Samarjith Biswas | Aerospace Engineer, Aerodynamics and Aeroacoustics

SEC 02 · Research

Current research areas

Aerodynamics and aeroacoustics first, then the wave physics behind them, from my PhD and the $30M NSF-funded NewFoS Science and Technology Center.

Focus areas: aircraft and engine aerodynamics, aeroacoustics, and noise reduction.

Turbofan section · stations highlight as you scroll; hover or tap for the landing configuration

Fan Moves most of the air; its blades set the blade-passing tone.
Bypass duct The cold bypass stream. Acoustic liners line these walls.
Compressor Raises the core pressure before combustion.
Combustor Burns fuel at roughly constant pressure.
Turbine Takes work out of the hot gas to drive the fan and compressor.
Nozzle Expands the core flow back to ambient pressure.

01 · Aerodynamics and CFD

Aerodynamics and CFD

I have done substantial ANSYS CFD work, including external-flow studies of aerofoils, alongside advanced graduate CFD coursework.

In AeroPhon I rigged the NASA CRM-HL in landing configuration, recovering the slat and flap positions from NASA’s positioning files (screw-motion kinematics, gap solved to 0.02 mm), and solved the flow with the lattice Boltzmann solver FluidX3D (D3Q19) on one laptop GPU: 38 million cells at 8° angle of attack, with Q-criterion views of the slat, nacelle and wing-tip wakes. The force path is validated against Stokes drag to 3.9 %. The grid Reynolds number is below flight, so it does not predict high-lift coefficients.

AeroPhon CFD → Try AeroPhon live → Inside a Turbofan film → Aerofoil and slat drawings →

For vehicle aerodynamics, neural surrogates trained on DrivAerNet CFD cars predict drag, surface pressure and wall shear; the drag model explains about 80 % of the drag variation on 825 held-out cars (R² 0.80, MAE 0.0083).

02 · Aeroacoustics

Acoustic Liners for Jet Engines

My PhD (Mechanical and Aerospace Engineering, Oklahoma State, 2019–2024) was on acoustic liners. I designed thermoacoustic liner test articles and test matrices for NASA Langley’s normal-incidence tube and grazing flow impedance tube, with grazing flow up to Mach 0.5, and analysed the results; Langley operated the facility.

A liner is a field of small resonators under a perforated face sheet. A quarter-wave cell absorbs best when its depth is a quarter of the wavelength, so the fan’s blade-passing tone sets the depth.

Mach 0.5Grazing flow

IJA 2025Int. J. Aeroacoustics

λ/4Quarter-wave cells

IJA paper → Liner calculator → Jet-engine noise film → Acoustic liners deep dive →

Biswas S., Kresl W., Manimala J. M., “On the relationship between acoustic absorption and temperature gradient in a thermoacoustic liner”, International Journal of Aeroacoustics, doi:10.1177/1475472X241306317.

Film · a liner cell that makes heat, tested at NASA Langley · the IJA paper

What the film shows. A single 3D-printed liner cell with a thermoacoustic stack inside, tested in the normal-incidence tube at NASA Langley. At 790 Hz and 140 dB the two faces of the stack settled 9.5 °C apart, and in sweeps from 400 to 1,000 Hz the cell absorbed 51 to 66 % of the incoming sound energy. Where the stack sits matters: the best heat came about 4 cm below the opening. At 150 dB the difference reached about 25 °C, and a linear model holds only up to a drive ratio of about 3 %. An illustrative simulation of one stack pore (not from the paper) shows the gas temperature in the pores swinging by about half a degree every cycle.

The machines that made me an engineer · rendered at my desk

SEC 03 · Software

Try the tools I built

Two browser-based design tools I wrote from the physics up. Both run in the browser, with no install and no API keys.

What AeroPhon’s network sees in a car: the real DrivAer reference body, sampled as 2,048 surface points, goes into a graph neural network, and the drag coefficient and surface pressure come back in seconds instead of hours. On 825 cars it never saw in training it explains about 80 % of the variation in CFD drag (R² 0.80). Try it live → · Watch in 1080p →

TOOL 01 · PhononIQ

PhononIQ

Phononic-plate bandgaps, designed from a plain-English prompt.

Surrogates: NN, PINN and physics-informed CNN
Data: 8,000-sample COMSOL dataset, geometry and material varied
Chat: optional on-device Llama-3.2-1B

Open PhononIQ →

TOOL 02 · AeroPhon

AeroPhon

Describe or sketch a car, then test it like an engineer.

Surrogates: neural drag, surface pressure and wall shear
Data: DrivAerNet CFD cars
Crash: analytical estimate

Try AeroPhon live →

03 · Computational Methods

Machine-Learning Surrogates

A finite-element band-structure calculation for one phononic-plate design takes minutes to hours. Neural surrogates trained on COMSOL data return the same quantities in milliseconds, so a whole design space can be searched.

Physics-informed terms in the training loss (PINNs) cut unphysical predictions, and the same models run in reverse for inverse design: from a target bandgap to a geometry. On a BOROFLOAT 33 glass plate the surrogate, FEM and measurement agree on a 195–214 kHz bandgap.

msPer prediction

195–214kHz bandgap, three ways

InverseDesign

Film · holes that stop waves

PINNsPyTorchInverse DesignNeural Surrogates

04 · Published 2026

Directional Interface Modes in Topological Acoustics

Where two acoustic lattices of different topology meet, waves can be bound to the interface. In our Scientific Reports paper these interface modes carry a pseudospin that locks their direction of travel: spin-momentum locking.

The platform is one plate of the phase-change material Ge₂Sb₂Te₅ (GST): crystalline inclusions in an amorphous host, near 200 MHz. The simulated interface signal stays more than 35 dB above the bulk over 15 wavelengths, round bends and junctions. Two forces a quarter-cycle apart pick the direction; swapping them reverses it. The results are computational (COMSOL); experiments are next.

On-chip RF routing one-way and two-way channels in the same plate
Isolators and duplexers direction picked by the source, not by moving parts
Frequency-selective filters one-way transport at chosen frequencies
Programmable acoustic circuits non-volatile patterns written into one film

Biswas et al., Scientific Reports (2026) → Topology explainer →

Simulated plate with an L-shaped interface: the wave from the source follows the single bend — L-shaped interface

Simulated plate with a Z-shaped interface: the wave follows both bends — Z-shaped, two bends

Simulated plate with a four-domain junction: the wave turns at the crossing — Four-domain junction

Simulated field on the plate (top) above its domain layout (bottom); the star marks the source. From Biswas et al., Scientific Reports (2026), supplementary material.

05 · Quantum Analogues

Acoustic Analogues of Quantum Systems

Phononic lattices built on the Su-Schrieffer-Heeger (SSH) model host edge modes that are protected against some kinds of disorder, and they do it at room temperature with ordinary sound.

Because the wave equations map onto each other, these lattices let us test topological ideas on the bench, with measurable displacements and phases, before anyone builds a quantum device.

SSHLattice model

Room temp.Operation

BenchScale of tests

Phi-bit film → SSH model explainer →

SSH chain · strong and weak bonds, edge mode · tap the drawing to compare a bulk mode

Phononic CrystalsEdge StatesSSH ModelQuantum Analogues

06 · RF Acoustics

Surface-Acoustic-Wave Devices

Surface-acoustic-wave (SAW) filters are how a phone separates one radio band from the next: the radio signal briefly becomes a sound wave on a piezoelectric chip.

I model how periodic structures on a piezoelectric substrate guide and filter these waves, with COMSOL simulations and machine-learning surrogates that make the design search faster.

SAWDevice platform

RFFilters

FEM + MLDesign method

RF FiltersSAW DevicesPiezoelectricCOMSOL

Film · a message’s journey to the SAW chip · watch with sound

Cross-Institutional Collaboration

UCLA

CUNY

SEC 04 · AeroPhon CFD

The aircraft, and the flow the solver computes

High-lift rigging on the NASA CRM-HL, landing configuration. My rigging tool recovers the slat and flap motion from NASA’s own positioning files; the same configuration then runs in GPU lattice-Boltzmann CFD (right).

The same case in CFD: Q-criterion isosurfaces coloured by velocity.

The same rigged aircraft as the film, on a finer grid: 424 × 636 × 141, so 38 million cells (the film’s sweep uses 15.6 million) at 8° angle of attack, on one laptop GPU. Drawn as isosurfaces of the Q-criterion coloured by velocity, which isolates where rotation dominates strain, so these are the vortices rather than the whole velocity field. Tubes shed from the slat gaps, the nacelle and the wing tip. The first 1.6 flow-throughs are discarded, so this is the developed wake and not the starting transient.

Three views: top-down, where the wing and slat wakes separate most cleanly; a wide three-quarter; and a close pass on the slat and nacelle. Solver: FluidX3D, D3Q19, FP16C with a Smagorinsky subgrid model, RTX 5070 Laptop. The force path is validated against the closed-form Stokes drag to 3.9 %.

On what this is and is not. The cell is 152 mm, so the vortices cannot be finer than roughly half a metre and the surface staircase is the voxelised geometry, not the flow. The 4th AIAA High-Lift Prediction Workshop puts the requirement above 250 million degrees of freedom, which this is 6.6× short of, and the grid Reynolds number of 10⁵ sits 331× below flight. This grid resolves separation but not attachment: the wing wake is many cells across and genuinely captured, while anything attached to a surface is not. It is a real solve at a resolution chosen so hundreds of riggings fit in the cost of one high-fidelity case, and it is not a prediction of high-lift coefficients.

SEC 05 · Projects

Thermoacoustic metastructures and structural energy storage

My PhD work on liners and barriers that absorb sound and harvest part of it as heat and electricity, and my M.S. work on car panels that store energy. All projects →

Thermoacoustic Meta-Structures (TAMS)

Resonator cavities with a thermoacoustic stack inside: the sound sets up a temperature difference across the stack, which a thermopile turns into a small voltage, while the cavity absorbs the sound.

Thermoacoustic Liner

TAMS Barrier

TAMS Unit Cell

Helical Resonator

Meta Wall Assembly

Spiral Stack Detail

Stack Geometry

Spiral Resonator

Turbojet Acoustic Liner

Thermopile

Thermoacoustic Liner

TAMS Barrier

TAMS Unit Cell

Helical Resonator

Meta Wall Assembly

Spiral Stack Detail

Stack Geometry

Spiral Resonator

Turbojet Acoustic Liner

Thermopile

Pie Slice Resonator

Wedge-shaped cavities that guide sound in one direction and keep neighbouring modes apart.

Spiral Resonator

A helical path that lengthens the resonator without enlarging it and spreads the acoustic energy evenly.

Spiral Stack

A spiral stack that packs a long thermoacoustic path into a small volume, for control over both the wave and the heat transfer.

Thermoacoustic metastructures: experiments and concepts

Concept films from my PhD work on thermoacoustic metastructures. The test rig and the metastructure are on the Projects page →

TAMS: concept film

Where thermoacoustic metastructures could be used: urban noise barriers, aircraft liners, and spacecraft with large temperature gradients. The liner concept is aimed at engine tones below about 1 kHz, where conventional liners struggle. It is a concept: the liner has been tested on the bench and in NASA Langley impedance tubes, and it has not been engine-tested.

The energy conversion chain

Sound sets up a temperature difference across the stack, heat exchangers carry the heat in and out, and a thermopile turns the gradient into a small voltage. On the bench the largest difference was 24.8 °C at 117.5 Hz with a resin stack; a separate experiment gave about 33 mV from the thermopile. The film below follows one gas parcel round the cycle.

View All Projects & Videos →

M.S. · Northern Arizona University 2017–2019 · electric vehicles

Structural supercapacitors

A structural supercapacitor is a composite panel that carries load and stores charge at the same time. I made carbon-fibre electrodes, functionalised them by CVD and plasma treatment, and tested the panels electrochemically and mechanically.

Measured on the composite: 53.58 mWh/kg energy density, 32.06 W/kg power density and 6.9 GPa flexural modulus. That makes it a power buffer that saves weight, not a replacement for the battery.

53.58mWh/kg energy density

32.06W/kg power density

6.9 GPaFlexural modulus

How it works. Two carbon-fibre electrodes carry the mechanical load and hold the electrical charge. A structural electrolyte between them passes load and ions from one to the other, and a thin separator keeps the electrodes from touching while the ions move through it.

M.S. research → Thesis →

SEC 06 · About

Sound and vibration in engineered structures

Aerospace engineer working on aerodynamics, aeroacoustics and vibration, using CFD, testing and machine learning to make design faster.

Cantilever beam · first two mode shapes

I am a Research Scientist III at the NSF New Frontiers of Sound (NewFoS) Science and Technology Center, University of Arizona. I work where flow, sound and structure meet: CFD on aerofoils, high-lift aircraft and vehicles (ANSYS, FluidX3D), acoustic liners tested at NASA Langley, and machine-learning surrogates that return a design answer in milliseconds.

My PhD at Oklahoma State University (2019–2024) developed thermoacoustic metastructures, liners that absorb engine noise and turn part of it into heat and a small voltage. The work is published in the International Journal of Aeroacoustics and filed as a PCT patent application (WO 2025/128348 A1, pending). From 2021 to 2024 I also managed the university’s vibration and acoustic test facility, more than $300K of equipment, and I have mentored seven students.

2017–2019M.S., Northern Arizona UniversityStructural supercapacitors
2019–2024PhD, Oklahoma State UniversityThermoacoustic liners, NASA Langley tests
2024–nowResearch Scientist III, NSF NewFoSPhononics, CFD, ML surrogates

Focus areas: aircraft and engine aerodynamics, aeroacoustics, and noise reduction.

Tucson, AZ · Open to relocation

Measured resultsREV 2026

±2 %FEM vs experiment, glass phononic plate

195–214 kHzBandgap matched by FEM, ML and measurement

Mach 0.5Grazing-flow liner tests, NASA Langley

2Papers published in 2026

Aerodynamics and aeroacoustics

External aerodynamics with CFD (ANSYS; GPU lattice Boltzmann)
High-lift configurations
Aeroacoustic liners and duct acoustics
Grazing-flow and impedance testing
Surrogate models for aerodynamic loads

Backed by the AeroPhon CRM-HL and DrivAerNet work and the NASA Langley liner tests above.

SEC 07 · Simulation and tools

Core Competencies

The simulation, test and software tools I use day to day.

PyTorch SolidWorks

R

Acoustic Engineering

Topological AcousticsPhononic CrystalsThermoacoustic SystemsRF DevicesAcoustic MetamaterialsSound AbsorptionNoise ControlUltrasonic Testing

AI & Machine Learning

Deep LearningTensorFlowPyTorchPINNsAutomated Design OptimizationNeural NetworksReal-time Optimization

Simulation & FEA/CFD

COMSOL MultiphysicsANSYS Fluent: external flow, aerofoilsFluidX3D (lattice Boltzmann)AeroacousticsANSYS WorkbenchDeltaECMATLAB/SimulinkModal AnalysisThermal-Structural Coupling

Programming & Software

PythonMATLABC/C++LabVIEWRFORTRANSignal ProcessingData Analysis

Advanced Engineering

Energy HarvestingMEMS DesignPhase-Change MaterialsThermoelectric ConversionPiezoelectric TransducersSmart Materials

CAD/CAM & Manufacturing

SolidWorksAutoCADG-code/CNCAdditive ManufacturingFDM/SLA PrintingProcess Optimization

View Full Competencies →

CFM56-7B · isometric cutaway, fan to exhaust

Cascade thrust reverser · hover or tap to deploy

SEC 08 · Research visualisation

Research projects in motion

Each film follows one project from physics to hardware, built from 3D renders of the real geometry and blueprint animations: aircraft and engine noise, turbofans and the Concorde intake, acoustic liners, thermoacoustics, phase bits and surface acoustic wave filters. The message-journey and supercapacitor films sit with their research above.

Concorde · front view, leading-edge vortices and intake section A-A

How a Phi-Bit Works

Can a classical sound wave carry information the way a qubit does? The logical phi-bit, the quantum-analogue idea at the heart of NSF NewFoS.

Thermoacoustics: Turning Sound into Heat and Electricity

One loud note makes one face of a stack hot and the other cold: the thermoacoustic cycle, our helical meta-stacks and the thermopile.

Why New Jet Engines Are So Hard to Quiet

Bigger, slower fans lower the blade-passing tone, and today's acoustic liners struggle below about a kilohertz.

Hear the Engine Start

A GE90 on a 777 from rest to take-off: starter, light-off, the fan's climbing tone and the jet's roar. Engine internals and start timing are illustrative.

Bigger Fans, Lower Tones

From the 727 to the 777-9: bigger, slower fans lower the blade tone, and the liner cells that absorb it have to get deeper.

How Engineers Make Airplanes Quieter

How engineers quiet airliners: high bypass ratios, shaped fans and vanes, chevrons, acoustic liners and ICAO limits.

Glass That Stops Sound

A laser-drilled honeycomb of holes turns a borosilicate glass plate into a phononic crystal. Model 195 to 214 kHz, measured 193 to 211 kHz.

Deep dives

Inside a Turbofan

Station by station through a modern turbofan: fan, compressors, combustor, turbines and nozzle; pressure ratio, bypass ratio, cooling and the geared fan.

Concorde's Engine and Intake

How Concorde's Olympus 593 breathed at Mach 2: variable intake ramps, shocks, a converging-diverging nozzle, reheat and fuel transfer for trim.

Road car vs race car: the air around a McLaren 720S

A McLaren 720S and its GT3 racing version in the same virtual wind tunnel: one lattice Boltzmann simulation on a laptop GPU, 73 million cells at about 250 km/h. Every flow shot comes from that simulation: swirl, arrows, oil-flow streaks and the slow wake. Wheels held still and a scaled Reynolds number, so it shows how the shapes treat the air, not drag numbers.

How the X-59 quiets the sonic boom

NASA's X-59 is built to fly faster than sound with a soft thump instead of a loud boom. A supersonic flow simulation (SU2, Mach 1.4) around a simplified X-59 shape: the shock cone, air carried along the solved flow, a slice sweeping down the cone and the pressure one body length below. Inviscid and simplified, so it shows the kind of shock pattern, not NASA's official signature.

Where landing noise really comes from

On final approach an airliner's engines are almost at idle, yet the airframe is about as loud. An A350 flies the real runway 26 approach into Phoenix, and my lattice Boltzmann simulations on NASA's CRM-HL research model show the landing gear, slat cove and flap edge making sound. The simulations are Reynolds-number scaled: they show the mechanisms, not certified levels.

How Acoustic Liners Work

How acoustic liners work: Helmholtz and quarter-wave cells, impedance, double layers, grazing-flow testing at NASA Langley, and why low tones need deep cells.

Inside a SAW Filter

From wafer to filter: how a surface-acoustic-wave filter is made and how its fingers pick one radio band.

All documentaries →

SEC 09 · Position

Current Position

Phononic lattice

Sept 2024 – Present

Research Scientist III

NSF NewFoS Center, University of Arizona

Topological acoustics: Lead author of the 2026 Scientific Reports paper on directional interface modes via spin-momentum locking in topological acoustic metamaterials.
Machine-learning surrogates: Build bandgap prediction models for BOROFLOAT 33 glass phononic plates; surrogate, FEM and measurement agree on a 195–214 kHz bandgap.
Cross-Institutional Leadership: Collaborate with CUNY, Wayne State, Caltech, UCLA, Spelman College, University of Alaska-Fairbanks, and UC Boulder on acoustics research.

View Full Experience →

SEC 10 · People

Team & Collaborations

Talks, conferences and team meetings.

Liner cell · quarter-wave

Presentation

Team Meeting

PhD Defense

ASA Presentation

ASA Conference

ASA Session

Presentation

Team Meeting

PhD Defense

ASA Presentation

ASA Conference

ASA Session

Boeing 747-400 · plan view, typical three-class main deck

Rocketdyne F-1 · major components, Saturn V first-stage engine · after the NASA SA-503 flight manual

SEC 11 · Drawing board

Drawing board

Outside work I draw the machines I grew up admiring, from the Wright Flyer to the Space Shuttle. Every plate is traced from a real 3D model, museum scan or NASA drawing.

Apollo Lunar Module · front and isometric, 9.4 m across the footpads · 3D model: NASA 3D Resources

Apollo 11 command module Columbia · cutaway from the Smithsonian 3D scan

Apollo-Soyuz · docked in 1975, side view at one scale · after the NASA ASTP press kit

Saturn V S-IC first stage · side, three-quarter and base views · 3D model: NASA 3D Resources

Rocketdyne J-2 · two views, 22 parts · after the NASA flight manual

See more plates →

SEC 12 · Contact

Contact

Open to roles and collaborations in aeroacoustics, vibration, acoustic liners, RF acoustic devices and simulation.

SAW transducer

Email samarjithbiswas@arizona.edu

Location Grand Challenges Research Building
Room 645
750 N Cherry Ave, Tucson AZ 85719

SamarjithBiswas

Designs from the lab