Report

Composition-matched molecular “solders” for semiconductors

See allHide authors and affiliations

Science  23 Jan 2015:
Vol. 347, Issue 6220, pp. 425-428
DOI: 10.1126/science.1260501

Soldering semiconductor nanoparticles

The optical and electronic properties of semiconductor nanoparticles can be tuned through changes in their size and composition. However, poor contact between interfaces can degrade nanoparticle performance in devices. Dolzhnikov et al. report the synthesis of a gel-like “solder” for metal chalcogenide nanoparticles, such as cadmium selenide and lead telluride, by cross-linking molecular wires of these materials.

Science, this issue p. 425

Abstract

We propose a general strategy to synthesize largely unexplored soluble chalcogenidometallates of cadmium, lead, and bismuth. These compounds can be used as “solders” for semiconductors widely used in photovoltaics and thermoelectrics. The addition of solder helped to bond crystal surfaces and link nano- or mesoscale particles together. For example, CdSe nanocrystals with Na2Cd2Se3 solder was used as a soluble precursor for CdSe films with electron mobilities exceeding 300 square centimeters per volt-second. CdTe, PbTe, and Bi2Te3 powders were molded into various shapes in the presence of a small additive of composition-matched chalcogenidometallate or chalcogel, thus opening new design spaces for semiconductor technologies.

Two pieces of metal, such as copper wires, can be mechanically and electrically connected by using soldering, a process in which metallic items are joined together by introducing a filler metal (solder) with lower melting point (1). Soldering and related processes (such as brazing) are very useful in many areas, from microelectronics to plumbing. Unlike the case for metals, there are no established methods for joining semiconductor pieces under mild conditions without disrupting semiconducting properties at the joint. A “semiconductor solder” could affect numerous fields, including printable electronics, photovoltaics, and thermoelectrics, because it would add new techniques for device integration.

The electronic properties of semiconductor interfaces are much more sensitive to impurities and structural defects than are metals. Metal-metal contacts always show ohmic behavior, whereas semiconductor-semiconductor interfaces are more complex. Misalignment of the Fermi energy levels or trapping charge carriers at the interface creates a Schottky barrier (2). An ideal “semiconductor solder” should be constituted of a precursor that, upon mild heat treatment, forms a semiconducting material structurally and compositionally matched to the bonded parts. In this study, we designed chalcogenidometallate “solders” for technologically important II-VI, IV-VI, and V-VI semiconductors, including CdSe, CdTe, HgxCd1−xTe, PbTe, (BixSb1−x)2Te3, and Bi2Te3.

The use of chalcogenidometallates as soluble precursors for inorganic semiconductors was first introduced by Mitzi et al., who used the ability of N2H4 to dissolve metal chalcogenides in the presence of elemental chalcogens (3), forming soluble hydrazinium chalcogenidometallates. These species cleanly decompose back into semiconducting metal chalcogenides upon heating. This simple approach works very well for the chalcogenides of Cu, Ga, Ge, In, Sb, Sn, and mixed phases such as Cu(In1−xGax)Se2 and Cu2ZnSn(S,Se)4 (46) but showed no success for Cd-, Pb-, and Bi-chalcogenides (7, 8), which are among the most widely used binary and ternary semiconductors. Here, we propose the origin of this limitation and a general solution, which substantially expands the list of solution-processed semiconductors.

In the above process, the formation of a soluble chalcogenidometallate begins with the reduction of an elemental chalcogen (Ch = S, Se, or Te) by N2H4 following Eq. 1 (3, 4)2nCh + 5N2H4 → 4N2H5+ + 2Chn2− + N2 (1)The Chn2− reacts with electron-deficient metal centers at the surface of a solid metal chalcogenide (MxChy), generating soluble MxCh(y+m)2m chalcogenidometallate ions:2nMxChy + 2mChn2− + 5(n − 1)mN2H4 → 2nMxCh(y+m)2m + (n − 1)mN2 + 4(n − 1)mN2H5+ (2)The progression of reaction in Eq. 2 is determined by the balance between the lattice energy of the metal chalcogenide and the free energy of formation and solvation of the chalcogenidometallate complex. Reactivity of Chn2− is highest at n = 1 and decreases as n increases. The reducing potential of N2H4 in the reaction in Eq. 1 is not sufficient to bring Se and Te into their most reactive Ch2− state. X-ray absorption near-edge spectroscopy (XANES) measurements show that Eq. 1 generates soluble Se or Te species with an oxidation state near zero, which is equivalent to large n (fig. S1) (9). To increase the driving force for Eq. 2, we used A2Se or A2Te (A = Na, K, and Cs) instead of elemental chalcogens and found that for n = 1, Eq. 2 proceeds smoothly for a number of metal chalcogenides previously considered unreactive, including CdSe, CdTe, PbS, PbSe, PbTe, Bi2S3, Bi2Se3, Bi2Te3, (BixSb1−x)2Te3, and some others, which are summarized in (9). These reactions could also be conveniently carried out in one pot—for example, by adding alkali metal hydride to the stoichiometric mixture of CdTe and Te in N2H4: CdTe + Te + 2NaH → Na2CdTe2 + gaseous products (9).

The structure of K2CdTe2·2N2H4 crystallized from the reaction mixture is shown in Fig. 1A. It contains molecular chains built of edge-sharing, slightly distorted [CdTe4] tetrahedrons with Te-Cd-Te angles of 99.4° and 119.1° (9). The bonds between Cd and Te are 2.81 and 2.83 Å in length, which are similar to that of zinc blende CdTe (2.81 Å). This structural motif with one-dimensional (1D) “molecular wires” has apparently not been reported for chalcogenidocadmates (10, 11). In the crystal lattice, [CdTe22−] chains are separated with N2H4 molecules, which leads to the facile solubility of the compound in various solvents, up to 600 mg/mL in N2H4 (Fig. 1B). The solubility of ditellurometallates could be further tailored by means of cation exchange (9). For example, Na+ or K+ was exchanged for alkylammonium cations such as didodecyldimethyl ammonium (DDA+) or tetraethyl ammonium (NEt4+), providing good solubility of DDA2CdTe2 in toluene and of (NEt4)2CdTe2 in acetonitrile (CH3CN) or N,N-dimethylfomamide (DMF), respectively (Fig. 1B). Exchange of K+ with hydrazinium cations (9) could be used to facilitate thermal decomposition of the resulting hydrazinium salts.

Fig. 1 Structure and properties of CdTe22– compounds.

(A) Crystal structure of K2CdTe2·2N2H4 determined by using single crystal x-ray diffraction. Hydrogen atoms are not shown for clarity. (B) Solubility of Na2CdTe2 in different solvents can be tuned by cation exchange from sodium (Na+) to tetratethylammonium (NEt4+) or didodecyldimethylammonium (DDA+). Polymeric [CdTe22−] ions can be cross-linked with Cd2+ ions, forming stable “CdTe-gel.” [CdTe22−] ions can serve as ligands for luminescent CdTe quantum dots (QDs). (C) Depending on solvent polarity and coordinating ability, [CdTe22−] can exist in two equilibrium forms.

X-ray diffraction and extended x-ray absorption fine structure (EXAFS) studies (figs S2 and S3 and tables S1 and S2) (9) of ditellurocadmates suggest that, in solution, [CdTe22−] chains can exist in the equilibrium forms outlined in Fig. 1C. In a strongly coordinating solvent such as N2H4, the equilibrium shifted toward the [CdTe(μ-Te)]n2n structure, where each Cd atom has 3 Te neighbors and Te has on average 1.5 Cd atoms in the first coordination shell. On the other hand, weakly coordinating solvents such as CH3CN shifted the equilibrium toward the structure with two bridging Te atoms per [CdTe2]2− unit. These variations in coordination environment were reflected by fully reversible shifts of the absorption bands in mixtures of N2H4 and CH3CN with various solvent ratios (fig. S4) (9).

Reacting Na2CdTe2 with a stoichiometric amount of CdCl2 in N2H4 formed a white amorphous gel, further referred to as “CdTe-gel” (Fig. 1B), which is an important addition to the chalcogel family (12). The gel was easily separated from its NaCl by-product by means of centrifugation and then redispersed in N-methylformamide (NMF) or N2H4, forming a stable gel of [CdTe2]2− polyions cross-linked with Cd2+. Upon heating above 250°C, the CdTe-gel transformed into crystalline CdTe, as shown in fig. S5 (9).

[CdTe22−] ions, similar to other chalcogenidometallates (13), were used as capping ligands for colloidal nanocrystals (NCs). [CdTe2]2−-capped CdTe quantum dots showed bright band-edge photoluminescence (Fig. 1B and fig. S6) (9). The presence of strong emission after the ligand exchange suggests that [CdTe2]2− ions did not introduce fast recombination centers at the CdTe surface.

Following similar approaches, we synthesized and characterized a series of soluble Cd-, Pb-, and Bi chalcogenidometallates and chalcogels (9). These chalcogenidometallates could be dispersed in polar solvents, including N2H4, ethylenediamine, NMF, dimethyl sulfoxide (DMSO), formamide (FA), DMF, and N2H4/water mixtures (9). Such chemical versatility allowed for the design of compositionally matched “solders” for practical semiconductors. These solders were used in various capacities—for example, to consolidate nano- or mesoscopic grains or to bond single-crystal wafers. In one example, we applied Na2Cd2Se3 solder to CdSe NCs: Colloidal 4.7-nm CdSe NCs capped with n-octadecylphosphonic acid (ODPA) were ligand-exchanged with [Cd2Se3]2− ions (9). The infrared (IR) absorption spectra confirmed a complete replacement of original ODPA ligands with [Cd2Se3]2− ions (fig. S7) (9). A stable colloidal solution of [Cd2Se3]2−-capped CdSe NCs was obtained, which preserved the original quantum dot excitonic features, with a small red shift of the first absorption peak (Fig. 2A). This shift was likely caused by the expansion of wave functions into the compositionally matched ligand shell. The solution of [Cd2Se3]2−-capped CdSe NCs was spin-coated on heavily doped Si substrates with a ~10-nm-thick ZrOx gate dielectric and annealed under nitrogen at various temperatures for 30 min. Mobility measurements were carried out in a standard field-effect transistor (FET) geometry (Fig. 2B) (14, 15). Even annealing at only 250°C was sufficient to achieve excellent output and transfer characteristics (Fig. 2, C and D, and fig. S8) (9) with an electron mobility of 210 cm2/V s, which exceeds the best reported values for any solution-processed inorganic semiconductor (table S3) (9). Annealing NC films at 300°C results in devices with FET mobility above 300 cm2/V s (fig. S9) (9). In control experiments, we found that FETs with channels made of pure CdSe NCs, Na2Cd2Se3, or (N2H5)2Cd2Se3 showed mobilities of <4 cm2/V s (figs. S10 to S12) (9).

Fig. 2 Transport properties of CdSe NCs after ligand exchange.

(A) Absorption spectra of 4.7-nm CdSe NCs capped with ODPA and [Cd2Se3]2− ligands. (Inset) Photograph of the solution of [Cd2Se3]2−-capped CdSe NCs. (B) Scheme of a FET device with 10-nm ZrO2 gate dielectric and Al source and drain electrodes. (C) Transfer characteristics and (D) output characteristics of an FET with a channel made of spin-coated 4.7 nm [Cd2Se3]2−-capped CdSe NCs annealed at 250°C for 30 min. FET channel width and length are 1500 and 30 μm, respectively. The inset in (C) shows the linear and saturation regime electron mobility extracted from FET transfer characteristics (9).

In recent years, various organic and inorganic ligands (such as 1,4-phenylenediamine, In2Se42−, and SCN) have been used as “electronic linkers” for CdSe NCs (1620). The highest reported FET mobilities of annealed CdSe NC samples were on the order of 35 cm2/V s (table S3) (9), which is ~5% of the mobility of CdSe single crystals at 300 K (21). In our case, [Cd2Se3]2− ions worked as a compositionally matched solder that improved carrier transport up to almost half of the single-crystal value, which is surprising given that the Scherrer size of CdSe grains was about 15 and 22 nm for samples annealed at 250°C and 300°C, respectively (fig. S13) (9). This grain size was also confirmed with transmission electron microscopy (TEM) measurements of annealed films (fig. S14) (9). Various chemical treatments have been successfully used to passivate interfaces and control doping in polycrystalline semiconductors and NC solids (22, 23). At the same time, it has proven difficult to achieve high mobility through sintering (24) or oriented attachment of NCs (25). Recent studies show that simply increasing CdSe grain size does not improve FET mobility (26). We therefore attribute this high mobility to exceptional electronic transparency of the grain boundaries, which did not introduce appreciable transport bottlenecks.

The molecular solders functioned not only for nanoscale but also for mesoscale grains. For example, PbTe, Bi2Te3, or CdTe films were obtained by soldering microparticles, obtained by ball-milling of ingots, by using Na2PbTe2, (N2H5)4Bi2Te5, or CdTe-gel as solders, respectively (Fig. 3, B to D, and fig. S15) (9). In the first case, we applied the sodium salt because Na+ is a commonly used p-type dopant for PbTe. A suspension of PbTe microparticles with a small amount [2 to 5 weight percent (wt %)] of Na2PbTe2 solder was drop-cast on a hydrophilized glass substrate and annealed at 500°C for 30 min, resulting in a continuous thin film with a uniform thickness of ~7 μm (9) with well-connected dense grains (Fig. 3B). The film exhibited a Hall mobility of 58 cm2/V s, which is comparable with that (~100 cm2/ V s) of optimized Na-doped PbTe thermoelectric pellets made at a similar temperature by means of spark-plasma sintering or hot pressing (27, 28). In control experiments, no continuous films were obtained from suspensions of PbTe and other microparticles without addition of solder. After annealing under identical conditions, the material remained powdery (Fig. 3A). The above examples show that chalcogenidometallates and chalcogels, mixed with nano- and microparticles, offer a viable path toward solution processing of thin-film semiconductors.

Fig. 3 Soldering various materials with novel chalcogenidometallates.

Scanning electron microscopy images of (A) ball-milled PbTe microparticles annealed at 500°C for 30 min. (B) Film of similar PbTe microparticles annealed in the presence of 5 wt % Na2PbTe2. (C) Film made of Bi2Te3 microparticles annealed in the presence of 5 wt % (N2H5)4Bi2Te5. (D) Film made of CdTe NCs annealed in the presence of 20 wt % “CdTe-gel.” The films were annealed at 500°C, 400°C, and 450°C for 30 min, respectively. (Insets) Film optical micrographs. (E) 3D blocks of PbTe, Bi2Te3, and CdTe cast by using graphite molds of various shapes from microparticles mixed with corresponding “solders”: Na2PbTe2 for PbTe, Na4Bi2Te5 for Bi2Te3, and Na2CdTe2 for CdTe. (F) A photograph showing two CdTe crystals with mechanically polished (111) faces soldered with Na2CdTe2 at 500°C.

The solder effect was used to consolidate microparticles (such as PbTe, Bi2Te3, Bi0.5Sb1.5Te3, and CdTe) into 3D blocks cast with a mold (Fig. 3E). The addition of a few weight percent of molecular or gel solder helped connect the microparticles and allowed the casting of solid pellets in graphite molds with no external pressure, similar to molding metal particles (29). Disk-, triangle-, square-, and bar-shaped blocks were obtained via annealing various combinations of nano- and microparticles (PbTe, Bi2Te3, Bi0.5Sb1.5Te3, or CdTe) with corresponding compositionally matched molecular chalcogenidometallates or chalcogels (9). In control experiments without solders, the microparticles molded under the same conditions remained powdery (fig. S16) (9). Comparison of Fig. 3, A and B, shows that the addition of molecular solder facilitates sintering and grain growth compared with pure components. The microscopic origin of this effect is likely related to improved mass transport along and across the grain boundaries (30).

Chalcogenidometallate solders were also used for bonding metal chalcogenide crystals under modest pressures and temperatures. Shown in Fig. 3F is an example in which two polished (111) CdTe surfaces (with an area of 0.25 cm2) were bonded at ~0.5 MPa and 500°C by using a minuscule amount (2 μL) of Na2CdTe2 solution in N2H4 (9). No physical bonding between CdTe crystals occurred in control experiments with similarly prepared crystal surfaces but no solder.

The above examples reveal the potential of bottom-up assembly of inorganic semiconductors with compositionally matched molecular solders. This approach enables the highest electron mobility achieved to date for solution-processed semiconductors, opening new opportunities for printable electronics and optoelectronics, and paves the way to future technological developments such as moldable or 3D-printable thermoelectric materials.

Supporting Online Material

www.sciencemag.org/content/347/6220/425/suppl/DC1

Materials and Methods

Figs. S1 to S16

Tables S1 to S3

References (3148)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. All software and sources of scattering factors are contained in the SHELXTL (version 5.1) program library (G. Sheldrick, Bruker Analytical X-ray Systems, Madison, Wisconsin).
  3. Acknowledgments: We thank S. Kwon and V. Zyryanov for help with EXAFS measurements and N. James for reading the manuscript. The work was supported by the II-VI Foundation, U.S. Department of Energy (DOE) SunShot program under award DE-EE0005312 and by NSF under award DMR-1310398. D.V.T. also thanks the Keck Foundation. This work used facilities supported by the NSF Materials Research Science and Engineering Center Program under awards DMR 08-20054 and DMR 14-20709. Use of the Center for Nanoscale Materials and Advanced Photon Source was supported by the DOE, Office of Science, Office of Basic Energy Sciences, under contract DE-AC02-06CH11357. MRCAT operations are supported by the DOE and MRCAT host institutions.
View Abstract

Subjects

Navigate This Article