Liquid metal on a copper cooler turns into solid CuGa2 instead of drying out
Thermal Grizzly boss der8auer's electron-microscope test finds liquid metal on copper turns into solid CuGa2, rated about 100 W/mK against copper's 400. He says nickel plating keeps that layer about 10 times thinner, Tom's Hardware reported.
The Engineer · Build desk
Key figures: on copper the CuGa2 layer reaches about 50 micrometers, up to 100. On nickel-plated coolers the layer is 1 to 5 micrometers, or 5 to 10 on his slide. CuGa2 conducts heat at about 100 W/mK, against roughly 400 for copper.
| Figure | Value | Note | Claim |
|---|---|---|---|
| CuGa2 layer on copper, about, at full growth | 50 micrometers | up to 100 micrometers | 9 |
| Layer on nickel-plated coolers, at least | 1 micrometers | up to 5; 5 to 10 on his slide | 13 |
| CuGa2 heat conductivity, about | 100 W/mK | copper: roughly 400 | 6 |
What happened
- He saw liquid metal after one hour, then no visible liquid and crystals everywhere at the two-month mark.
- On the cold plate of Thermal Grizzly's new stainless steel water block, the film could not be wiped off with isopropyl alcohol or acetone after less than five minutes.
- On aluminum the gallium travels along grain boundaries, and a scrap 2 mm sheet swelled after about five minutes and turned brittle within half an hour.
- Nickel-plated coolers get NiGa5 instead, a compound of five gallium atoms per nickel atom that likewise stays solid at room temperature.
Why it matters
- What it limits constraint Heat will not undo the film on a copper cooler, because CuGa2 stays solid up to about 254 degrees C.
- Who has to decide decision Choosing nickel trades a worse-conducting compound for a thinner film and much less absorbed gallium, and der8auer judges it "most of the time superior".
- What it costs cost Gallium supply adds cost: Thermal Grizzly pays almost 10 times what it did five years ago, EUR 2,500-3,000 per kilogram against EUR 300.
Nothing evaporates. On copper, the gallium reacts with the metal to form CuGa2, and the leftover indium and tin form other solid areas [6]. The film looks like it is solidifying because it is. Der8auer said the liquid metal "doesn't really dry out" and that it "reacted chemically and it formed a solid compound" [5]. Of the two-year sample he said: "After a total of two years, this honestly did not change further" [3].
He stored the setup in a vacuum or a nitrogen-filled box between sessions to avoid oxidation [4]. A cooler in a running PC gets no such treatment, so the test took oxidation out of the picture.
The conductivity figure is a material rating. CuGa2 is rated at about 100 W/mK against roughly 400 for copper [6], a ratio of 0.25 [20]. Tom's Hardware's account of the video does not include die temperatures before and after, so what a thick film costs a given chip is not measured here. For the quarter-conductivity number to transfer, the film would have to reach full thickness inside the joint, and the chip would have to be limited at that interface.
The growth rate fits the timeline. His example of pure gallium between copper plates at 70 degrees C shows about 4 micrometers over 96 hours [10]. That is 1 micrometer a day [19], the rate he gives for the layer in general, with a ceiling of about 50 to 100 micrometers that depends on temperature and how much gallium is available [9]. At that rate the ceiling arrives in roughly 50 to 100 days [21]. In our view the two-month check, with crystals everywhere, and the unchanged two-year sample fit that [2].
Nickel plating moves the layer to 1 to 5 micrometers, or 5 to 10 on one of his slides [13]. The compound it forms conducts heat worse than the copper-gallium one [14]. The thinner film is therefore a trade against a worse material.
The tester sells both the liquid metal and the remover. Thermal Grizzly sells its own liquid metal as Conductonaut [16]. Its TG Metal Polish & Restore costs $22.11 direct for 30 g of paste, a polishing pen and microfiber cloths, and in a Sept. 30 blog post the company says the paste works as a surfactant, binding the gallium so it cannot stick to the surface again [17]. A PC Games Hardware forum user tried it on an Aqua Computer block that had been cooling a delidded Skylake-X with liquid metal since 2018, and removal took about an hour [18].
Der8auer recommends that liquid metal "stay in the enthusiast bubble" [22].
What to watch
- Thermal Grizzly's replacement for nickel plating, which der8auer says it hopes to have next year.
- Measured chip temperatures on copper and nickel coolers after months of aging, which the video as reported does not supply.
- The new KryoSheet graphene pad for bare dies, which he puts at maybe three or four degrees worse than liquid metal.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence50
- Adoption
- Not scored
- Hype gap−10
- Incentives55
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
YouTuber and Thermal Grizzly boss der8auer tested a two-year-old liquid metal TIM application with an electron microscope and found it does not dry out; the gallium bonds with the copper to form a solid compound.
- [2]
His test showed the TIM as liquid after one hour, no liquid visible with crystals everywhere at the two-month mark, and then unchanged after two years.
- [3]
"After a total of two years, this honestly did not change further," he said.
- [4]
The test relied on storing the setup in a vacuum or a nitrogen-filled box between sessions to avoid oxidation.
- [5]
The liquid metal "doesn't really dry out"; it "reacted chemically and it formed a solid compound," so the metal is still there in another form.
- [6]
With copper, gallium bonds with copper to form CuGa2, one copper atom to two gallium atoms, with leftover indium and tin forming other solid areas, which makes the film look like it is solidifying. CuGa2's heat conductivity is rated at about 100 W/mK versus copper's roughly 400.
- [7]
CuGa2 stays solid up to about 254 degrees Celsius, so "you will not be able to melt it again."
- [8]
With copper, the liquid metal on the cold plate of Thermal Grizzly's new stainless steel water block cannot be wiped off with isopropyl alcohol or acetone after less than five minutes.
- [9]
The CuGa2 layer grows about 1 micrometer a day until it reaches about 50 to 100 micrometers, depending on temperature and how much gallium is available.
- [10]
His example of pure gallium between copper plates at 70 degrees Celsius shows about 4 micrometers over 96 hours.
- [11]
With aluminum, gallium travels along the grain boundaries once the oxide layer is broken, weakening the metal; his demo on a scrap 2 mm aluminum sheet showed surface swelling after about five minutes and brittleness within half an hour.
- [12]
On nickel-plated coolers, the gallium forms NiGa5 instead, one nickel atom to five gallium atoms, which is also solid at room temperature.
- [13]
The nickel plating stops the layer at about 10 times thinner, at 1 to 5 micrometers, or 5 to 10 micrometers on his slide.
- [14]
NiGa5 conducts heat worse than the copper-gallium compound.
- [15]
Nickel is "most of the time superior" in his estimation because much less gallium gets absorbed into the surface and it protects the copper underneath.
- [16]
Thermal Grizzly sells its own gallium-based liquid metal TIM as Conductonaut.
- [17]
TG Metal Polish & Restore costs $22.11 direct for 30 g of paste, a polishing pen, and microfiber cloths; the company's Sept. 30 blog post describes it as a surfactant that binds gallium to stop it from re-sticking.
- [18]
A PC Games Hardware forum user used it to clean an Aqua Computer block that had run liquid metal on a delidded Skylake-X since 2018, with removal taking about an hour.
- [19]
4 micrometers over 96 hours is about 1 micrometer a day, matching the stated general growth rate.
- [20]
CuGa2's rated conductivity is one quarter of copper's.
- [21]
At about 1 micrometer a day, a 50 to 100 micrometer layer takes roughly 50 to 100 days to form.
- [22]
He recommends that liquid metal should "stay in the enthusiast bubble," intended for very high power density components like a manually overclocked CPU.
- [23]
Thermal Grizzly aims to replace nickel with new coatings, "hopefully next year."
- [24]
Thermal Grizzly plans to release a new version of its KryoSheet graphene pad for bare dies shortly, which he says is maybe three or four degrees worse than liquid metal.
- [25]
His company pays almost 10 times as much for gallium after five years, from EUR 300 per kilogram to EUR 2,500-3,000, and prices have more than doubled since the start of 2025.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- tomshardware.comExperiment shows liquid metal TIM doesn’t dry out; it turns solid
1 article · October 16, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Entities
- ConductonautFollow
- der8auerFollow
- Tom's HardwareFollow
- Thermal GrizzlyFollow