The first time you see a coolant sample come out of a system black as coffee, you think it’s a lab error. It has to be. Secondary refrigerants — the heat transfer fluids that circulate through reactor jackets, cold storage evaporators, and process chillers — are supposed to be clear. Pale yellow, maybe. Light blue. Something that looks like it belongs in a clean industrial loop. Not something that looks like it was drained from a crankcase.

But you shake the bottle. Hold it up to the light. Nothing. Opaque. Black. And then you smell it.

That’s when you know you’re not dealing with a routine fluid degradation issue. You’re dealing with corrosion. The kind of corrosion that doesn’t just eat your piping — it eats your production schedule, your maintenance budget, and sometimes your reputation.

In 2025, our technical team investigated six corrosion incidents across China that collectively cost the operators somewhere north of 4 million yuan in direct damages. The indirect costs — lost production, missed delivery deadlines, emergency contractor fees — probably doubled that. But the real cost, the one that doesn’t show up on any spreadsheet, is what these incidents did to the way those operators think about their cooling systems. Trust, once broken, is hard to rebuild.

This is the story of those six incidents. Not as case studies in a sales presentation. As honest accounts of what went wrong, what we found, and what it means for anyone running a secondary refrigerant loop.

Data centers are not where you expect to find corrosion problems. They run at relatively mild temperatures — inlet around 10°C, return around 18°C. The piping is typically stainless steel or treated carbon steel. The coolants are water-based, usually LM-4 or equivalent, with a full inhibitor package. Everything about the operating conditions says “benign.”

Xinhua Insurance’s Beijing facility had been running LM-4 for three years. The fluid was tested annually. TDS, pH, inhibitor concentration — all within spec. Then, in April 2025, during a routine maintenance window, a technician pulled a sample from the low point of the loop. The fluid was black. Not dark brown. Black. And it was foaming. The sample bottle was half liquid, half foam, and the foam wasn’t breaking down. It sat there for 20 minutes, a thick layer of gray-brown bubbles that looked like a cappuccino gone wrong.

We got the sample in our Shanghai lab two days later. The pH had dropped from 8.8 to 6.1. The inhibitor — a nitrite-based package — was depleted. The iron content was 47 mg/L. Normal is under 5. Something was corroding the carbon steel piping at an alarming rate, and whatever it was, it was also generating gas.

The gas turned out to be hydrogen. The corrosion mechanism was acidic attack on the steel, producing iron ions and hydrogen gas. The hydrogen was what caused the foaming. The black color was suspended iron oxide particles — magnetite, mostly, which forms under low-oxygen conditions. The system was corroding from the inside out, and the only visible sign was a fluid that looked like crude oil.

Root cause? The system had a small freshwater makeup line that was supposed to be valved off during normal operation. The valve seat had degraded over three years, allowing a slow, continuous trickle of untreated city water into the loop. City water in Beijing is hard — around 250 mg/L as CaCO₃. The calcium and magnesium were precipitating in the heat exchangers, but that wasn’t the real problem. The real problem was the dissolved oxygen in the makeup water. A constant stream of oxygenated water was overwhelming the inhibitor’s capacity to passivate the steel surfaces. The inhibitor was being consumed faster than it could be replenished, and once it dropped below the critical threshold concentration, the corrosion rate went exponential.

We drained the system. Acid-cleaned the piping. Replaced the makeup valve. Recharged with fresh LM-4. The system has been stable since. But the data center manager told me something that stuck: “I looked at that fluid every quarter for three years and never saw a problem. Until I did. And by then, it was almost too late.”

Four days after the Beijing case, we got another black fluid sample. This time from Xinjiang. A cold storage facility that was converting from ammonia direct expansion to a secondary refrigerant loop with LM-4. The conversion was supposed to be straightforward: remove the ammonia evaporators, install a plate heat exchanger, circulate LM-4 through the cold storage coils. The existing coils were copper. The rest of the piping was carbon steel. LM-4 is compatible with both. No problem.

Except the fluid turned black within two weeks of commissioning.

The system was a 3,000-liter loop. The LM-4 was fresh. The piping was flushed and passivated before filling. The water used for dilution was deionized. Everything was done by the book. And yet, the fluid was black and the copper coils were showing signs of pitting.

We sent a metallurgist to the site. She took sections of the copper tubing and examined them under a microscope. The pits were deep — 0.3 to 0.5 mm in some places — and they had the characteristic morphology of ammonia-induced stress corrosion cracking. But the system was supposed to be ammonia-free. That was the whole point of the conversion.

Here’s what had happened: the ammonia removal process hadn’t been complete. The old evaporator coils had absorbed ammonia into the copper grain boundaries over years of service. When the coils were flushed with water during the conversion, the flush water removed the surface ammonia but not the ammonia that had diffused deep into the metal. When the new LM-4 was introduced, the inhibitor package — specifically the triazole component — began to mobilize that residual ammonia. The ammonia leached out of the copper and into the fluid, where it formed copper-ammonia complexes. Those complexes are dark blue-black in solution. They also catalyze further corrosion of the copper by disrupting the protective oxide layer.

The solution was counterintuitive: we had to let the process run its course. The system was drained and refilled with fresh LM-4 three times over a period of six weeks. Each drain cycle removed more of the residual ammonia. By the fourth fill, the fluid stayed clear. The copper surfaces repassivated. The pitting stopped.

The lesson: converting from ammonia to a secondary refrigerant system isn’t just a plumbing job. It’s a chemistry problem. The old refrigerant leaves a footprint in the metal that can take months to fully erase.

Two days after Xinjiang. April was not a good month.

Shandong Huatai operates a batch chemical process that uses LM-4 as a secondary refrigerant on the shell side of a glass-lined reactor. The process chemistry involves hydrochloric acid. The reactor jacket is separated from the process by the glass lining and a PTFE gasket at the nozzle. Under normal conditions, there’s no path for HCl to enter the coolant loop.

But conditions stopped being normal sometime in late March. A gasket at the bottom outlet nozzle developed a hairline crack. Not enough to leak process fluid visibly. Just enough to allow HCl vapor to diffuse through the crack and into the jacket space. HCl vapor is aggressive. It dissolves in the water phase of LM-4, dropping the pH and attacking the carbon steel jacket walls.

By the time the operators noticed the problem, the pH of the coolant had dropped to 3.2. The jacket walls had lost 0.8 mm of thickness in localized areas. The reactor was still operable — barely — but the jacket was now a safety concern. The plant shut down for two weeks while the jacket was repaired and the gasket was replaced.

The coolant, of course, was destroyed. Not just contaminated. The HCl had reacted with the inhibitor package, producing chlorinated organic compounds that couldn’t be separated from the base fluid. The entire 800-liter charge had to be disposed of as hazardous waste. The disposal cost alone was 18,000 yuan.

What makes this case instructive is that the operators had a pH alarm on the coolant loop. It was set at 6.5. The alarm never triggered. When we checked the pH probe, it was coated with a layer of iron chloride precipitate that had insulated the glass electrode from the bulk fluid. The probe was reading pH 7.8 while the actual fluid was at pH 3.2. The alarm system was giving a false sense of security.

Fujian was different. Not a chemical compatibility issue. A galvanic one.

The cold storage facility used aluminum evaporator coils in a system filled with LM-4. Aluminum and LM-4 are compatible — the inhibitor package includes silicates that protect aluminum surfaces. The system had been running for five years without problems. Then, in May 2025, a coil perforated. The hole was about 2 mm in diameter, surrounded by a white, crusty deposit of aluminum oxide. The perforation caused a loss of about 200 liters of coolant before the operators could isolate the coil.

When we examined the failed coil section, we found the perforation was at a point where the aluminum tube passed through a steel support bracket. The bracket had been replaced during a maintenance shutdown six months earlier. The original bracket was aluminum. The replacement was galvanized steel. The two metals, in contact through the condensed moisture on the coil surface, formed a galvanic couple. Aluminum is anodic to steel. The aluminum was sacrificially corroding to protect the steel bracket.

The corrosion was external to the coolant loop. It had nothing to do with the fluid chemistry. But it perforated the tube from the outside in, releasing coolant and shutting down the cold room. The repair cost was modest — about 5,000 yuan for the coil section and labor. The real damage was the 18 hours of downtime in a facility that operates 24/7 during the peak season. They lost a shipment of frozen seafood that was supposed to go out that night. The customer — a major export distributor — took their business elsewhere.

The China Aerospace case was the most technically interesting of the six. Not because of the damage — it was relatively minor — but because it exposed a subtle flaw in how we think about “closed” loops.

The facility in Huanghua uses LM-8, a water-based secondary refrigerant rated from -50°C to 120°C. The system is a closed loop. No makeup water. No open expansion tank. A bladder-type expansion tank with a nitrogen blanket. By every conventional definition, this was a sealed system with no oxygen ingress path.

And yet, the carbon steel piping showed classic oxygen pitting. Shallow, wide pits with tubercles of iron oxide. The kind of corrosion you see in open cooling towers, not in closed loops. The fluid analysis showed dissolved oxygen at 2.1 mg/L. In a properly sealed system, dissolved oxygen should be below 0.1 mg/L.

We traced the oxygen source for three days. Pressure-tested the entire loop. Checked every gasket, every valve stem, every instrument fitting. No leaks. The expansion tank bladder was intact. The nitrogen blanket was holding pressure. Everything was sealed.

The oxygen was coming from the LM-8 itself. Specifically, from the water used to dilute the concentrate during the initial fill. The plant had used deionized water — good practice — but the deionized water had been stored in an unblanketed tank for two weeks before use. During that time, it absorbed oxygen from the air. When the LM-8 concentrate was diluted with this oxygenated water, the dissolved oxygen was carried into the system. The inhibitor package in LM-8 includes an oxygen scavenger, but it was exhausted within the first 48 hours because the initial oxygen load was so high. After that, the residual oxygen was free to attack the steel.

The fix was simple: we added a supplemental dose of oxygen scavenger, and the system stabilized. But the real fix was procedural: the plant now specs that dilution water must be used within 4 hours of deionization, or be stored under nitrogen blanket. A small change. A big difference.

This was actually the first corrosion case of the year, chronologically. It just took us the longest to fully understand.

Sanchuan Chemical operates an ammonia refrigeration system with a brine secondary loop. The brine is LM-4, chosen for its anti-corrosion properties. The system uses a shell-and-tube heat exchanger where ammonia evaporates on the shell side and the brine is cooled on the tube side. The tubes are carbon steel. The tube sheet is carbon steel. Everything is compatible with both ammonia and LM-4.

In March, the operators noticed a pressure drop on the brine side of the heat exchanger. The flow rate was down about 15% from design. They suspected fouling. They backflushed the heat exchanger. The flow improved temporarily, then dropped again. This cycle repeated three times over two weeks.

When they finally opened the heat exchanger, they found the tube side was coated with a gray-white scale. The scale was analyzed. It was ammonium bicarbonate — a compound that forms when ammonia, carbon dioxide, and water react. The ammonia was coming from a pinhole leak in one of the tubes. The carbon dioxide was dissolved in the brine from atmospheric exposure. The water was, well, the brine.

The pinhole leak was so small that it wasn’t detectable by pressure testing. The shell side operated at a lower pressure than the tube side under normal conditions, so ammonia shouldn’t have been able to leak into the brine. But during defrost cycles, the shell-side pressure spiked above the tube-side pressure, and for 15-20 minutes at a time, ammonia was being pushed through the pinhole and into the brine.

The ammonium bicarbonate scale was self-limiting — it plugged the pinhole after a certain thickness built up. But by then, the tubes were already partially blocked. The heat exchanger had to be retubed. The cost: 120,000 yuan in parts and labor, plus a week of downtime.

Six incidents. Six different corrosion mechanisms. Acidic attack. Ammonia leaching. HCl ingress. Galvanic coupling. Oxygen pitting. Ammonium salt scaling. Only one of them — the Beijing data center — was a straightforward case of inhibitor depletion. The other five required investigation, metallurgical analysis, and a willingness to look beyond the obvious.

If you’re operating a secondary refrigerant system, the single most important thing you can do is pull a sample and look at it. Not send it to a lab. Not run it through an analyzer. Just look at it. Hold it up to the light. Is it clear? Is it the right color? Does it smell right? Does it foam when you shake it? These are not sophisticated tests. They’re the kind of thing an experienced operator does without thinking. But they catch problems before the pH meter does, before the corrosion coupon does, before the thickness gauge does.

The second most important thing: know what your fluid is touching. Not just the piping. The gaskets. The valve seats. The pump seals. The expansion tank bladder. The instrument fittings. Every material in the loop is a potential corrosion participant. A single incompatible component can take down an entire system.

Corrosion is never just a chemistry problem. It’s a systems problem. Treat it like one.