Iron filings between two magnets. Photo by MikeRun, Vera Wurmsdobler, CC BY-SA 4.0

All workPhysics and maths, foundations

Eigen physics clubat KYPT

Three tournament problems, from climbing magnets to falling coins and twisted spaghetti

Turning strange everyday phenomena into equations, then testing those equations against a stopwatch, a camera and a force sensor.

Role: Captain
When: 2024 to present
Grades: G10 to G12
Recognition: KYPT 2025 bronze award (grade 10 team member)
KYPT 2026 silver award (captain, leading two teams)
KYPT poster excellence award

Best model error: 0.66%Best model errorClimbing magnets at 975 RPM: 17.5 cm/s measured versus 17.62 cm/s predicted
Top climbing speed: 21 cm/sTop climbing speed15 magnets (6 mm × 3 mm) on the rod spun at 1625 RPM
Coin drops filmed: 240Coin drops filmed6 coin and angle combinations × 20 trials each, analysed in Tracker
Tumble-only at 90°: ~78%Tumble-only at 90°Share of coins released vertically that tumbled straight to the bottom
Measured optimal twist: 40° / 50° / 60°Measured optimal twistSpaghetti bundles of 20, 30 and 40 strands
Teams fielded in 2026: 2Teams fielded in 2026The club grew from one KYPT team to two under Wonjun's captaincy (project brief)

In brief

The problem: The Korean Young Physicists' Tournament (KYPT) sets open-ended problems with no single right answer: a team must explain a phenomenon, build a model, run its own experiments and defend the result against opposing teams. Wonjun reported on one problem each season for the Eigen physics club, first as a grade 10 team member and then as captain. Each problem demanded a theory that made quantitative predictions, and an experiment precise enough to show where that theory held and where it broke.
The spark: In all three problems the interesting physics sits at a threshold: the point where friction gives way to slipping, where a coin's restoring torque loses to the fluid and it starts to tumble, or where extra twist stops locking a bundle together and starts weakening it. Writing a model whose parameters move that threshold, and then locating it experimentally, became the club's recurring method.
The outcome: Across the three problems the models matched experiment to within roughly 1–19 percent. The climbing-magnet model tracked measured climbing speeds to within 0.66–8.28 percent once its constants were fitted in Python; the coin model's critical speeds came within 4.0–19.4 percent of measured values; and the spaghetti model correctly predicted that the optimal twist angle rises with the number of strands, with the measured optimum 11.1–17.4 percent below theory. The club took bronze at KYPT 2025 and silver at KYPT 2026, the second year with Wonjun as captain leading two teams.

Method

How it was done, step by step.

1Unpack the problemIdentify the phenomenon and the relevant parameters (rotation speed and magnet count; coin mass and release angle; strand count and twist angle).
2Qualitative pictureName the competing effects: stick versus slip, restoring versus fluid torque, inter-strand locking versus helical pre-stress.
3Quantitative modelWrite force or torque balances (Newton's second law, torque equations, beam bending theory) and derive a prediction: a climbing speed, a critical falling speed, an optimal twist angle.
4Controlled experimentBuild a dedicated rig (motor with speed controller, water tank with laser reference, clamps with a PASCO force sensor) and vary one parameter at a time.
5Video and sensor analysisTrack motion frame by frame in Tracker software or log fracture loads, repeating trials (20 drops per coin condition, 3 iterations per spaghetti angle).
6Theory versus experimentCompare predictions with measurements, compute error rates and explain the deviations physically.

01The tournament

Physics as an argument, not an answer key

KYPT is the Korean stage of the young physicists' tournament format: teams receive a list of open problems months in advance, research them independently, and then present and defend their solutions in debate-style rounds. There is no reference solution. What wins is a clear explanation of the phenomenon, a model that predicts something measurable, and data good enough to survive an opponent's questions. For the Eigen physics club this meant treating every problem as a small research project, with its own apparatus, its own analysis code and an honest account of where the model falls short.

2025: climbing magnets, as a grade 10 team member
2026: autumn coin and twisted spaghetti, as captain

022025 · Climbing magnets

Why a magnet climbs a spinning rod

A short rod of stacked neodymium magnets is stuck sideways to a vertical steel rod. Spin the steel rod and the magnets walk upward. Wonjun's model combined stick-slip friction (the magnets alternately grip and slide), eddy currents induced by the rotation (Faraday's law and the Lorentz force), and gravity. Setting the net force to zero for steady climbing gave a closed-form speed that rises with rotation speed but saturates, and grows only logarithmically with the number of magnets. Experiments from 330 to 1625 RPM confirmed both the rise and the saturation.

Speed climbed from about 8 cm/s to 21 cm/s across the RPM range
Low RPM: irregular, stick-slip dominated motion; high RPM: smooth climbing

032025 · Fitting the model

Letting Python find the constants

The climbing-speed formula has several constants that are hard to measure directly: an overall magnetic force scale, an attenuation coefficient for the saturation, static and kinetic friction, and a stick-slip loss term. Wonjun wrote a Python script that averaged the measured speed at each RPM, computed the theoretical speed for candidate constants, and searched for the set that minimised the error rate. The optimised model reproduced all five RPM settings within 8.28 percent, and within 3 percent at four of them, turning a qualitative story into a usable predictive formula.

042026 · Autumn coin

Steady, flutter or tumble

A coin dropped into water rarely falls straight. It may glide steadily, flutter side to side like a leaf, or tumble end over end. Wonjun modelled the coin's tilt as a damped oscillator whose stiffness is the gravitational restoring torque minus a destabilising fluid torque that grows with the square of the falling speed. When the two cancel, the equilibrium disappears and the coin must tumble, which defines a critical speed. Twenty drops of each coin (100 won and 500 won) at each release angle (0°, 45°, 90°) were filmed and classified by regime.

0° release: flutter that collapses into tumbling was most common (~65%)
90° release: tumble-only reached ~78%

052026 · Autumn coin

Heavier coins settle down faster

Both coins mostly ended up tumbling, but they differed in how messy the path was. The lighter 100-won coin switched regimes repeatedly, with nine tumble-flutter-tumble sequences across its 60 drops, while the heavier 500-won coin had only four and was the only one to settle from tumbling into a steady glide (twice). The model's critical speeds were compared with Tracker measurements: for the 500-won coin released flat the prediction was within 4.0 percent, while the 100-won coin showed larger deviations of around 18–19 percent, consistent with its more unstable dynamics.

062026 · Twisted spaghetti

How much twist makes a bundle strongest

A twisted bundle of spaghetti can resist a sideways push better than a straight one, but only up to a point. Twisting presses the strands together so they share load like a solid rod, raising the effective bending stiffness. Yet every twist also pre-bends each strand into a helix, using up part of its fracture strength before any load is applied. Wonjun combined beam bending theory with this helical pre-stress, wrote the breaking load as a function of twist angle, and differentiated it to find an optimum that should increase with the number of strands.

072026 · Twisted spaghetti

Testing 0 to 160 degrees

Bundles of 20, 30 and 40 strands, each 250 mm long and held by rubber bands, were clamped at both ends and twisted to set angles measured against reference lines. A PASCO force sensor pushed down at the centre until the bundle snapped, with three repetitions per angle across 0–160°. Every bundle showed the predicted rise-then-fall in breaking force. The measured optimum (40°, 50°, 60°) increased with strand count exactly as predicted, though it sat consistently below theory (45°, 57.5°, 72.6°), which Wonjun traced to imperfect alignment and friction lowering the real effective stiffness.

08Leadership

From one team to two

After competing as a grade 10 member on the 2025 bronze-winning team, Wonjun became captain of the Eigen physics club and expanded its tournament programme from one team to two for KYPT 2026. Alongside presenting his own problems, he coordinated both teams' preparation, and the club returned with a silver award and a poster excellence award. The three problem reports here are the ones he personally presented as reporter.

The problems

Three problems, argued live.

Attach a rod assembled from cylindrical neodymium magnets horizontally to a vertical ferromagnetic rod. Limit the motion of the magnets to the vertical direction. When the ferromagnetic rod is spun around its axis of symmetry, the magnetic rod begins to climb up. Explain this phenomenon and investigate how the rate of climbing depends on relevant parameters.

Theory: The magnets move in a stick-slip cycle. In the stick phase static friction (up to μs·N) holds them in place while energy builds up; once the driving force exceeds that limit they slip, kinetic friction (μd·N) takes over, they jump upward, slow down and stick again. The driving force comes from the spinning steel rod: rotation induces eddy currents (Faraday's law) whose field pushes on the magnets (Lorentz force). The induced current scales with rotation speed but saturates, modelled as A·(1 − e^(−α·ω)), and the field grows only logarithmically with the number of magnets N, modelled as ln(N + 1). Applying Newton's second law at steady climbing (a = 0) to magnetic force, static friction, kinetic friction, gravity and a stick-slip loss term gives the climbing speed: v = [A·(1 − e^(−α·ω))·ln(N + 1) − (μs·N + m·g) − F_ss] / (m + μd).
Experiment: A 1 m steel rod, 10 mm in diameter, was coupled to a motor with a speed controller (330–1625 RPM) and held in a custom wooden frame. Neodymium disc magnets 6 mm in diameter and 3 mm thick were stacked horizontally against the rod so they could only move vertically. Climbing was filmed and analysed in Tracker for displacement, velocity and acceleration. Series 1 fixed 15 magnets and varied speed (330, 650, 975, 1300, 1625 RPM); series 2 fixed 1300 RPM and varied the number of magnets from 12 to 20. Theoretical speeds were then fitted to the data in Python by searching for constants that minimised the error rate.
Findings: Climbing speed rose with rotation speed, quickly at first (about 8 to 13 cm/s between 330 and 650 RPM, where stick-slip is strong) and then levelling off (about 17.5 to 21 cm/s from 975 to 1625 RPM), a saturation the exponential term captures. Low RPM gave irregular stick-slip motion; high RPM gave smooth climbing. More magnets also meant faster climbing, with diminishing returns: roughly 2.5 to 4.5 cm/s from 12 to 14 magnets but only 7.1 to 8.0 cm/s from 18 to 20, attributed to magnetic saturation, extra friction and magnet-magnet interactions. With optimised constants, theory matched experiment within 0.66–8.28 percent across all five speeds.

Rotation speed: 330, 650, 975, 1300, 1625 RPM
Magnet count: 12–20 (15 in the speed series)
Magnet size: 6 mm diameter × 3 mm thick
Steel rod: 1 m long, 10 mm diameter
Analysis: Tracker video analysis + Python constant fitting

Climbing speed versus rotation speed15 magnets (6 mm × 3 mm). Measured averages from Tracker against the Python-fitted model.Climbing speed (cm/s)

Experiment
Theory (fitted)

Model error at each speedError rate between measured and theoretical climbing speed.Error rate (%)

Climbing speed versus magnet count (1300 RPM)Only the approximate range endpoints are given in the deck; the per-count values were shown graphically.

Section	Magnets	Approximate speed change	Interpretation
Initial	12 → 14	2.5 → 4.5 cm/s	Cumulative magnetic force grows quickly
Middle	15 → 17	Rising, but more slowly	Saturation and friction start to bite
Saturation	18 → 20	7.1 → 8.0 cm/s	Magnetic saturation, friction and magnet interactions

Fitted model constantsConstants found by the Python optimisation to minimise the error rate.

Constant	Value
Magnetic force scale A	3.86
Attenuation coefficient α	0.0175
Static friction μs	0.106
Kinetic friction μd	0.414
Magnet mass m	0.000632 kg
Gravitational acceleration g	9.81 m/s²
Stick-slip friction F_ss	0.0052

Try it

An interactive model of the work.

The data

Results, rebuilt from the original slides.

Model accuracy across three problemsError rates between each problem's theoretical prediction and the measured value, as reported in the decks.

Problem	Quantity predicted	Error range
Climbing magnets (2025)	Climbing speed vs RPM	0.66% – 8.28%
Autumn coin (2026)	Critical falling speed	4.0% – 19.4% (reported cases)
Twisted spaghetti (2026)	Optimal twist angle	11.1% – 17.4%

The club's research loopThe same sequence structured all three problem reports.

1Problem statement
2Phenomenon
3Qualitative analysis
4Quantitative model
5Hypothesis
6Experimental setup
7Tracker / sensor data
8Theory vs experiment
9Conclusion

Plain English

Terms worth knowing.

Stick-slip: Jerky motion where an object alternately grips (static friction) and suddenly slides (kinetic friction).
Eddy current: Swirling electric current induced in a metal when the magnetic field through it changes, for example when it spins next to a magnet.
Saturation: When adding more of something (speed, magnets, strands) produces smaller and smaller gains.
Flutter and tumble: A falling coin rocking side to side like a leaf, versus flipping end over end.
Critical speed: The falling speed at which the fluid's twisting push overpowers the coin's natural stability.
Reynolds number: A ratio of inertial to viscous effects in a fluid that tells you what kind of flow to expect.
Second moment of area: A measure of how a beam's cross-section resists bending; spread-out, locked-together material bends less.
Tracker: Free video-analysis software used to measure position, velocity and acceleration frame by frame.

Reflection

Across three seasons the lesson was the same: a model is only as good as its contact with data. Fitting the magnet constants in Python showed how a qualitative story becomes a predictive tool; the coin's 4 to 19 percent errors showed that a clean equation can still miss chaotic, regime-switching motion; and the spaghetti's consistently low optimum showed that a systematic error is itself a clue about the physics, here imperfect locking between strands. Becoming captain added a second lesson about scaling a team's ambition responsibly, from one squad to two.

SourcesEigen physics club, 2025 KYPT problem 4 'Climbing magnets' report deck, reporter Wonjun Lee (Google Drive PDF 1o-bIMx69n_x26AcRAIpKzYWwbe9jT_im); Eigen physics club, 2026 KYPT problem 12 'Autumn coin' report deck, reporter Wonjun Lee (Google Drive PDF 1UaOgqMl_eGjCPppsse8FnX6IEXRf8poB); Eigen physics club, 2026 KYPT problem 16 'Twisted spaghetti' report deck (final), reporter Wonjun Lee (Google Drive PDF 19WDG0qPZsovMj9lkqrr9SpwqBBJxjH8N); Project brief: award results, captaincy and team growth