Close
0%
0%

Semisolid Metal Printing

Print metal just like plastic.

Similar projects worth following
Ever since the start of RepRap, we have collectively dreamed of printing metal the same way we print plastic. I was inspired by Lulzbot's experiments to start a student project at WPI, which eventually became a full-scale collaboration with Lawrence Livermore National Lab - and then the team gave up because it was hard. But I didn't give up. Now almost 8 years of tinkering later, the time for SMP has arrived.

SMP uses the properties of semisolid alloys and mixtures as well as wettability and temperature boundary conditions to precisely and passively control the fluid flow properties of metal alloys to enable stable extrusion. This project makes the technology compatible with common RepRap hardware and software, aiming for a total retrofit cost under $500.

-no more arduous post processing
-no more struggling with misshapen or brittle sintered parts
-no more $500K+ metal printing machines
-no more dealing with expensive, messy and hazardous metal powders

This technology can be used to print conducting traces and antennas like this:

Which was done on an earlier prototype of this technology. Since the tin-zinc alloy melts at 240C it is compatible with PETG printing and as long as your PETG doesn't offgas water vapor (dry it well before use), the mechanical interface should be strong between the materials. Solid metal parts can also be printed with this machine, which do not require any difficult post-processing steps like sintering.

System features:

  • Print bed which can go up to 200C (E3D's high temperature heat bed works fine for this)
  • Wettability-patterned hotend components
  • Optimized printing materials
  • liquid glass gasketing and liberal use of X-Pando to seal the liquid metal during extrusion
  • brush to guide metal onto substrate

You can find the Github page here.

The license for this project is Apache-2.0, and there is a provisional patent for an industrial version of the technology which can print aluminum, copper, and steel alloys, but anything that doesn't glow red hot and that a hobbyist would feel safe about using at home is fair game and open-source.

IMG_2039.JPG

An early test of this printing technology: not enough force could be developed to get additional layers to flow from the nozzle, but the first layer adhered just fine, and it could draw wires up from the bed. The new technology is capable of providing enough extrusion force.

JPEG Image - 1.10 MB - 10/26/2021 at 23:19

Preview
Download

BridgingTest.mov

An early test of the bridging capability of this technology on an earlier prototype in years past.

quicktime - 33.44 MB - 10/26/2021 at 22:50

Download

  • 1 × M6 threaded rod 20 cm of rod is plenty. Medium strength steel is sufficient.
  • 1 × Steel block, 60X60X30mm Any HSLA or mild steel will do, molybdenum or tungsten-bearing steels work the best and last a very long time
  • 1 × Aluminum block, 60X60X30mm I used 6061 aluminum but others may work
  • 1 × X-Pando Their standard product is fine: https://www.xpando.com/xpando.php and it can be purchased on McMaster. Seals way better than Teflon at high temperatures
  • 1 × E3D Thermistor Semitec 104GT when sold elsewhere

View all 16 components

  • A separate jamming issue

    Michael Perrone11/25/2021 at 18:01 0 comments

    Careful application of the X-Pando to the male threads, and replacement of the glycerol-borax flux with glycerol-zinc chloride flux removed the clogging issue at the nozzle, but now it appears that there is a new issue:

    As the wire approaches the heatbreak it softens and widens , apparently becoming semisolid. It then apparently mechanically sticks to the wall of the heatbreak. The length of material that does this seems a bit longer than the heatbreak itself, suggesting that most of the heat is being conducted along the metal wire. This means that the thermal gradient will be less steep than we would expect from thermoplastic printing, which means E3D's design would ultimately need to be re-optimized for metal printing; something I had hoped to avoid. I increased the clamping force to try and avoid stripping, but it still got stuck this way. Perhaps I should also reduce maximum acceleration of the extruder, to help prevent stripping, but the better thing would be to figure out the cause of the jamming

    I poked around in the nozzle with my 100 micron tipped engraving bits but found only metal this time: no nonmetallic blockages this time. If the temperature gradient at the nozzle tip is too steep, then perhaps granular jamming is occuring at the tip, but the melting point of this alloy is at least 20 degrees below the steady state extruder temperature when not in contact with a surface. When subsequently lifting off the surface, the blockage should re-dissolve readily in the alloy, allowing for extrusion again, but that's not what I observed: once it's clogged, if you cancel the print and move the extruder up in the Z axis, it remains clogged. There is a chance I am wrong about the temperature at the nozzle tip, so I will investigate this further

    I also noticed something else: the glycerol-borax solution I've been using had some aluminum hydroxide powder I mixed in then forgot about for a while. If it were a glassy material we would expect it to be clear

    The zinc chloride/glycerol deep eutectic solvent was clear for example, though highly viscous with the 50/50 wt% combination I used. I bet this deep eutectic would be excellent for rechargeable zinc batteries: that viscosity would really help reduce dendrite formation.

    But back to the jamming issue, I realized that the zinc chloride could have caused the tin-zinc to solder to the stainless steel, so I tested it:

    And indeed, the glycerol-zinc chloride acted as a flux which made the stainless steel solderable by 80/20 wt% tin-zinc. I'm sure this is useful for a number of projects here on Hackaday. It seems to allow soldering to freshly abraded surfaces better than surfaces that have had time to develop a stable oxide layer.

    With the knowledge that metallurgical bonding might be happening instead of just mechanical bonding, I realized that the molten metal was probably slowly creeping up the heat break and solidifying. Moreover, in thermoplastic printing it is standard to decrease the print temperature after printing the first layer: this could also cause jamming by allowing some of the print material to solidify higher up in the heatbreak as the temperature decreases. In all subsequent tests, I increased the printing temperature instead of decreasing it, but the jamming still occured. It's possible that long retracts could also pull liquid metal further up the heat break and allow it to solidify. In any case, zinc chloride is definitely a bad choice for this design, so I replaced it with just glycerol, which is still viscous enough to seal the surface when in the cold end.

    Unfortunately the X-Pando bonds too well between the heatsink and the heat break to be able to take them apart without damage, so I tried heating up the nozzle and wicking up the tin-zinc, but it was apparently fully solid in the heat break while at 300C. I drilled out the old one and tried to use it again with glycerol and an increased printing temperature after the first layer instead of a decrease, but it just jammed...

    Read more »

  • Jamming issue fixed

    Michael Perrone11/15/2021 at 23:06 2 comments

    Last project log, I ended with another jamming issue. This time it turned out to be pretty standard jamming compared to the last jamming issue. Even just looking at the nozzle under a microscope, it's easy to see a bunch of nonmetallic stuff there.

    And just to be thorough, I tried removing the filament while the extruder was hot, and it removed and re-inserted nice and cleanly, with a clear and stable interface between the glass gasket-wettable section and the metal-wettable section, enforced by the steep temperature gradient and patterned wettability at the interface. It wasn't so easy to capture with my camera, but was perfectly visible to the naked eye. You can kind of make out the smooth surface of the liquid metal though.

    And I was in for a little surprise too: apparently this tin zinc alloy expands a little bit as it cools: it must have expanded enough to produce 5000 PSI of pressure in order to leak through the X-Pando like that and form droplets. Later on when printing again though, even with no servicing to some of those joins, the leaks did not continue at lower pressures. I imagine the capillary pressure developed at those tiny radii plus in conjunction with the surface tension of the metal in the cracks is immense enough to keep them sealed under normal operating conditions. What surprised me though is that metal came out here, but nothing at all came out of the nozzle. It must have been extremely clogged.

    Speaking of the nozzle, it was time to figure out what went wrong with it. I slowly machined down into it, checking things out as I went

    As I got deeper into the nozzle, it became obvious that there were little nonmetallic specks throughout it. I was pleasantly surprised to see so few air pockets, because one failure mode I was worried about was getting a large air pocket lodged in the nozzle, which would have made extrusion very difficult to control. Maybe it was just because the thing was at at least 5000 PSI though?

    From looking at the chips directly, it's obvious that there are little grey or black nonmetallic specks in there. I hypothesized that these were bits of X-Pando that had leaked into the metal, because I had applied the X-Pando to the female pipe threads instead of the male pipe threads, which would have tended to push the X-Pando into the liquid metal instead of out to the nozzle.

    This was easy enough to test: I just added some wet X-Pando to the female threads, threaded the nozzle on, then removed it, and Viola! The X-Pando was all on top of tinned surfaces that would have liquefied when the nozzle was heated next, releasing that material to clog the nozzle. I thought about other mechanisms that could have caused a clog, and three came to mind: particulates in the actual metal wire, a reaction between borax from the liquid glass gasket and the zinc chloride in the Sta-Clean flux I had used to make the inside of the nozzle so wettable, and standard semisolid granular jamming. Particulates in the metal wire would be unlikely, because I kept it well away from wettable metals and metal powders like copper, iron and nickel as I made the wire, and because surface chemistry would have caused any organic contaminants to keep to the surface of the metal, where they would have popped out or at least become visible during wire rolling. I have no reason to suspect that the clog was related to borax reacting with zinc chloride at the moment, but perhaps I will consider changing the glass gasket recipe in the future to use lewis acids instead of bases. But for now, it seems all the liquid gasket material floats nicely on top of the molten metal and doesn't get pulled down through it appreciably. Nevertheless, it's something to watch out for. If the X-Pando hadn't caused the clog, granular jamming probably still would have: I realized as I was going over these things that the alloy I made had 30% weight zinc, when I needed 30 atomic percent, which would have been 19.1% by weight. So I...

    Read more »

  • Progress?

    Michael Perrone11/10/2021 at 21:48 0 comments

    Got slightly further before it jammed? I'll take it! It still jammed, and I'm pretty sure I know why: Prusa Slicer starts with the fan off for better adhesion between layers but I was using that same fan on the cold end of the extruder to keep that side from getting melty soo...

    Things got a little too melty. Also It looks like the default fan on the E3D nozzle is 24v, so I'll get a 12v one with ideally more CFM to get more cooling power on the cold end. I'll probably do a full teardown and diagnosis while waiting for those parts, in case I can think of a way to get this running before those arrive. The default oozebane settings might be way too high, causing liquid metal to jump into the cold end, among a few other possible issues.

    Really feeling this web comic right now...

  • Issue identified

    Michael Perrone11/04/2021 at 22:29 0 comments

    During the last test the metal filament kinked and the extruder clogged. I initially thought it was just due to the bend in the wire, but while the system was hot, the wire could spin freely, although it wouldn't budge when pushed in or pulled out.

    As soon as I noticed that the wire could spin freely I was pretty sure I knew what was going on, but I wanted to make sure, so I took apart the clogged extruder and sectioned the clogged cold end.

    As I suspected, the nozzle and heatbreak performed just fine, wetting and not wetting where expected. The anodization on the aluminum heatsink also performed admirably: the interface between the feedstock alloy and the aluminum is very clear and there is no metalurgical bond between the aluminum and the tin-zinc alloy

    Thankfully, just as I suspected, this was an easy thing to fix. Some metal flowed back up towards the cold end and solidified in a cylindrical cavity behind the back of the heat break. Then, because it was in the cold end, it could not melt again and flow through the nozzle. Some thermoplastic extruders encountered the same issue during their development in the RepRap project. It will just require some 3D modeling on my part and maybe some waiting for a few more components in the mail, but implementing the fix for this will be nothing new. But for those who are not familiar, I'll explain:

    If you look at the heatbreak of most modern nozzles, they all have this extra unthreaded section in the cold end.

     This design feature is key because when you want to tap threads into a hole that doesn't go all the way through the part, it's hard to get threads all the way to the back of the hole. Indeed as you can see above, the filled section doesn't really have threads. Bottoming taps can do a little better, but still aren't perfect. In practice, this means a fully threaded heatbreak can't thread all the way to the back of the hole, leaving a cylindrical cavity behind it. I thought my liquid glass would still be viscous enough to prevent metal from flowing back, but without a cooling fan on the heatsink, apparently the temperature gradient was not steep enough for that effect. Without active cooling, and with the cavity in a position where the feedstock alloy started getting melty, it's no surprise that this region quickly filled with liquid metal from further below, which subsequently froze and locked the wire in place.

    The unthreaded section of E3Ds heatbreaks is made to fill up the unthreaded or poorly threaded space at the bottom of the hole and prevent that area from filling up with plastic. It even has a lip at the back of it which bites into the aluminum to seal it well. This leaves no cavity where plastic or metal could potentially get stuck, reducing the possible failure modes of the hotend. So all I need to do is implement the same idea on my extruder. Since I've destroyed my heatsink to demonstrate the failure mode, I think I'll cannibalize the E3D extruder I have, or get one shipped quickly via Amazon, since this E3D nozzle was an old one meant for 3mm filament and I have 1.75mm metal filament instead. I would need to anodize the aluminum and black oxide coat the stainless steel depending on the grade, or if I get one of the new titanium heatbreaks then I would anodize that as well. I'll probably need to redesign and reprint most of the extruder system as well.

    And just to show that feedstock alloy has not welded onto the aluminum, we can remove it with no damage to the surrounding aluminum. What's especially fascinating about this is that the solidified melt around the filament actually detached cleanly from the filament wire. This demonstrates that it backflowed from further down in the extruder then solidified further up where it was cold. Indeed, the small finger of metal that shot up really far into the cold end may have been the thing actually jamming the nozzle, because further down I would have expected the metal in the cavity to at least...

    Read more »

  • The history of SMP and some shoutouts

    Michael Perrone11/02/2021 at 18:16 0 comments

    The first attempt at this type of printing was done by someone on the Lulzbot forum back in 2013. Back then they didn't give any particular consideration to the material properties of the molten metal, which is why they couldn't get past the clogging issues they were having.

    Later on in 2013 or 2014, some high school student who's name I forget tried semisolid metal printing with an antimony alloy. Were they the one who eventually went on to start Vader Systems? I forget but if anyone still knows where that info is, please do mention it in the comments.

    In 2015 I started a project at WPI which over the following years became a full scale collaboration with Lawrence Livermore National Lab and produced some papers. But the technology was abandoned: they attempted to print in the semisolid state, but semisolid metals undergo granular jamming under compression and shear, because they are not one homogenous phase. This meant that either the nozzle had to be very wide as with a concrete printer, or they had to implement something complicated to avoid granular jamming. Ultrasonic, induction and direct mechanical stirring were proposed, but each had its pitfalls, either making the system more expensive than other metal printing technologies, or outright not being feasible. I've heard whispers that Desktop Metal also attempted Semisolid Metal Printing, but evidently they got no further than LLNL.

    While at Voxel8 in 2015 and again in 2016, I attempted to further develop semisolid metal printing in order to print traces for use in 3d printed circuit boards. This time, it was abandoned by a fluke more than any engineering bottleneck: the company pivoted to printing shoes, and therefore didn't need the capability to print metal anymore. I learned what I could from that last prototype, and continued thinking of a way to solve all of its flaws.

    And so we find ourselves at today, with this project. I've iterated a few more times on my own, and we're finally closing in on a real engineering solution. It's been a long road.


    Many thanks to Ninja-Robot for their affordable and customer-focused CNC machining service! Go get your stuff machined with him

    Shoutout to Johnny at Ultimachine: I've never managed to brick a RAMBO dude, and I know a ton of that is thanks to your diligence.

  • Last couple bugs

    Michael Perrone11/02/2021 at 02:51 0 comments

    I tried hard oxide coating the aluminum: unlike normal anodization, this process requires high voltage and active chilling, as well as a ramp-up in voltage that is gradual so that the surface isn't porous. Also, it evidently requires a fresh surface to start with:

    I machined this feature on the top to retain the liquid glass gasket material and prevent it from getting stuck in the threaded holes. The hard oxide coating was somewhat microporous because I ramped up the current too fast, but on this freshly cut surface it was stable.

    The rest of the surface was already anodized, so I assumed nothing would happen there. I guess it got scraped over time from handling. It also seems like fingerprints somehow destabilized the alumina coating under additional anodizing. Doesn't look pretty; oh well. The thick oxide coating was still very protective and abrasion resistant. The dark areas are hard oxide coated, and rose off the surface kind of like braille: the oxide coating got surprisingly thick!.


    Setting up the liquid glass gasket was a little finicky with this design. I'll have to figure out something better in future iterations. First we prime the metal wire with some melted borax/glycerol (50/50 wt %) that has been cooked at 220C for 1 hour to fully diffuse the components into each other. This solidifies on the wire in a film as it cools, almost like molten sugar.

    I then heat up the hot end and squish it in through the top.

    I may have added a bit too much borax/glycerol mix.

    To correct somewhat for the overheating in my previous test, I added wooden belt clamps and tacked on an infrared reflector on the bottom of my extruder cold end to help keep it cooler. It was just some extra graphite gasket I had lying around. It wasn't perfect, but it worked well enough for now, because I was finally ready for my first true test print. After a false start with some bad slicer settings, here's the result of my first attempt:

    The high temperature heat bed worked perfectly: it kept the metal molten long enough to wet to the print bed (Print beds at the normal temperature of around 60C are nowhere near hot enough in my experience). But then almost immediately, the extruder jammed! It's always something, isn't it? The issue was easy enough to diagnose:

    The wire had buckled in my crappy prototype extruder because it wasn't constrained properly. And moreover, when I opened up the nozzle, the molten metal had forced its way up into a cavity in the cold end and solidified, locking the filament in place. It will spin around because the cavity is circularly symmetric, but it won't push in or pull out. This means I need a steeper temperature gradient and more careful control of the internal geometry of the extruder. Luckily, I have an extra E3D nozzle I can cannibalize to test that out. More info will be available as soon as that's done.

    Also something important to note: If you look at the Prusa Slicer config, you will notice that the print temperature is set to 300C instead of 240: this is because of the thermal gradient from the nozzle to the print bed: while printing, we want the very tip of the nozzle to be at about 240C, so to achieve that in practice, we need to increase our print bed temperature substantially, and need to do the same with the hotend. For the extruder geometry I chose, a midpoint rule can conveniently be used to roughly estimate the extruder and bed temperatures required: the extruder set point is 60 degrees above 240, and the bed temperature is 70 degrees below that temperature. A bed temperature of 180 degrees would work just as well. for an even split. All that temperature gradient falls across just the small nozzle, so it's feasible to control the temperature gradient precisely with a large heat source and sink: the heater block and the print bed. In that sense, the thermal boundary conditions are controlled.

  • Quick update before the Hackaday Prize deadline

    Michael Perrone10/27/2021 at 03:11 0 comments

    I wanted to post a video of my printer actually printing tonight, but I ran into a couple unexpected bottlenecks. First off, my Y-axis belt clamps straight up melted off!

    I guess I should have expected something like that now that I can heat the print bed up to 200C. No biggie: I can make new belt clamps out of laser cut wood and epoxy that will hold up to the forces and temperatures.

    And secondly, there were 3 issues I need to fix with the extruder.

    1. There needs to be a liquid glass gasket reservoir on the top to enable easy application of the gasket material, and so that it will pour into the extruder after being applied. Otherwise it is quite difficult to actually apply the gasket material
    2. The heat break I used was filled in with molten solder to keep the inner wall wettable by the feedstock alloy, but the heatsink worked too well and cooled the upper end of the heat break until the solder there was always solid, all the way up to 300C measured in the heater block. Moreover the drill through-hole wasn't perfectly straight because the old drill press I used wasn't great, so the metal filament wasn't even hitting the hole in the heat break. This can be fixed by a bunch more dremeling, anodizing, black oxide coating and other processes I'll explain in more detail once I have pictures to show of them.
    3. The kinematic coupling to the extruder seemed loose when the extruder was heated up to temperature: this means that the plastic components were heating up too much. In the future I can fix this with a better heatsink geometry, but for now I can laser cut some wood parts to replace more of the plastic.

    On the bright side, the design does not leak or form gas pockets, unlike old designs from the past. This means as long as the gasket forms a good seal, it will be capable of controllable extrusion. These problems may take a day or two to fix, but after that we'll be all set to print!

    On an unrelated note, when reviewing my documentation I noticed I hadn't talked about my print bed modifications yet!


    They're quite straightforward and I'm not entirely certain they are necessary to get Semisolid Metal Printing functionality, but they've been very helpful in prototyping things so far, and I'm sure they can help other people with other projects down the line I'll add 3D models at some point but mine are hand-machined because it was easy to do with a dremel and a drill press. It's just a 1/4 inch by 8 inch square 6061 aluminum plate, some drilled plywood, a 200X200mm E3D high temperature heat bed (requires a beefy solid state relay and some wire), some 0.2 inch by 3/8 inch SmCo magnets, a .042 X 8 X 12 inch spring steel sheet, and a little epoxy. I just drilled out holes slightly larger than the SmCo magnets and epoxied them in, then drilled holes as needed in the plywood (The Prusa Mendel I2 Mk2 documentation includes a printable drill guide/laser cutting DXF)) and drilled then dremeled the corners of the aluminum plate so that the bolts would fit sunk into the surface. I also added some thermal compound between the heated bed and the aluminum to help conduct the heat from the bed into the aluminum.

  • A brief aside on conversion coatings

    Michael Perrone10/22/2021 at 14:54 2 comments

    One of the key details for semisolid metal 3D printing technology is controlling and patterning wettability to passively control the flow of the liquid metal, and mitigating solubility of the nozzle in the molten metal. With steel and aluminum alloys, the easiest way to do this is with black oxide coating and anodization respectively. Over the past few weeks I have run many experiments to implement these in my process. In my first experiment I learned that black oxide coating of steel doesn't really happen much in the absence of nitrates and nitrites.

    With my first test in sodium hydroxide using mild steel, I got the following result:

    There is a thin dark layer of presumably magnetite nanoparticles in the dipped section after running for an hour at the recommended 140C, but it's barely distinguishable and definitely not thick enough to resist abrasion. Moreover, partially submerging the parts led to rust at the air-liquid interface, where the solution didn't limit oxygen diffusion and control the surface chemistry of the iron It may also be that the iron exposed to salt and air galvanically protected the region below the fluid from forming magnetite at a faster rate. I was hoping for a cleaner interface; perhaps there is an electroplating process that can do that. Worst case scenario I could black oxide coat the whole part, then machine away the coating where it isn't needed. Later on I thought of another solution described further below, which works well for my purposes.

    So then began a series of additional tests with anodization, and in the process I discovered that TSP-borax electrolytes are perfect for anodizing titanium, in case that becomes useful later on. The electrochemical attempt at black oxide coating steel shown above ended in relative failure: the oxide coating was nice and dark, but did not withstand much abrasion and wasn't electrically insulating or corrosion resistant. In the end I opted for standard hot tank bluing with a solution of 2 parts sodium hydroxide, one part sodium nitrate by weight, and just enough water to get it mostly dissolved at 140C.  Running that process for about an hour ended up with a workable result, but in the future I may need to optimize further.

    For anodizing, I just followed this guy's instructions for aluminum anodizing here and it worked like a charm. I think I used 12 grams of 70% sulphuric acid per 1000 grams of distilled water, so it was a little bit overly concentrated compared to what other people use. I didn't even wire brush or sand blast the parts ahead of time, they just had their typical machined surface finish, and the surface finish at the end reminded me of a Macbook Pro's surface finish, with the machining marks hardly noticeable.:

    That said, I'd like to go even further in the future and try hard oxide coating, where you chill the sulfuric acid bath to near freezing then dump many amps through it at 100 volts or more. Hard oxide coatings on aluminum and titanium can be extremely abrasion and corrosion resistant.

    I realized that my printing alloy melted around 240C, well above the temperatures required for black oxide coating steel. To selectively keep surfaces wettable, I filled or covered those regions with my printing alloy as a mask. Like these drilled vias in these components about to be black oxide coated.

    An M2 tool steel extruder with a rough attempt at patterned wetting properties at the tip. M2 tool steel offers excellent wettability but low solubility in molten zinc alloys, so it makes for an excellent nozzle for this kind of printing. All the threads required anodization  or black oxide coating because, as I have learned, you do NOT want to encourage the molten metal to wick into the threads and squeeze out via capillary action. That happened to me once at a previous company: once the liquid metal wicks in and wets the threads, it's virtually impossible to stop and you generally have to throw away the whole component...
    Read more »

  • Quick update: the new design

    Michael Perrone09/26/2021 at 13:55 0 comments

    While hiking up Mt Ranier I kept thinking about how I should redesign my metal extruder, and eventually near Camp Muir I arrived at a solution.

    Printing metal filament is nothing like printing plastic filament: plastic is a viscous liquid which forms its own sealing surface as it passes through its glass transition temperature. Metal may have semisolid states where there are solid and liquid phases are mixed together, but it is possible to separate these phases, causing granular jamming and clogging nozzles. This can be avoided by printing above the liquidus, having material only pass into the semisolid state as it goes over a deposition brush.


    If one attempts to print metal in a fully liquid state, the thin liquid will leak out of any hole provided: the metal does not go through a glass transition, and therefore it does not form a sealing gasket surface. It is possible to have extremely tight tolerances on the metal filament and on the extruder, but then any variation in your filament would induce a clog in the nozzle. The required precision would make metal filament far too expensive for the average maker. Rubber gasketing could in theory take the place of the self-forming sealing surface on the metal, but it would be limited to low temperatures and the trapped air would tend to expand when heated, causing the nozzle to drip.


    This is how I arrived at using a liquid glass gasket: design low melting-point glasses which surround the metal wire as it enters the hotend and before it itself melts, to restore the gasketing behavior that thermoplastic normally has. Serendipitously, I discovered a few such compositions myself long ago when playing around with a hotplate, and a quick look through google scholar shows plenty of options for these low-melting glasses even below 400C. The vast majority of these glasses are less dense than the metal alloys I plan to work with, except maybe in the case of aluminum, where care must be taken to pick the proper glass. This is important because the liquid glass gasket remains in place due to buoyancy on the molten metal in the nozzle and a few other factors, and the molten metal remains in place because of the capillary forces in the brush, as I learned all too well when playing around with the MHD system. I'm getting some key parts to test this concept with machined at the moment, but it won't be long before this method gets tested.

  • Rethinking designs

    Michael Perrone09/23/2021 at 21:52 0 comments

    Hi everyone! Here's how things have been going since the last project update:

    Firstly, I finished up the cold end of the extruder for attaching the MHD extrusion system to a bondtech extrusion system and my old Prusa Mendel printer, for prototyping purposes. (little did I know this effort would prove useless) Then I got back to testing my MHD extrusion system:

    In the previous update, I had this extruder built as well, but there was a problem: the capillary force/surface tension of the liquid metal was just too high to be overcome by the MHD forces plus gravity. Even if it were enough, liquid metals are so thin that if it overcame the capillary forces, metal would shoot out of the tip very fast. This would require some very involved control electronics to stabilize, pushing the price well outside the range I was hoping for. I did a few tests with different magnetic yoke configurations and was able to get steady-state fields as high as 1.15 Tesla, but ultimately the chip shortage going on right now was the straw that broke the camel's back for this design, and I had to rethink things. The controls problem was workable if I could boost the field or the current enough, but if the control electronics were going to be prohibitively expensive or outright unavailable, it would defeat the purpose of making highly available metal printing at a cost makers can actually afford.

    Here were some of the other test configurations to see how much magnetic field I could get out of the yoke. Perhaps having small volumes of high magnetic field will help someone else down the line, but I ultimately couldn't enhance the field enough for MHD to make sense for metal extrusion. I could have made up the difference than upping the current on the driving circuitry, but it was already going to be way too expensive, if available at all, due to the chip shortage. These digital tesla meters from China are pretty good by the way.

    Since I couldn't get enough extrusion force with MHD, I threw together a new "nozzle" out of Cotronics Silicon carbide ceramic, plus high temperature epoxy for added tensile strength, in order to test pressure-based extrusion. With a stainless steel tube as a standoff for the tubing, I attached it to an air tank, pressure regulator, and foot pedal switch to open and close the pressure source. With this I was able to slowly increase the pressure to directly measure how much more pressure I needed, compared to how much the MHD system produced. The ceramic/epoxy held up surprisingly well to the pressure, without any leaks during testing at temperature. The secret was epoxying the stainless steel tube at the top: this allowed the different thermal expansion coefficients to just stress the high temperature epoxy, which had enough elasticity to it, and not the silicon carbide cemented ceramic. The extrusion pressure was about 4 times higher than what the MHD system could produce, which was in agreement with my calculations. Once capillary forces were overcome though, extrusion was not very controllable. I was able to stack material moving the thing around by hand as with a hot glue gun, but this print wasn't going to win any prizes.

    Even so, this was a key step forward relative to my previous work at Voxel8: we were able to make nice first-layer traces and bridge wires for long distances, but we couldn't build subsequent layers on top of previous layers because not enough pressure could be built up to keep the liquid flowing.

    Extruding wire to demonstrate bridging capabilities with the the old system at Voxel8 in 2016. The capillary properties help extrusion stability in this case instead of acting against it, making extrusion control easier.

    The same thing was true for producing traces with the Voxel8 machine: the substrate was stable enough; even printed substrate performed decently. It was perfect for printing electronic traces to replace the silver ink we were using previously, but producing solid metal...

    Read more »

View all 12 project logs

  • 1
    Getting Started

    The first thing you'll need to do is purchase the components. The bill of materials can be found on the github page

    This metal extruding system is meant to be an upgrade to existing 3D printers, so it is assumed that you already have a 3D printer to modify. On the Github page you will find 3D models for attaching the extruder to various common printers (for prototyping purposes at the moment, only Prusa I2 MK2 is supported, but feel free to submit a pull request for your type of printer), and configuration files for various firmware and slicing settings. Eventually the plan is to clean and package these up and submit pull requests on the respective projects.

    The printer I am testing with has 200 step motors for the Z axis, and 400 step for all other axes, so you may need to multiply my steps per rev by 2 if using Marlin. My setup uses Prusa Mendel I2 MK2, Rambo, Marlin, Prusa Slicer and Octoprint, all of which have been around for a while and have excellent documentation, and some have large communities around them where you can potentially find support if something's not working properly. I've been out of the loop for quite a while with Marlin 2.0 updates, but it seems like more recent versions of Marlin are no longer compatible with Arduino IDE, and since I don't have time to figure out all the new features before the end of the Hackaday Prize, I'll be sticking with Marlin 1.8.15 for now. I'll sit down and add Marlin 2 compatibility afterwards.

    While you wait for your materials to arrive, you can make sure your printing system has everything it needs to be compatible with the metal extruder.

  • 2
    Machine the extruder

    For low temperature metals, aluminum or steel extruder components will work. Molybdenum and Tungsten tool steels can be used to print silver and its alloys at around 900C. 3D models for the extruder can be found on Github. The long bore that goes through the center is meant to be drill pressed all the way through by hand after CNC machining because I was unsure if the CNC I was using could do it. I can add a model later that has the bore actually going all the way through instead of just with pilot holes. Generally I have found that for mild steels and HSLA steels, a high spindle speed and low pressure with machining oil and titanium nitride drill bits on a hand-controlled drill press does fine, but low speed high torque kills 2mm drill bits if they get even a little stuck. High spindle speeds aren't great for drill bits either, but the nitride coating and the oil mitigate any abrasion or accidental annealing that would otherwise occur. Drilling down from the top is better, because the melt chamber further down can handle a bit less precision. Drilling from the top and bottom to meet in the middle is probably best though.

    The two holes on the top of the heat sink off center are meant to be tapped with an M3 tap. The hole on the bottom of the heat sink and the centered holes on the heater blocks are meant to be tapped with an M6 tap.

    You will then need to anodize or black oxide coat some of the parts: that process is described in detail in this project log. More detail on the way in the future.

  • 3
    Assemble the extruder and printer

    This is all specific to the printer you want to use. In the future, instructions will be included with each calibration file or set of 3D prints. For the time being, refer to the project logs while I get something better sorted out. It should be pretty obvious which screws go where when fitting everything together from the pictures provided, but feel free to reach out with questions in the comments section.

    To make the liquid glass gasket material, simply mix about 18 grams borax (hydrated) with 20 grams glycerol and bring to a boil until the bubbles become less frequent. Cooking at 300C for 30 minutes after mixing well seems to do the trick.

    Do not add the liquid glass gasket to the extruder until the Xpando has had time to set in all the M6 screw threads: it will mix with the Xpando and prevent it from setting properly. After the Xpando is cured they are fine to use together though.

    At present I also use a Bondtech extruder and added a custom heated bed, which might not ultimately be necessary. It makes the prototype system more reliable and versatile for testing, but also more expensive than a final solution would be. They're also a bit ad-hoc so if it becomes clear they are necessary, I'll improve the documentation about them. With all those modifications it seems like it would be difficult to keep the upgrade cost below $500 like I'm targeting, so hopefully at least the Bondtech extruder is overkill, if not both features.

    Since I don't think I've mentioned it elsewhere, in case they are needed, bristles can be added to the nozzle aperature with one of those new portable spot welders Tesla uses to build their battery packs. They can be spot welded to the back of the nozzle and then run through the tip of the nozzle and cut to the preferred length. In practice they shouldn't stick out of the nozzle more than a millimeter or so: the thermal gradient outside the nozzle tends to be steep.

View all 5 instructions

Enjoy this project?

Share

Discussions

dekutree64 wrote 06/01/2021 at 20:01 point

How do you get the layers to bond if the metal is never melted? Or is layer bonding done as a post-processing step, using plaster or sand or something to make a temporary mould while you melt the whole thing?

  Are you sure? yes | no

Michael Perrone wrote 10/27/2021 at 03:08 point

The metal is melted and the thermal gradient is controlled so that a slight melt pool forms on the previous layer, allowing the layers to bond. Also a few metal bristles can be used to break through the layer of surface oxide if needed.

  Are you sure? yes | no

John wrote 05/20/2021 at 14:35 point

This is a real step forward in my opinion.

  Are you sure? yes | no

Michael Perrone wrote 10/27/2021 at 03:06 point

It sure is! I'm looking forward to documenting it better!

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates