Materials
Waste office papers (Deli A4), tissue papers (Vinda gentle facial tissue), newspapers and cardboards (Express box) were collected in work places and were used directly. Polyacrylamide (PAM, MW = 3,000,000) was purchased from Sinopharm Chemical Reagent Co., Ltd (China). Waterborne polyurethane (PU) was purchased from King Chemical Co., Ltd (China). All chemicals were used without further purification.
Preparation of waste paper pulps
To prepare printable paper inks, we first created a paper pulp composed of lignocellulosic fibres from waste papers ((Supplementary Fig. 1). First, waste papers were cut into small sheets by a paper shredder with a shred width of 4 ~ 5 mm (Deli GA311), and was ground in a mortar grinder (RM 200, Retsch, Germany) with excess water for 3 hrs. This prolonged grinding time ensures the fiber particles reached the minimal achievable particle size of the grinder. The resulting mixture was resuspended and filtered through a 150-mesh sieve to remove any large paper particles. The suspension was loaded into 50-ml conical tubes and centrifuged at 1000 rpm for 3 min to separate the paper fibers from the solution. The supernatant was then removed, leaving a paper pulp. For every batch of the solution, we measured the paper content of the paper pulp. This involved drying 0.1 g of the paper pulp at 100°C for 60 minutes, and then weighting the resulting dried pulp to calculate the solid content of the pulp for the following ink preparation. The same protocol was used to create all types of waste paper pulps, including office papers, newspapers, cardboard and tissue papers used in this study.
Preparation of 3D printable paper inks: (Supplementary Fig. 1 denotes the ink preparation procedure. The waste paper inks were composed of PAM (MW = 3,000,000), deionized water and paper fibre in a weight ratio of 0.1:8.9:1, unless otherwise specified. We added paper pulp, which contains water, to the inks. Therefore, the amounts of deionized water and paper pulp added into the inks were determined based on the measured paper content and water content of the paper pulp. All types of paper inks, including office paper, newspaper, cardboard and tissue paper inks, were prepared using the same formulation ratio.
For the paper filler-reinforced PU inks, two inks were prepared in this study. The ink formulation contains paper fibres, PAM, PU and water. The water content in both inks is 80% by mass. The remaining 20% is composed of paper fibres, PAM and PU at mass ratios of 1:0.15:2 and 1:0.15:2, respectively. The amounts of deionized water and paper pulp added were determined based on the paper content and water content of the paper pulps. All inks were homogenized and degassed using a planetary mixer (AR-100, Thinky) at 1000 rpm for 2 mins prior to printing. The prepared inks were stored at room temperature in a sealed container when not in use.
3D Printing of the pulp inks.
All 3D printing experiments were conducted using a multi-material direct ink writing printer (3D Bio-Architect WS, Regenovo). The printed structures were designed using a 3D modelling software (Cinema 4D, MAXON), and converted into G-code using a slicing software (Regenovo). Prior to printing, the ink was drawn into a syringe barrel and degassed. The barrel was then loaded into the syringe holder of the printer. A printability test was performed to optimize the printing parameters, including printing speed, extrusion pressure and nozzle-to-stage distance. The optimized parameters were found to be an extrusion pressure of 100 kPa, a printing speed of 20 mm/s and a nozzle-to-stage distance of 0.3 mm. All experiments were carried out using nozzles with an outlet diameter of 0.34 mm.
Solidification of the 3DP constructs.
After printing, the printed structures were dehydrated and solidified. Three drying methods were tested in this study, including freeze-drying, air-drying and ethanol exchange drying. Unless otherwise noted, all 3D structures were freeze-dried, and all 2D prints were air-dried. For freeze-drying, the 3D constructs were placed in a refrigerator at -20°C for 12 hr to ensure they were frozen prior to loading them to a freeze-dryer (Scientz, China). The constructs were then freeze-dried at -80 oC under a pressure of 1 Pa for 12 hr. For air-drying, the 2D prints were left at ambient conditions until fully dried, which took 4–12 hr in this study, depending on the size of the print. Samples dried via ethanol exchange drying were also prepared for mechanical tests. For ethanol exchange drying, the constructs were transferred to an anhydrous ethanol solution and soaked for 12 hrs. The constructs were then dried at ambient temperature on a sieve that enables ease of removal after drying. We noted that shrinkage of the 3D printed constructs happened with the ethanol exchange drying method. To precisely preserve the designated shape, a scale factor of 2.8 was applied to the print model for shrinkage compensation.
Construct recycling
To recycle the printed constructs, a process similar to the paper ink preparation was adopted. This process involved mechanical shredding, grinding, filtration, and centrifugation to obtain a pulp precipitate. The yield of the lignocellulosic paper fibre was around 98% for each recycling cycle. The resulting paper pulp was then mixed with PAM in a mass ratio of 10:1 to obtain a recycled paper ink. The printed samples created using the recycled paper ink were then air-dried at room temperature and subsequently subjected to mechanical testing.
Finite element modelling
All numerical simulations in this work were carried out using a commercial finite-element software, ABAQUS (Version 2022, SIMULIA, France). Two finite element models, the Miura-origami model and tessellated model, were developed to simulate the stress distribution or strain distribution within the models under z-axis compression. The models, consisting of stiff domains and soft domains, were built according to the geometrical dimensions stated in Fig. 4d and Fig. 4g, with the thickness of the Miura-origami model as 2 mm. The experimentally-measured stress-strain curves and Poisson’s ratio of the soft domains (paper ink mixed with PU) and the hard domains (paper ink) were imported as the material parameters of the respective domains for further simulation.
For simulation of the Miura-origami model, an implicit dynamic analysis was adopted where a slow loading (1 mm/min) was employed to emulate the quasi-static compression used in the experiments. For simulation of the tessellated model, a non-linear static finite element simulation was adopted which activating nonlinear configurations. A compression speed along the z direction was set as 1 mm/min which was applied to the top surface of the model. A mesh convergence test was conducted to obtain an optimal mesh size for accurate simulation results at an acceptable simulation time.
Optical microscope observation
Optical microscope (DM 2700 M, Leica, Germany) was used to visualise the paper fibre length before and after grinding. The fibres in the pulp were dispersed in water and a dried droplet of the dispersion was observed using the microscope.
Scanning electron microscopy
Before carrying out imaging using a field-emission SEM (Merlin, Zeiss), the samples were fractured under liquid nitrogen, and sputtered with platinum using a sputter coater (Q150TES, Quorum). SEM was performed at an acceleration voltage of 5 kV and an electric current of 100 pA.
Fiber length statistics
To evaluate the distribution of length of the paper fibres before and after the mechanical treatments, 0.01 g of paper pulp was dispersed in 5 mL of water. The solution was placed in an ultrasonic bath for 20 minutes to ensure proper dispersion, and then a droplet of the solution was placed on a dust-free glass slide and allowed to dry under ambient air. Micrographs of the samples were taken using an optical microscope. The length of the fibres in the micrographs were measured using ImageJ software (Version 1.54h). The number of fibres in each group of samples is n > 500.
Confocal microscopy: To evaluate the geometry deviation between the computer-aided design (CAD) model and the 3D printed model, a staircase sample (dimensions: 5.5 mm in height, 10 mm in length, and 10 mm in width) was fabricated using a paper ink, and its dimensions were acquired using a laser scanning confocal microscope (VK-X1000, Keyence).
Rheometry
All rheological measurements were carried out using a rheometer (DHR rheometer, TA instruments) equipped with a parallel-plate geometry (20 mm in diameter) at room temperature, and a gap size of 1000 µm was used. Shear viscosity measurements were performed in a steady-state flow mode over a shear rate range of 0.01–500 s− 1. Dynamic oscillatory strain amplitude sweep measurements were conducted at a frequency of 10 rad/s in an amplitude range of 0.01% − 200%. Oscillatory thixotropy and rotational thixotropy tests were also performed to determine the time-dependent moduli and viscosity recovery behaviours of the inks. In oscillatory thixotropy tests, an altered shear rate was applied stepwise between 0.1% and 200%, while in rotational thixotropy tests, an altered shear stress was applied between 0.1% and 200%. The results were analysed using TA Instruments TRIOS software.
Mechanical testing
All tensile tests were conducted using a universal testing machine (CMT2000K, SANSI, China) equipped with a 500 N load cell. Samples (10 mm X 10 mm X 0.2 mm) were clamped between two grippers, and were tested at deformation rates of 1 mm min− 1 (waste paper samples) and 20 mm min− 1 (pure PU) at room temperature. Compression tests were conducted using an electronic universal testing machine (MTS, 100 kN load cell, USA) at a deformation rate of 1 mm min− 1.
Photograph and video recording
Unless otherwise specified, all photographs in this work were taken with a SONY camera (A7RIV, Japan), and all videos were recorded using a Nikon camera (D7500, Japan).
Statistical Analysis
All the results in this study were presented as mean ± S.D., and all the mechanical properties presented in this study were measured from at least three parallel samples. Data distribution was assumed to be normal for all the parametric tests, but not formally tested, and no significant difference analysis was performed. The statistical analyses were carried out with the OriginPro 2021 software.