The Influence of Relative Humidity on Particle Adhesion – a Review of Previous Work and the Anomalous Behaviour of Soda-lime Glass

The adhesion of fine particles plays a significant role in the performance of particulate processes and in the quality of particulate products. The extent of adsorption of water on the particles from the surrounding atmosphere is governed directly by the relative humidity of the air. Published evidence suggests that changes in relative humidity of the air can have a profound inf luence on the adhesion of individual particles. However, there are numerous conf licting reports in the literature suggesting that adhesion can increase, decrease or pass through a maximum as the relative humidity increases. A thorough review of the relevant literature is presented in which the experimental evidence relating to the inf luence of relative humidity on particle adhesion is gathered and discussed. Apart from the amount of water adsorbed on the surface, it is clear that the adhesion depends upon the surface roughness which prevents the formation of a complete capillary meniscus at the particle contact point until a critical relative humidity is reached. At this point, sufficient water is adsorbed to engulf the asperities and a marked increase in adhesion is noted. Many surfaces undergo some physical or chemical change in the presence of adsorbed water, e.g. solubility, softening, or phase change. This can lead to complex adhesion behaviour. Hysteresis of adhesion with increase/decrease of relative humidity is commonly observed. Theoretical approaches correctly recognise the role of the Laplace pressure in the capillary bridge as being dominant in controlling adhesion. However, evidence suggests that contributions from the solid-solid interaction, surface tension of the bridging film and disjoining pressure can also be important under certain circumstances. As an illustration of complex behaviour, original data for the adhesion of a glass microsphere on a f lat glass surface are presented. The data were obtained using a custom-built AFM instrument. Upon desorption, a critical relative humidity lying between 30% and 40% was observed. At this point, a singular peak in adhesion occurred which was accompanied by long-range repulsion between the surfaces prior to contact. The long-range repulsion was found to be an exponential function of separation distance. This phenomenon has been attributed to the spontaneous formation of glass corrosion products from the liquid layer on desorption when the two surfaces approach each other. According to published literature, these corrosion products can take the form of needle-like or dendritic structures. The resulting steric repulsion is observed to be exponential with respect to distance, in agreement with the established trend. Once contact is made between the surfaces, the strong adhesion may be due to either sintering or entanglement of the corrosion products. Further work is required to determine the precise chemistry of the process. At high humidities, the force-separation curves exhibit an extended pull-off region, suggesting that the adsorbed film is mobile. This effectively reduces the adhesion by increasing the volume of the liquid bridge as the surfaces are pulled apart. This may explain why some reports in the literature show decreases in adhesion at high humidity.


INTRODUCTION
When ostensibly dry powders are exposed to the surrounding atmosphere, adsorption of water molecules takes place on the particle surface.The extent of adsorption will naturally depend upon the partial pressure of water vapour, the system temperature and the affinity of the particles for water molecules.The presence of an adsorbed water layer modifies the interactions between individual particles and has an effect on a number of process operations, e.g.f luidisation, compaction, agglomeration, conveying of powders and hopper f low.Adsorbed water can also inf luence the interaction between particles and a surface leading to unpredictable behaviour for processes such as gas filtration, coating, and deposition.The driving force for adsorption is usually expressed in terms of the relative humidity (RH) defined as follows: where p is the partial pressure of water vapour in the air, and P o is the vapour pressure of water.Relative humidity therefore accounts for variation in temperature through the associated variation in P o .
The inf luence of RH on various processes is discussed by Harriman and Simkins [1] and Bracken [2].They advocate a pragmatic approach, recommending RH limits for certain processes and discussing the relative merits of various methods of RH control.Whilst this approach may solve practical problems, a fundamental understanding of the inf luence of adsorbed water on particle interactions will lead to the development of more efficient processes and more effective products.All too frequently, adsorbed water is seen to be problematic in processes, but it must be borne in mind that there are potential benefits which can be identified through understanding the fundamental mechanisms involved.
The significance of air humidity in powder processing is stressed by Harnby et al. [3].They comment that small changes in RH can produce a drastic change in powder cohesion, leading to loss of process control.They review some experimental attempts to characterise the effect of RH on cohesion by studies on single particles and on bulk powder samples.The main conclusion of their study is that there are large discrepancies between the findings of various workers.Apart from differences in measurement techniques, they attribute the discrepancies to the large number of variables, the most significant of which p P o were thought to be the chemical nature of the surface, the particle size, shape and surface roughness.
The inf luence of adsorbed moisture on particle adhesion is not straightforward, and is expected to depend on the thickness of the adsorbed layer, the surface roughness, the surface chemistry, the contact geometry and any dissolution, intra-particle absorption or chemical changes that might arise due to the presence of water.To further complicate the issue, one can expect that some changes may happen over a long time scale whilst others may be instantaneous.There has therefore been much evidence in the literature that is apparently conf licting, whilst theoretical approaches have tended to oversimplify the issues.In the following sections we review the published evidence relating to the inf luence of humidity on particle adhesion, starting with pioneering experimental efforts in the first half of the last century, following the progress through to the present day.We report some original data for adhesion between glass surfaces that exhibit an anomalously high value of adhesion at a critical humidity.The singular peak in adhesion is accompanied by strong repulsion prior to contact that is not observed at any other humidity.It is this kind of behaviour that typifies the complexity of the inf luence of adsorbed water on particle adhesion.

A HISTORICAL PERSPECTIVE OF EXPERIMENTAL STUDIES
The first clues as to the influence of humidity on the force between surfaces came to light towards the end of the 1920s.Stone [4] made a qualitative study of the adhesion between pairs of glass spheres (diameter 1 Ҁ 2 mm) suspended on silk threads.He noted a significant decrease in adhesion when conducting the experiment in dry air, compared with the adhesion observed in ambient air.The difference was attributed to the presence of adsorbed water.From his adhesion observations he also inferred that adsorbed moisture could be removed from the spheres by a current of dry air provided the beads are separated.If the beads are in contact then adhesion still persists in spite of the dry air f low.This observation has implications for the storage and drying of static bulk powders.It is unfortunate that Stone made no report of relative humidities or force values for his otherwise carefully conducted experiments.
Tomlinson [5,6] conducted studies of the adhesion between pairs of quartz fibres and pairs of glass spheres by measuring the deflection of the fibres.He studied freshly formed surfaces in a vacuum, and noted that on exposure to the atmosphere, the surfaces lost most of their adhesion in a few hours.He attributed the loss in adhesion to the accumulation of contaminating matter from the air, reducing the surface energy of the solid surfaces.The striking difference in the trend between the observations of Tomlinson and Stone no doubt stem from Tomlinson's use of freshly prepared surfaces in vacuum, which will maintain much higher surface energies compared with the dry air conditions of Stone.
Bradley [7] used a spring technique to study adhesion between two spheres of quartz.In contrast to the findings of Tomlinson, Bradley reported that the adhesion was independent of whether the measurement was made in a vacuum or in air.Experiments were also conducted with freshly fused sodium pyroborate (Na 2 B 4 O 7 ) spheres.The adhesion in air was found to be about twice that measured for quartz.Furthermore in a vacuum, the adhesion was observed to drop to about a third of its value measured in ambient air.The high value of adhesion in air was attributed to the adsorption of water.The difference between the adhesion of quartz surfaces and borate glass surfaces is most likely due to the glass being more hydrophilic than the quartz.In the work of Stone, Bradley and Tomlinson, it is unfortunate that no values of relative humidity were quoted.
A more systematic study of the influence of relative humidity was made by McFarlane and Tabor [8] for the interaction between a spherical glass bead on a glass plate using a pendulum technique similar to that of Tomlinson [5].They produced direct verification of the expression for the adhesive force due to capillary action (F c ) between a sphere of radius R and a f lat with a thin film of liquid interposed, having a specific surface energy g and a contact angle q.Note that equation ( 2) is an approximation which is strictly valid only when the film thickness is very much smaller than the sphere radius (R) and the contact angle is small [9].
McFarlane and Tabor found equation (2) applied at or close to 100% RH.The value of the surface tension for the air-water interface was calculated from the graph of adhesion versus R to be 67.3 mJ.m Ҁ2 , which is less than the recognised value of 72.7 mJ.m Ҁ2 .The discrepancy was attributed to a systematic experimental error.For McFarlane and Tabor's system, the adhesion was negligible at low RH values, only showing a notable increase at an RH of around 80%.This increase was shown to correspond closely to a notable increase in experimental values for the adsorbed film thickness of water on glass.It is interesting that the maximum adhesion was observed to occur at RH values of 88% for glass on glass, even though the adsorbed film thickness is shown to carry on increasing until complete saturation is attained.
McFarlane and Tabor [8] studied the effect of surface roughness in the presence of adsorbed water, and concluded that increasing the roughness will decrease the adhesion.They proposed that adhesion falls as soon as the mean asperity height becomes comparable to the thickness of the adsorbed moisture film.At this point, multiple asperity contact was thought to occur with an associated decrease in adhesion.
A range of detailed studies of the influence of relative humidity on adhesion was made by Zimon in the 1960s and early 1970s, and these have been translated and summarised [10].Zimon characterised the adhesion between glass spheres coated on to f lat glass surfaces by the extent of particle detachment either in a centrifuge or by applying force via a pulsed vibration.This yielded an adhesion number, defined as the ratio of the number of particles remaining on the surface after an applied load to the number of particles originally present.
Zimon's experiments [10] showed that there was little change in adhesion over an RH range of 5 Ҁ 50%.Above an RH of 50%, Zimon noted a marked increase in adhesion.By comparison of experimentally determined values of adhesion with values predicted from equation (2), Zimon concluded that over an RH range of 50 Ҁ 65%, capillary forces are only beginning to take effect, whilst above 65% they prevail over other force mechanisms.This observation can be interpreted in terms of McFarlane and Tabor's inference [8] that capillary forces dominate once the adsorbed layer is sufficiently thick to engulf the surface asperities.
Zimon [10] modified the hydrophobicity of one or both contact surfaces by adsorption of various chlorosilanes.Compared with uncoated glass (adhesion numbers 90%), the adhesion number dropped drastically to between 2% at an RH of 25% and 40% at an RH of 90% when both surfaces were rendered hydrophobic.When only one of the contact surfaces was coated, the adhesion number lay roughly midway between the two above extremes.In all cases, however, the adhesion number was observed to increase monotonically with increasing RH.Values of the contact angle for water on uncoated and coated glass were given by Zimon as 30°and 80°, respectively.This gives a factor of 0.2 between the two adhesive forces predicted by equation ( 2).This factor is in broad agreement with the difference in the adhesion number for the uncoated and coated systems presented by Zimon.
Zimon observed in many cases that the adhesive force at saturation was significantly less than the corresponding theoretical values calculated from equation (2).He explained this phenomenon by invoking the concept of disjoining pressure exerted by the trapped film of water at the contact site.The disjoining pressure is a consequence of repulsive van der Waals forces that arise when the dielectric permittivity of the water is intermediate in value between the dielectric permittivity of the two solid surfaces, [9].Zimon proposed that a finite separation distance will be established at mechanical equilibrium, the magnitude of which is determined by the disjoining pressure.He embodied this in the following equation for the force of adhesion F ad where F c is the force due to capillary bridging and F disj is the disjoining pressure.It is possible that disjoining pressure could account for the difference between the measured adhesion and the adhesion predicted from capillary bridging at saturation.However, there are many other factors that can influence the adhesion, e.g.mobility of the adsorbed film, viscous effects and uncertainty of the contact angle.It is worth noting also that disjoining pressure is only manifest for specific combinations of interacting components.Furthermore, disjoining pressure does not arise for the interaction between surfaces of the same material, irrespective of the interposing medium.The final observation of Zimon [10] that is noteworthy relates to adhesion hysteresis with relative humidity.He reported higher adhesion on desorption than on adsorption for spherical glass particles interacting with a f lat quartz surface.This is most likely due to the reduced tendency for water to evaporate from a capillary meniscus once it has formed.Thus for a given humidity, more water will be held at a particle contact point when desorbing compared with the amount of water present when adsorbing.
The centrifuge technique employed by Zimon [10] has been used with good effect by Podczeck et al. [11,12,13] to study the effect of relative humidity on the adhesion of pharmaceutical powders to various surfaces.In general, their findings were specific to each system studied, which were by nature very complex; their irregularly shaped particles ranged from hydrophobic to hydrophilic, with varying degrees of solubility.Adhesion of the hydrophilic particles at high RH could be described by capillary bridging.This could not be applied to hydrophobic surfaces.They examined the effect of press-on force and contact time for their material.In some instances the press-on force had a profound effect, depending on the prevailing humidity.This set of work from Podczeck et al. successfully characterises the dependence of adhesion on humidity for these specific materials.It also gives an indication of the complexity of the underlying mechanisms, implying that broad generalisations regarding the effect of RH on adhesion cannot necessarily be transposed to other systems with success.
Further evidence relating to the fundamental nature of capillary menisci was gathered by Fisher and Israelachvili [14] and Christenson [15] using the surface force apparatus.This instrument and its applications to force measurement have been reviewed by Luckham [16].The contacting surfaces comprise two crossed mica cylinders, which give an almost perfectly smooth contact with a geometry equivalent to a sphere on a f lat surface.Christenson [15] found that the conclusions of Fisher and Israelachvili were in error in some instances by their use of a spring that permitted rolling and shearing of the surfaces in contact.The general trend observed by Fisher and Israelachvili, and Christenson, was that adhesion increased monotonically with increasing relative humidity to a maximum value corresponding to the value predicted by equation (2).Christenson identified this to occur at an RH of 70%, however, this is very system-specific, depending on the hydrophilicity and surface roughness, and the value is therefore somewhat arbitrary.The agreement between theory and experiment implied that the adhesion was dominated by the reduced Laplace pressure in the meniscus.Any other possible contributions to adhesion were negligible, i.e. solid-solid interaction, electrostatic double-layer force, disjoining pressure or the resolved component of the surface tension in the liquid neck.None of the authors attempted to address the adhesive mechanisms prevailing at lower RH values, prior to the onset of capillary formation.In a virtually saturated atmosphere, Fisher and Israelachvili observed that upon contact of the surfaces, the adsorbed moisture films become displaced from the contact region, leaving a residual monolayer of molecules sandwiched between the solids.This evidence must be taken into account when modelling the interaction between surfaces.It must be borne in mind that the surfaces used in these experiments had a smoothness verging on the atomic, thus the effect of surface roughness is not apparent.
The issue of the prevailing mechanism of adhesion prior to the onset of capillary bridging has been addressed by Chikazawa et al. [17].They used an electrobalance to measure the adhesion between surfaces of Pyrex glass, soda-lime glass, KCl and KBr.The potassium halide samples first showed notable adhesion at an RH of 20%.The configuration for the tests was sphere-on-f lat.As the RH was increased, a sudden increase in adhesion was observed at around 65%, dropping sharply off again at an RH of 70%.A similar event occurred for the glass samples although the onset of adhesion was not detected until RH҃ 40%, and the peak adhesion occurred between an RH of 70% and 85%.
Chikazawa et al. complimented their adhesion studies with measured adsorption isotherms of water on samples of the various materials studied.They concluded that there was insufficient coverage of water on the surfaces to facilitate capillary condensation at the onset of adhesion at an RH of 20% for the alkyl halides (40% for the glass samples).Between this point and the point at which adhesion was observed to peak, the authors proposed that hydrogen bonding was responsible for the interaction.The peak adhesion was proposed to coincide with the adsorption of sufficient water to allow capillary formation.No explanation was offered for the subsequent sharp drop in adhesion on increasing the RH.In the work of Chikazawa et al., the issue of surface dissolution of the KCl and KBr is expected to add to the complex interplay of surface roughness, microporosity and hydrophobicity.
A novel technique for measuring the adhesion force between an array of individual particles and a f lat surface has been developed by Harnby et al. [18].Several thousand particles of either glass ballotini or sand were painstakingly positioned in the apertures of an electroformed microsieve.The adhesive force between the array and a f lat surface of brass, glass or stainless steel was measured to within 0.1 mg using a top pan balance.The force between individual particles was then inferred by dividing the measured force by the total number of particles in the array.For example, the measured value for adhesion of a single glass ballotini particle (diameter҃155 µm) on a glass surface was 1.7҂10 Ҁ7 N, at conditions approaching 100% RH.This is over two orders of magnitude lower than the result predicted from equation (2), i.e. 6.3҂10 Ҁ5 N, using Harnby et al.'s measured contact angle of 26°.This is in contrast to the majority of other workers who find at least relatively good agreement close to saturation.It is therefore most likely that in spite of the careful efforts of Harnby et al. in the preparation of their array, there would have been some deviation in the height of the protruding particles.This would cause a significant reduction in the number of contacts compared with the number of particles present in the array, and hence is the most likely explanation for the low adhesion force measured.
In spite of the unrealistically small values of single particle adhesion reported by Harnby et al. [18], the overall trends are worthy of mention.No significant effect on adhesion was obser ved for different times of contact or predetermined loading values, implying that the contacts did not involve plastic deformation.However, the separation velocity was seen to exert a strong inf luence on adhesion, with the lowest separation velocities providing the strongest adhesion.It was concluded that low separation velocities allow the water sufficient time to f low into the liquid bridges, thereby strengthening them.In all conditions they observed an increase in adhesion with increase in RH, with no significant hysteresis.The critical value of RH at which the adhesion increased markedly was found to lie between 60% and 90%, which is broadly in agreement with the observations of Zimon [10].
The effect of surface roughness was also studied by Harnby et al.At close to 100% RH, sand particles contacting polished brass had only half the adherence of the same particles contacting roughened brass (4.4 µm).Harnby et al. consider this to be due to an increase in the contact area for the angular sand particles with the roughened surface.This is thought to be unlikely because the difference only manifested itself at close to 100% RH, whilst at all other values of RH, there is no discernible dependence of adhesion on roughness.It is more likely that the roughened surface promotes water condensation compared with the polished surface, and that the increase in adhesion is purely a result of more water present to form capillary bridges.The reverse trend was observed for glass ballotini on stainless steel.Smooth stainless steel resulted in significantly higher adhesion than roughened stainless steel.This trend was observed between RH values of 50% and 100%, suggesting that the roughened surface prevented effective contact between the glass ballotini and the adsorbed water layer, as previously identified by McFarlane and Tabor [8].
By analogy, the interaction between particles in a humid environment can be extended to the contact between the read/write head and computer disk drives.This issue has been addressed by Li et al. [19], and Tian and Matsudaira [20], who examined the friction force as a function of lubricant addition and relative humidity.Both groups reported an increase in static friction coefficient with increasing thickness of adsorbed water film.Attempts have been made to derive a theoretical model for the contribution to static friction between the head and disk from capillary forces, taking into account surface roughness and adsorbed film thickness [21,22].The model by Gao et al. [22] predicted critical conditions for the minimisation of static friction, but unfortunately the authors did not support their analysis with experimental observation.

ADHESION STUDIES USING ATOMIC FORCE MICROSCOPY
The study of particle adhesion entered a new phase with the development of the atomic force microscope (AFM).The high sensitivity of this instrumentation enabled the measurement of nano-scale forces and separations between particles with diameters of only a few micrometres.Sugawara et al. [23] used an AFM to measure the adhesion between a standard Si 3 N 4 probe tip and a mica surface.They measured the adhesion under UHV conditions and under ambient pressure with relative humidities of between 23% and 65%.Their data revealed a smooth monotonic increase in adhesion with increasing humidity.The maximum adhesion they recorded was in the region of 1.7҂10 Ҁ8 N (at RH҃65%), whereas the adhesion calculated from equation ( 2) is 2.3҂10 Ҁ8 N, given their quoted value of 25 nm for the probe tip radius.The authors attribute the discrepancy to the surface roughness of the cantilever probe.It is indeed likely that the effective tip radius is less than quoted due to irregularities or roughness.However, the agreement is close given that perfect wetting has been assumed.
Binggeli and Mate [24] used an AFM to study the adhesion and sliding friction between an AFM probe tip and a hydrophilic surface of silicon oxide as well as a hydrophobic surface of amorphous carbon.They assessed the amount of water adsorbed as a function of RH by studying the break-free length of the capillary meniscus when the probe tip is retracted from the surface.This method provides qualitative information regarding the film thickness.For more reliable measurements, account must be made of the film mobility and the fact that the presence of the probe tip on the surface will promote the localised conden-sation of water [25].The enhanced adsorption of water on the hydrophilic surface was shown to act as a lubricant, reducing the coefficient of friction.No reduction in friction coefficient was observed for the hydrophobic carbon surface, explained by the relative absence of water molecules on the surface.It is interesting to note that the adhesion for both surfaces was shown to decrease with increasing RH, over the range of 75% to 95%.The authors explained this reduction using a simple thermodynamic argument involving the chemical potential of the water film.This can be interpreted in terms of the decrease in adhesion that is predicted by a number of capillary bridge models as the volume of liquid between a sphere and a f lat plate increases, e.g.[26,27].The principle can be understood in terms of the reduction in Laplace pressure inside the bridge as the radius of curvature increases.It is worth noting that the opposite trend is predicted for conical geometry contacting a flat surface as the bridge volume (i.e. the RH) increases, e.g.[28,29].This is due to the lessened dependence on the Laplace pressure for adhesion for this geometry, compared with the force required to overcome the surface tension of the film.
Tang et al. [30] also studied the adhesion between an AFM probe tip and various surfaces including mica, calcium fluoride (CaF 2 ) and potassium chloride (KCl).In contrast to Sugawara et al. [23], Tang et al. found that the adhesion for both mica and CaF 2 did not vary with humidity over the range 20% to 80% RH.The authors conclude that the surface tension of the adsorbed film is responsible for the adhesion.This is likely to be the case for these very smooth surfaces.Studies with the KCl revealed adhesion values of nearly an order of magnitude higher than those for CaF 2 .In addition, they observed an increase in adhesion between 60% and 80% RH.The authors explain the difference by the increased surface tension promoted by the dissolution of KCl which is readily soluble compared with the other components.It is likely that there are other effects occurring here.The surface tension of water does not increase by an order of magnitude when adding KCl [31].Furthermore, there is evidence that at high concentrations, a solution of KCl in water becomes "capillary inactive", refuting the fundamental mechanism underlying equation (2).It is more likely that the high adhesion observed with the KCl arises due to the tip embedding into the surface that has been softened by the presence of adsorbed water.
Sugawara et al. [23], Binggeli and Mate [24], and Tang et al. [30] used the AFM probe tip as a surro-gate particle for studying interactions.However, it is possible to mount an individual particle on to an AFM cantilever using adhesive.Interactions between the probe particle and other surfaces or particles can then be measured directly.A number of researchers have used this principle to good effect.A study by Berard et al. [32] has looked at the interaction between the anti-inf luenza drug Zanamivir and αlactose monohydrate.The interaction between active drug particles and carrier or excipient particles is important for the effective operation of dry powder inhalation devices.They reported a significant increase in adhesion as the humidity was increased from 0% to 85% RH.They attributed this to capillary bridging.However, they also observed modification of the surface of lactose due to adsorbed water.Smaller surface features preferentially dissolve and the resulting supersaturation promotes growth on larger features.This process, known as Ostwald ripening, has also been observed for boric acid particles [33].It results in a significant change in surface roughness which is expected to modify adhesion.It may also lead to the production of a different phase on the particle surface which will modify its properties.
Using an AFM, Jones et al. [34] conducted a thorough study of adhesion between flat surfaces of glass or silicon and silicon AFM tips or glass microspheres.Adhesion as a function of relative humidity was observed over the range 5 Ҁ 90% RH.The f lat substrates were treated to render them hydrophilic or hydrophobic.The interaction between the silicon probe tip and hydrophilic glass or silicon showed a steady increase in adhesion with RH that was reversible.The magnitude of adhesion agreed well with the Laplace-Kelvin theory of equation ( 2).The interaction between hydrophilic surfaces and glass microspheres (radius ȁ20 µm) also showed a steady and reversible increase in adhesion with increasing RH.However, these data fell at least three times below the adhesion predicted by equation ( 2), implying that full capillary formation was prevented by the surface roughness.Measurements of the surface profile of a glass microsphere revealed asperities of the order of 15 nm high and 50 nm across.Using the Kelvin equation, Jones et al. calculate the critical meniscus radii to range from 0.2 to 5 nm over the RH range 10 Ҁ 90%.This further supports the notion that full capillary meniscus formation was not achieved.
With hydrophobic f lat surfaces, Jones et al. [34] observed a lower force, as expected, for all surfaces studied.However, they observed an anomaly for experiments with glass (glass microsphere on glass f lat, and silicon probe tip on glass f lat) in the form of a high peak in adhesion at a relative humidity of between 20% and 40%.Their corresponding data are reproduced here in Figure 1.This adhesion peak was accompanied by a long-range attraction prior to contact, effective over a distance of around 2 µm.The effect disappeared on repeating experiments at the same contact spot, but reappeared when moving to a new region on the surface.It was also shown to occur both during adsorption and desorption.The authors attribute the anomalous attraction to an electrostatic effect and acknowledge that it is likely to be a transient and dynamic in nature, rather than being a true equilibrium phenomenon.The presence of electrostatic charges on surfaces at this level of humidity is somewhat surprising.Visser [35] stated that humidity eliminates or greatly reduces the effect of electrical forces.This statement has been corroborated by Wan et al. [36], who measured adhesive forces and surface charges for surfaces of mica and silica.High values of adhesion were observed during experiments in dry nitrogen when the relative humidity was between 1 and 2%.Upon increasing the RH slightly (still with RH 10%), the adhesive force fell by about an order of magnitude.The drop in adhesion coincided with a decrease in the time taken for electrical charge dissipation from the surfaces; the exponential decay time constant fell by about two orders of magnitude.
In addition to their experimental study, Jones et al. [34] give a review of theoretical attempts to describe Fig. 1 Anomalous adhesion between a glass microsphere and a f lat glass surface, both of which were rendered hydrophobic.Reproduced from Jones et al. [34].
particle adhesion, acknowledging the various contributions from solid-solid interaction forces, (van der Waals forces), capillary bridging and long-range electrostatic (Coulomb) forces.In a later paper, Jones et al. [37] used an AFM to measure adhesion for powders of commercial interest.They observed a mix of behaviour with some powders exhibiting a marked increase in adhesion with increase in RH, whilst others showed no effect.This again confirms the complex nature of the prevailing phenomena.It is interesting to note that for the majority of powders tested, the adhesion was seen to increase with increasing maximum applied load (push-on force), suggesting that permanent deformation was occurring at contact areas.
A recent paper by Rabinovich et al. [38] addressed the transition between asperity contact and full capillary condensation for glass microspheres on a range of f lat surfaces having varying degrees of roughness.The authors developed a model for the prediction of adhesion comprising two components.The first component accounts for the van der Waals forces between dry asperities, and incorporates the RMS roughness and the peak-to-peak distance between the asperities.The second component predicts the adhesion due to capillary bridging in an equation similar to equation ( 2) with a modification for the effective separation between the particle and the average surface plane, allowing for the presence of asperities on the surface.The first component was considered to be dominant prior to engulfment of the asperities by the capillary meniscus.Once engulfment occurred, the second component was considered to dominate.The transition between the components was shown to occur at a critical humidity.Experimental measurements gave excellent agreement with the theoretical approach.The critical humidity was observed to increase with increasing roughness.For the smoothest surfaces considered (RMS roughness 0.2 nm), the transition occurred at an RH of approximately 25%, rising to a value of 65% for a surface with an RMS roughness of 3 nm.Rabinovich et al. acknowledge the work of Coelho and Harnby [39,40] in predicting this critical humidity from a thermodynamic perspective involving the BET model for water sorption coupled with the Kelvin equation for liquid bridge stability.However, this model typically results in a critical RH in the region of 70 Ҁ 99%.Rabinovich et al. commented that a critical RH of 25% observed for smooth surfaces corresponds to a critical meniscus radius of approximately 0.4 nm as predicted by the Kelvin equation.This is an order of magnitude smaller than the accepted lower limit for this macroscopic analysis [14,15], indeed it is of the same order as the nominal diameter of a water molecule.Doubt is therefore cast on the model of Coelho and Harnby for predicting critical humidity.Rabinovich et al. propose that a capillary meniscus existing at this low level of RH may well not be in thermodynamic equilibrium with the surrounding vapour, but may be in a metastable state.The work of Rabinovich et al. elegantly demonstrates the profound inf luence of nano-scale surface roughness on adhesion, and their modelling approach incorporating surface roughness shows great promise.Of course, the implication is that roughness data can be readily obtained in a form that realistically represents the specific physical situation.
The development of adhesion measurement as a function of relative humidity has been tracked, and the key experimental techniques have been identified.For fine powders, the centrifuge techniques give useful results that relate to the powder assembly.However, it is the AFM that enables detailed study of the relevant mechanisms at the level of the single particle.The disadvantage with the AFM technique is the large statistical variability that arises between individual measurements.In order to obtain statistically meaningful data, it is important to take sufficient measurements and apply some statistical criterion of confidence.The experimental evidence regarding the inf luence of humidity on particle adhesion has been presented.It is clear that the mechanisms responsible for adhesion are a complex interplay of surface roughness, amount of adsorbed water, electrical and mechanical properties of the solid, applied load and strain rate for separation.Further complexity is introduced for surfaces that interact with water in terms of solubility or plasticization.The modelling of particle adhesion as a function of relative humidity is still in its infancy.

ANOMALOUS BEHAVIOUR OF GLASS SURFACES
Jones et al. [34] have reported anomalously high values of adhesion for glass systems at certain values of relative humidity.Their data are reproduced here in Fig. 1.Note how a singularly high value of adhesion was observed at a relative humidity of 40%.We now report original data that are reminiscent of this anomaly.The study was conducted on a custom-built instrument based on the principle of an AFM with vertical displacement but no facility for scanning.A full description of the development and operation of the instrument has been given in a previous publication [41], including details of cantilever spring constant determination and methodology for mounting particles on to the cantilever.
Adhesion was studied using single particles interacting with a flat surface.The particles repeatedly cycled in and out of contact with the surface at a frequency of 0.1 Hz, corresponding to an approach and retraction velocity of ȁ0.5 µm.sҀ1 .The interaction between glass on glass and gold on gold was studied.The glass surfaces were cleaned in iso-propyl alcohol prior to measurement, to remove any organic contaminants.The gold surfaces were formed by sputtercoating gold on to a glass microsphere and glass slide.Sufficient time was allowed for sputtering to ensure a layer of gold in excess of 50 nm thick.Humidity control during an experiment was achieved by placing the entire instrument in a Perspex chamber and incorporating a beaker of distilled water or a tray of silica gel to raise or lower the RH accordingly.In practice, this gave a sufficiently slow change in RH such that instantaneous equilibrium could be assumed.This was tested experimentally by inducing a step-change in RH and observing that the adhesion force reached its equilibrium value directly.
Typical force-separation curves are shown in Figures 2 and 3 for glass-on-glass at different values of RH.Each force curve, comprising 1000 data points, has been corrected to account for the cantilever def lection [41] to give true separation between the surfaces.The force data are therefore vertical for zero separation, as expected for rigid body contact.Both figures show the presence of attractive force prior to contact.The adhesion force was obtained from the minimum in the cantilever retraction curve, corresponding to the maximum deflection experienced by the cantilever prior to detachment from the surface.Note that at the moderate humidity shown in Fig. 2, there is little or no elongation of the liquid bridge when the particle is retracted from the surface.However, at the high humidity shown in Fig. 3, there is substantial elongation of the liquid bridge prior to complete detachment at a separation distance of ȁ1500 nm, suggesting that the liquid film is mobile at 83% RH.This mobility effectively increases the volume of the liquid bridge, and consequently reduces the measured adhesion for this geometry.If the film were not mobile, the adhesive force should correspond closely to that predicted by equation (2).It would be interesting to study the effect of strain rate on adhesion at this level of humidity to assess the contribution to adhesion from the vis-cosity of the mobile film.
A typical plot of adhesion as a function of relative humidity is shown in Figure 4 for glass-on-glass.Three data points were taken at each humidity, hence the scatter in the graph.Also superimposed on Fig. 4 are the adhesion data for gold-coated surfaces.Adhesion values for the glass system and gold systems generally lay in the range 2000 to 6000 nN.This is well below the value of 17000 nN predicted by equation (2) for a sphere-on-f lat configuration.On increasing the RH, little change in adhesion was observed with relative humidity for all surfaces studied.However, on desorption, a singular peak in adhesion was repeatedly observed to occur between a relative humidity of 40 and 30%.This is apparent in Fig. 4 at a relative humidity of 32.4%.The bizarre feature of this singular peak is that the strong adhesion is accompanied by a long-range (ȁ300 nm) repulsion between the surfaces.The force curve corresponding to the peak in Fig. 4 is shown in Figure 5. Repulsion is seen as the positive force prior to contact between the surfaces.
The nature of the strong adhesion and strong repulsion at this peak force is of great interest.Regard-ing the repulsion prior to contact, the possibility of electrostatic repulsion was considered.Experiments were therefore repreated with an α-radiation source (Americium 241) in close proximity to the surfaces.This technique is advocated by Tomlinson [5] and Zimon [10] amongst others, to dissipate electrostatic charge.The strong repulsion pervaded even in the presence of the α-radiation source, suggesting that the force was not electrostatic in nature.Further evidence against electrostatic repulsion was gained by analysing the dependence of repulsive force on separation for a power law dependence.A log-log plot of repulsive force versus distance yielded a curve rather than a linear relationship, suggesting that electrostatic charging was not responsible for the observed repulsion.
Further analysis of the repulsive section of the curve revealed that the force decayed exponentially with separation distance.It is significant that others have also observed this dependence for repulsive forces between surfaces.Vigil et al. [42], examining the adhesion of silica surfaces in air, describe an exponentially decaying repulsion which they explain by the presence of a brush-like structure on the surface formed by the interaction between the silica surface and water vapour.Exponentially decaying steric repulsion has been observed by others working with polymer systems [43,44].
Vigil et al. [42] explain their brush-like structure on silica by growth of silanol and silicilic acid chains on the surface which give rise to the steric repulsion on the surface and have a sintering effect when contact is made, giving rise to high values of adhesion.It is likely that a similar process can account for the repulsion and strong adhesion observed in the data presented here.A further possibility comes from the presence of glass corrosion products.It is well-established that soda-lime glass can undergo a range of water-induced surface reactions [45,46].For example, soda (Na 2 O) can react with water to give sodium hydroxide, which will in turn react with any carbon dioxide present to yield sodium bicarbonate and ultimately sodium carbonate [46].We postulate that this kind of reaction can occur at high humidity in the relatively thick adsorbed liquid layer.Upon desorption, the water layer becomes supersaturated and the corrosion products precipitate out as a needle-like structure on the surface.Needles of this type up to 1 µm in length have been reported by Arman and Kuban [47] for soda-lime glass stored in a humid environment.Figure 6 shows a scanning electron microscope image of the contact region of a glass microsphere taken immediately after it had experienced repulsion and peak adhesion.The features on the surface are thought to be corrosion products that have crystallised from the supersaturated water layer.They have subsequently been flattened by the forced contact between the surfaces.It is not clear at this stage exactly what reactions are occurring on the surface.Further work is required in this area to identify the precise chemistry that occurs on the surface of glass surfaces contacting in the presence of water vapour.The anomalous adhesion for glass is reminiscent of the work of Jones et al. [34] who studied a similar system, see Fig. 1.However, Jones et al. observed this behaviour for glass surfaces that had been rendered hydrophobic.No anomalous peak was observed by them for untreated (hydrophilic) glass surfaces, on increasing or decreasing the relative humidity.They also obser ved similar behaviour to that in Fig. 1 for a silicon probe tip interacting with hydrophobic glass.Note that Jones et al. observed their anomalously high force during both increases and decreases in relative humidity.In all cases, their peak in adhesion was accompanied by a strong attractive force prior to contact.This is in contrast to the work presented here, in which the peak adhesion was accompanied by strong repulsion prior to contact.In spite of the ostensible similarity in the adhesion peak between the work of Jones et al. and this present work, the large number of differences suggests that the responsible mechanisms may be different in each case.

CONCLUSION
A review of the literature relating to the influence of relative humidity on particle adhesion reveals that the behaviour is exceedingly complex.Adhesion is shown to depend on the thickness of the adsorbed layer of water, which in turn depends on the surface chemistry of the surfaces.Surface roughness plays a significant role, preventing the formation of a complete capillar y meniscus at the particle contact point until a critical relative humidity is reached.At this, point sufficient water is adsorbed to engulf the asperities.Some systems exhibit adhesion hysteresis with relative humidity.More complex behaviour can be expected if the surfaces experience some modification induced by adsorbed water, or if the surfaces are partially soluble.Theoretical approaches correctly recognise the role of the Laplace pressure in the capillary bridge as being dominant in controlling adhesion.However, evidence suggests that contributions from the solid-solid interaction, surface tension of the bridging film and disjoining pressure may also be important under certain circumstances.There is also evidence that water in the contact region may not necessarily be in thermodynamic equilibrium, existing instead in a metastable state.Adhesion data also have reportedly shown dependence on the separation strain rate in certain cases.
Original data have been presented for the adhesion of a glass microsphere on a f lat glass surface.At high humidities, there is strong evidence that the adsorbed film is mobile, effectively reducing the adhesion by increasing the volume of the liquid bridge as the surfaces are pulled apart.Upon desorption, a critical relative humidity has been observed that yields a singular peak in adhesion, which is accompanied by longrange repulsion between the surfaces prior to contact.The long-range repulsion is an exponential function of separation distance.This phenomenon has been attributed to the formation of glass corrosion products which crystallise spontaneously from the liquid layer on desorption when the two surfaces approach each other.These are suspected to be in the form of needle-like structures that give rise to steric repulsion.Once contact is made, the strong adhesion may be due to either sintering or entanglement of the corrosion products.Further work is required to determine the precise chemistry of the process.

Fig. 2 Fig. 3
Fig.2Force-separation curve for the interaction between a glass microsphere (diameter҃40 µm) and a flat glass surface at RH҃54%.

Fig. 4 Fig. 5
Fig.4 Adhesion as a function of relative humidity for sphere (diameter 37 µm) interacting with a flat surface.The glass interactions are shown for decreasing humidity.The same peak in adhesion was not observed during increasing humidity.The data for gold-coated glass were obtained whilst increasing and decreasing humidity.

Fig. 6
Fig.6SEM image of the contact site of a glass sphere directly after the repulsion and peak adhesion were observed.The features are proposed to be glass corrosion products that have spontaneously formed by crystallisation from the supersaturated liquid layer on decreasing the humidity.

Table 1
Summary of experimental studies of the influence of relative humidity on particle adhesion.