Effect of Different Dopants on Dislocation Defects in VGF-InP and Inhibition Mechanisms Authors: Hua Wei, Hongfei Gu, Bin Yang, Xiaoda Ye, Xingkai Zhao, Hanbao Liu, Feng Qiu, Jie Yang, Tinglong Liu Published: 2026 ABSTRACT In this study, we integrate first-principles calculations with experimental statistics to systematically elucidate the fundamentally distinct mechanisms of S and Fe doping on dislocation defects in vertical gradient freeze (VGF)-grown indium phosphide (InP) crystals. Atomic-scale analysis reveals, for the first time, that S substitution at P sites forms shorter and stronger In–S bonds (bond length shortened from 2.54 to 2.42 Å), inducing significant lattice hardening (elastic modulus increased by ∼15%). The resultant local strain field effectively pins dislocations, significantly lowering the dislocation density. Experiments confirm the dislocation density hierarchy: heavily S-doped < lightly S-doped < nondoped. Conversely, Fe doping reduces the material’s yield strength and stacking fault energy. Furthermore, due to its extremely low solid solubility, Fe leads to supersaturation and precipitation of FeP2 secondary phases during VGF. Crystal quality characterization (X-ray diffraction (XRD) and Raman) indicates that high S-doped samples exhibit superior crystalline integrity, while Fe-doped samples (especially the tail) show the poorest quality. Based on these mechanisms, we optimized the VGF + VB process to control the solid–liquid interface and solute segregation, effectively suppressing Fe precipitation and improving dislocation distribution and electrical uniformity. This research provides a crucial theoretical foundation and a clear process pathway for the rational design of InP single crystals with both excellent electrical properties and ultrahigh structural integrity. Introduction Indium phosphide (InP), as a core substrate material for second-generation semiconductors, holds an irreplaceable position in cutting-edge fields such as high-speed optoelectronic communication devices, high-frequency microwaves, and terahertz technology due to its outstanding properties, including high electron mobility, high saturated electron drift velocity, and direct band gap. − Among various crystal growth techniques, the vertical gradient freeze (VGF) method has become a mainstream technology for producing large-diameter, high-quality InP single crystals due to its stable thermal field, low thermal stress, and relatively low energy consumption. , However, defects like dislocations, induced during growth by factors such as thermal stress and constitutional supercooling, remain a key bottleneck limiting the further improvement of crystalline quality and device performance. Dislocations not only act as nonradiative recombination centers, degrading the efficiency and lifetime of optoelectronic devices, but also cause carrier scattering, deteriorating the performance of high-frequency devices. Therefore, suppressing the dislocation density in InP crystals is a core scientific and technological challenge for advancing their high-end applications. Current research primarily focuses on optimizing crystal quality from a macroscopic process control perspective. Miyazaki and Kuroda revealed the influence pattern of doping on dislocation distribution in 8-in. InP single crystals through computer simulation. Duan et al. optimized the solid–liquid interface shape during 4.5-in. InP growth by establishing a finite element physical model, effectively reducing the probability of twin formation. Zhu et al. further demonstrated the importance of precisely controlling the cooling process for reducing dislocations and twins by analyzing the InP growth cooling curve. These studies provide essential guidance for the crystal growth optimization. Elemental doping is a key technique for simultaneously tailoring the electrical and structural integrity (e.g., dislocation density) of materials. − However, the mechanisms by which different dopants affect dislocations are complex and vary significantly. For the n-type dopant sulfur (S), some experimental studies have shown that it can effectively reduce dislocation density in InP. , Its mechanism is often speculated to involve “solute pinning” or “lattice hardening” effects. Nevertheless, the microscopic physical essence of S doping-induced lattice hardeningincluding quantitative relationships of atomic-scale bonding strength, changes in elastic constants, and their interaction with dislocationsstill lacks systematic first-principles computational support and in-depth mechanistic explanation. Conversely, the impact of iron (Fe) doping, essential for achieving semiconducting properties, on the structural integrity of InP crystals is more complex. Numerous experiments have found that dislocation density in Fe-doped InP crystals is often higher than in undoped crystals. , This phenomenon is related not only to the weakening of the material’s mechanical properties after doping but also, more critically, to the extremely low solid solubility of Fe in InP. The latter leads to its supersaturation and precipitation during cooling, forming secondary phases (e.g., FeP2). The lattice mismatch between the precipitates and the matrix generates severe localized stress concentration, which becomes a new source of dislocations. However, specific formation conditions, morphology, and the dynamic process of defect generation induced by FeP2 precipitates still lack direct experimental observations and in-depth theoretical correlation analysis. In summary, systematically elucidating the opposing mechanisms of S and Fe doping on dislocation defects in InP is crucial for understanding the general rules of doping-defect interactions and for targeted optimization. Addressing the aforementioned research gaps, this paper innovatively combines first-principles calculations with systematic experimental statistical analysis of crystal defects, aiming to achieve two core objectives: (1) to quantitatively reveal the microscopic mechanism of S doping-induced lattice hardening at the atomic scale and (2) to clarify, through direct experimental observation, the defect generation mechanism where Fe doping leads to high dislocation density due to precipitate formation. On the basis of this mechanistic understanding, we further implemented process optimizations to effectively suppress defects. Calculation Methods and Model Establishment First-Principles Calculations This study employs first-principles calculations based on density functional theory (DFT), using the VASP software package for all computations. The strain-energy method was employed, wherein finite strains were applied to the supercell models and the resulting total energies were computed to derive the elastic constants (C 11, C 12, C 44). To ensure the reliability of our first-principles calculations, we conducted a rigorous benchmarking test by using pristine InP as a reference. The calculated elastic constants for intrinsic InP exhibit excellent agreement with previously reported standard values. By constructing supercell models of intrinsic InP, S-doped InP, and Fe-doped InP, the elastic constants (C 11, C 12, C 44), Young’s modulus, shear modulus, yield strength, and stacking fault energy were calculated and compared. This approach investigates, from the atomic scale, the changes in chemical bonding induced by doping and their effects on the material’s elastic properties and dislocation behavior. Experimental Design and Characterization Crystal Growth and Sample Preparation The experimental raw materials included self-synthesized 6N purity InP polycrystals, a PBN crucible, an InP seed crystal with <100> orientation, boron oxide with water content ≤ 200 ppb, quartz tubes, 5N indium trisulfide, and 5N iron powder as dopants. After polishing, etching, and cleaning, the PBN crucible was sequentially loaded with the seed crystal, 5 kg of polycrystalline material, 45 g of boron oxide, an appropriate amount of elemental phosphorus (to generate phosphorus vapor pressure suppressing InP decomposition), indium trisulfide, and iron powder. The sealed crucible was then placed in a 4-in. VGF-InP single-crystal furnace. Crystal growth was performed under conditions of an axial temperature gradient of 25 K/cm and a growth rate of 1.8 mm/h. Performance Characterization Methods Carrier concentration and resistivity measurements were performed at room temperature (21–25 °C) and ambient humidity (20–65% RH). Samples were classified according to carrier concentration or resistivity test results (Table ) into different doping categories. Dislocation density (EPD) was statistically determined following the GB/T20230-2022 standard, using a metallographic microscope to observe etch pits revealed by corrosion. X-ray diffraction (XRD) and Raman spectroscopy (Raman) were employed to characterize crystal quality. XRD full width at half-maximum (FWHM) was used to assess crystal integrity, while Raman spectroscopy’s transverse optical (TO) mode peak position and FWHM were used to analyze lattice stress and disorder. For Fe-doped samples, after cross-sectional polishing, scanning electron microscopy (SEM) equipped with an energy dispersive spectrometer (EDS) was used for microscopic morphology observation and precipitate phase composition identification. 1Doping Types and Electrical Parameters of Experimental Samplesno.doping typecarrier concentration/resistivitysemiconductor type1nondoped2–8 × 1015 cm–3 N-type2lightly S-doped1–3 × 1017 cm–3 N-type3heavily S-doped2–8 × 1018 cm–3 N-type4Fe-doped∼1 × 107 Ω·cmsemi-insulating type Doping Types and Electrical Parameters of Experimental Samples Results and Discussion Lattice Hardening Mechanism of S-Doping First-principles calculation results show that S doping significantly enhances the elastic properties of InP. As shown in Figure A, after S doping, the elastic constants C 11, C 12, and C 44 of InP reach 111.4, 65.9, and 86.9 GPa, respectively. Accordingly, the calculated Young’s modulus increased from the intrinsic 110.8–126.6 GPa, and the shear modulus increased from 44.7 to 51.4 GPa, both representing an increase of approximately 15%. This enhancement in elastic properties originates from structural changes at the atomic scale. S atoms occupy P sites in InP. Due to the differences in atomic radius and electronegativity between S and P, stronger and shorter In–S bonds are formed, with bond lengths shortening from the intrinsic In–P bond of 2.54–2.42 Å. This bonding change causes lattice distortion around the S atoms, generating a local tensile strain field. This local strain field interacts elastically with the stress field of the dislocations themselves, creating significant resistance to dislocation motion, i.e., the “pinning effect.” Macroscopically, this manifests as an increase in the critical resolved shear stress required for plastic deformation, i.e., a “lattice hardening” effect, effectively inhibiting dislocation slip and multiplication. (A) Calculation of mechanical constants for S-doped InP. (B) Calculation of yield strength and generalized stacking fault energy for Fe-doped InP. To visualize the impact of this mechanism on dislocation distribution, we first examined the etch pit patterns across the cross sections of different doped wafers, illustrated in Figure A–L. Experimental statistical results (Figure M) are consistent with the first-principles calculations, showing that dislocation density increases in the order: heavily S-doped < lightly S-doped < nondoped < Fe-doped. This indicates that higher S doping concentrations lead to a stronger lattice hardening effect and more significant inhibition of dislocations, experimentally validating the mechanism of dislocation suppression through solid-solution strengthening by S doping. Weakening of Mechanical Properties and Precipitate-Induced Defect Mechanism in Fe-Doping To investigate the effect of Fe doping, theoretical calculations were first conducted. The calculation results (Figure B) show that the yield strength of intrinsic InP is 0.439 GPa, which decreases to 0.327 GPa after Fe doping, indicating that Fe doping weakens the material’s resistance to plastic deformation. Furthermore, Fe doping significantly reduces the stacking fault energy of InP (intrinsic ISF and ESF energies of 1.25 and 1.20 meV/Å2 decrease to 0.92 and 0.85 meV/Å2, respectively), meaning decreased lattice stability and a greater tendency for stacking fault and dislocation generation. However, the primary cause of the high dislocation density in Fe-doped InP lies in its thermodynamic behavior. Fe has a high segregation coefficient but extremely low solid solubility in InP. During the cooling phase after crystal growth, especially in the crystal tail, Fe atoms initially dissolved in the lattice or remaining in the melt become supersaturated and precipitate. To determine the nature of these precipitates, a cross-sectional SEM analysis was conducted. As shown in Figure A, the micrographs revealed that dispersed, irregularly shaped secondary-phase particles and elongated areas were distributed within the InP matrix. The EDS analysis performed on these areas indicated that they were mainly composed of iron and phosphorus. Crucially, in region 2, the atomic ratio derived from the EDS spectral data is consistent with the FeP2 phase. A significant lattice constant mismatch exists between FeP2 and the InP matrix. This mismatch generates substantial localized stress concentration at the precipitate–matrix interface. This stress concentration region is sufficient to act as a potent dislocation source, continuously nucleating new dislocations. Additionally, the precipitates themselves can directly induce dislocation multiplication during their formation. Therefore, for Fe-doped InP, these Fe precipitates become the main source of dislocation defects, leading to higher dislocation densities. SEM image of Fe-doped InP (A), total element mapping (B), elemental content table (C), Fe element mapping (D), In element mapping (E), and P element mapping (F). Comparison of Dislocation Distribution Characteristics under Different Doping Conditions After clarifying the macroscopic influence patterns of S and Fe doping on dislocation density, this study further analyzed the axial and radial dislocation distributions in the crystal boule (Figure A–L), with statistical comparisons shown in Figure M. Dislocation mapping of heavily S-doped InP (head, middle and tail wafers: A–C), lightly S-doped InP (head, middle and tail wafers: D–F), nondoped InP (head, middle and tail wafers: G–I), Fe-doped InP (head, middle and tail wafers: J–L), and a statistical box plot of dislocation densities for InP with different dopant types (M). S-doped crystals exhibit a typical dislocation density distribution characterized as “high at the head, low at the tail, and concentrated at the four corners” (Figure A–F). This means that the dislocation density is relatively high at the boule head and gradually decreases along the growth direction toward the tail; on the cross section, dislocations tend to cluster in the four corner regions. This phenomenon is primarily driven by the thermal stress. Thermal stress arises from the lattice mismatch at the initial seed contact and radial temperature gradients during cooling. When local stress exceeds the material’s critical resolved shear stress, dislocations activate along the ⟨111⟩⟨110⟩ slip system. The ⟨110⟩-oriented edges/corners of the (100) wafer experience maximum thermal stress concentration due to their geometric positions, leading to dislocation aggregation there. The “low tail” phenomenon is mainly attributed to the solid solution strengthening effect introduced by S doping. Although thermal stress still exists in the crystal tail, the strong lattice hardening effect effectively pins dislocations, preventing their further multiplication and thereby achieving low dislocation density in the tail. Fe-doped crystals exhibit a dislocation distribution characterized as “low at the head, high at the tail, with a sharp increase in the tail” (Figure J–L). Dislocation density is acceptable at the boule head but increases sharply along the growth direction toward the tail. The root cause lies in the extremely low equilibrium solid solubility of Fe in InP and its significant segregation effect. In the crystal tail, Fe becomes enriched due to segregation and precipitates extensively as FeP2 phases during cooling because of supersaturation. The massive formation of these precipitates in the tail, with their severe lattice mismatch with the matrix, generates intense localized stress, becoming efficient dislocation nucleation sources and directly causing a sharp increase in dislocation density in the tail. Therefore, the high dislocation density problem in Fe-doped crystals is essentially determined by the tail Fe impurity enrichment and the accompanying precipitation behavior. Characterization and Analysis of InP Crystal Quality under Different Doping Conditions X-ray Diffraction (XRD) Analysis: XRD is a key technique for characterizing crystal structural integrity, lattice strain, and defect density. Diffraction peak shifts directly reflect changes in interplanar spacing. For the InP (400) plane, the standard diffraction angle 2θ is 63.345°. A peak shift to the right (higher angle) indicates decreased interplanar spacing, i.e., compressive stress; a shift to the left indicates tensile stress. S doping introduces compressive stress. The covalent radius of S atoms (∼1.02 Å) is slightly smaller than that of P atoms (∼1.06 Å). Substitution of P by S causes local lattice contraction around the S site, introducing compressive stress. As shown in Figure A, the diffraction peak of S-doped samples shifts to the right, consistent with theoretical prediction, experimentally confirming lattice distortion (shortened In–S bond) caused by S doping. XRD patterns of head and tail sections of InP with different dopant types (A) and Raman spectra (B). Unlike S doping, which induces measurable lattice contraction via substitutional solid solution, the effect of Fe doping on the matrix lattice parameters is negligible. This is because Fe has an extremely low solid solubility in InP. Consequently, instead of acting as a substitutional dopant to introduce uniform lattice strain, the vast majority of Fe atoms precipitate out of the lattice during the VGF, forming FeP2 secondary phases. Since these precipitates exist as discrete secondary phases rather than being uniformly distributed within the lattice, they do not contribute significantly to the macroscopic lattice strain of the InP matrix. Therefore, no substantial peak shift is observed in the XRD patterns of Fe-doped samples (Figure A), which is consistent with the expectation that Fe does not act as a simple substitutional shrinkage agent. XRD FWHM is a core parameter for assessing crystal quality; a narrower FWHM generally indicates higher crystal integrity. As shown in Table , S-doped and undoped samples exhibit comparable XRD FWHM values (∼18.3 to 18.8 arcsec). This indicates that substitutional S doping introduces a relatively small atomic size mismatch and does not cause severe lattice distortion. 2FWHM of XRD and Raman for Head and Tail Sections of InP with Different Dopant Types head FWHMtail FWHMdoping typeXRDRamanXRDRamannondoped18.3213.7818.5012.46lightly S-doped18.7613.0518.9013.80heavily S-doped18.7611.4719.3710.48Fe-doped19.2613.5219.7314.24aNote: XRD FWHM in arcsec; Raman FWHM in cm–1. FWHM of XRD and Raman for Head and Tail Sections of InP with Different Dopant Types Note: XRD FWHM in arcsec; Raman FWHM in cm–1. Fe-doped samples show significantly larger XRD FWHM (head 19.26 arcsec, tail 19.73 arcsec), substantially higher than undoped and S-doped samples. This directly confirms that the quality of the Fe-doped InP crystal deteriorated, which may be related to the formation of FeP2 precipitates. Raman spectroscopy analysis: The transverse optical (TO) phonon mode in Raman spectroscopy is sensitive to lattice vibrations, making it an ideal probe for assessing InP crystal quality and internal stress. The characteristic TO peak for an intrinsic InP single crystal is at 304 cm–1. Compressive stress causes a blue-shift, while tensile stress causes a red-shift. The TO peak positions of all samples are significantly lower than the intrinsic value (Figure B), showing an overall red-shift, which mainly originates from residual thermal stress from the crystal growth process. Compared to the peak position, the FWHM of the Raman peak is a more sensitive parameter for evaluating the crystal quality and lattice integrity. Broadening of the FWHM arises from lattice disorder, high defect density, and nonuniform internal stress distribution. The measured data are shown in Table : heavily S-doped samples exhibit optimal crystal quality, their TO mode FWHM is the narrowest (head 11.47 cm–1, tail 10.48 cm–1), and the peak shape symmetry is the best (Figure B), indicating the most uniform lattice vibration and the lowest stress and distortion. Lightly S-doped and undoped samples have comparable quality: Their FWHM values are close. Fe-doped samples have poorer crystal quality: their FWHM is wider, especially reaching 14.24 cm–1 in the tail, significantly higher than other samples. This directly reflects severe lattice distortion and high defect density in the tail region due to the presence of abundant FeP2 precipitates, causing a strong phonon scattering in this area. XRD and Raman spectroscopy results collectively confirm that heavily S-doped samples possess the best crystal structural integrity, while Fe-doped samples (especially the tail) exhibit significantly degraded crystal quality due to precipitate phases. Inhibition Technique for Doping Homogeneity Based on Mechanism Building upon the clarified mechanism of Fe-doping-induced high dislocation density, this study systematically optimized the conventional VGF process, aiming to suppress Fe precipitation, improve dislocation distribution, and enhance electrical homogeneity. The optimization core involves dynamic control of the solid–liquid interface shape and solute transport. By optimizing the thermal field (multizone temperature regulation) and introducing crucible rotation along with programmable pulling speeds, the solid–liquid interface shape during growth was effectively controlled from a “concave” shape to a more “flat” shape. A flat solid–liquid interface helps reduce the radial temperature gradient, thereby lowering the tendency for thermal stress-driven dislocation nucleation. More importantly, for Fe-doped crystals, a flat interface combined with an optimized cooling rate can delay the supersaturation process of Fe atoms, inhibiting the massive formation of FeP2 precipitates in the tail, mitigating the “precipitation-induced” dislocation multiplication at the source, and helping to improve the “sharp tail increase” dislocation distribution. Simultaneously, by adjusting the crucible rotation speed to enhance melt convection, breaking the solute boundary layer, and dynamically matching the growth rate along the axial direction, the effective segregation coefficient of Fe was regulated, making the axial distribution of Fe in the crystal more uniform and thereby significantly reducing the head-to-tail resistivity difference. Ultimately, process optimization achieved the synergistic control of structural defects and electrical inhomogeneity. On the one hand, low dislocation density (especially effective suppression of tail dislocations) ensures good structural integrity of the InP wafers. On the other hand, improved resistivity uniformity directly enhances the performance consistency of large-area devices fabricated on these wafers. This proves that targeted process regulation based on a mechanistic understanding can simultaneously tackle the core challenges of structural integrity and electrical homogeneity in large-diameter InP single crystals. Conclusion Through first-principles calculations and experimental analysis, this paper reveals the opposing mechanisms of S and Fe doping on dislocation defects in VGF-InP crystals and proposes the corresponding defect inhibition methods: (1) S doping inhibits dislocations through lattice hardening: S atoms substituting for P sites form shorter and stronger In–S bonds, inducing lattice hardening (elastic modulus increased by ∼10%). The resulting local strain field effectively pins dislocations, significantly reducing dislocation density, with the effect increasing with concentration. Its dislocation distribution shows a “high head, low tail, concentrated at four corners” pattern, driven by thermal stress and suppressed by solid solution strengthening. (2) Fe doping induces dislocations through precipitation phases: Fe has extremely low solid solubility in InP, precipitating as FeP2 secondary phases during VGF. The lattice mismatch between precipitates and the matrix causes severe stress concentration, forming efficient dislocation sources, leading to higher dislocation density in Fe-doped InP. Its dislocation distribution shows a “low head, high tail, sharp increase in the tail” pattern, dominated by tail impurity enrichment and precipitation behavior. (3) Crystal quality comparison: XRD and Raman characterization show that heavily S-doped samples have the best crystal quality, while Fe-doped samples (especially the tail) exhibit the poorest quality due to precipitate phases. (4) Process optimization pathway: Based on the above mechanisms, optimizing the VGF/VB process (controlling the solid–liquid interface, enhancing melt convection, etc.) can effectively inhibit Fe precipitation, improve dislocation distribution, and resistivity uniformity, providing a clear theoretical basis and process route for preparing high-quality, high-homogeneity InP single crystals. The contrasting behaviors of S and Fe underscore a critical principle for defect engineering in compound semiconductors: the necessity to balance electronic functionality with structural integrity, where dopant selection must be guided by its thermodynamic solubility and specific lattice interactions.