Controlling Data Centers’ Air Pollution, Environmental Control to Ensure Equipment, Systems Reliability

From eSociety, February 2020

Air pollution and environmental control in data centers is critical to the reliability of equipment and systems, says the principal investigator of a recently released ASHRAE research report. 

The goal of 1755-RP, Impact of Gaseous Contamination and High Humidity on the Reliable Operation of Information Technology Equipment in Data Centers, is to improve the understanding of the effects of gaseous pollution combined with higher relative humidity, temperature, fluctuating relative humidity and electrical current on the corrosion of copper and silver materials.

Jianshun (Jensen) Zhang, Ph.D., Fellow ASHRAE, of Syracuse University discusses the research.

ASHRAE members can access the research report for free on ASHRAE’s Technology Portal.

1. What is the significance of this research?

Air pollution and environmental control in data centers is critical to the reliability of equipment and systems. ASHRAE provides guidance on the recommended range of temperature and humidity conditions that would limit the severity of copper and silver corrosion to acceptable levels. The recommendation is based on the assumption that if the temperature and humidity are within the recommended thermal envelope and the 30-day corrosion thickness is below 300 Å and 200 Å for copper and silver, respectively, datacom equipment should not have premature failures. The severity of copper and silver corrosion in the data center environment is impacted by gaseous pollution in combination with temperature and relative humidity (RH). 

However, limited data exists to support what level of contamination most affects these limits, which in some cases may have placed over-stringent requirements that limit the use of free cooling for saving energy. This research addresses the following question: can the recommended thermal envelope be expanded to promote reduced cooling energy consumption if air pollution conditions are better understood and controlled?

2. Why is it important to explore this topic now?

Data centers have the highest energy usage density among all building types, and their energy consumption has increased dramatically and will continue to increase as more and more servers are needed around the world. It is vitally important and urgent to find ways to reduce the cooling energy consumption while ensuring high reliability of IT equipment in data centers.

3. What lessons, facts, and/or guidance can an engineer working in the field take away from this research?

This study investigated how increasing RH and temperature from the recommended range would affect the corrosion of copper and silver coupons exposed to “realistic indoor” concentrations. Tests were conducted first at a reference temperature and RH (21°C and 50% RH) within the current recommended thermal envelope, and then at elevated temperature (28°C) and/or elevated RH (70% and 80% RH). 

Experiments were also conducted under fluctuating RH conditions and under voltage-biased conditions to understand their impact. The concentrations for O₃, NO₂, SO₂, Cl₂, and H₂S were selected to be 60 ppb, 80 ppb, 40 ppb, 2 ppb and 10 ppb, respectively, based on the results of an extensive literature review. Test specimens were analyzed by coulometric reduction to determine the total corrosion thickness and quantities of corrosion products. A selection of test specimens was analyzed by Scanning Electron Microscopy and Energy Dispersive Spectrometry, Atomic Force Microscopy and X-ray Photoelectron Spectroscopy to understand the formation of corrosion products and the corrosion mechanisms. A computer imaging method was developed to qualitatively correlate the severity of the corrosion based on a color index that reflected a change of color relative to a clean specimen that was not exposed to pollutants. 

This research found that: 

1. O₃, NO₂ and SO₂ can be pervasive pollutants in a data center, while the presence of Cl₂ and H₂S are local phenomenon depending on the activities inside and surrounding specific data centers. 

2. A logarithmic relationship exists between the corrosion thickness and exposure time for copper. Silver exhibits a linear relationship.

3. Under the reference temperature and RH (21°C and 50% RH), which is inside the ASHRAE recommended environment, significant copper corrosion occurred only when Cl₂ was contained in the gas mixture, suggesting that Cl₂ was the most corrosive to copper. For silver, significant corrosion occurred only when H₂S was present. 

4. Increasing the RH from 50% to 70% at the reference temperature of 21°C, increased the rate of corrosion on copper when Cl₂ was present. Increasing the RH to 80% resulted in significant copper corrosion regardless of the pollutant combination, suggesting that a critical RH exists for copper between 70% and 80%. For silver, increasing the RH did not cause a significant increase in corrosion thickness. 

5. For copper, increasing the temperature from 21°C to 28°C at the reference 50% RH reduced corrosion. It is likely that at a higher temperature, a lower amount of pollutant can be adsorbed or absorbed on the surface to cause corrosion. For silver, significant corrosion thickness was still detected at the higher temperature for the H₂S containing mixture conditions.

6. The impact of humidity fluctuation was consistent with the effects of the average RH on the corrosion of copper and silver.

7. Voltage bias on the PCBs reduced the level of corrosion likely due to the electrical current carrying copper traces raising the local temperature slightly.
   
   

4. How can this research further the industry's knowledge on this topic?

The following recommendations are made to the industry:

1. Biannually operators should perform testing to determine if their site has low (less than 300 Å/month for copper and 200 Å/month for silver) or high levels of corrosion: 

   1. For environments with low levels of both copper and silver corrosion the industry can consider increasing the upper moisture limit to 70% RH for temperatures between 18°C and 21°C. Furthermore, if the dew-point limit could be relaxed to 17°C, then the temperature limit could be extended from 21°C to 22.5°C at 70% RH.

   2. For environments with high levels of both copper and silver corrosion, then the upper moisture level should be kept below 60% RH, possibly lower than 50%.

2. The RH level should not exceed 70% at 21°C given the increase in corrosion for copper at 80% RH regardless of pollutant composition. 

3. Due to heating as the air moves through the server, manufacturers might benefit from placing copper containing components at the rear of the server to reduce the susceptibility to corrosion. 

4. When H₂S exists any components that use silver are susceptible to corrosion regardless of the thermal environment. 

5. Data center operators should take actions to remove the highly corrosive catalysts type pollutants using chemical filtration.

6. Data center site selection should include extensive coupon testing or measurements of the pollutant concentrations in and around the area to determine the possibility of high levels of corrosive pollutants.

7. The color index method for the standard copper coupon surface when exposed in the data center environment, can be used as a screening tool to determine if more quantitative environmental testing is needed to confirm the pollution or corrosion levels. 
   
   

5. Were there any surprises or unforeseen challenges for you when preparing this research?

One surprising fact found in the project was that a higher temperature reduced the corrosion instead of increasing it. It highlights the complexity of the corrosion process in which the synergistic effects between multi pollutants, moisture content and temperature need to be considered in assessing the corrosion levels. There was also a lack of field-monitored data on pollutant concentrations in data centers and we had to rely on the outdoor air concentration data to define the “realistic worst case” scenarios for the corrosion tests in the laboratory experiments. 

Literature review also showed the lack of simulation models that can be used to predict the corrosion levels in the field based on the laboratory experimental results. We hence used the existing “recommended thermal zone” as the reference condition to determine the reference corrosion level and made sure that the new recommended thermal zone will have lower corrosion level than the reference level. Further research is needed to investigate the mechanism of the corrosion process, and to develop a model that would enable the prediction of the corrosion level based on the pollutant concentrations and thermal conditions found in data centers.