated in the Materials Project database,[27] is also shown. The Cu–Cu rattling mode (∼100 cm–1) believed to limit the thermal conductivity of CuP2[26] is indicated.

It might be tempting to conclude that single-phase CuP2+ can be grown over a wide x range, in which defect formation is favored over secondary phase precipitation. However, secondary phases in amorphous form (not detected by XRD) are likely to be present in our samples for the following reasons. First, amorphous secondary phases were identified in a previous study on the phase equilibrium between Cu3P and CuP2 in powder form.[32] This observation rendered XRD-determined phase boundaries incorrect. Second, our short annealing times may not be sufficient to crystallize phases with a significantly different composition than the original precursors. In fact, Cu3P can only form if CuP2 loses more than 80% of its original P in some regions of the film. If the annealing process is stopped before these losses can take place, and if there are no stable phases between CuP2 and Cu3P, the likely result is formation of amorphous phases with intermediate composition. Third, the electrical conductivity of our films is roughly constant in the CuP1.0–CuP2.2 composition range (Figure a). It is improbable that the high defect densities required to accommodate this nonstoichiometry do not have any effect on the electrical properties. Thus, electrically inactive secondary phases (such as the disconnected Cu3P islands shown in Figure a) are very likely to coexist with point defects in highly nonstoichiometric CuP2.

Figure 6

Room-temperature electrical and optical properties of postannealed CuP2+ films. (a) Electrical conductivity as a function of composition (measured after annealing) and annealing conditions. Films are of comparable thickness except for the P-poor purple data points, which have about half the thickness as the other ones. A zoomed-in view around the CuP2 stoichiometry is available in Figure S4. (b) Absorption coefficient α of a postannealed CuP2.0 film together with a α1/2 plot versus photon energy. Inset: photograph of a film with increasing P/Cu ratio from bottom left to top right. When the P/Cu ratio decreases below roughly 1, the appearance of the film changes from dark red (characteristic of semiconducting CuP2) to gray (characteristic of metallic Cu3P). (c) Refractive index and extinction coefficient of the same CuP2.0 film shown in (b).

The Raman spectrum of the same sample used for XRD characterization is plotted in Figure b. The phonon density of states (DOS) of CuP2, as calculated by density functional perturbation theory in good agreement with recent experiments,[26] is also shown for comparison.[27,33,34] Because Raman spectra of bulk CuP2 are not available in the literature, we briefly discuss some qualitative aspects here. Raman features originating from the phonon bands centered around 300 and 450 cm–1 can clearly be seen in the experimental spectrum. In particular, the most intense Raman peak at 425 cm–1 probably arises from one of the lowest-energy phonon branches within the highest-energy band in the calculated DOS. All modes in this band essentially involve vibrations of P atoms with nearly static Cu atoms. The lower the phonon energy, the larger the contribution from Cu vibrations, as expected from the larger mass of Cu.

Because the film is polycrystalline, there are selection rules for Raman-active phonon modes and the Raman spectrum will not directly reflect the phonon DOS. Specifically, all atoms in CuP2 are at 4e Wyckoff positions of the P21/c space group, so only the Ag and Bg modes are Raman-active according to the character tables.[35] With a 12-atom unit cell, 18 Raman-active modes are predicted in total.[35] Eight peaks can be identified the experimental spectrum (Figure b). The Cu–Cu rattling mode identified by Qi et al. as an important scatterer of heat-transporting phonons[26] is either symmetry-forbidden or too low in intensity to be distinguished by Raman spectroscopy.

Electrical and Optical Properties

The room-temperature electrical conductivity of postannealed polycrystalline films in the CuP2.0–CuP2.2 composition range is between 0.5 and 1.0 S/cm at room temperature, without a clear dependence on the P/Cu ratio (Figure a). The conductivity slightly increases with increasing annealing temperature, regardless of annealing time (Figure S4). Previously reported conductivities of CuP2 single crystals range from 0.01 to 30 S/cm, presumably due to differences in the crystal quality.[6,8−10] Films with severe P losses have significantly higher conductivities (Figure a), probably due to percolation paths between highly conductive Cu3P phases.[6] The Seebeck coefficient measured on a freshly annealed CuP2.0 film is +390(10) μV/K (Figure S3), indicating native p-type doping. All previously reported CuP2 single crystals were also p-type with higher Seebeck coefficients in the 690–820 μV/K range. The work function, measured with a Kelvin probe in air on a freshly annealed CuP2.0 film, is 5.0(1) eV.

CuP2 is a relatively strong absorber of light. Its absorption coefficient α reaches 105 cm–1 at a photon energy hν = Eg + 0.6 eV above its band gap Eg = 1.5(1) eV (Figure b). This compares favorably even with the most efficient direct gap photovoltaic absorbers such as GaAs, CdTe, and CH3NH3PbI3 (MAPI).[36] In fact, the absorption coefficient is as high as in some exciton-enhanced photoabsorbers such as BiI3 and Cu2BaSnS4,[37,38] indicating that CuP2 may deserve more detailed optoelectronic characterization.

We find that α1/2 is linear in photon energy over a 2 eV spectral range above the band gap (Figure b), indicating that α ∝ (hν – Eg)2. Both the estimated band gap and the spectral dependence of the absorption coefficient are in agreement with previous work on CuP2 single crystals.[8−10] Because the α ∝ (hν – Eg)2 behavior is often associated with an indirect gap in conventional semiconductors,[39] an indirect gap was previously assumed for these CuP2 crystals.[8,10]

However, there are at least two other factors to consider. (1) The absorption strength of CuP2 is high even for a direct gap material, so indirect transitions are unlikely to be responsible for it. (2) According to the calculated band structure of CuP2,[27] the fundamental gap should be direct and located between the Γ and the Y point of the Brillouin zone. Two indirect gaps with slightly higher energies exist due to additional valence band pockets at the X point and between the Y and H points.[27] Even though we observe an α ∝ (hν – Eg)2 behavior, care should be taken when employing the absorption characteristics typical of group IV and III–V semiconductors to interpret the nature of the optical transitions of other semiconductors with substantially different band structures. A clear difference between CuP2 and conventional semiconductors is that the former has many valence and conduction band pockets at different points of the Brillouin zone. Hence, many different optical transitions can contribute to the overall absorption coefficient.

The refractive index of CuP2 is 3.3–3.4 in the transparent region (Figure c). Extrapolation of the real part of the dielectric function to zero photon energy (Figure S5) yields a high-frequency permittivity ε = 10.5 ± 1.0. Interestingly, there seems to be a critical P/Cu ratio close to 1, where the electrical and optical properties shift from being “CuP2-like” (semiconducting and IR transparent) to being “Cu3P-like” (metallic and opaque). This transition is manifested by an abrupt change in conductivity (Figure a) and visual appearance (inset of Figure b).

Thermoelectric Characterization

We conducted DC and double AC Hall effect measurements as well as temperature-dependent thermoelectric characterization of three films. They have the following compositions: Cu2.50P (labeled “Cu3–P”), Cu1.61P, and CuP1.35 (labeled “CuP2–”). We use these labels to emphasize similarity to Cu3P and CuP2 as discussed in the previous section. This set of films was deposited on Si3N4 membranes as part of a microchip-based thin-film transport characterization platform.[40] Differences between this set of films and the films deposited on glass characterized in the rest of the article are listed in the Supporting Information. Because these films have intermediate compositions between CuP2 and Cu3P, their properties may be influenced by inhomogeneity, as exemplified by the dual-phase morphology shown in Figure . Nevertheless, important qualitative trends in the transport properties of these films as a function of composition can still be discerned.

The temperature dependence of the electrical conductivity (Figure a) suggests that CuP2– is a nondegenerately doped semiconductor and that the two other films are either metallic or degenerately doped semiconductors. Hall effect measurements at room temperature confirm this interpretation (Figure ), with high carrier concentrations measured in Cu3–P and Cu1.61P (above 1020 cm–3) and a moderate carrier concentration measured in CuP2– (1015–1017 cm–3). All films have a positive Hall voltage confirming their p-type conductivity. Note that the conductivity of Cu1.61P and CuP2– after one month of storage is appreciably lower (Figure a), highlighting possible stability issues as discussed in the previous sections.

Figure 7

Thermoelectric properties of three postannealed Cu–P films as a function of temperature T. The compositions after annealing are indicated. (a) Electrical conductivity σ, which was also remeasured at room temperature one month after the temperature-dependent measurement (star markers). (b) Absolute Seebeck coefficients S ≡ SCu–P, with linear trends indicated. (c) Thermal conductivity κ. (d) Thermoelectric figure of merit zT = σS2T/κ.

Figure 8

Hole mobility and concentration by double AC Hall effect measurements on the same Cu–P films shown in Figure . The film labeled “Cu3P ref.” is a continuous polycrystalline Cu3P film deposited by reactive sputtering at 360 °C and used as a reference.

Because of the inverse relationship between carrier concentration and thermovoltage,[41] the Seebeck coefficient is highest in CuP2– and lowest in Cu3–P (Figure b). Interestingly, the Seebeck coefficient increases linearly with temperature in all three films (Figure b). This behavior is often a sign of a temperature-independent carrier concentration,[41] a typical feature of materials with nonzero density of states at the Fermi level (i.e., metals and degenerate semiconductors such as Cu3–P and Cu1.61P). However, a linear increase of the Seebeck coefficient with temperature is not readily explained for a more weakly doped semiconductor like CuP2–. In an ideal scenario, we would expect the carrier concentration to increase with temperature due to increasing defect ionization and the Seebeck coefficient to decrease accordingly. The reason for this discrepancy is unclear. One could invoke the role of film inhomogeneity due to the presence of Cu3P secondary phases (Figure ) or assume that the increase of electrical conductivity with temperature (Figure a) is due to mobility changes rather than to the hole concentration changes. Yet, a simultaneous increase in hole concentration and Seebeck coefficient with temperature was reported for CuP2 single crystals,[8] where inhomogeneity effects can be excluded. Multiband transport could also cause an unusual temperature behavior due to increasing contributions from the additional valence band pockets of CuP2 with increasing temperature. However, application of the Boltzmann transport equation[42] on the calculated CuP2 band structure[27] reveals that a significant decrease in the Seebeck coefficient is expected in the 200–300 K range assuming a concurrent increase in hole concentration by 1 order of magnitude (Figure S6).

As another hypothesis, one could assume that CuP2– is highly compensated by donor defects at low temperatures, but its p-type character becomes more dominant at higher temperatures due to activation of a deeper acceptor. If this hypothesis is correct, one would expect both the electrical conductivity and the Seebeck coefficient to increase with temperature as we experimentally observe—the former due to an increase in the concentration of ionized acceptors and the latter due to a decreasing contribution from the (negative) n-type Seebeck coefficient.[41] Previous work also suggested the possibility of charge compensation in CuP2 single crystals based on the temperature dependence of their carrier mobility.[10] The position of the acceptor level in our CuP2– film can be estimated as 121(3) meV above the valence band from an Arrhenius plot of the electrical conductivity in the 230–300 K temperature range (Figure S7).

The room-temperature thermal conductivity of CuP2– is 1.1 W/(K m) (Figure c). This value is lower than in CuP2 single crystals (3.6–4.7 W/(K m) depending on lattice direction)[26] as may be expected for a polycrystalline sample. Our measured conductivity is, however, in excellent agreement with the calculated 1.12 W/(K m) amorphous limit for bulk CuP2.[12] The increasing thermal conductivity with increasing temperature is unlike the ∝1/T behavior typical of crystalline semiconductors in this temperature range. Instead, it is often observed in amorphous or highly disordered materials, consistent with the observation that our measured conductivity is very close to the amorphous limit. On the basis of these results, we assume that the phonon mean free path in the CuP2– film is low due to disorder[43] and/or phonon boundary scattering.[44] The latter is likely enhanced by the small grains, low thickness, and porous morphology of the film.[43−45] The electronic contribution to the thermal conductivity is negligible due to the low hole concentration of CuP2– (Figure ). The thermal conductivities of Cu3–P and Cu1.61P are only slightly higher and their temperature dependences are similar to the case of CuP2–. Thus, we conclude that the thermal conductivity is phonon-mediated and strongly limited by film morphology in all three films. In fact, scattering of charge carriers (holes) is also morphology-limited. The hole mobility of the present Cu3–P film (0.27 cm2/(V s)) is 2 orders of magnitude lower than in a continuous Cu3P film on glass with about the same carrier concentration (28.8 cm2/(V s), see Figure ).

CuP2 has recently been proposed as a potential thermoelectric material.[12,13] Our measurements on a CuP2– film confirm that the lattice contribution to its thermal conductivity is indeed sufficiently low for thermoelectric applications. However, the thermoelectric figure of merit zT at room temperature is still low for all investigated compositions (Figure d) due to low power factors (Figure S8). In the vicinity of the Cu3P stoichiometry, the main issue is a low Seebeck coefficient. In the vicinity of the CuP2 stoichiometry, the main issue is low electrical conductivity. Even taking the more favorable properties of our CuP2 films on glass (Table ) or of previously reported CuP2 single crystals,[8] the zT value at room temperature would only be 0.004 and 0.05, respectively. It might be possible to optimize the hole concentration of CuP2 by extrinsic doping to obtain higher zT values. Nevertheless, phosphorus losses at moderate temperatures and potential stability issues under ambient conditions are likely to limit its practical applicability in thermoelectric devices. Similar issues might exist in other phosphorus-rich phosphides.

Table 1

List of Electrical, Optical, and Thermal Properties Measured in This Study on Postannealed CuP2+ Films at Room Temperaturea

electrical conductivity	0.5–1.0	S/cm
Seebeck coefficient	+390 ± 10	μV/K
thermal conductivity	1.1 ± 0.1	W/(K m)
band gap	1.5 ± 0.1	eV
work function	5.0 ± 0.1	eV
dielectric constant (ε∞)	10.5 ± 1.0

The film composition is CuP2.0 for all properties except for the thermal conductivity (CuP1.35).

Conclusion

We deposited amorphous CuP2+ thin films with a wide range of x (positive and negative) by reactive sputtering in a PH3/Ar atmosphere. By annealing these films above their crystallization temperature in an inert atmosphere, we observed that the CuP2 phase was thermodynamically unstable with respect to the Cu3P phase. However, it was possible to kinetically stabilize polycrystalline CuP2 by satisfying all the following conditions: (1) amorphous precursors mixed on the atomic level (rather than a heterogeneous mixture of amorphous components) to ensure the correct local bonding environment; (2) initial composition sufficiently close to the ideal P/Cu ratio of 2, also to ensure the correct local bonding environment; (3) annealing temperature just high enough to allow for solid-state diffusion; (4) annealing time just long enough for crystallization to be completed, but not long enough for a large fraction of P to diffuse to the surface.

Remarkably, amorphous films that were either too P-poor or too P-rich quickly decomposed into Cu3P and gaseous phosphorus upon heating. This “compositional lock-in” behavior highlights the importance of pre-existing short-range order for kinetic stabilization of materials under conditions where decomposition and crystallization are in competition with each other.

Polycrystalline CuP2+ films are semiconductors with native p-type conductivity. Their electrical properties are rather insensitive to elemental composition in the vicinity of the stoichiometric point and only moderately affected by the annealing conditions. The thermal conductivity of a P-poor CuP2 film is 1.1 W/(K m) at room temperature, confirming its potential applicability as a thermoelectric material. However, the hole conductivity of CuP2 is too low to achieve a high power factor (and therefore a high zT value) without extrinsic doping. Furthermore, decomposition of CuP2 into Cu3P and gaseous phosphorus at around 400 °C hinders high-temperature applications. Although stability issues are not mentioned in the CuP2 single-crystal literature, our polycrystalline CuP2+ films were only stable in ambient conditions for a few days. It is currently not clear if this issue is related to the porous morphology of our films or if it is an intrinsic behavior of CuP2.

Finally, CuP2 is a stronger light absorber than many established photovoltaic materials, with absorption coefficient rapidly rising to 105 cm–1 above its 1.5 eV band gap. Combined with a native doping density in the optimal range for a photovoltaic absorber in a pn junction solar cell (1015–1017 cm–3), we conclude that CuP2 may deserve more detailed optoelectronic characterization.

Experimental Details

Film Growth

Amorphous CuP2+ thin films were deposited on Corning Eagle XG borosilicate glass by reactive radio frequency (RF) sputtering over a 10 × 5 cm2 area. A Cu target and a Cu3P target were cosputtered at 2 Pa total pressure in a 5% PH3/Ar atmosphere without intentional heating and without substrate rotation. The targets were oriented so that one short side of the substrate would mainly be coated by the Cu target and the other short side by the Cu3P target.

Immediately after deposition, CuP2+ films were cut into smaller pieces and annealed in a lamp-based rapid thermal annealing (RTA) furnace in a N2 atmosphere. Because of the sputtering target geometry and differences in their applied power, small gradients in P/Cu ratio and film thickness were obtained across the substrate. These gradients enabled us to characterize several data points (“samples”) for each annealing run, each with a distinct composition and thickness. More details on film deposition and annealing are available in the Supporting Information.

Film Characterization

All measurements except for nanocalorimetry and thermoelectric/Hall effect characterization were performed within 24 h after annealing to avoid sample degradation. The combinatorial characterization data arising from compositional gradients in the films were managed with the COMBIgor tool[46] and the Research Data Infrastructure[47] and integrated into the High-Throughput Experimental Materials Database.[48]

Elemental composition and film thickness were determined by X-ray fluorescence (XRF) calibrated by Rutherford backscattering spectrometry (RBS, composition) and spectroscopic ellipsometry (thickness). X-ray diffraction (XRD) measurements were conducted by using Cu Kα radiation, a 2D detector, and a fixed incidence angle of 10°. Raman spectra were measured with 532 nm excitation wavelength and 4 W/mm2 power density. Scanning electron microscopy (SEM) images were taken at 5 kV beam voltage.

Sheet resistance was measured with a collinear four-point probe directly contacting the film. The Seebeck coefficient of a CuP2 film on glass was measured in a custom-built setup by using In contacts. The work function was measured with a Kelvin probe calibrated with a standard Au sample. The absorption coefficient and optical functions were extracted by spectroscopic ellipsometry. Because of higher porosity in the upper part of the film, we modeled the system as a glass substrate of known optical functions, a CuP2 layer with a linearly increasing fraction of air from bottom to top,[49] and a roughness layer treated with Bruggeman effective medium theory.

For nanocalorimetry and thermoelectric/Hall effect characterization, CuP2+ films were deposited on previously described microfabricated chips designed for calorimetry[50] and in-plane thermoelectric characterization[40] of thin-film samples. In both types of chips, CuP2+ was deposited on a free-standing Si3N4 membrane. Because of the fragility of the membrane, thinner CuP2+ films (90–120 nm) were employed for these studies.

Nanocalorimetry experiments were conducted in a N2 atmosphere on an as-deposited amorphous film with initial CuP2.5 composition, with an average heating rate of roughly 5000 °C/s. Temperature-dependent thermoelectric characterization (electrical and thermal conductivity and Seebeck coefficient) was performed in a vacuum on three films with different compositions after annealing. The electrical conductivity was measured by using the van der Pauw (vdP) method.[51] The Seebeck coefficient was measured with respect to platinum metals lines by using an internal four-probe platinum thermometer.[40] The thermal conductivity was derived from the current–voltage characteristics of membrane heaters/thermometers in the self-heating regime.[40,52] The hole concentration and mobility were measured on the same samples by double AC Hall.[53] More details on all measurements are available in the Supporting Information.