Searcharxiv⌕ Search

arXiv subjects

Pubudu Wijesinghe

Publications and source records attributed to Pubudu Wijesinghe.

2 recordsLinked to original sources

Electric-Field-Induced Second Harmonic Generation at a Reconfigurable LaAlO$_3$/SrTiO$_3$ Nanojunction

Electrically tunable nonlinear optical responses at the nanoscale remain challenging to achieve because conventional nonlinear materials lack the combination of large susceptibility, nanoscale confinement, and in situ reconfigurability. Here we report electric-field-induced second harmonic (EFISH) generation from a nanoscale tunnel junction defined by conductive atomic force microscope lithography at the LaAlO$_3$/SrTiO$_3$ interface. A conducting channel written at the interface is interrupted by a nanoscale insulating gap, across which applied DC bias produces local electric fields exceeding $10^7$ V/m. The SHG signal is spatially localized at the junction, exhibits a quadratic bias dependence described by $I(2ω) \propto |χ^{(2)}_\mathrm{0} + χ^{(3)} E_\mathrm{DC}|^2$ with no hysteresis, a modulation depth exceeding 380% at $|V_\mathrm{DC}| = 1$ V, and shows a two-lobed input-polarization pattern aligned with the junction axis, consistent with EFISH from a centrosymmetric host. Calibration against a BBO reference crystal gives $|χ^{(3)}| \approx 1\times10^{-19}$ m$^2$/V$^2$ at 6 K. These results establish cAFM-written oxide nanojunctions as a reconfigurable platform for nanoscale nonlinear optics in which the junction geometry sets the symmetry of the response and the large field-induced $χ^{(2)}$ of SrTiO$_3$ provides the optical nonlinearity. Because the nonlinearity is both generated and read out within the same nanoscale gap, the junction operates simultaneously as a subwavelength source and a near-field detector of optical nonlinearity.

cond-mat.mes-hall↗

Typologically-Informed Candidate Reranking for LLM-based Translation into Low-Resource Languages

Large language models trained predominantly on high-resource languages exhibit systematic biases toward dominant typological patterns, leading to structural non-conformance when translating into typologically divergent low-resource languages. We present a framework that leverages linguistic typology to improve translation quality without parallel training data or model retraining. The framework consists of two components: the Universal Metalinguistic Framework (UMF), which represents languages as structured profiles across 16 typological dimensions with divergence-weighted scoring, and the Computational Engine, which operates through linguistic disambiguation during generation and typological compliance scoring during selection. Evaluation across nine language pairs demonstrates intervention rates strongly correlating with typological distance from English. In experiments on 341 English sentences each having different morphological and syntactic phenomena, the framework shows an intervention precision of 48.16% for conservatively treated languages, 28.15% for morphologically dense languages, and 86.26% for structurally profiled languages. The framework requires no parallel training data and operates with any LLM capable of producing multiple candidate outputs, enabling practical deployment for under-resourced languages.

cs.CL↗