Mechanical Properties of Water-Assembled Graphene Oxide

Subscriber access provided by UNSW Library

Article

Mechanical Properties of Water-Assembled Graphene Oxide Langmuir Monolayers: Guiding Controlled Transfer Katharine L. Harrison, Laura Butler Biedermann, and Kevin R. Zavadil Langmuir, Just Accepted Manuscript • DOI: 10.1021/acs.langmuir.5b01994 • Publication Date (Web): 24 Aug 2015 Downloaded from http://pubs.acs.org on August 31, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Langmuir is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

Mechanical Properties of Water-Assembled Graphene Oxide Langmuir Monolayers: Guiding Controlled Transfer Katharine L. Harrison1,2, Laura B. Biedermann2, and Kevin R. Zavadil1,2* *Address correspondence to [email protected] 1

Joint Center for Energy Storage Research

2

Sandia National Laboratories, MS 0888, 0889, and 0892, Albuquerque, New Mexico 87185,

United States Keywords: graphene oxide, reduced graphene oxide, monolayer transfer, oscillatory barrier measurement, shear modulus, hydrophobic surfaces, Langmuir-Blodgett Abstract Liquid-phase transfer of graphene oxide (GO) and reduced graphene oxide (RGO) monolayers is investigated from the perspective of the mechanical properties of these films. Monolayers are assembled in a Langmuir-Blodgett trough and oscillatory barrier measurements are used to characterize the resulting compressive and shear moduli as a function of surface pressure. GO monolayers are shown to develop a significant shear modulus (10-25 mN/m) at relevant surface

ACS Paragon Plus Environment

1

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 29

pressures while RGO monolayers do not. The existence of a shear modulus indicates that GO is acting as a two-dimensional solid driven by strong interaction between the individual GO sheets. The absence of such behavior in RGO is attributed to the decrease in oxygen moieties on the sheet basal plane, permitting RGO sheets to slide across one another with minimum energy dissipation. Knowledge of this two-dimensional solid behavior is exploited to successfully transfer large-area, continuous GO films to hydrophobic Au substrates. The key to successful transfer is the use of shallow-angle dipping designed to minimize tensile stress present during the insertion or extraction of the substrate. A shallow dip angle on hydrophobic Au does not impart a beneficial effect for RGO monolayers, as these monolayers do not behave as two-dimensional solids and do not remain coherent during the transfer process. We hypothesize that this observed correlation between monolayer mechanical properties and continuous film transfer success is more universally applicable across substrate hydrophobicities and could be exploited to control the transfer of films composed of two-dimensional materials. Introduction Graphene-based materials on metal current collectors represent ideal constitutive building blocks for electrode architectures. Such intricate architectures comprising various structural and chemical forms of graphene ensure conductivity and electrolyte access, thus maximizing site (intrinsic and extrinsic) electroactivity and density.1, 2 Developing an understanding of the charge storage chemistry and its impact on graphene stability requires the ability to interrogate these carbon/metal interfaces at sufficient spatial and temporal resolution. Uniform assemblies of graphene-based nanosheets on metal substrates allow interfacial probe access to these structures while under electrochemical control. A means to assemble uniform, large-area graphene-based monolayer assemblies on metal current collectors is needed.


2

Page 3 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

Graphene oxide (GO) can be assembled as a monolayer and can be subsequently converted to reduced graphene oxide (RGO) for use in electrode architectures. GO comprises graphene functionalized with epoxides and hydroxyls on the basal plane and carboxylic acid on the prismatic plane.3, 4, 5 These oxygen modalities create an amphiphilic nanosheet stable in aqueous solutions and many alcohols. Where GO has many applications including non-volatile memory devices, biosensing, and water filtration,6, 7, 8 chemical or physical reduction of GO to RGO yields a useful 2D electrode material. Chemical, thermal, or optical treatments primarily remove the basal plane oxygen functional groups of GO, restoring conductivity but decreasing hydrophilicity.9, 10 GO may be reduced to RGO either in solution or after deposition on the target substrate. Both GO and RGO are pH-tunable materials; their carboxylic acid perimeter groups deprotonate at ~pH 10.11 Langmuir-Blodgett and Langmuir-Schaefer (LS) depositions have been used to assemble and transfer large-area monolayer films of 2D nanomaterials on fluid-fluid interfaces.12 Control of the nanomaterials’ amphiphilicity, solvent interactions, and substrate surface energy allows for optimization of the transfer results.12 Large-area LB monolayer transfer of GO onto hydrophilic substrates was demonstrated by Cote et al.13, 14 As has been observed for LB deposition of other 2D nanosheets,15, 16, 17 Hidalgo et al. recently demonstrated that LS dips onto hydrophilic surfaces resulted in lower coverage than LB dips under the same conditions; coverage for LS dips increased from 30% to 80% with increased Csp3/Csp2 ratio.18 Key deposition challenges remaining include creating continuous (≥90%) large area GO monolayer films on arbitrary (e.g. hydrophobic) surfaces and developing an LB-strategy for RGO films.13, 19, 20

LB and LS deposition of RGO requires the use of surfactants or nanosheet functionalization,21,

22, 23

both of which change the inter-sheet interactions and improve the stability of the RGO


3

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 29

Langmuir monolayer. Understanding the viscoelasticity of Langmuir monolayers can guide transfer of continuous, controlled-structure LB films. Continuous film transfer requires monolayers with high mechanical strength and elastic interfacial assemblies.24 An ideal, elastic monolayer quickly strains upon applied stress and relaxes equally rapidly upon removal of the stress.25 Elastic Langmuir monolayers can be reproducibly assembled and have the mechanical strength to remain cohesive during transfer. Specifically, transfer of nanosheet films requires that the films act as two-dimensional (2D) solid monolayers, which we define as an assembly comprising nanosheets bound at their perimeters. A 2D-solid nanosheet assembly behaves as a planar structure with long-range order rather than as a collection of independent sheets and exhibits an energy barrier to nanosheet rearrangement. This energy barrier indicates the presence of a significant shear modulus.26 Oscillatory barrier measurements allow quantification of the complex compression and shear components of the elastic moduli in situ with the same conditions used for Langmuir-Blodgett depositions.26 Understanding the viscoelasticity of GO and RGO monolayers can guide transfer of these monolayer graphitic films onto diverse substrates. This work characterizes the mechanical properties of GO and RGO monolayers assembled on water and capitalizes on understanding these properties to improve transfer of GO monolayers to arbitrary surfaces. Oscillatory barrier measurements reveal critical viscoelastic properties of these monolayers including the presence of a shear modulus for GO and the absence for RGO. Thus, GO sheets interact to form a 2D-solid monolayer, so we hypothesize that minimizing strain on the monolayer during transfer will maintain the continuity of the assembly upon transfer. We demonstrate that a shallow LB dip angle minimizes strain and allows GO transfer to hydrophobic surfaces with 99% coverage, greatly expanding the versatility


4

Page 5 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

of the LB technique. To the best of our knowledge, angled dipping is a novel strategy for improving 2D material LB transfer, though angled transfer of non-2D materials has been discussed in the literature in other configurations.27, 28 In contrast to GO, we are unable to LB transfer continuous RGO films; we attribute this failure to the absence of a shear modulus. With subsequent reduction after transfer of GO monolayers, LB could be a valuable process for creating graphitic electrode structures for electrochemically-based energy storage. Experimental Preparation of GO and RGO Langmuir monolayers. GO was synthesized following the ‘improved’ method of Marcano et al.5 We repeated these experiments with GO oxidized by another method29 and observed similar trends, so we do not think that the oxidation procedure is critical. The GO spreading solution was a 1:5 mixture of 0.1 mg/mL GO in water:methanol for which centrifugation was used to remove the smallest GO flakes and any agglomerates (see Supporting Information). 10-20 mL of the GO spreading solution was deposited dropwise on a clean water subphase Langmuir trough (KSV Nima medium trough, 364 x 75 x 4 mm) at a rate of 6 mL/h. Prior to deposition, contaminates were aspirated from the subphase until the variation in surface pressure (Π) during an isotherm was < 0.2 mN/m. To allow time for the methanol to evaporate, the monolayer was allowed to rest overnight before beginning experiments. The GO monolayer on a basic subphase was prepared in exactly the same way; the subphase pH was adjusted using ammonium hydroxide after the initial GO deposition on the trough. GO was partially reduced using ascorbic acid as proposed in Fernando-Merino et al.30 and was suspended in basic water (pH ~10.5 via ammonium hydroxide). Then the RGO was diluted to 1/5 of the initial concentration with water and methanol such that the final spreading


5

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 29

solution consisted of 1:1 methanol:water and had pH ~10. Prior to the spreading solution deposition and throughout the experiment, the water subphase was adjusted to a pH of ~10 using ammonium hydroxide to minimize aggregation. Typically only a few mL of RGO spreading solution was required to achieve the isotherms presented. Similar results were achieved using a more reduced RGO sample, but only partially reduced RGO allowed us to reproducibly achieve a spreading solution and monolayer that was not as prone to agglomeration of the sheets. RGO reduction and characterization are discussed in more detail in the Supporting Information. Oscillatory Barrier Measurements. Oscillatory barrier measurements were performed on the asprepared GO and RGO monolayers. Isotherms were run until there were negligible changes from one isotherm to the next. A paper Wilhelmy plate was positioned alternatingly parallel and perpendicular to the barriers in the Langmuir trough (Figure 1a). The barriers were compressed to a target area (A0) and then the position was held for two minutes to allow the monolayer to stabilize. A series of 10-20 barrier oscillations were then administered at a rate of 48 mHz with a strain of 5% (∆A/A0 = 5%) and the Π-A measurements were recorded. These measurements were repeated several times on different GO films and with the order of parallel and perpendicular measurements varied. The equations used to calculate the mechanical properties from the oscillatory barrier data are defined below and are derived in Cicuta et al.26 Langmuir-Blodgett Depositions. After monolayer deposition on the Langmuir trough as described previously, the monolayer was compressed at a rate of 50 mm/min to Π = 15 mN/m.

Π was held constant while transferring the GO monolayers to gold substrates. Surface treatments to prepare hydrophilic and hydrophobic gold are detailed in the Supporting Information. Dipping was performed using conventional techniques to deposit on the dip upstroke and downstroke.31


6

Page 7 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

The angle was varied between 30o and the standard 90o by attaching the substrate to an appropriately bent steel plate with a magnet, allowing for angled dipping while the dipping arm moved vertically. The surface of interest on the substrate was facing up away from the subphase, opposite of the orientation in a LS dip. To ensure equivalent transfer times between dips at different angles, the dipping rate was adjusted according to , = , ∗ (sin ), where rdip,90 is the 0.5 mm/min rate used for the 90o dips and θdip is the dip angle. Materials Characterization. An FEI Magellan was used for all Scanning Electron Microscope (SEM) images. A KSV Nima Brewster Angle Microscope (BAM) was used to create movies of the GO and RGO monolayers (see Supporting Information for more details). Results and Discussion Oscillatory Barrier Measurements. Langmuir-Blodgett deposition requires that the nanosheets form a stable and strongly assembled monolayer at the subphase-air interface.24 Understanding the nature of the GO monolayer, including the elastic moduli, is critical to optimizing LB transfer to various substrates. Although some viscoelastic properties of monolayers can be determined from isotherms and shear rheometry, these methods fall short by not allowing calculation of dissipative components32 or by not performing measurements in situ on the Langmuir trough geometry.24 Oscillatory barrier measurements were performed to quantify the compression (ε*) and shear (G*) elastic moduli to determine conditions suitable for LB deposition.26 Surface pressure area (Π-A) isotherms were measured with a paper Wilhelmy plate alternately parallel and perpendicular to the barrier sweeps of a Langmuir trough. Representative (Π-A) isotherms in Figure 1b exhibit three distinct compression regimes that will be discussed in more detail subsequently: percolation, high stiffness, and overlapping. Π anistotropy with respect


7

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 29

to the Wilhelmy plate orientation indicates the presence of a finite shear modulus. From the isotherms, the compressional elastic modulus, εeq, (Figure 1c) was calculated by equation 1.26, 33

=

≅

(1)

Figure 1. (a) Schematic of Langmuir trough showing Wilhelmy plate configurations. (b) Π-A isotherm and (c) εeq-A as measured with the Wilhelmy plate positioned in the parallel (▲) and perpendicular (●) configurations. Oscillatory barrier measurements were performed at areas corresponding to εeq ~ 10 mN/m in the percolation regime (●), to the maximum εeq (●) in the high stiffness regime, and to εeq ~ 5 mN/m in the overlapping regime (●).

Calculation of εeq highlights three target areas (A0) of interest at which to probe viscoelastic behavior as indicated by the circular markers in Figure 1c. Figure 2 shows the Π


8

Page 9 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

responses to programmed A oscillations for each regime. With increasing compression, ∥ becomes significantly larger than ! , confirming the presence of a shear modulus (q.v. Eq. 10 in Cicuta et al.26). Physically, the shear modulus indicates that the monolayer acts as a 2D solid in which there is an energy barrier associated with the GO sheets rearranging to accommodate stress.26, 34 With the Wilhelmy plate parallel to the barriers, the wide plate dimension directly interferes with the monolayer compression, forcing monolayer rearrangement around the plate. Increased compression (higher Π) is necessary to achieve the rearrangement and some hysteresis can occur in the isotherm due to the irreversible overlap, wrinkling, or buckling of the sheets that occurs with compression. Conversely, when the Wilhelmy plate is perpendicular to the barriers, the thin dimension of the plate contacts the GO monolayer and the 2D solid can easily slide around the plate to compress with the barrier motion. If the monolayer acts as a 2D solid, as in the case of GO, ∥ " ! and there is hysteresis in the parallel measurement. A monolayer’s elastic moduli can be calculated through oscillatory barrier measurements by quantifying the anisotropy in the surface pressure response to area oscillations.26

Figure 2. Area (black lines) and surface pressure (symbols) oscillations with the Wilhelmy plate in the perpendicular (●) and parallel (▲) orientations in the (a) percolation, (b) high stiffness,


9

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 29

and (c) overlapping regimes. A subset of the oscillations is shown here. refers to the surface pressure offset applied so that ∆ oscillates about zero (see Supporting Information). The complex compression, ε*, and shear, G*, moduli comprise storage (real) components

ε’ and G’ and dissipative (imaginary) components ε” and G”. ∗ = & + 0 && . ∗ = . & + 0. &&

(2) (3)

The storage components represent the energy input upon compression that is returned upon relaxation of the stress. Conversely, the dissipative components represent the energy input lost to friction, which arises due to the presence of finite viscosity.25, 26 These moduli components are calculated26 from the magnitude of the surface pressure oscillations, ∆Π, and the phase shifts, %, between the barrier motion and surface pressure response, as measured for oscillations of magnitude ∆A about trough position A0. 1 ∆∥ ∆! 1 ∆∥ ∆! & = ) *+,(%∥ ) + *+,(%! ) . & = ) *+,(%∥ ) *+,(%! ) 2 ∆ ∆ 2 ∆ ∆

(4)

1 ∆∥ ∆! 1 ∆∥ ∆! ′′ = ) ,0(%∥ ) + ,0(%! ) .′′ = ) ,0(%∥ ) ,0(%! ) 2 ∆ ∆ 2 ∆ ∆

(5)

For a purely elastic monolayer, the surface pressure and area oscillations are in phase; a purely viscous monolayer exhibits a π/2 radian phase lag. Thus, upon barrier oscillation, a lag in surface pressure (stress) response to changes in area (strain) indicates viscous behavior. Determining quantitative values for elastic moduli ε’ and G’ allow inferences of GO interactions in the three compression regimes, as presented in Figure 3 along with SEM images on hydrophilic gold surfaces and cartoons depicting the GO sheet interactions. In the percolation regime, ε’ is small because the GO sheets are only beginning to form a loose percolation network. In the high stiffness regime, ε’ increases as the sheets approach one another closely


10

Page 11 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

enough to interact. The GO sheets repel one another in this regime, causing the monolayer to be very stiff, giving rise to a large ε’ at the area associated with the maximum εeq. Finally, further compression of the monolayer forces the sheets to accommodate the compressive stress by buckling, wrinkling, or sliding over one another.11 This stress accommodation results in a more elastic monolayer and, in turn, a smaller ε’. Additional moduli data (e.g., dissipation terms) are presented in the Supporting Information.

Figure 3. Storage moduli components ε’ and G’ measured with target areas corresponding to the interaction regimes designated in Figure 1. Solid columns depict ε’, patterned columns depict G’ and error bars indicate standard deviations of the measurements. Below the column graph are cartoons showing GO sheets interactions as well as representative SEM images of LB dipped GO monolayers on hydrophilic Au surfaces in each regime. Like ε’, G’ increases from the percolation to the high stiffness regime as the sheets begin to interact. However, unlike ε’, G’ is relatively constant between the high stiffness and the overlapping regimes. The presence of a significant shear modulus is important as it indicates that the monolayer interacts as a solid, rather than a network of independent particles.26 GO is an


11

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 29

amphiphilic, 2D nanosheet due to the presence of hydrophobic islands of sp2-conjugated graphene surrounded by hydrophilic sp3-conjugated functionalized regions.4, 14 GO’s protonated –COOH edge groups present in the slightly acidic pH of GO in DI water may interact by hydrogen bonding, causing the wrinkling and buckling of the sheets evidenced on transferred monolayers.11, 35 We argue that the G’ develops due to the energy barrier associated with sheet edge interaction and overlap. Activation energy must be overcome to drive overlap between sheets due to the repulsion and steric hindrance of bulky oxygen moieties, thus creating an energy barrier to sheet rearrangement, i.e., a shear modulus. Barrier compression provides the compressive force needed to overcome the friction associated with sheet overlap. However, activation energy is also required to reverse the overlap, again because of friction associated with the sheets sliding on one another. Barrier expansion does not provide the tensile force directly on individual sheets that is needed to completely reverse the edge overlap. Thus, we posit that some sheets remain overlapping after barrier expansion, leading to hysteresis in the isotherms (Figure 1). BAM movies reveal the formation of large-scale GO networks that move as units and rearrange during compression and expansion (Supporting Information). Rearrangement of these units to fill gaps in the monolayer during compression and expansion provides a further energy barrier for rearrangement of the monolayer. The shear modulus measured here is two orders of magnitude higher than that measured through rheology experiments.24 We attribute the shear modulus discrepancy to differing stress accommodation mechanisms, which could arise due to varied sheet size, morphology, or chemical composition. The GO monolayer in Imperiali et al.24 accommodates stress primarily by wrinkling significantly, whereas our GO monolayer exhibits minimal wrinkling (as demonstrated


12

Page 13 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

in the BAM movies). These differences in monolayer behavior reinforce the importance of measuring the shear modulus in situ prior to deposition. Because sheet size, pH, and C/O ratio have also been shown to affect sheet overlap in transferred films, it would be illuminating to expand these measurements to the parameter space including varied pH, sheet size, and GO synthesis strategies to better understand how the edge interactions contribute to G’.13, 14, 35, 36, 37 In addition to the storage terms, oscillatory barrier measurements also allow calculation of the dissipative moduli terms. Minimal phase shifts between A and Π in all three regimes (Figure 2) and dissipative complex moduli terms (Supporting Information) approximately an order of magnitude smaller than the storage terms show that GO behaves primarily as an elastic monolayer in agreement with Imperiali et al.24 At higher Π, we do observe some dissipation in the compression modulus. This dissipation represents loss of energy, which could manifest itself as friction between the GO sheets as they slide upon one another. Standard deviations in phase shifts of

Mechanical Properties of Water-Assembled Graphene Oxide

Recommend Documents