Electropermanent Magnets: Programmable Magnets with Zero Static Power Consumption Enable Smallest Modular Robots Yet

Electropermanent Magnets Enable Programmable Matter Robots (Robot Pebbles)

Take a moment and envision an electromagnet: a simple coiled wire driven by a hefty electrical current gives a fully-programmable magnetic field strength (on, off, and everything between).  Electromagnets are ubiquitous, but it turns out that there is a little-known device with similar functionality yet zero static power consumption -- they are called electropermanent magnets, and they've been around and in use since the 1960's!  A 2010 PhD thesis by MIT Media Lab's Ara Knaian examines the physics, scaling, trade-offs, and several new actuator designs (eg. stepper motors) using these little-known wonders.  Recently, electropermanent magnets facilitated an innovation in "programmable matter," where they were instrumental in creating the world's smallest self-contained modular robots to date (12mm/side).  Read on for details about this fascinating technology, along with discussions about existing and possible robotic applications.


Electropermanent (EP) Magnets for Programmable Matter:


Electropermanent magnets are an innovation in "programmable matter."  Specifically, they are instrumental for actuation in the world's smallest self-contained (energy storage, actuation, communication, computation) modular robots to date -- measuring just 12mm per side.  These robots were presented in an ICRA 2010 paper titled, "Robot Pebbles: One Centimeter Modules for Programmable Matter through Self-Disassembly" by Kyle Gilpin, Ara Knaian, and Daniela Rus. Externally, EP magnets look very similar to electromagnets (a coil of wire), but they differ in some fundamental ways (rather technical and non-robotic -- more materials science) that will be examined in later.  But first, let's take a look at the Robot Pebbles.

Electropermanent Magnets Enable Programmable Matter Robots (Robot Pebbles)


Electropermanent Magnets Enable Programmable Matter Robots (Robot Pebbles) Electropermanent Magnets Enable Programmable Matter Robots (Robot Pebbles)

Programmable Matter using Electropermanent Magnets  Programmable Matter using Electropermanent Magnets


Electropermanent magnets are instrumental for modular robots of this size, allowing Robot Pebbles to (programmatically) self-adhere in a scalable fashion (as devices are miniaturized) with relatively low power consumption.  This is in contrast to competing techniques that either use electromagnets (power-hungry) or permanent magnets with a mechanically actuated disconnect latch (bulky and poor scaling). 

While plenty of people are working on nano-robots (usually computation-less devices operating in externally-controlled electric/magnetic fields), it is encouraging that progress is still being made in modular robot miniaturization.  Robot Pebbles have onboard computation (microcontroller), energy storage (via surface mount capacitor), actuation (via EP magnets), and communication / power transfer (via iron electrical contacts on the EP magnets) for fully self-contained operation -- this a small but definite step towards true programmable matter.

Realistically, true programmable matter is still relegated to the distant future; however, I'd really like to have a few of these as a desktop toy.  Unfortunately, I doubt that 1cm EP-magnet modular robots will become commercially available any time soon.  However, Cubelets offer a nice near-term commercial option if you're willing to settle for a bit more bulk -- they're supposed to start shipping in January 2011.

Cubelets Modular Robots



Electropermanent Magnet Operating Principles:

I first learned about EP magnets from Ara Knaian's PhD Thesis, titled "Electropermanent Magnetic Connectors and Actuators: Devices and Their Application in Programmable Matter," which also included the Robot Pebbles from the above.  Ara's thesis was advised by Dr. Daniela Rus (Robotics) and Dr. Neil Gershenfeld (Physics) from the MIT Media Lab -- so it has a nice balance in perspective between the two disciplines.  To summarize the operating principle:

An electropermanent magnet is a solid-state device which allows an external magnetic field to be modulated by an electrical pulse. No electrical power is required to maintain the field, only to do mechanical work or to change the device's state. The electropermanent magnets described in this thesis contain two magnetic materials, one magnetically hard (e.g. Nd-Fe-B) and one semi-hard (e.g. Alnico), capped at both ends with a magnetically soft material (e.g. Iron) and wrapped with a coil. A current pulse of one polarity magnetizes the materials together, increasing the external flow of magnetic flux. A current pulse of the opposite polarity reverses the magnetization of the semi-hard material, while leaving the hard material unchanged. This diverts some or all of the flux to circulate inside the device, reducing the external magnetic flux.


Electropermanent Magnet Design  Electropermanent Magnet Design


For those with a bit of physics background, these B-H curves tell the story: 

B-H Curves for Electropermanent Magnets


For electrical engineers (like myself), an electrical circuit analogy is useful as I rarely work with magnetic circuits.  Imagine that you have two parallel resistors (magnets) of approximately equal value.  One resistor's value is easily increased or reduced (the low-coercivity "semi-hard" magnet) by applying a temporary directional magnetic field (eg. by a coil of wire), and this resistor will retain its new value after the field is removed (magnetic remanence).  The other resistor's value cannot be changed (the high-coercivity "hard" magnet).  By programmatically supplying current to the coil of wire (pulse strength and duration), we can achieve any desired resistance (within reason) and it will remain constant even after the current is disconnected from the coil of wire.  This is what is happening in the electropermanent magnet, only the resulting effect is a programmable magnetic field strength with zero static power consumption -- cool!  The resulting magnetic field can be used to apply static forces and/or do additional work -- such as clamping (left), gap-closing (middle), or side-drive operation (right). 

Electropermanent (EP) Magnet Electropermanent magnet Forces


It is important to note that physical work is being done to to alter magnetic domains in the semi-hard magnet (magnetic remanence).  A detailed analysis of the thermodynamic cycle of these devices is presented in Ara Knaian's thesis, but in effect, there is no free lunch: electropermanent magnets are only superior to electromagnets if the time between switching is not too short. To quote:

Electropermanent magnets use higher instantaneous power but lower overall energy than electromagnets, with break-even times in the milliseconds at centimeter scale and in the microseconds at millimeter scale according to our analysis. Electropermanent magnets are less sensitive than electromagnets to lower winding fill fractions and lower conductivity wire, making them more amenable to microfabricated coils. 

EP magnets aren't exactly new.  They were developed in the 1960's to build "ferreed switches" (left) used in telephone switching equipment (middle).  Today, they are still used for work-holding chucks (right) and some large "electromagnet" cranes.  You can find a wealth of information about these designs in Ara's thesis as well.

Ferreed Switch Ferreed Switch Telephone equipment  Work-holding chuck using EP magnets


The main contribution of Ara's thesis was the physics, scaling, trade-offs, and several new actuator designs using EP magnets.  We've already seen his work on Robot Pebbles, but he also worked on electropermanent stepper motors!


Electropermanent Magnet Stepper Motors:


One very interesting outcome of Ara's thesis was the development of (patent pending) EP magnet stepper motors.  This tiny "wobble" stepper motor uses EP magnets in place of traditional stator windings (electromagnets).

Electropermanent magnet stepper motor  Electropermanent magnet stepper motor

Electropermanent magnet "wobble" stepper motor


By switching the state of nearby EP magnets, the motor will step through various (rotational) configurations.   

Electropermanent magnet "wobble" stepper motor


Owing to the unique properties of EP magnets, this type of stepper motor offers some interesting benefits.

  1. At low RPM (very little on-off switching), EP magnet motors have higher efficiency compared to electromagnetic motors.
  2. At zero RPM ("stall"), EP magnet motors still outputs constant force / torque.  In electromagnetic motors, stall conditions result in massive power consumption -- this is a major problem for motor startup.


Personally, I think EP magnet stepper motors could be very well-suited to hub motors (like those done by Charles Guan), as they are notoriously low-RPM.  What sort of applications can you imagine?




Straight from Ara Knaian's thesis:

Electropermanent magnets can have their holding force switched on and off by the application of a momentary electrical pulse. Electropermanent magnets have low power consumption and temperature rise compared to electromagnets, especially at small length scale. Electropermanent magnets require energy proportional to their volume and hold with force proportional to their area, so fundamental scaling favors their low-energy operation at small dimensions.

At centimeter scale, electropermanent magnets can hold hundreds of times their own weight, can exert force comparable to their weight from a distance comparable to their length, and switch is a fraction of a millisecond. At millimeter scale, scaling laws predict each of these metrics should improve.

Electropermanent magnets use higher instantaneous power but lower overall energy than electromagnets, with break-even times in the milliseconds at centimeter scale and in the microseconds at millimeter scale according to our analysis. Electropermanent magnets are less sensitive than electromagnets to lower winding fill fractions and lower conductivity wire, making them more amenable to microfabricated coils.

Electropermanent magnets are stronger than breakdown-limited electrostatic plates in air, operate at much lower, more practically switched voltages, and allow larger air gaps for higher resistance to dust and contamination. On the other hand, they have a higher profile and use more energy to switch.

All of the above conclusions make electropermanent magnets a promising building block for actuators in the next generation of batch-fabricated, millimeter-scale robotic systems.


Incidentally, Ara's thesis is one of the most comprehensive and impressive I have ever read (I've read it twice).  Aspiring to make my own as impressive, I am shifting most of my focus onto my thesis for the coming months.  It is likely that my Hizook output will taper off in the meantime.  I will try to get guest posters to fill in occasionally (want to help?!), but I'll be back...




Thanks for a fantastic post, I just discovered this great blog (I like several other posts to, like the one about Paco Spiralift telescoping linear actuator, great stuff!)

I am very interested in (Self-reconfiguring) modular robots and I actually attended ICRA and saw Kyle's presentation and I interviewed him on the programmable matter pebble and the connector for my podcast. That episode hasn't been released yet but you can check out the episodes that have been released here: http://itc.conversationsnetwork.org/series/flexible.html and I will get back to you when Kyle's episode is released).

I am also very interested in Modular Robotics Cublets and have ordered one of the alpha kits (sadly they are now out of stock) and I will do lots of videos and tests when I get them. I also plan hold a competition on my blog where the units will be mailed, for free, to the winners for some "tinkering time" for a month or more!

As I am interested in Self-reconfiguring modular robots, I think a lot about the connectors and I think that the electropermanent magnetic and the electroadhesive connectors can drastically reduce the complexity of the requirements we have on the connectors. And I think that if you design around there weaknesses they can make great connectors for Self-reconfiguring modular robotics systems.

So again, thanks for a great blog!

Thanks for the many kind words Per. I look forward to hearing the podcast.

Oh, and for those not familiar with the notion of Programmable Matter, a video mockup by Intel Research explains the vision:


—Travis Deyle

Travis - this is getting interesting...

It looks like Per Sjoborg just got around to posting the podcast about Electropermanent Magnets with Kyle Gilpin -- the fellow predominantly responsible for the modular robots (if memory serves).  Go check it out on Per's website: FlexibilityEnvelope.

—Travis Deyle
As in the article I can also recomend Ara Knaian's PhD Thesis, titled
"Electropermanent Magnetic Connectors and Actuators: 
Devices and Their Application in Programmable Matter.
It is a great indeepth look at this.

can i buy this little ep Magnets?

There is a really cool article in Science titled "Cameleon Magnets" that discusses another way to tune a material's magnetic properties and (potentially) create "programmable magnets" -- this time turning a paramagnetic material (titanium dioxide) into a ferromagnet.  Here's a quote:

Tunable ferromagnetism (magnet)

Tunable ferromagnetism. Cobalt-doped titanium dioxide, (Ti,Co)O2, is paramagnetic. Yamada et al. show that applying a voltage to this material in an electrolytic cell causes it to become a ferromagnet. (A and B) With no applied gate voltage VG between the electrolytic top contact and the (Ti,Co)O2, the carrier density (electrons, e–) is low, and the Co2+ spins (violet arrows) interact weakly and are not aligned. (C and D) For VG at 3.8 V, the carrier density increases by nearly a factor of 10 and the electrons act as magnetic messengers, aligning the Co2+ spins. (E) Magnetization can be detected because it creates an imbalance in the electron spin populations.  For an electron current fl owing in the y direction, the spin-up electrons will tend to scatter to the left (–x direction) and the spin-down electrons to the right (+x direction). With more spin-up electrons, there will be a net accumulation of electrons on the left side of the sample, which in turn can be measured as an anomalous Hall voltage VH or Hall conductivity σ AH. Reversing VG to zero transforms the ferromagnet back into a paramagnet.


—Travis Deyle
This is an excellent report/blog/news it has gave me an idea for a project i'm working on brilliant work!
—Will C

Where can I find small NdFeB and AlNiCo as you use in small quantity? 

Any plan to release a Hacker DIY kit for the electropermanent magnets?



@HVP,  Yep!  We've been developing an open-hardware / open-software EP-magnet for hobbyists...  Keep an eye on Hizook for announcements!

—Travis Deyle

I had a chance to speak with Kyle Gilpin during my March 2012 visit to MIT (for Rod Brooks' retirement gathering).  He was very close to graduating, and had built a ton of EP magnet pebbles.  Well, about a month ago, his project was re-covered by IEEE Spectrum:

EP Magnet Smart Pebbles


Maybe I should be more proactive about covering all the cool robotics projects I get to see.  ;-)

—Travis Deyle

Hi there! This was a great post. I read Ara's paper on the Electropermanent magnets and their uses. 

I was just curious about some of the things that go into making them work.

While I understand a current is required to flip the polarities, I'm having trouble getting a reasonable answer for how much current. I'm getting something in the amps. Is that reasonable and is there a good way to minimize that?

Secondly, why is DC current able to switch polarities? Because it just looks like large DC pulses to me. 

I'm a student so some of my concepts might be off, but trying to use these for a project and would appreciate any guidance.




Your calculations are correct.  The current is on the order of amps, but for a very short duration.  Reading through Ara's thesis (specifically, section 3.5) you'll note that he's charging a 100uF capacitor to 20V.  To "flip" the AlNiCo magnet's polarity, he discharges the capacitor through a 20 uH (3.8 ohm) coil... which results in a 5.3 amp "DC" current surge for a mere 20 usec.   This should be well within the specs for various surface mount transistors -- as evidenced by the fact that Ara accomplished this using all surface mount parts.

As for switching the polarity of DC pulses... you need to look at an H-Bridge: a simple 4-transistor circuit that lets you control the direction of current through a load.  These are used to drive motors (forwards or backwords) and are exceedingly common.  

—Travis Deyle

Hey Travis,

Thanks so much for your response! It  was very helpful and I'm glad to hear I'm on the right track. 

Actually, my second question was a little more theoretical. I think I didn't express it right though. I was wondering why a DC pulse is able to change the magnetic properties ( i.e the magnetic domain orientations)? For some reason, I was under the impression that most magnetism is and its properties can be controlled by AC. Or is a really fast DC pulse simulating an AC current in a sense? 

Thanks again,



A basic description... during fabrication, all magnets are polarized by applying a strong magnetic field to align their internal magnetic moments -- and that magnetic field is often applied using electromagnets.  The direction of the applied field determines the polarity of the magnet.  Some magnets (eg. "hard magnets" like Nd-Fe-B) require much greater magnetic fields to align the polarity.  Others (eg. "semi-hard magnets" like Alnico) require a lower applied magnetic field to change their polarity.  

In the electropermanent magnet, when you apply a short pulse to the coil (ie. an electromagnet), the more susceptible "semi-hard" Alnico magnet's polarity is flipped whereas the "hard" Nd-Fe-B magnet's polarity remains unchanged.  When the fields of the two magnets (inside an EP magnet) are aligned, the EP magnet is "on".  When they're opposing one another, the EP magnet is "off".  That's all there is to it.

The physics behind this are mostly described by a magnet's susceptibility, coercivity, and hysteresis loop (this is described briefly in Section 2.7 and Figure 2-21 in Ara's thesis).  If you want a more detailed treatment, this UCSD article is a good start.

—Travis Deyle

Ahhh. Interesting. Ok that makes a lot of sense. the article was great too.

I promise I'm going to stop bothering you soon, but I have another question about the practical implementation of such a system. 

In Ara's system, he uses 20V to generate his 5.3 A current to switch polarities, whereas the OpenGrab Kickstarter project that links your website uses a voltage as high as 300V. I was wondering if there was any reason why the voltage needed to be that much higher? (His system is larger, but if everything were in series, would it matter?) 

Thanks so much for all your help!



For those who don't know, Andrew Jochum launched a (funded) Kickstarter campaign to build an EP Magnet "cargo lifter" for Unmanned Aerial Vehicles (UAVs).  It's actually a pretty clever application -- UAVs don't have the payload capacity to carry extra batteries for standard electromagnets, so EP magnets offer the best of both worlds.

@Yash, I'm not sure why he's using 300V.  There are a lot of interrelated variables in this system that a designer has play around with.... possible explanations: (1) The coil he's using has higher resistance either from the length of the windings or the wire gauge, so he needed more voltage to get higher currents; (2) If he's using thicker magnets, he may need to up the voltage to generate more magnetic flux to switch the magnets?  Frankly, I'm just speculating....  You could probably email Andrew Jochum if you want to understand his design.  

Of course, the more effective way to explore all this would be to setup some benchtop experiments. You shouldn't need much more than a decent power supply, some capacitors, switches, and a driving transistor.  Eg. this circuit (via CircuitLab):

EP Magnet Experimentation Drive Circuit

 Use the power supply to charge the capacitor (through the switch).  Disconnect the switch.  Apply a "on" signal to the control transistor.

—Travis Deyle

That actually is indeed my next step! Just ordered some parts for it today. 

Thanks so much Travis. I really appreciate all your help! 


Could an electropermanent magnet be made to multivibrate if both opposing directions of flux were made unstable? Could a low coercivity material placed between two NIB magnets with opposing flux directiion cause AC current in a coil wrapped around both magnets since BH delay would not hold one direction very long before pressure from the opposing magnet would shift flux. If AC results than something more stable might allow the electropermanent magnet to shift flux at lower power and higher speed.


@Anonymous: I'm hesitant to provide a "negative" or "affirmative" response to your questions, since I'm afraid they're a bit outside my depth.  Can the EP magnet be integrated into a multivibrator (eg. one-shot or oscillator) -- possibly.  I'm intrigued by the idea of tweaking the composition or adding more materials to the "stack" to permit faster switching at lower power.  

However, it seems like many of the properties are sufficient for steady state operation (ie. infrequent switching) already.  Looking at my previous comments (extracting details from Ara's thesis), the 100uF supply capacitor at 20V is only discharging a maximum of 20 milli-Joules, and it's switching the EP magnet's polarity in a mere 20 micro-seconds.

—Travis Deyle

Trying to source the small electropermanent magnets for a project that i'm working on. Does anyone know where i could get them from?

—Caleb Bramwell


Andrew Jochum built some EP magnets (for drones) via a successful Kickstarter campaign.  It looks like he's also launched a web store that sells small magnets: NicaDrone.

—Travis Deyle


It is so nice to read. I am looking for strong magnet, super fine around 200 micro diameter rod. Could anyone kindly help?  Many thanks.



Hi sir..I need details of electro permanent magnet connections for 36*36*20 sized Alnico5 magnet.. I need details of ampere turns, voltage, wiring details and all...plz do help me anyone...this is my 1st project to prove myself... mail id : abineshmech92@gmail.com

Dear sir,

how we can calculate force and no of magnet to lift 12000*2500*50mm plate.

we are looking  electropermanent magnet to lift plates from 6500to12000length*1500to3000mm*8to60mm

we want to lift with singal beam only.

if you guide us please inform

wating for youer reply

sandesh shah



i saw the research on the stepper motor . but there is no mension of how many times in a second can the magnet be turned on and of. I need a small 3mm diameter magnet to switch on and off 1 - 2 times per 5 seconds.. Is that possible to do?


There I can't think of any practical limitations on switching speed beyond power consumption.  I believe Ara's PhD thesis addresses this in terms of duty cycles: For lower duty cycles, the EP magnet is better; for higher duty cycles, the electromagnet is better.

In fact, at higher duty cycles, an EP magnet is (effectively) the same as an electromagnet -- except that you also have losses to overcome coercivity.

—Travis Deyle

Dear all,

I'm curious to test the tiny ep magnets for several applications. How can I get some of them?

Best regards