A smart friction control strategy enabled by CO2 absorption and desorption

Intelligent control of friction is an attractive but challenging topic and it has rarely been investigated for full size engineering applications. In this work, it is instigated if it would be possible to adjust friction by controlling viscosity in a lubricated contact. By exploiting the ability to adjust the viscosity of the switchable ionic liquids, 1,8-Diazabicyclo (5.4.0) undec-7-ene (DBU)/ glycerol mixture via the addition of CO2, the friction could be controlled in the elastohydrodynamic lubrication (EHL) regime. The friction decreased with increasing the amount of CO2 to the lubricant and increased after partial releasing CO2. As CO2 was absorbed by the liquid, the viscosity of the liquid increased which resulted in that the film thickness increased. At the same time the pressure-viscosity coefficient decreased with the addition of CO2. When CO2 was released again the friction increased and it was thus possible to control friction by adding or removing CO2.

Friction is the force resisting the relative motion of solid surfaces or fluid layers sliding against each other. Often, we try to minimize friction but there are also many situations where high friction is desirable. While in some cases something in between, i.e. optimum friction is desirable. Nowadays both ultrahigh friction 1 and ultralow friction [2][3][4][5] have been studied extensively. Ultrahigh interlayer friction was observed when Niguès et al. 1 investigated the friction performance of multiwalled boron nitride nanotubes (BNNTs), caused by the structural reorganization of the BNNTs layer. On the other side, not only at nano-or microscale [2][3][4] , but also at macroscale 5 , ultralow friction can be attained. It was shown that highly oriented pyrolytic graphite, diamondlike carbon films and molybdenum disulfide can lead to ultralow friction under special lubrication conditions. Compared with achieving extremely high or low friction, the intelligent control of tribological interactions is also attractive [6][7][8][9] . For example, friction control by light can be used to manipulate the gripping friction of a robotic finger 9 , in which the friction drops by a factor of four to five when the light is switched on.
Some other methods have also been employed to control friction 10,11 . Meng et al. 10 found that the friction coefficient of the brass/silicon dioxide couple sliding material pair was controllable between 0.06 to 0.21 by applying a certain range electric voltage on the material pair measured by a self-made friction tester under a load of 9.8 N (Hertzian contact pressure ≈ 610 MPa). The friction of poly methylacrylic acid sodium (PMAA) brushes can be controlled via pH value under a constant load of 0.5 N (Hertzian contact pressure ≈ 0.23 MPa) 11 . PMAA showed ultralow friction coefficient (μ ∼ 0.006) under both neutral (pH ≈ 7) and basic (pH ≈ 12) solutions whereas at low pH value (pH ≈ 2), the negative carboxylate groups were rapidly protonated and the PMAA brush fully collapsed and dehydrated, which led to ultrahigh friction (μ > 1). However, one of the limitations of these studies is that the loads in these experiments are quite low. Instead, running experiments with full size from the real application has the advantage of giving realistic results in terms of power losses depending on lubricant type, load, speed and operating temperature 12 .
The atmosphere at which experiments are conducted may affect the friction in dry lubrication 13,14 , thus it has been demonstrated that it is possible to control friction via altering the atmosphere. Mishina et al. 13 studied the effect of the composition of the gas atmosphere on friction and adhesive wear of six different pure metals and Zhang et al. 14 found that the diamond-like carbon (DLC) film shows significantly different friction characteristics under different atmosphere. The influence of the atmosphere on the friction behavior of fluid lubrication has been studied as well. Lee et al. 15 revealed that friction coefficient and the amount of wear of sliding surfaces in the compressor became larger due to the increased pressure of CO 2 , when polyol ester was used as lubricant. Their studies, however, did not address the possibility of friction control by atmosphere environment.
It is reported that the viscosity of the switchable ionic liquids can be increased by up to an order of magnitude after absorbing CO 2 16 .Viscosity is the most essential feature of lubricants, as it enables them, given the right conditions, to separate two solid bodies in relative motion 17 . Switchable ionic liquids /CO 2 binding organic liquids(CO 2 BOLs) are the mixtures of an alcohol and an amidine or guanidine, based on Jessop's switchable solvent [18][19][20] , which can capture and release CO 2 efficiently. The mechanism of capturing CO 2 is that an amidinium or guanidinium alkylcarbonate salt is formed after absorbing CO 2 . By heating or bubbling another gas, the CO 2 will be reversibly released and the low viscosity liquid is obtained again. Inspired by the controllable change in viscosity of switchable ionic liquid under CO 2 atmosphere, it seems that CO 2 has the potential to be used to control tribological performance. Pohrer et al. 21 found a significant change in viscosity after absorbing CO 2 but the change of viscosity could not affect the friction property. The possible explanation is that the friction test in their work was operating in boundary lubrication where viscosity of lubricants has almost no significant impact on friction. Elastohydrodynamic lubrication (EHL) is a more appropriate lubricating regime to evaluate the influence of viscosity on friction, since the viscosity is considered to be one of the best single parameters for assessing the performance of fluid film lubrication.
In this article, switchable ionic liquids, able to capture CO 2 in a controlled manner, were used as lubricants. The idea was to investigate if it would be possible to adjust friction by controlling viscosity in a lubricated contact between a ball and disc operating under elastohydrodynamic lubrication (EHL) conditions. The ability to adjust the viscosity of switchable ionic liquids was exploited by studying mixtures of glycerol and 1,8-Diazabicyclo (5.4.0) undec-7-ene (DBU) while CO 2 was added. It was shown that the viscosity of the mixture increased when CO 2 was absorbed. The higher viscosity mixtures were shown to increase film thickness and reduce friction under EHL. The friction reduction is partly attributed to the decrease of the fluids pressure-viscosity behavior with increasing CO 2 -content. The effect was reversed when CO 2 was released. Altogether it is shown that the absorption and release of CO 2 in a switchable ionic liquid can be used for online control of friction and fluid film under EHL.

Results
co 2 absorption and desorption. The absorption loading of CO 2 in the studied mixtures was calculated using the following equations: where P 0 and P t are the initial and instantaneous pressures, respectively. Comparing with the operating pressure, the saturated vapor pressure of the mixtures can be neglected. V A and V L represent the volumes of the absorption vessel and mixtures. Z 1 and Z 2 are the compressibility factors corresponding to the initial and instantaneous pressure, respectively. The generalized second virial coefficient correlation was used to get an approximation of compressibility factor Z. In equation (2), B is the correlation to critical properties of CO 2 . n CO 2 is the number of moles of CO 2 absorbed by the switchable ionic liquids. n DBU is the molar amount of DBU, equal to the theoretical amount of absorbed CO 2 and, finally, x is the absorption loading. As shown in Fig. 1(a), the CO 2 loading of DBU/glycerol mixture with 3:1 molar ratio increases as a function of time. The CO 2 loading increased rapidly when the loading was below 20%, then the absorption rate decreased gradually. This is mainly because the viscosity of mixtures at high loading is too large to be stirred magnetically, resulting in a mass-transfer limitation 22 . This is the main reason we choose the low CO 2 loading for friction study in the coming part. Figure 1(b) shows the TGA curve for the mixture of DBU and glycerol in 3:1 molar ratio as a function of time. After 1 hour's absorption, the maximum CO 2 loading achieved 32% of theoretical loading and change of absorption rate appeared to be similar with that present in Fig. 1(a). The decomposition temperature of DBU/glycerol/CO 2 is 60 °C 23 , and the desorption only took approximately 15 minutes at 65 °C. The weight loss was even larger than the weight of CO 2 absorbed after 10 minutes' desorption, which indicated that part of the DBU was evaporated during the TGA measurement.

Viscosity. The most accurate three-parameter equation for describing the viscosity, named
Vogel-Fucher-Tammann-Hesse (VFTH), was used to describe the viscosity of each component 24 , see equation 4. Then a simplified two-component model was used to predict the viscosity via Eq. (5). Figure 2 shows how the viscosity increases as a function of CO 2 loading. Like all studied switchable ionic liquids, an increase in viscosity can be observed after absorbing CO 2 since mass-transfer limitations are easily introduced 16 . According to Ikenna's study, the highest viscosity of DBU/glycerol mixture at 25 °C is 160 Pa · s 23 , which is much lower than the results presented here. This is mainly because the high viscosity results in an absorption rate so low that higher CO 2 loading (>90%) could not be achieved in short time by using magnetic stirrer. There is one thing needs to be pointed out here that when CO 2 loading is above 30%, DBU/glycerol/CO 2 mixtures become too viscous for the pump to cycle them for the friction test.
www.nature.com/scientificreports www.nature.com/scientificreports/ Considering the absorption/desorption and viscosity properties of DBU/glycerol/CO 2 mixtures, four mixtures with lower CO 2 loadings (0%, 7%, 14% and 21%) were prepared to investigate the tribological properties. Photos of these samples are shown in Fig. 3(a). It is obvious that absorption of CO 2 has a strong effect on viscosity. The viscosity of the liquid increases with increasing CO 2 loading ( Fig. 3(b)), from 0.15 Pa · s to 1.3 Pa · s, which stays in a good viscosity range for the EHL friction study. The viscosity is relatively insensitive to shear rate in the measured range as illustrated in Fig. 3(b), which will ensure a good friction measurement. friction test. The first EHL study was a qualitative analysis with the aim of capturing general trends of the friction coefficient with and without CO 2 . Different slide-to-roll ratios (SRR) and entrainment velocities were used and the result is presented in Fig. 4(a). Friction coefficients versus SRR and entraining velocities were measured with fluids without CO 2 and with approximately 7% CO 2 loading. The results clearly show that introducing CO 2 to the mixture reduced friction at all investigated entrainment speeds.
EHL friction is greatly influenced by the pressure-viscosity behavior of the lubricant. When CO 2 is added, it is likely that the pressure-viscosity coefficient is reduced and thus also friction. The film thickness increase itself may also alter the friction since the average Couette shear rate is reduced when the film becomes thicker.
To better understand the role of CO 2 in controlling friction, the prepared DBU/glycerol mixtures with different CO 2 loading (0%, 7%, 14% and 21%) were used for studying the relationship between the friction coefficient  www.nature.com/scientificreports www.nature.com/scientificreports/ and CO 2 loading. The slide-to-roll ratio was varied whereas the contact pressure and entrainment speed were kept constant at 1.35 GPa and 3 m/s, respectively. The recorded friction coefficients as a function of CO 2 concentration in the lubricants are shown in Fig. 4(b). At low SRR it is again shown that the friction coefficient is reduced with increasing CO 2 concentration. The trend was very similar at high SRR's with decreasing friction with increasing concentration of CO 2 , except for the specimen of 21% CO 2 loading where the friction remained more or less constant at SRR higher than 75%. Here, the friction becomes larger than the mixtures with 7% and 14% CO 2 loading. This may be explained by the contradicting effects of CO 2 and SRR.  The friction behavior of specimens after CO 2 desorption was investigated as well. In order to enhance the desorption of CO 2 , heating, stirring and bubbling with N 2 gas were applied. The regeneration temperature was set to 65 °C, same as the desorption temperature in TGA. After half an hour's desorption, the CO 2 loading decreased from 14% to 10%, and the friction behavior before and after desorption are shown in Fig. 4(c). For the mixture with partial desorption CO 2 , the expected increase of friction occurred at low SRR.

Discussion
When absorbing CO 2 , the friction of the specimen of 21% CO 2 loading becomes larger than the mixtures with 7% and 14% CO 2 loading at high SRR. It may be explained by the contradicting effects of CO 2 and SRR. The viscosity becomes higher as CO 2 loading increases but at higher SRR the high shearing may reduce the viscosity at the inlet to the EHL contact and thus also reduce the film thickness, with an increase in friction as a result. Yet, another possible explanation is that the switchable ionic liquids decompose at high shear rate. The decomposition temperature of DBU/glycerol/CO 2 is 60 °C 23 , which would indicate that the energy required in this decomposition reaction to break down the bonds present in the substance is not very high. It is possible that this temperature can be reached locally in the high-pressure contact 25 .
In order to understand more about the lubricating mechanism of DBU/glycerol/CO 2 mixture, the central film thickness of the lubricants as a function of the entrainment speed was investigated. This requires the refractive index of the lubricants to be known. Consequently, the refractive index of the lubricants was measured and the values obtained are presented in Supplementary Table 1. As shown, the effect of CO 2 is very small and can be neglected.
The central film thickness for the different lubricants is shown in Fig. 4(d). At a CO 2 loading below 14%, the central thickness increases linearly with increasing entrainment speed when plotted on a logarithmic scale. For the sample with 14% CO 2 loading, the central thickness shows a linear relationship when the entrainment speed is lower than 100 mm/s. As it can be seen in Fig. 3(a), the opacity of the fluid is increased with higher CO 2 loading. This creates a problem with the interferometry measurements and the mixture with 21% CO 2 loading could not be measured.
The experimental data were fitted using the following power function which is derived from the Hamrock-Dowson equation 26 where H C is the central film thickness (nm), U e is the entrainment speed (mm/s), a and b are constants. For 0% CO 2 loading, 7% CO 2 loading and 14% CO 2 loading, the film thickness increased with the entraining speed at a rate of U e 0 69 . , U e 0 70 . and . U e 0 72 respectively. This indicates that the lubrication film formation of DBU/glycerol mixture is slightly more sensitive to the entrainment speed at higher CO 2 loadings.
As mentioned, the pressure-viscosity relationship has a large impact on the viscosity inside the EHL contact, and thus also the EHL full film friction. Let us, for simplicity, use the concept of a pressure-viscosity coefficient, see for example the work of Hamrock and Dowson 26 . According to their film thickness equation, the pressure-viscosity coefficient is one of the most important factors that affect the film thickness in the EHL regime. Based on Van Leeuwen's work 27 , the Hamrock-Dowson equation can be used to calculate the pressure-viscosity coefficient, α, from film thickness data and this equation is considered to be superior to other equations used. Therefore, the Hamrock-Dowson equation 26 was employed here to calculate the α-value of the different DBU/ glycerol/CO 2 solutions. The Hamrock-Dowson central film thickness equation is defined as follows: where R x is the radius of curvature of the ball in -x-direction, R y is the radius of curvature of the ball in -y-direction, U is the dimensionless speed parameter, U 1 is the ball speed (mm/s), U 2 is the disc speed (mm/s), G is the materials parameter, W is the load parameter, C o is the ellipticity influence, η 0 is the viscosity at atmospheric pressure (Pas), E′ is the equivalent Young modulus (Pa), F N is the normal force (N) and α is the pressure-viscosity coefficient (Pa −1 ). Since all parameters, except the pressure-viscosity, are known it is possible to compute α. One must, however, understand that effects of inlet shear thinning and heating are not taken into account and this (2019) 9:13262 | https://doi.org/10.1038/s41598-019-49864-w www.nature.com/scientificreports www.nature.com/scientificreports/ implies that the absolute value of α may be affected by relatively large errors. Still it is possible to compare the computed values obtained in the same setup.
The calculated pressure-viscosity coefficients for the lubricants are listed in Supplementary Table 2. The pressure-viscosity coefficient decreases with increasing CO 2 loading, which may be one of the reasons for the observed trend with lower friction coefficient with increasing CO 2 loadings. A lower pressure-viscosity coefficient will lead to lower local viscosity in the high-pressure zone of the contact and therefore lower friction. Low pressure-viscosity has been given as an explanation for low friction in other fluids, for example glycerol 28,29 .
In conclusion, a novel lubricant with controllable friction properties has been investigated. The friction control is obtained by adding CO 2 to the lubricant. The modulation of friction is based on the switch of viscosity depending on CO 2 loading. It was demonstrated that the viscosity can be altered and that the lubricant does not suffer the problem of high mass transfer resistance at room temperature when the CO 2 loading is less than 21%. The EHL friction in ball-on-disc test switches to a lower value when CO 2 is absorbed, and it turns higher when CO 2 is released. This study has also shown that in pure rolling condition, with higher CO 2 loading, the film thickness will be increased and the pressure-viscosity coefficient will be decreased, which are believed to be the main reasons for the observed reduction of friction.
The DBU glycerol carbonate was prepared via bubbling of CO 2 through a 3:1 molar ratio solution of DBU and glycerol. DBU (127.38 g) and glycerol (25.42 g) were used when preparing a 3:1 molar ratio of the mixture. Thereafter, a narrow tube was inserted and CO 2 (AGA, 99.7+%, H 2 O < 100 ppm) was bubbled through the liquid. The reaction was exothermic and the mixture was stirred mechanically throughout the bubbling cycle. The CO 2 loading was calculated by the weight increment. The accuracy of the analytical balance used was 0.01 g. The water content was checked with a Karl Fisher titration device to be less than 0.5 wt% before and after each test. co 2 absorption and desorption. The experimental apparatus for measuring the gas absorption loading has been described previously 33 . It contains a gas reservoir, an absorption vessel, a magnetic stirrer and two pressure transducers. The water bath was heated to the set temperature (25 °C) and maintained for 1 hour. An accurate amount (3.3 ml) of a 3:1 molar ratio solution of DBU and glycerol was added into the absorption vessel. The dissolved gas was removed by a vacuum pump. Thereafter an accurate amount of pure CO 2 gas was pumped in to absorption vessel, and the pressure decrease was recorded.
In order to evaluate the performance of DBU/glycerol mixtures on CO 2 absorption and desorption, a thermogravimetric analyzer (TGA, Perkin Elmer 8000) was used. The initial activation of the sorbent was carried out by heating 5.127 mg of sample loaded in the platinum pan of the thermal analyzer, to 25 °C in pure nitrogen (99.99% purity) at a flow rate of 20 ml/min. Pure CO 2 (99.99% purity) was then introduced at the same flow rate and the temperature was kept at 25 °C. The CO 2 absorption rate of the sample was calculated from the mass gain after holding it at 25 °C in CO 2 for 60 minutes. Subsequently, the desorption of CO 2 was carried out by increasing the temperature to 65 °C for 15 minutes in an atmosphere of pure N 2 at a flow rate of 20 ml/min. Viscosity. The viscosities of the DBU/glycerol/CO 2 mixtures were measured at 25 °C using a CVO Bohlin Rheometer. The low viscosity samples were measured with the bob/cup geometry, i.e. cylinder in cylinder geometry with a clearance of 1.25 mm.
The shear rate was increased logarithmically from 0 to 100 s −1 . The high viscosity samples with the cone/plate geometry, cone-angle = 1°, cone radius R = 10 mm, at a constant shear rate of 10 s −1 . friction tests. The EHL friction behavior of the DBU/glycerol/CO 2 mixtures was investigated in a WAM (Wedeven Associates machine) ball-on-disc test apparatus, model 11A, at room temperature (ca 25 °C), at a load of 100 N (1.35 GPa maximum Hertzian pressure), with an entrainment speed between 1-5 m/s and a slide-to-roll ratio (SRR) between 0-100%. Before each test, the device and specimens were thoroughly cleaned with acetone and ethanol. Thereafter, the specimens, the ball and the disc were assembled. Then the relative position of the ball and disc was corrected to pure rolling and the desired slide-roll ratio was set. The disc was constantly lubricated using a recirculation pumping system attached to the test rig. A full description of this equipment has been reported previously 34 .
All specimens used in the tests, both balls and disc were made from AISI 52100 bearing steel. The balls were taken directly from the factory and the disc was processed in the same way as bearing raceway material. The balls were grade 20 with a surface roughness (Ra) of 30 nm, an outer diameter of 20.637 mm, and a hardness of about 60 HRC. The disc had a surface roughness (Ra) of 90 nm, 101.6 mm as outer diameter, a circumferential grind and were through hardened to about 60 HRC. film thickness investigation. The WAM-11 was used to investigate the formation of a lubricating film for the DBU/glycerol/CO 2 mixtures. A super-polished steel ball with a roughness of 10 nm (Ra) was brought into contact and loaded against a glass disc coated with a chromium semi-reflecting layer with a roughness of 1 nm (Ra). The chromium layer acts as a beam splitter to enable interferometric measurements of film thickness. The disc and the ball were driven by separate motors. During the tests, the steel ball was partially submerged in the lubricant. The lubricant was entrained into the contact by the ball and formed a lubricating film. A full description of this process has been reported previously 35 .