Connected Masses & Pulley Systems · Guide · glidepathscience.org
GuideConnected Masses & Pulley Systems
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round, in the order a textbook would take them. Part 1 prepares the Intro track and Part 2 the Advanced track; each takes about twenty minutes to read. Each section ends with a worked example that is not one of the sim's own problems.
Part 1 · The essentials
One string, one system
A dropped ball falls 1.2 m in half a second, too fast to time by hand. In 1784 George Atwood slowed gravity down without changing it: he hung two nearly equal masses from a string over a pulley, and the pair sped up gently enough to time. Every setup in this sim is a set of bodies tied together by strings, and the first step is to see why such a set behaves like one object.
Start with the string. A light string that does not stretch has a fixed length, so over a fixed pulley, if $m_2$ goes down 10 cm, $m_1$ goes up 10 cm: the two share one speed and one acceleration along the string. This is a constraint, a fact of geometry that ties the motions together before any force is considered, and it lets one acceleration a describe the whole machine.
Now treat both masses and the string as one system. Newton's second law says that the net force along the string from outside the system equals the total mass times a. Gravity pulls $m_2$ forward along the string with $m_2 g$ and $m_1$ backward with $m_1 g$; the string's pulls are internal and cancel. So
$a = (m_2 - m_1)g/(m_1 + m_2)$
Here g = 9.81 m/s² and a is the acceleration of every part of the system along the string. Only the difference of the weights drives the motion, but the whole mass has to be accelerated. That ratio is Atwood's trick.
Equal masses have no driving force at all. Nudge one and let go: the weights balance at every position, so the pair keeps the speed you gave it until one lands, with nothing slowing it or pulling it back to level. That is Newton's first law for the whole system.
The usual slips are dividing by one mass and letting the heavier mass fall freely. The next question is how hard the string pulls.
The tension, from one body at a time
The system's equation hides the string, whose pulls cancel inside the system. To find the tension T, the string's pull on what it is tied to, look at one body alone. Over a light, freely turning pulley a light string pulls with the same T at both ends.
The lighter mass $m_1$ feels its weight $m_1 g$ down and the pull T up, and it speeds up going up: $T - m_1 g = m_1 a$. The heavier mass speeds up going down: $m_2 g - T = m_2 a$. Either gives T once a is known:
$T = m_1(g + a) = m_2(g - a)$ $T = 2m_1 m_2 g/(m_1 + m_2)$
The second form comes from putting the system's a into the first. Read the first form carefully: while the masses accelerate, the tension is more than the lighter weight, or $m_1$ could not speed up on its way up, and less than the heavier weight, or $m_2$ could not speed up on its way down. It lies between the two weights and equals neither. The tension equals a hanging weight only when that mass is not accelerating: at rest, or moving at a constant speed. What decides is the direction of the acceleration, not of the motion: a mass that falls but slows down is accelerating upward, so the string pulls it with more than its weight.
Test the formulas at their extremes. If $m_2$ is much heavier than $m_1$, then a approaches g: the heavy mass falls almost freely, and the string has to pull $m_1$ upward at nearly g, so T approaches $m_1(g + g) = 2m_1 g$, not $m_2 g$. If the masses are equal, a = 0 and T = mg. The pulley's support carries 2T, less than the total weight while the masses accelerate, because the system's center of mass is speeding up downward.
Saying that the string "holds up the heavier mass" is the most common mistake here: it would give $T = m_2 g$, and then $m_2$ could not fall at all. With one string understood, we can add more of them.
A train of carts: each link pulls what is behind it
Link several carts with light strings on a level track and pull the front one with a force F. The whole train moves as one system, so $a = F/\Sigma m$, where $\Sigma m$ is the total mass of the carts. Each link has its own tension, and the tensions are not equal.
Cut the train just behind a link: the only horizontal force on the rear part is that link's pull, and the rear part accelerates at a, so T = (mass behind) × a. The front link carries the most, the last link only accelerates the last cart, and the pull is neither passed along whole nor shared out equally.
Friction changes the acceleration but, surprisingly, not the pattern. If every cart has the same coefficient $\mu_k$, each cart's friction is proportional to its mass, and the tensions come out as F × (mass behind)/$\Sigma m$, the same with friction as without it.
A hanging mass can do the pulling instead of a hand, and then gravity and the table share the work.
Pulled across a table, with and without friction
Set a block $m_1$ on a table and tie it over a pulley at the edge to a hanging mass $m_2$. The string bends around the pulley, but the block still moves exactly as far as the hanging mass drops. The only driving force along the string is the hanging weight $m_2 g$, since the table holds up the block's weight. On a frictionless table,
$a = m_2 g/(m_1 + m_2)$ $T = m_1 a$
The tension comes from the block alone, which has only the string pulling it along. It is always less than the hanging weight, so the hanging mass speeds up on its way down, and a is always less than g, however light the block.
Now give the table friction. While nothing moves, static friction on the block matches the hanging weight exactly, pointing away from the pulley, up to a limit $\mu_s N = \mu_s m_1 g$, where $\mu_s$ is the coefficient of static friction for the two surfaces. The system starts only if the hanging weight beats that limit, $m_2 > \mu_s m_1$. Once the block slides, friction is kinetic, a steady $\mu_k m_1 g$ against the motion, with $\mu_k$ usually smaller than $\mu_s$:
$a = (m_2 - \mu_k m_1)g/(m_1 + m_2)$
Static friction is not a fixed force: below the limit it is exactly what is needed, and only at the edge of slipping is it $\mu_s N$.
Using $\mu_k$ to decide whether the block starts, or $\mu_s$ once it slides, is the mistake to avoid. Tilt the table into a ramp and friction gets a second job: deciding which way the block goes.
Up the ramp, down the ramp, or stuck
Put the block on a ramp at angle θ, tied up the slope over a pulley at the top to a hanging mass $m_2$. The block's weight splits into a part $m_1 g\sin\theta$ down the slope and a part $m_1 g\cos\theta$ into the surface, which the normal force balances. So the hanging weight competes with $m_1 g\sin\theta$, not $m_1 g$. Without friction, the hanging mass wins if $m_2 > m_1\sin\theta$, and then
$a = (m_2 - m_1\sin\theta)g/(m_1 + m_2)$
With friction, the normal force is $N = m_1 g\cos\theta$ and static friction can hold up to $\mu_s m_1 g\cos\theta$, pointing either way along the slope: down the slope if the hanging mass is winning, up the slope if the block is. The system stays at rest for every hanging mass in a band, from where the block just holds against sliding down to where it just holds against being dragged up:
$m_1(\sin\theta - \mu_s\cos\theta) \le m_2 \le m_1(\sin\theta + \mu_s\cos\theta)$
Inside the band nothing moves, even when the hanging weight is clearly larger than the slope's pull: friction decides whether anything moves, and which way.
Two mistakes are common: comparing $m_2$ with $m_1$ instead of $m_1\sin\theta$, and using $N = m_1 g$ on a slope. So far every pulley has been light. A real pulley has mass, and spinning it up takes some of the effort.
A pulley with mass
A real pulley turns with the string, and if the string does not slip, its rim moves as fast as the string: α = a/R, with R its radius. Spinning it up takes a net torque, which only the two tensions at opposite sides of the rim can supply, so they must differ: $(T_2 - T_1)R = I\alpha$, where I is the pulley's moment of inertia. With α = a/R,
$T_2 - T_1 = (I/R^2)a$ $a = \Delta m\,g/(\Sigma m + I/R^2)$
where $\Delta m = m_2 - m_1$ and $\Sigma m = m_1 + m_2$. The pulley behaves like extra mass, $I/R^2$, that the string must accelerate but gravity does not pull along the string. For a uniform disk of mass M, $I = \tfrac{1}{2}MR^2$, so $I/R^2 = \tfrac{1}{2}M$; for a thin ring with all its mass at the rim, $I/R^2 = M$. The radius cancels: only the pulley's mass and how it is spread matter. Each tension still comes from its own mass, now with the slower a.
Adding the whole pulley mass rather than $\tfrac{1}{2}M$ for a disk, or leaving the pulley out, are the usual slips. A measurement settles which is right.
Timing a fall, and Atwood's measurement of g
With a constant acceleration, the distance from rest after a time t is $d = \tfrac{1}{2}at^2$, so a timed drop gives the acceleration,
$a = 2d/t^2$
A measured acceleration includes everything that really happens: the pulley's inertia, its bearing, the exact masses. Solving the machine's equation for g turns it into a value of g:
$g = a(\Sigma m + I/R^2)/\Delta m$
This form neglects bearing friction. Nearly equal masses make a small and the fall slow, at a price: the small $\Delta m$ in the denominator magnifies every error, in the masses, in the time (which enters squared) and in anything the formula leaves out: a pulley's inertia, which belongs in $\Sigma m + I/R^2$, or its bearing's drag, which comes off the driving force (Part 2). A careful measurement therefore fits a line through many runs.
Forgetting the 2 in $a = 2d/t^2$, which halves every result, is the slip to watch for. Part 2 opens up the pulley and the bearing, and then goes on to strings that slide, pulleys that move and a chain that runs away.
Part 2 · Going further
Body by body: the rotating pulley
The heavy-pulley result is best derived one body at a time, the method that works for every harder problem. There are four unknowns: a, $T_1$, $T_2$ and α. Newton's second law gives one equation for each mass, $T_1 - m_1 g = m_1 a$ and $m_2 g - T_2 = m_2 a$. The rotational form, $\Sigma\tau = I\alpha$, gives one for the pulley, $(T_2 - T_1)R = I\alpha$. The no-slip constraint, a = αR, is the fourth. Adding the two linear equations to the rotational one divided by R eliminates both tensions and leaves $a = \Delta m\,g/(\Sigma m + I/R^2)$.
Every pulley has $I = kMR^2$, with k set by its shape: ½ for a uniform disk, 1 for a thin ring. So $I/R^2 = kM$, and a ring slows the masses more than a disk of the same mass, since its mass is farther from the axle. In energy terms, the $\Delta m\,gh$ released in a drop h is shared between $\tfrac{1}{2}(\Sigma m)v^2$ in the masses and $\tfrac{1}{2}I\omega^2 = \tfrac{1}{2}(I/R^2)v^2$ in the spin, so the pulley's share is $(I/R^2)/(\Sigma m + I/R^2)$ at every moment.
The tensions on the two sides can differ only because the rim grips the string. Take the wheel away and the rope must slide, and then friction itself sets the ratio of the tensions.
A rope over a fixed rod: the capstan rule
A rope over a fixed rod or a branch cannot turn it, so it slides. A short piece of rope touching the rod over a small angle dφ is pressed against it by its two end tensions with a force T dφ, so friction can change the tension across it by at most dT = μT dφ. Friction grows with the tension it acts on, so the tension grows exponentially around the contact:
$T_{big} = T_{small}\,e^{\mu\varphi}$
where φ is the wrap angle in radians (π for a rope over the top). At rest, the rope holds while $T_{big}/T_{small} \le e^{\mu_s\varphi}$. Sliding, the ratio is fixed, $T_{big} = e^{\mu_k\varphi}\,T_{small}$, and with $m_2$ heavier the masses move at
$a = g(m_2 - e^{\mu_k\varphi}m_1)/(m_2 + e^{\mu_k\varphi}m_1)$
This is the Euler–Eytelwein formula: each extra turn multiplies the grip by the same factor, which is how a few turns around a bollard hold a ship, and the same grip lets a string turn a real pulley.
Treating the grip as growing in proportion to the angle, or putting the angle in degrees, are the errors here. A turning pulley avoids sliding friction, but its bearing has friction of its own.
Bearing friction, and g from a line
A real pulley's bearing resists with a small torque: up to $\tau_s$ while it is still and a steady $\tau_k$ while it turns, acting at the rim like a force $\tau/R$. A nearly balanced machine does not start unless the imbalance's torque beats $\tau_s$, and once it moves the drag comes off the driving force:
$\Delta m\,gR > \tau_s$ $a = (\Delta m\,g - \tau_k/R)/(\Sigma m + I/R^2)$
Friction is not inertia: inertia stores energy as $\tfrac{1}{2}I\omega^2$ and gives it back, while bearing friction turns it into heat, and it matters most for the small imbalances Atwood's machine runs on.
The way out is to measure many runs and fit a line. Keep the total $\Sigma m$ fixed and move small washers from one side to the other, so that only $\Delta m$ changes. Then $M = \Sigma m + I/R^2$ is the same for every run and a is a straight line in $\Delta m$, $a = (g/M)\Delta m - \tau_k/(RM)$. The slope gives g/M, so g = slope × M, and the intercept gives the bearing: $\tau_k = -\text{intercept} \times RM$. The line crosses a = 0 at $\Delta m_0 = \tau_k/(gR)$, the imbalance that just matches the drag. Changing $\Sigma m$ between runs would change M and break the line.
Reading the slope as g itself, without multiplying by M, is the error to avoid. Ramps have their own map of where a system holds, and it has surprises at both ends.
The full incline map, and a launch against the drive
The holding band of Part 1, $m_1(\sin\theta - \mu_s\cos\theta) \le m_2 \le m_1(\sin\theta + \mu_s\cos\theta)$, has a lower edge that can be negative. That happens when $\tan\theta < \mu_s$: the ramp is gentle enough that the block would sit on it with no string at all, so no hanging mass, however light, lets it slide down. The lower edge is then zero, and the band runs from 0 to its upper edge.
Now let the drive, $A = m_2 g - m_1 g\sin\theta$ (up the slope positive), point down the slope, and launch the block up the slope anyway. On the way up, kinetic friction, $f_k = \mu_k m_1 g\cos\theta$, points down the slope too, so the drive and friction add and the block slows quickly, $a = -(|A| + f_k)/(m_1 + m_2)$. At the top of its path it stops for an instant, and static friction takes over: if $|A| \le \mu_s m_1 g\cos\theta$ it stays there for good. Otherwise it slides back, now with friction pointing up the slope, against the drive, so its acceleration is smaller, $a = -(|A| - f_k)/(m_1 + m_2)$. The v(t) graph is two straight lines with different slopes, steep and then gentle, meeting at zero.
Expecting the same slope both ways, or a block that stays wherever it stops, misses the role of friction's direction. The next tools change the geometry instead: strings that run through moving pulleys.
Movable pulleys, constraints and tackles
Run the block's string over the edge pulley, down under a movable pulley carrying the hanging mass, and up to a fixed hook. Two strands now hold the hanging pulley and both shorten when it rises, so when the block moves x the hanging mass moves x/2: $a_2 = a_1/2$. Two tensions pull it up, $m_2 a_2 = m_2 g - 2T$, while the block on a frictionless table gives $T = m_1 a_1$. Together,
$a_2 = a_1/2$, $a_1 = 2m_2 g/(4m_1 + m_2)$
A block and tackle extends the idea. A gun tackle has two single blocks, a luff tackle a double and a single, and a double tackle two doubles. Rove to disadvantage, the strands holding the moving block equal the sheaves; rove to advantage, there is one more. Ideally the n holding strands share the load, so the hauler pulls W/n and takes in n meters of rope per meter of lift: force is traded for distance. Raising a load slowly, with the moving block's own weight neglected, each sheave passes on only a fraction η of the tension, so counting from the hauling part (j = 0), strand j carries $P\eta^j$ and $W = P\sum\eta^j$ over the holding strands: j = 0 to n − 1 when rove to advantage (the hauling part leaves the moving block and holds), j = 1 to n when rove to disadvantage. The lost work becomes heat in the sheaves.
Forgetting the constraint, so that the hanging mass moves as fast as the block, is the classic error. A chain over a peg breaks a different assumption: its drive is not constant.
A chain over a peg: a force that grows
Hang a uniform chain of length L and mass per meter λ over a smooth peg much smaller than the chain (the short arc over the peg is neglected), its long side y past halfway. The extra length 2y on the long side, weighing 2λyg, drives the whole chain, of mass λL, and every link moves along the chain with acceleration $y^{″}$:
$y^{″} = (2g/L)y$ $y = y_0\cosh(\sqrt{2g/L}\,t)$
for a chain released from rest at $y_0$. The drive grows as the chain moves, so the chain runs away exponentially and $d = \tfrac{1}{2}at^2$ does not apply.
The cosh solution has an edge. The chain has to turn around the peg at speed v, and bending a moving chain takes an inward force $\lambda v^2$ that only the tension at the top can supply. The peg's push is $N = 2(T - \lambda v^2) = \lambda gL - 2\lambda(y\,y^{″} + v^2)$, and it falls as the chain speeds up. It reaches zero at $y^2 = (L^2/4 + y_0^2)/2$, before the chain could slide off at y = L/2, and from there the chain lifts away from the peg. A model carries its domain with it: this one holds while N > 0.
Treating the chain like a constant-force problem is the trap. The last system has two degrees of freedom, and a better method than force-by-force bookkeeping.
The double Atwood, by Lagrange
Hang $m_1$ on one side of a fixed pulley and, on the other, a light movable pulley carrying $m_2$ and $m_3$. Two coordinates describe it: x, how far $m_1$ has dropped, and y, how far $m_2$ hangs below the movable pulley. The Lagrangian, K − U written in x and y, gives two equations with no tensions at all, and their solution is neatest with the reduced mass $\mu = m_2 m_3/(m_2 + m_3)$:
$a_1 = g(m_1 - 4\mu)/(m_1 + 4\mu)$, $\mu = m_2 m_3/(m_2 + m_3)$
with down positive for $m_1$. The lower pair acts on $m_1$ like a single mass $4\mu$, which is less than $m_2 + m_3$ unless the two are equal. If the movable pulley were held still, the unequal pair would accelerate relative to it with its center of mass speeding up downward, and its string would carry only $4\mu g$, less than the pair's weight. Free to move, the string above the pulley carries $T_1 = 2T_2$ with $T_2 = 2\mu(g + a_1)$, so $T_1 = 4\mu(g + a_1)$: just what a single mass $4\mu$ rising at $a_1$ would need. That is less than the pair's weight in the example below, but a much heavier $m_1$ pulls the pulley up fast enough that $T_1$ exceeds it.
Comparing $m_1$ with $m_2 + m_3$ instead of $4\mu$ gives the wrong direction here. Every result so far rests on idealizations, and the last section asks where they stop being good enough.
Where the model ends
The sim's strings are massless, never stretch and never slip; its masses hang straight; its bearing friction is a fixed torque; there is no air. A real string stretches, so the two sides start a moment apart. It slips on its pulley if the rim's grip, $e^{\mu\varphi}$, cannot supply the tension ratio the pulley needs. Real bearing friction grows with the load, $T_1 + T_2$, and with speed. A heavy rope's tension changes along it, a swinging mass needs extra tension, and air drag matters for light, fast masses. None of this makes the ideal model wrong: it is a model with a domain, excellent for slow machines and light strings, and the start to which each correction is added.
Reference
Common mistakes
- The tension as a weight. While the masses accelerate, T lies between the hanging weights and equals neither.
- Dividing by one mass. The net force accelerates the whole system: divide by $m_1 + m_2$, plus $I/R^2$ for a heavy pulley.
- The same pull in every link. Each link carries (mass behind) × a.
- Comparing $m_2$ with $m_1$ on a ramp. Compare it with $m_1\sin\theta$, and use $N = m_1 g\cos\theta$.
- Mixing up the coefficients. $\mu_s$ decides whether it starts; $\mu_k$ acts once it slides; a capstan's grip is $e^{\mu\varphi}$, with φ in radians.
- The whole pulley mass. A disk adds $\tfrac{1}{2}M$, a ring M; a massless pulley adds nothing.
- Losing the 2. From rest, $a = 2d/t^2$.
- Ignoring the constraint. A movable pulley moves half as far as the string pulled through it.
- Balancing a double Atwood by weight. Compare $m_1$ with $4\mu$, not with $m_2 + m_3$.
Equation sheet
| Part 1 | |
|---|---|
| Atwood | $a = (m_2 - m_1)g/(m_1 + m_2)$ |
| tension | $T = m_1(g + a) = m_2(g - a)$ |
| train | $a = F/\Sigma m$; each link: T = (mass behind) × a |
| table | $a = (m_2 - \mu_k m_1)g/(m_1 + m_2)$; it starts only if $m_2 > \mu_s m_1$ |
| ramp | $m_1(\sin\theta - \mu_s\cos\theta) \le m_2 \le m_1(\sin\theta + \mu_s\cos\theta)$ |
| heavy pulley | $a = \Delta m\,g/(\Sigma m + I/R^2)$; disk $I/R^2 = \tfrac{1}{2}M$, ring $I/R^2 = M$ |
| from rest | $d = \tfrac{1}{2}at^2$, so $a = 2d/t^2$; $g = a(\Sigma m + I/R^2)/\Delta m$ |
| Part 2 | |
|---|---|
| tensions | $T_2 - T_1 = (I/R^2)a$ |
| capstan | holds while $T_{big}/T_{small} \le e^{\mu_s\varphi}$; sliding $T_{big} = e^{\mu_k\varphi}\,T_{small}$ |
| bearing | starts only if $\Delta m\,gR > \tau_s$; at fixed Σm, $a = (g/M)\Delta m - \tau_k/(RM)$ |
| movable | $a_2 = a_1/2$, $a_1 = 2m_2 g/(4m_1 + m_2)$ |
| tackle | strand j carries $P\eta^j$; $W = P\sum\eta^j$ over the holding strands |
| chain | $y = y_0\cosh(\sqrt{2g/L}\,t)$, while $N > 0$ |
| double Atwood | $a_1 = g(m_1 - 4\mu)/(m_1 + 4\mu)$, $\mu = m_2 m_3/(m_2 + m_3)$ |
Symbols: $m_1$, $m_2$, $m_3$ masses (kg) · $\Sigma m$ their total, $\Delta m = m_2 - m_1$ · g = 9.81 m/s² · a acceleration along the string · T tension (N) · $\mu_s$, $\mu_k$ coefficients of static and kinetic friction · θ ramp angle · I/R² a pulley's inertia as a mass (kg) · M a pulley's mass, or $\Sigma m + I/R^2$ in the fit · τ torque (N·m; $\tau_s$ and $\tau_k$ for a bearing) · φ wrap angle (rad) · η the fraction of tension a sheave passes on · P the hauler's pull, W the load's weight · L, λ, y a chain's length, mass per meter and offset · μ (double Atwood) $m_2 m_3/(m_2 + m_3)$.
Further reading
Samuel J. Ling, Jeff Sanny and William Moebs, University Physics Volume 1 (OpenStax, 2016), free online under CC BY-NC-SA 4.0: Section 6.1, Solving Problems with Newton's Laws, and Section 6.2, Friction, for connected bodies and Atwood's machine; Sections 10.4–10.7 for moments of inertia and rotation about a fixed axis. J. R. Taylor, Classical Mechanics (2005), chapter 7, for Lagrange's equations and the double Atwood machine. G. Atwood, A Treatise on the Rectilinear Motion and Rotation of Bodies (1784). The Field Manual lists every source and value the sim uses.
Connected Masses & Pulley Systems v1.3 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org