Sn Ratio in the Synthesis of Cu2ZnSnS4 Films

Mar 12, 2015 - We find that the Cu/Sn ratio in CZTS is self-regulating and approaches 2, regardless of the initial composition of the precursor films,...
1 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/cm

Self-Regulation of Cu/Sn Ratio in the Synthesis of Cu2ZnSnS4 Films Melissa C. Johnson, Cody Wrasman, Xin Zhang, Michael Manno, C. Leighton,* and Eray S. Aydil* Department of Chemical Engineering & Materials Science, 421 Washington Avenue, University of Minnesota, Minneapolis, Minnesota 55455, United States S Supporting Information *

ABSTRACT: There has been a tremendous recent surge of interest in copper zinc tin sulfide (Cu2ZnSnS4, CZTS) as a photovoltaic material, because its optical and electronic properties are well-suited for solar cells, and its elemental constituents are abundant in the earth’s crust. Here we have studied the formation mechanisms of CZTS films, and the factors that control the cation stoichiometry during ex situ sulfidation of precursor Cu−Zn−Sn alloy films in a closed isothermal system. We find that the Cu/Sn ratio in CZTS is self-regulating and approaches 2, regardless of the initial composition of the precursor films, provided that adequate Sn is available in the sulfidation system. If precursor films are initially tin rich, excess Sn evaporates in the form of SnS. If precursor films are initially Sn-deficient, the inclusion of solid Sn in the sulfidation ampule readily generates SnS vapor, which mitigates the films’ Sn deficiency to return the Cu/Sn ratio to 2. When sulfidized for sufficiently long times at sufficiently high temperatures (e.g., 600 °C, 8 h), films with similar Cu/Zn ratios exhibit similar phase compositions, such that if Cu/Zn >2, a Cu2SnS3 impurity phase is present in addition to CZTS, and if Cu/Zn < 2, a ZnS impurity phase occurs. To achieve phase-pure, void-free films, Sndeficient precursor films with Cu/Zn in the desired range (typically close to, but slightly less than 2) can be sulfidized with excess Sn in a closed system, or a system that maintains a SnS vapor pressure over the film. Time-dependent sulfidation experiments were performed to elucidate the mechanism of this Sn self-regulation. During the formation of CZTS, almost all of the Sn is found to leave the film as SnS, later reincorporation of the Sn occurring through reactions between SnS vapor and CuS to form Cu2SnS3. The ZnS and Cu2SnS3 phases within the films then interdiffuse to form CZTS. Because Cu/Sn is 2 in both Cu2SnS3 and CZTS, the Cu/Sn ratio tends to 2 when sufficient Sn is included in the system to consume all Cu. This strategy is useful for avoiding Cu−S minority phases, provided the films are sulfidized to the point of equilibrium phase composition.

■

INTRODUCTION

strongly on the elemental composition (stoichiometry) of the CZTS layer. Solar cells made from stoichiometric Cu2ZnSnS4 films perform significantly worse than those made from Cu poor films. Specifically, the highest efficiency cells tend to be Cupoor ([Cu/(Zn+Sn)] ∼ 0.8) and Zn-rich (Zn/Sn ∼ 1.1).3,8 However, synthesizing phase-pure CZTS thin films with a specific off-stoichiometric composition is nontrivial. For instance, the small cation composition window for CZTS phase stability is often blamed for the facile formation of secondary phases when the cation stoichiometry differs from 2:1:1 (Cu:Zn:Sn).9 Tuning the composition is also complicated because Sn may leave the film at elevated temperatures if the composition of the vapor phase above the film is not carefully controlled. For example, low S partial pressures above the film are known to favor Sn loss.10−12 This problem is exacerbated when films are annealed or sulfidized in open systems, where vapor composition and pressures are not well controlled and Sn loss through the vapor phase may be irreversible.

The global solar-to-electric energy conversion capacity has doubled every 2.5 years between 1975 and 2013.1 Maintaining this growth toward terawatt levels with current thin film solar cells based on CdTe or CuIn1−xGaSe2 (CIGS) may be difficult because In and Te are scarce within the earth’s crust.2 In contrast, Cu2ZnSnS4 (CZTS) and Cu2ZnSnSe4 (CZTSe) are comprised of earth-abundant elements and are under consideration for use as the absorber layer in thin film solar cells.3 These p-type semiconductors have high absorption coefficients (>104 cm−1 for λ 2) are sulfidized with elemental Sn (1 mg) in the sulfidation ampule, in addition to the S charge, the Sn is incorporated into 2508

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514

Article

Chemistry of Materials the film until the Cu/Sn ratio falls to ∼2. Thus, the Sn content in the film is self-regulating such that the Cu/Sn ratio becomes ∼2 at equilibrium, provided that adequate Sn is present in the system. Figure 1 shows this visually: the scattered green ○ represent the compositions of the precursor alloy films prior to sulfidation, collapsing to a line of blue squares around the dotted Cu/Sn = 2 line (i.e., the final composition of the sulfidized films). Inclusion of Sn in the annealing system was shown by Redinger et al.12 to be an effective strategy to counter SnS loss from the film10 via the reaction Cu2ZnSnS4(s) ⇔ Cu2S(s) + ZnS(s) + SnS(g) + 1/2 S2(g). Redinger et al. proposed that the presence of SnS vapor drives the decomposition reaction in the reverse direction. Our experiments show that Sn from SnS vapor can be incorporated into the film even when the precursor films are significantly Sn poor to begin with. Moreover, including excess Sn in the precursor films is sufficient to prevent detrimental amounts of Sn loss, as long as the Sn leaving the film as SnS vapor remains above the film (as in a closed system) and is not removed by flow. Zn and Cu loss from the films was negligible, and the elemental composition moved along lines of constant ZnS and Cu2S. The latter is expected because Cu vapor pressure at 600 °C is very low. Zn vapor pressure, however, is high and rises from 1.76 × 10−14 Torr at 25 °C to 11.4 Torr at 600 °C. The absence of significant Zn loss from the film thus suggests that as the temperature increases, Zn is sulfidized rapidly before it can leave the film as Zn metal vapor.14 This observation is consistent with that reported by Mousel et al., who studied cation composition during selenization.15,16 Figure 2 (panels a−f) show SEM images of six representative films with different cation compositions (labeled A−F). Figures

Figure 3. XRD patterns collected from films shown in Figure 2. The bottom panels show the expected powder diffraction patterns for Cu3SnS4 and CZTS.

Figure 4. Raman spectra collected from films shown in Figure 2. The vertical dotted lines indicate the locations of the Raman scattering peaks from CZTS (284 and 338 cm−1), Cu2SnS3 (305 and 355 cm−1) and ZnS (351 cm−1).

the precursor films leaves the film as SnS until the Cu/Sn ratio in the film increases to ∼2 (Table 1). Second, stoichiometric CZTS can be obtained, even if there is excess Sn in the precursor film, as long as the Cu/Zn ratio is ∼2. Third, when the Cu/Zn ratio is lower than 2, excess Zn precipitates as ZnS, and when Cu/Zn is much higher than 2, Cu2SnS3 precipitates form. Finally, significant material loss (e.g., Sn as SnS) from the film creates voids. These points are illustrated below using films A, B, and C as examples. For example, prior to sulfidation, film A was significantly Snrich and significantly Zn-poor. Upon sulfidation, the film remained Zn-poor, but the correct quantity of Sn left the film such that the Cu/Sn ratio in the sulfidized film increased to ∼2, the value for stoichiometric CZTS. While the precursor film was continuous, the sulfidized film (Figure 2a) is discontinuous, with voids separating large grains. We attribute this morphology to significant material loss from the film (i.e., Sn as SnS vapor). Prior to sulfidation, film B was also Sn-rich, but the Cu/Zn ratio was nominally the stoichiometric one (∼2). Upon sulfidation, a nominally stoichiometric CZTS film was obtained: the Cu/Zn ratio remained at ∼2, while, again, the correct quantity of Sn left the film as SnS such that the Cu/Sn ratio in the sulfidized film increased to ∼2. The plan view SEM (Figure 2b) of this film also showed voids between the grains, again indicative of material loss. Finally, prior to sulfidation, film C was slightly Zn- and Sn-rich (Cu/Zn < 2 and Cu/Sn < 2).

Figure 2. Plan-view SEM images of selected films A (a), B (b), C (c), D (d), E (e), and F (f). The same scale bar applies to all images.

3 and 4 show the corresponding XRD and Raman scattering data for these films. The precursor alloy film compositions and compositions after sulfidation are listed in Table 1. Films A, B, and C were synthesized by sulfidizing initially Sn-rich precursor films (i.e., Cu/Sn < 2), while films D, E, and F were synthesized by sulfidizing initially Sn-poor precursor films with elemental Sn in the sulfidation ampule. Prior to sulfidation, in addition to being Sn-rich, film A was Zn-poor (Cu/Zn > 2), film B had a stoichiometric amount of Zn (Cu/Zn ∼ 2), and film C was Zn-rich (Cu/Zn < 2). These three compositions were chosen as examples to illustrate four important observations. First, during sulfidation, excess Sn in 2509

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514

Article

Chemistry of Materials

Table 1. Precursor and Sulfidized Film Compositions of Six Selected and Representative Films (from Over 30) As Determined from EDSa precursor film

sulfidized film

film

Cu

Zn

Sn

Cu/Sn

Cu

Zn

Sn

S

Cu/Sn

A Cu2.56Zn0.46Sn1.13S3.83 B Cu2.18Zn0.99Sn1.04S3.79 C Cu1.84Zn1.30Sn0.98S3.88 D Cu2.42Zn0.64Sn1.15S3.80 E Cu2.08Zn0.89Sn1.07S3.95 F Cu1.65Zn1.36Sn0.94S4.06

45.8 (29.7) 40.6 (25.5) 44.1 (28.3) 66.2 (49.4) 58.3 (41.1) 46.2 (30.1)

9.6 (12.5) 19.2 (24.1) 29.6 (37.9) 13.5 (20.2) 23.3 (32.9) 34.9 (45.4)

44.6 (57.9) 40.3 (50.5) 26.4 (33.8) 20.3 (30.4) 18.4 (26) 18.9 (24.5)

1.03

32.0 (44.4) 27.3 (34.9) 23 (28.8) 30.2 (40.3) 26.0 (34.7) 20.6 (26.5)

5.8 (16.1) 12.4 (31.8) 16.3 (40.8) 8.0 (21.4) 11.1 (29.6) 17 (43.5)

14.2 (39.5) 13 (33.2) 12.2 (30.5) 14.4 (38.3) 13.4 (35.7) 11.7 (30.0)

47.9

2.25

47.38

2.1

48.53

1.89

47.5

2.1

49.4

1.94

50.8

1.76

1.01 1.67 3.26 3.17 2.44

a The precursor composition that corresponds to stoichiometric Cu2ZnSnS4, in the absence of metal loss from the film, is 50% Cu, 25% Zn, and 25% Sn. The stoichiometric sulfidized film composition should be 25% Cu, 12.5% Zn, 12.5% Sn, and 50% S. The composition within parentheses is the composition normalized to the corresponding binary sulfides (e.g., Cu2S, ZnS, and SnS2) and are the values plotted in Figure 1. The final composition of the sulfidized films is also given in the form CuwZnxSnySz in column 1.

Upon sulfidation, Sn loss from the film increased the Cu/Sn ratio to ∼2. However, excess Zn formed ZnS, visible in the SEM in Figure 2c, as small insulating grains: these grains appear brighter than the CZTS grains. Figure 2 (panels d−f) shows SEM images of three films (labeled D, E, and F) all of which were synthesized by sulfidizing initially Sn-poor precursor films (i.e., Cu/Sn > 2, see Table 1). These films were sulfidized with ∼1 mg of Sn present in the sulfidation ampule. Prior to sulfidation, in addition to being Sn-poor, film D was Zn-poor (Cu/Zn > 2), film E had a stoichiometric amount of Zn (Cu/Zn ∼ 2), and film F was Znrich (Cu/Zn < 2). Despite starting out Sn-poor, all three films have a Cu/Sn ratio of ∼2 after sulfidation, indicating that additional Sn was incorporated into the films. It is also clear from the plan-view SEM images that film D is much less porous than film A because there is no material loss from the film D; in fact, there is net material deposition onto film D through reaction of SnS vapor to increase the Sn content. This is in contrast to film A, which lost Sn during sulfidation. Finally, the Zn-rich film F, like C, exhibits small insulating domains of ZnS on the surface. We identified the phases present in films A−F using XRD, in conjunction with Raman scattering (Figures 3 and 4). The XRD patterns from CZTS, Cu2SnS3, and ZnS are virtually indistinguishable, except for two weak diffraction lines at 2θ = 36.9° and 37.9°, which are unique to CZTS.17 The XRD pattern collected from film A shows that these unique CZTS diffraction peaks are absent. That the low intensity identifying peaks for CZTS are absent indicates that a substantial fraction of the film may in fact be Cu2SnS3. We also detect weak diffraction lines that can be assigned to Cu3SnS4. This is consistent with the Raman spectrum, which shows a broad feature in the 290−305 cm−1 range, in addition to a peak at 336 cm−1 and another broad feature at 355 cm−1.17−19 The latter two overlap with Raman scattering peaks for CZTS and ZnS but the feature at 290−305 cm−1 indicates the presence of Cu2SnS3. There is no indication of the formation of binary Cu sulfides in XRD or Raman data. Raman peaks characteristic of Cu3SnS4 are expected at 295, 318, and 248 cm−1. While two of these peaks overlap with those from Cu2SnS3 and CZTS, the broad feature at 318 cm−1 in the Raman spectrum of film A

(Figure 3) is consistent with Cu3SnS4. We conclude that in the absence of adequate Zn to convert all Cu to CZTS, Cu−Sn−S phases form and excess Sn leaves the film, as SnS. The Cu/Sn ratio in both CZTS and Cu2SnS3 is 2, while it is 3 in Cu3SnS4. The measured Cu/Sn ratio in film A is 2.25, close to 2, which indicates that just enough Sn has remained in the film to form mostly Cu2SnS3 and CZTS, the excess evaporating as SnS. The Cu/Sn ratio being slightly higher than 2 is consistent with the formation of small amounts of Cu3SnS4. When the Cu/Zn ratio in the Sn-rich precursor film is decreased to 2 by increasing the amount of Zn (film B), the XRD of the sulfidized film becomes consistent with that expected from phase-pure CZTS: the unique CZTS diffraction peaks at 2θ = 36.9° and 37.9° appear, while the weak Cu3SnS4 diffraction lines disappear. Moreover, the Raman spectrum exhibits peaks characteristics of CZTS at 284 and 338 cm−1. The broad features for Cu2SnS3 (290−305 cm−1) and Cu3SnS4 (318 cm−1) have also disappeared. We conclude that there is now enough Zn in the precursor film to convert all Cu to CZTS. The Cu/Sn ratio is 2.1, very close to stoichiometric and increased significantly from the Cu/Sn ratio of 1 in the precursor film because excess Sn has left the film as SnS. Increasing the Zn content in Sn-rich films further (film C) lead to the formation of ZnS, which is difficult to detect by XRD, and even with Raman spectroscopy, but is most obvious in the SEM images as bright insulating Zn-rich regions (e.g., Figure 2c). The Cu/Sn ratio in film C is 1.9, again very close to stoichiometric and increased from the Cu/Sn ratio of 1.67 in the precursor film. In summary, the Sn-rich precursor films lose Sn such that the final Cu/Sn ratio in the films becomes ∼2 after sulfidation. When these precursor films are Zn-poor, Cu−Sn−S phases form in addition to CZTS. When the precursor films are Zn-rich, ZnS forms in addition to CZTS. If the precursor films are Sn-poor, additional Sn can be incorporated into the film by including Sn in the sulfidation ampule. The solid Sn reacts with S vapor to form SnS vapor, which can then react with the film to incorporate Sn and bring the Cu/Sn ratio down to ∼2. For example, the Cu/Sn ratio in film D decreased from 3.25 before sulfidation to 2.1 after sulfidation when Sn was included in the ampule. Like film A, film D was also Zn-poor. Like film A, the CZTS diffraction 2510

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514

Article

Chemistry of Materials peaks expected at 2θ = 36.9° and 37.9° are undetectable in film D, indicating that only a fraction of the film is CZTS (Figure 3). Like film A, the broad Raman scattering feature in the 290− 305 cm−1 range indicates the presence of Cu2SnS3 in film D. Remarkably, even though films A and D have significantly different concentrations of Sn before sulfidation, the Cu/Sn ratio in both films is ∼2 after sulfidation and both films are a mixture of Cu−Sn−S phases and CZTS. Despite the fact that film D is 66% Cu before sulfidation, there is no indication that the sulfidized film contains any binary Cu sulfides. Apparently SnS can readily react with any Cu−S binaries that may have formed to convert them to Cu−Sn−S. Unlike film A, however, we do not detect any Cu3SnS4, perhaps suggesting that this phase forms when the precursor films are Sn-rich, and/or there is no mechanism to form this phase while incorporating Sn as SnS from the vapor phase. When the amount of Zn in the precursor films is increased (films E and F), the unique CZTS diffraction peaks at 2θ = 36.9° and 37.9° appear, and XRD indicates phase pure CZTS. However, close examination of film F under SEM, like film C, show ZnS precipitates because the precursor film is slightly Zn-rich. The simple view that emerges from our experiments is that if Cu−Zn−Sn alloy precursor films are sulfidized with excess Sn in the system, the amount of Sn in the film is self-regulated such that the Cu/Sn ratio in the sulfidized film is ∼2 after sulfidation. The excess Sn can be in the alloy precursor itself or added into the sulfidation ampule. However, care must be taken so that the sulfidation system is either closed, or excess Sn does not leave the immediate vicinity of the film as SnS vapor. The relative amounts of Cu and Zn in the film determines the secondary phases that may form during the synthesis. When the films are Zn-deficient, Cu−Sn−S phases form in addition to CZTS. When the films are Zn-rich, ZnS forms in addition to CZTS. The most significant difference between the films that were initially Sn-rich and films that were Sn-poor was the microstructure. Films that underwent significant Sn loss tended to have more voids between the grains due to material loss. There is a limit, however, to the Sn deficiency in the precursor film that can be made up by including Sn in the sulfidation ampule. Specifically, the Sn-deficit could not be overcome in Zn-rich precursor films with Cu/Sn ratio greater than 3.6 (blue □ in the lower right corner of Figure 1). Figure S1 of the Supporting Information shows the corresponding SEM image and XRD and Raman spectra for one such film. Sulfidation of this Zn-rich and very Sn-poor precursor resulted in a film comprising CuS, ZnS, and CZTS. The film had a trilayered structure with large (5−10 μm) CuS grains scattered on a layer of ZnS, which in turn formed on top of a layer of CZTS. According to the pseudo binary phase diagram,9 as much as 10% ZnS may be dissolved in Cu2S at 600 °C, the relevant sulfidation temperature. Upon cooling, ZnS solubility decreases sharply and Cu−S and ZnS are expected to segregate at room temperature. We believe that the layered structure results as a consequence of this segregation. Moreover, there is no stable Cu−Zn−S compound that can form. The reaction between SnS and Cu−S may be impeded by the solubility of ZnS within the Cu−S phase, as it has been shown elsewhere that Sn diffusion through ZnS is a particularly slow process.20 The presence of a uniform ZnS layer over the surface of the film would further limit the incorporation of Sn. Finally, the Cu−S phase grows into very large crystals, driven perhaps by the same thermodynamic driving force that segregates Cu−S and ZnS. SnS adsorption, reaction, and diffusion may have been

slowed by this microstructure because these crystals are large and have a small surface-to-volume ratio.20 These observations suggest strategies for synthesizing CZTS films suitable for solar cells. First, our results suggest that sulfidizing precursor films with Zn and Cu compositions reasonably near stoichiometric CZTS in the presence of excess Sn may be an effective strategy to manipulate Cu- and Znrelated defects such as Cu vacancies and antisite defects. If there is excess Sn in the film, or if excess elemental Sn is included in the closed sulfidation ampule, the Cu/Sn ratio will approach 2 while the Cu/Zn ratio will remain essentially unchanged. Second, our experiments suggest that, in order to make dense and void-free films, it is preferable to start with Sn-deficient precursor films and add Sn as SnS via the vapor phase. This approach eliminates a potential void formation mechanism. Synthesizing void-free films for solar cell devices will require further optimization of grain growth by controlling growth times, temperatures, and impurity content, but we have clearly found here that mitigating material loss is also a major consideration. Mechanism of Sn Self-Regulation. There is mounting evidence that during ex situ sulfidation synthesis of CZTS from Cu−Zn−Sn alloys, Cu2SnS3 is an intermediate phase.14,21,22 Thus, it is plausible that after all the Cu is consumed to form the intermediate phase Cu2SnS3, the unreacted Sn leaves the films as SnS vapor. The Cu/Sn ratio is 2 in both CZTS and in Cu2SnS3, meaning that after all the Cu is converted, the Cu/Sn ratio becomes 2. To elucidate the mechanisms that regulate the Sn content in our sulfidized films and to determine the mechanism of Cu2SnS3 formation, we studied the timedependence of the films’ elemental and phase composition during sulfidation at low temperatures, where reaction intermediates can be captured before conversion to CZTS. Specifically, precursor films with nominally identical initial elemental composition were sulfidized at 300 °C. Four ampules were loaded into the furnace, and the furnace was heated to 300 °C at 6.5 °C/min. The first ampule was removed from the furnace as soon as the temperature reached 300 °C (t = 0 min). The second and third ampules were removed 60 and 240 min later (t = 60 min and t = 240 min, respectively). After 480 min (t = 480 min), the furnace was turned off and the last ampule was allowed to cool to room temperature naturally. Figures 5 and 6 summarize the temporal evolution of these films’ atomic and phase composition as a function of sulfidation time. Figure 6 summarizes both the XRD (the colored bars) and the Raman scattering data (phases listed above the bars). The details of the analysis used to construct the phasecomposition bar graph in Figure 6 has been described elsewhere.14 Briefly, the fraction of each phase is determined from the integrated intensity of an XRD peak unique to that phase, after normalizing these values by the film thickness, Xray exposure time, and the expected relative intensity for the peak. The phase fraction is determined from the ratio of this normalized value for each phase to the sum of all values for all the phases detected in the film by XRD. This gives a reasonable estimate of the phase composition, provided that the films are not textured. Indeed, sulfidized films were not textured: the 2D area detector images of the XRD from the sulfidized films showed uniform intensity arcs, as expected for diffraction from randomly oriented grains. The individual XRD and Raman data used to construct Figure 6 are included in the Supporting Information as Figures S2 and S3, respectively. 2511

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514

Article

Chemistry of Materials

Figure 7. Summary of important reactions leading to the formation of CZTS from the precursor metal alloy films. See text for additional discussion. Figure 5. Atomic composition of copper−zinc−tin alloy films sulfidized at 300 °C as a function of sulfidation time.

basis of XRD, the rest of the film appeared to be entirely alloyed and comprised of Cu6.26Sn5 and Cu5Zn8. This alloying, and low extent of sulfidation, may be described via the overall reactions 5Sn(s) + 6.26Cu(s) → Cu6.26Sn5(s)

(1)

8Zn(s) + 5Cu(s) → Cu5Zn8(s)

(2)

and

Zn(s) + S(g) → ZnS(s)

(3)

The film sulfidized for 60 min at 300 °C is nearly completely sulfidized, and the S composition has reached 47%. The cation composition has also changed dramatically: the Cu/Sn ratio has increased significantly to 6, and the Cu/Zn ratio has increased to 3.3. While nearly all the metals have been sulfidized, XRD data shows the presence of a small fraction of elemental Sn and Cu5Zn8. This elemental Sn is textured and has been omitted from Figure 6, but, based on the composition analysis, we believe elemental Sn comprises a rather small amount of the film. A weak shoulder remains at 2θ = 43.3°, which we attribute to the most intense diffraction from Cu5Zn8, though this diffraction overlaps with a lower intensity peak from CuS (2θ = 43.10°; relative intensity of 6%). If we assume that this peak is due entirely to Cu5Zn8, included in Figure 6, its contribution (1.6%) would be too small to see. Most importantly, XRD shows the presence of SnS, CuS, as well as diffraction lines that can be assigned to CZTS, Cu2SnS3, or ZnS. These last three phases are indistinguishable by XRD and are represented by the blue bar labeled ∑CZTS/ZnS/Cu2SnS3.14 Raman scattering confirms the presence of all three, although the fraction of each phase cannot be easily quantified from the Raman spectrum. These observations can be understood by considering the reactions that must occur to convert the metal precursor alloy film to the thermodynamically stable metal sulfides. Elemental Cu, Zn, and Sn phases can sulfidize directly to form binary sulfides via

Figure 6. Phase composition of Cu−Zn−Sn alloy films sulfidized at 300 °C as a function of sulfidation time. The heights of the colored bars represent the fraction of the phases present in the film. The phases listed above each bar are those identified by Raman spectroscopy.

The precursor film is Sn-rich and slightly Zn-poor such that Cu/Sn = 1.6 and Cu/Zn = 2.3. This film consisted of elemental Zn, Cu, and Sn phases, as well as the Cu6.26Sn5 alloy phase. Note that the phase composition of the precursor film is not plotted in Figure 6 because Zn and Cu diffractions overlap, and the film is textured.14 Figure 7 summarizes the reactions that lead to the formation of CZTS from the precursor metal alloy films. We deduced these reaction pathways by analyzing the time dependence of the films’ elemental and phase composition during sulfidation at low temperature (300 °C). We now explain our rationale and describe these reactions, and the evidence for them, in detail. The elemental composition of the film removed immediately after the furnace has reached 300 °C (t = 0) is nearly identical to that of the precursor film, with very little S (∼7%) incorporation. Similar to films sulfidized at 200 °C and discussed elsewhere,14 we found by SEM and EDS small insulating grains on the surface of the film, which we attribute to the early stages of ZnS formation (see Supporting Information, figure S4); the volume fraction of these ZnS grains is below the detection limits of Raman and XRD. On the

Sn(S) + S(g) → SnS(s) ⇄ SnS(g)

(4)

Cu(s) + S(g) → CuS(s)

(5)

and Zn(s) + S(s) → ZnS(s)

(6)

The presence of significant CuS and loss of Sn from the film suggest that in the presence of the S vapor, the Cu−Zn and 2512

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514

Article

Chemistry of Materials Cu−Sn alloys segregate as the elements comprising these alloys sulfidize (Figure 7). This is expected for Cu5Zn8 via Cu5Zn8(s) → 5Cu(s) + 8Zn(s)

SnS may adsorb on the CuS crystals to react or condense onto solid SnS and then react with CuS via 2CuS(s) + SnS(s) → Cu 2SnS3(s)

(7)

The reaction in the solid phase would shift the solid−vapor equilibrium (SnS(g) ⇌ SnS(s)) toward the solid phase to incorporate more of the SnS vapor into the film. Our experiments do not reveal whether a heterogeneous reaction takes place or whether SnS condenses on the surface of the film (SnS domains) and then undergoes a solid-state reaction with CuS to form Cu2SnS3, or possibly both. Either way, SnS lost from the film returns to the film in a closed system. The Zn loss from the film after 240 min at 300 °C is significant: the Cu/Zn ratio in the sulfidized film has risen to 6.6, indicating that Zn continued to sublime. We infer that during the 240 min period, there was still unsulfidized Zn in the film, and that at 300 °C, the Zn sublimation rate is comparable to the ZnS formation rate. The ZnS that does remain within the film interdiffuses with Cu2SnS3 to form CZTS via

(because there is no stable Cu−Zn−S phase), but apparently the Cu6.26Sn5 alloy also dissociates into Cu and Sn via Cu6.26Sn5(s) → 6.26Cu(s) + 5Sn(S)

(8)

Sn melts at 231 °C and can sulfidize via reaction 4. According to the Cu−Sn phase diagram, the Cu phase will be Cu-rich but can still include up to 4% Sn at 300 °C and as much as 9.5% Sn at 600 °C. Some of the Sn remains unsulfidized and crystallizes upon cooling to form the textured elemental Sn observed in the XRD. While the reported Gibbs free energy of formation for SnS, ΔGf,SnS, is lower than that for CuS, ΔGf,CuS, the formation of CuS appears to be more facile as we found no evidence of elemental Cu remaining in the film, even when we detected elemental Sn. Similar observations have been reported elsewhere.23 This may possibly be due to diffusion limitations. Within Cu6.26.Sn5, Cu diffusion is faster.24 The Cu may diffuse to the surface of the film where it readily reacts with the S vapor, while Sn remains within the film and melts. Formation of a SnS crust on liquid Sn would also limit sulfidation of Sn. A small amount of Cu5Zn8 alloy remaining in the film indicates that it does not completely dissociate within the 60 min. The Cu5Zn8 alloy does have a lower ΔGf at 300 °C compared to the Cu6.26Sn5 alloy and is more stable. Again, anion diffusion through the sulfide phases may also be limiting the dissociation and sulfidation of Cu5Zn8 as well. (See the Supporting Information Table S1 for relevant Gibbs free energies of formation.) The changes in the elemental composition of the film are consistent with Zn and SnS intermediates, which have high vapor pressures and can leave the film (Figure 7) via Zn(s) → Zn(g)

Cu 2SnS3(s) + ZnS(s) → Cu 2ZnSnS4(s)

(12)

Within the accuracy of the EDS, the elemental composition of the film sulfidized at 300 °C for t = 480 min is similar to that of the film sulfidized for t = 240 min. The film is comprised entirely of sulfides, although the fraction of ∑CZTS/ZnS/ Cu2SnS3 has increased at the expense of CuS and SnS. The increase in the ∑CZTS/ZnS/Cu2SnS3 fraction is attributable solely to the increase in the Cu2SnS3 phase because there is no increase in the amount of Zn within the film to make ZnS or CZTS. Hence, after t = 240 min, the film is slowly converting to Cu2SnS3 via the reaction of CuS with SnS (Figure 7). Comparison of the compositions of the films before and after sulfidation at 600 °C shows that Cu/Zn does not change significantly when the heating rate is 6.5 °C/min. It appears that this heating rate is fast enough to avoid large amounts of Zn sublimation during heating to temperatures greater than 300 °C. These experiments demonstrate the importance of the vapor phase composition over the film during sulfidation of Cu−Zn− Sn alloys to form CZTS. The favorable Gibbs energy of formation of CuS and SnS drives the segregation of the Cu6.26Sn5 alloy and subsequent rapid sulfidation of Cu and Sn to form the binary sulfides. Moreover, there is not enough Cu in the Cu6.26Sn5 alone to form single-phase Cu2SnS3. Consequently, CuS and SnS remain as intermediates. The formation of significant amounts of Cu2SnS3 relies on the dissociation of Cu5Zn8 for more Cu, which appears to be a slower process. This gives ample time for a considerable fraction of the SnS within the film to evaporate. Thus, before CZTS is formed, the majority of the Sn may be lost to the vapor phase. As more CuS is formed, SnS vapor and/or solid react with CuS to form Cu2SnS3 (Figure 7). This is the same mechanism by which the Sn content of an initially Sn-deficient film may be increased by including elemental Sn in the sulfidation ampule. This pathway is the main factor behind the self-regulating nature of Sn: the Sn concentration in the film is governed by the availability of CuS to form Cu2SnS3. While, we cannot determine the rates at which Sn leaves or returns to the films as SnS vapor, we can infer from the final composition that just enough Sn eventually ends up in the films to form Cu2ZnSnS4 and Cu2SnS3 (phases with Cu/Sn = 2). It appears that Sn can leave the film as SnS if CuS has not formed yet or is slow to form. However, Sn can then reincorporate into the film

(9)

and reaction 4. The vapor pressure of Zn at 300 °C is significant (∼1.4 mTorr), and some Zn evaporates (Figure 7) before it is converted to low vapor pressure ZnS. This evaporation is responsible for the rise in the Cu/Zn ratio. In most high temperature sulfidation (e.g., 500−600 °C) treatments, the film is heated quickly through 300 °C, the temperature at which sulfidation and evaporation may be in competition to determine the Zn content in the film. If the sulfidation of Zn is slow, which is the case at low temperatures, the Zn intermediate may evaporate, and Zn may be lost from the film. This observation underscores the importance of the heating rates in sulfidation. If the heating rate is slow and the film is exposed to a long enough time period at intermediate temperatures (e.g., ∼300 °C), Zn may leave the film. After t = 240 min at 300 °C, the S concentration reaches 50%. Remarkably, the Cu/Sn ratio is now 2.75. While still higher than the precursor film, it is significantly lower than the film that was sulfidized for 60 min. Surprisingly, the Sn that left the film during the first 60 min has now been reincorporated. The film is comprised entirely of sulfide phases, including SnS, CuS, ZnS, Cu2SnS3, and CZTS. In particular, the fraction of ZnS, Cu2SnS3, and/or CZTS has increased (Figure 6). This suggests that SnS from the vapor phase reacts with CuS (Figure 7) to form Cu2SnS3 via 2CuS(s) + SnS(g) → Cu 2SnS3(s)

(11)

(10) 2513

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514

Article

Chemistry of Materials

C.L.). Parts of this work were carried out in the Characterization Facility, University of Minnesota, which receives partial support from NSF through the MRSEC program.

via reactions 10 and 11 after copper sulfide has formed. The kinetics of SnS evaporation and reincorporation during sulfidation is complex because of the competition between reactions 4 and 11. Previous studies hypothesized that the Sn loss from the films occurs after CZTS formation and as a consequence of decomposition via 2Cu 2ZnSnS4 → 2Cu 2S + 2ZnS + 2SnS + S2

■

(13)

In contrast, our experiments show that that Sn loss, as well as incorporation into the film, occurs during the formation of CZTS. The sulfur pressure in our closed system at 600 °C is ∼36 Torr, well above the necessary partial pressure for CZTS stability.11 If this reaction were to take place in an open system, or a poorly controlled pseudoclosed system (e.g., the commonly used loosely sealed-graphite-box method), the film would be unable to recover from the initial Sn loss via the mechanisms described here.

■

CONCLUSION During synthesis of CZTS films via ex situ sulfidation of Cu− Zn−Sn alloy films in a closed system, the final Sn content in the sulfidized film is self-regulating. If the metal alloy precursor film is Sn rich, the Sn concentration increases until Cu/Sn ∼ 2. If the metal precursor film is Sn-deficient, the Sn content can be increased by including elemental Sn in the sulfidation ampule. Elemental Sn readily sulfidizes to generate SnS vapor which then incorporates into the film until the Cu/Sn ratio becomes ∼2. Material loss from Sn-rich precursor alloy films results in CZTS films with voids, while sulfidation of Sn-poor precursor films with Sn added into the sulfidation ampule eliminates the voids and leads to less porous CZTS films. A detailed investigation of the time dependence of sulfidation at 300 °C showed that SnS is lost from the film during the formation of CZTS, not as a result of its decomposition. A significant fraction of the Sn can evaporate from the film as SnS, but SnS is later reincorporated into the film via reactions with CuS to form Cu2SnS3. The reactions between SnS vapor and Cu2SnS3 are also the mechanism by which Sn can be added to Sndeficient films. In summary, these findings significantly inform synthesis approaches for optimization of stoichiometry and microstructure for CZTS solar cells.

■

ASSOCIATED CONTENT

S Supporting Information *

SEM images; XRD; Raman spectroscopy; and thermodynamic data for the formation of binary sulfides and alloys. This material is available free of charge via the Internet at http:// pubs.acs.org.

■

REFERENCES

(1) Norris, D. J.; Aydil, E. S. Science 2012, 338, 625−626. (2) Wadia, C.; Alivisatos, A. P.; Kammen, D. M. Environ. Sci. Technol. 2009, 43, 2072−2077. (3) Mitzi, D. B.; Gunawan, O.; Todorov, T. K.; Wang, K.; Guha, S. Sol. Energy Mater. Sol. Cells 2011, 95, 1421. (4) Ito, K.; Nakazawa, T. Jpn. J. Appl. Phys. 1988, 27, 2094. (5) Nakayama, N.; Ito, K. Appl. Surf. Sci. 1996, 92, 171−175. (6) Wang, W.; Winkler, M. T.; Gunawan, O.; Gokmen, T.; Todorov, T. K.; Zhu, Y.; Mitzi, D. B. Adv. Energy Mater. 2014, 4, 1301465. (7) Chirilă, A.; Reinhard, P.; Pianezzi, F.; Bloesch, P.; Uhl, A. R.; Fella, C.; Kranz, L.; Keller, D.; Gretener, C.; Hagendorfer, H.; Jaeger, D.; Erni, R.; Nishiwaki, S.; Buecheler, S.; Tiwari, A. N. Nat. Mater. 2013, 12, 1107−1111. (8) Katagiri, H.; Jimbo, K.; Tahara, M. MRS Online Proc. Libr. 2009, 1165. (9) Olekseyuk, I.; Dudchak, I. V.; Piskach, L. V. J. Alloys Compd. 2004, 368, 135−143. (10) Weber, A.; Mainz, R.; Schock, H. W. J. Appl. Phys. 2010, 107, 013516. (11) Scragg, J. J.; Ericson, T.; Kubart, T.; Edoff, M.; PlatzerBjörkman, C. Chem. Mater. 2011, 23, 4625−4633. (12) Redinger, A.; Berg, D. M.; Dale, P. J.; Siebentritt, S. J. Am. Chem. Soc. 2011, 133, 3320−3323. (13) Johnson, M.; Baryshev, S. V.; Thimsen, E.; Manno, M.; Zhang, X.; Veryovkin, I. V.; Leighton, C.; Aydil, E. S. Energy Environ. Sci. 2014, 7, 1931−1938. (14) Johnson, M.; Manno, M.; Zhang, X.; Leighton, C.; Aydil, E. S. J. Vac. Sci. Technol., A 2014, 32, 061203. (15) Mousel, M.; Schwarz, T.; Djemour, R.; Weiss, T. P.; Sendler, J.; Malaquias, J. C.; Redinger, A.; Cojocaru-Mirédin, O.; Choi, P.; Siebentritt, S. Adv. Energy Mater. 2014, 4, 1300543. (16) Mousel, M.; Redinger, A.; Djemour, R.; Arasimowicz, M.; Valle, N.; Dale, P.; Siebentritt, S. Thin Solid Films 2013, 535, 83−87. (17) Cheng, A.-J.; Manno, M.; Khare, A.; Leighton, C.; Campbell, S. A.; Aydil, E. S. J. Vac. Sci. Technol., A 2011, 29, 051203. (18) Fernandes, P. A.; Salome, P. M. P.; da Cunha, A. F. J. Phys. D 2010, 43, 215403. (19) Berg, D. M.; Djemour, R.; Gutay, L.; Siebentritt, S.; Dale, P. J.; Fotane, X.; Izquierdo-Roca, V.; Perez-Rodriguez, A. Appl. Phys. Lett. 2012, 100, 192103. (20) Thimsen, E.; Baryshev, S. V.; Martinson, A. B. F.; Elam, J. W.; Veryovkin, I. V.; Pellin, M. J. Chem. Mater. 2013, 25, 313−319. (21) Schurr, R.; Hölzing, A.; Jost, S.; Hock, R.; Voβ, T.; Schulze, J.; Kirbs, A.; Ennaoui, A.; Lux-Steiner, M.; Weber, A. Thin Solid Films 2009, 517, 2465−2468. (22) Han, J.; Shin, S. W.; Gang, M. G.; Kim, J. H.; Lee, J. Y. Nanotechnology 2013, 24, 095706. (23) Yin, X.; Tang, C.; Sun, L.; Shen, Z.; Gong, H. Chem. Mater. 2014, 26, 2005−2014. (24) Kumar, S.; Handwerker, C. A.; Dayananda, M. A. J. Phase Equilib. Diffus. 2011, 32, 309−319.

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Science Foundation (NSF) through Award CBET-0931145 and in part by the NSF MRSEC under DMR-0819885 and DMR-1420013 (M.M.), as well as UMN IREE (C.W.), and DMR-1309642 (X.Z. and 2514

DOI: 10.1021/acs.chemmater.5b00108 Chem. Mater. 2015, 27, 2507−2514