SCIENCE ADVANCES | RESEARCH ARTICLE

APPLIED SCIENCES AND ENGINEERING Copyright © 2020

The Authors, some
Additive manufacturing of cellulose-based materials rights reserved;
exclusive licensee
with continuous, multidirectional stiffness gradients American Association
for the Advancement
of Science. No claim to
P. A. G. S. Giachini1*, S. S. Gupta1*, W. Wang2, D. Wood1†, M. Yunusa2, E. Baharlou1,3, original U.S. Government
M. Sitti2,4†, A. Menges1† Works. Distributed
under a Creative
Functionally graded materials (FGMs) enable applications in fields such as biomedicine and architecture, but their Commons Attribution
fabrication suffers from shortcomings in gradient continuity, interfacial bonding, and directional freedom. In NonCommercial
addition, most commercial design software fail to incorporate property gradient data, hindering explorations of License 4.0 (CC BY-NC).
the design space of FGMs. Here, we leveraged a combined approach of materials engineering and digital processing
to enable extrusion-based multimaterial additive manufacturing of cellulose-based tunable viscoelastic materials
with continuous, high-contrast, and multidirectional stiffness gradients. A method to engineer sets of cellulose-­
based materials with similar compositions, yet distinct mechanical and rheological properties, was established. In
parallel, a digital workflow was developed to embed gradient information into design models with integrated
fabrication path planning. The payoff of integrating these physical and digital tools is the ability to achieve the

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

same stiffness gradient in multiple ways, opening design possibilities previously limited by the rigid coupling of
material and geometry.

INTRODUCTION additional challenge for all MMAM methods is how to incorporate

Functionally graded material (FGM) has gradual change of composition continuous material gradients throughout the design-to-manufacturing
or structure in either a continuous or a stepwise manner, giving rise to a workflow in which the gradient design is no longer restricted by the
corresponding change in the property of the material (1). Its principle traditional boundary representation of common computer-aided design
is similar to that of many natural materials: The concentration of a (CAD) software (17, 25). Specifically, although voxel-based software,
specific functional component or the distribution of a particular structural such as Monolith (26), enables complex material arrangement, it lacks
arrangement changes gradually across space, often adapting to the automatic distribution of materials based on performance requirements
functions at multiple locations of a single organism (2, 3). This adapt- and strong integration between design and manufacturing processes
ability of composition and structures at the local scale endows FGM for the fabrication of objects with continuous material gradient (25).
with the capability to fulfill multiple, sometimes conflicting, design Here, we show an FGM manufacturing method that combines
requirements and promises applications in diverse fields, such as thin materials engineering and digital processing and uses both constructive
films (4), coatings, biomedical engineering (5, 6), and architecture (7). and mass transport processes in creating continuous gradients (Fig. 1).
A typical class of FGMs is materials with stiffness gradients; they can Specifically, the solutions of a cellulose derivative were engineered
better distribute stress at interfaces (3, 8–10), program deformation of to give tunable viscoelastic properties that enabled controlled extrusion
soft actuators (11–15), and influence the speed of cell migration (16). (Fig. 1A), and a digital workflow was developed to embed gradient
Materials with functional stiffness gradients can be fabricated by information in the designs and generate custom G-code to control
constructive processes, where requisite material parts are directly stacked the three-dimensional (3D) printer and syringe pumps (extrusion
together to form gradients, or mass transport processes, where physical and positioning systems in Fig. 1B). Filaments of varying composi-
processes, such as molecular diffusion or fluidic flows, create the gradients tions and/or cross sections were extruded in a solution phase to
(1). Multimaterial additive manufacturing (MMAM) (17), a type of facilitate molecular diffusion across filament boundaries in creating
constructive process, enables simultaneous geometric and compositional continuous gradients (Fig. 1C). To highlight the relevance of com-
variations during fabrication (18, 19), producing less waste than direct bining materials engineering with custom manufacturing strategies,
coupling of materials, layering, or subtractive manufacturing (20–22). this research uses a biopolymer-based fabrication material that is
Among the several methods of MMAM that offer various tradeoffs environmentally friendly and abundant, with diverse applications
among complexity, printing resolutions, scalability, and material from tissue engineering (16) to the construction industry (27). The
types, the extrusion-based method is noted for its simplicity, low cost, synthesis of these physical and digital tools is the ability to create
scalability, and wide range of material choices (18, 23, 24). Despite multidirectional, continuous stiffness gradients in a variety of ways,
these advantages, extrusion-based methods suffer from low gradient expanding design possibilities for FGMs.
resolution and weak bonding between dissimilar materials (17). An

1
RESULTS
Institute for Computational Design and Construction, Faculty of Architecture and
Urban Planning, Stuttgart University, Stuttgart, Germany. 2Physical Intelligence
Printing materials with tunable rheological and
Department, Max-Planck Institute for Intelligent Systems, Stuttgart, Germany. mechanical properties
3
School of Architecture, University of Virginia, Charlottesville, VA, USA. 4School of Our approach to engineer a printing material combines the preparation
Medicine and School of Engineering, Koc University, Istanbul, Turkey. of a base material solution that provides tunable rheological properties
*These authors contributed equally to this work.
†Corresponding author. Email: sitti@is.mpg.de (M.S.); achim.menges@icd. for extrusion printing and the addition of mixable additives that
uni-stuttgart.de (A.M.); dylan.wood@icd.uni-stuttgart.de (D.W.) extend the range of properties and property gradients of the final

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 1 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

A C
HEC acqueous Controlled solution mixing
solution with additives and deposition

Molecular structure of HEC

Blending of filaments to
RO
achieve continuous gradients
O R = H or
RO OR
O O O R * H
OR OR O
x
RO OR n

B Control system II
IV
I Extrusion system
III

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

Computer III Positioning System I

II V
I Extrusion system
II Material reservoir
III Positioning system
IV Delivery tubes
V Printed object

Fig. 1. Schematics of the fabrication process for printing continuous gradients. (A) Schematic of the preparation of printing solutions. Hydroxyethyl cellulose (HEC)
in powder form is dissolved in water and mixed with additives in a beaker before being transferred into a syringe. (B) Diagram of the control system and schematics of the
3D printing system. The diagram (left) shows the communication workflow that synchronizes the (I) extrusion system and (III) positioning system through a feedback
loop. The 3D printing system (right) consists of (I) an extrusion system (one or two syringe pumps), (II) the reservoirs (syringes), and (III) a positioning system (customized
low-cost 3D printer TEVO Tarantula i3). Delivery tubes (IV) are equipped with extruding tips and transport the printing solutions and deposit filaments of the solutions on
the printing platform (V). (C) Illustration showing the blending of filaments after deposition. Neighboring filaments blend into each other through the diffusion at the
molecular scale, thus creating seamless objects and continuous gradients.

printed material (table S1). Hydroxyethyl cellulose (HEC), a thickening as shown in Fig. 2B, extending the gelation point to around 261 min
and gelling agent in various applications in the food and cosmetics (172% increase) and enabling satisfactory extrusion consistency.
industry (28), was chosen as the base material. HEC is a derivative Figure 3A shows the characterization of the printed material and
of cellulose, the most abundant biopolymer on Earth, and it is the effect of additives. The addition of lignin increased the stiffness
nontoxic, biodegradable (29), and environmentally friendly (30). Its (Fig. 3B) and tensile strength (Fig. 3C) the most, while the inclusion
solution exhibits shear thinning properties (31) and pH-dependent of CA decreased these mechanical properties the most. The former
gelation time, which allows extrusion-based printing and further result is due to the fact that lignin is a natural reinforcer to cellulose
adjustment with additives. We have selected and tested five different in Nature, serving as a binder that cross-links polysaccharides to confer
nontoxic additives compatible with HEC base solutions that vary strength. The CA was most effective in decreasing the mechanical
the rheological and mechanical properties (table S1). properties because the HEC hydrogel is formed via physical cross-­
The gelation point of HEC, marking its transition from an aqueous linking of hydrogen bonds. The presence of sufficient H+ ions in the
solution to a solid hydrogel, was experimentally determined to acidic solution perhaps disrupts the hydrogen bonding between the
occur at 96 min (Fig. 2A), and the associated increases in solution HEC polymer chains and is replaced by bonding between HEC and
viscosity during gelation challenged the syringe pumps’ ability to CA. This would result in a less strong polymer network. These lignin-
successfully and consistently deliver the fluids, limiting the printing and CA-differentiated solutions served to provide a range of mechanical
time window. We studied solution parameters such as HEC molecular properties from which objects with property gradients could be printed,
weight (fig. S1A), HEC content by weight (fig. S1B), and choice of either by sequential deposition of each printing material or in situ
acidifying agent (fig. S1C) to minimize the rate of increase in solution mixing (see “Design-to-fabrication workflow” section). Another
viscosity. We observe the following general trends: Increases in HEC interesting kind of property variation that was noted was the decrease
molecular weight (near pH ~2) and decreases in HEC weight content in stiffness (fig. S2A) and increase in size (fig. S2B) and weight
led to decreases in the gelation rate, and for acidic solutions of the (fig. S2C) of printed samples, with increasing relative humidity due
same starting pH (2 to 3.5 region), citric acid (CA) is more effective to the hygroscopic nature of cellulose, which could be explored for
than hydrochloric acid (HCl) in decreasing the rate. Adding CA to the other applications involving shape-changing structures (32) or
solution to reach a pH of 3.0 proved to slow the gelation rate the most, temporary, water-soluble objects (see movie S1). Varying the viscosity

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 2 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

A A 80
104
HEC HEC

Shear moduli (Pa)

(MPa)
solution hydrogel HEC 10%
Gelation point HEC 10% + MCC 5%
HEC 10% + MFC 3%
40 HEC 10% + CA 4%
102

Stress,
HEC 10% + LIG 5%
Physical
bonding G′, Storage modulus
G″, Loss modulus
0
1 0 0.04 0.08
0 5 10 15 Strain,
Time (103 s)
B 2500
B
300

E modulus E2% (MPa)

Peak of fabrication
Gelation point (min)

time window
200 1500

100

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

500

0
1.0 2.0 3.0 4.0 0
pH HEC 10% HEC 10% HEC 10% HEC 10% HEC 10%
+ + + +
Fig. 2. Rheological properties of the printing solution. (A) A schematic on the MCC 5% MFC 3% CA 4% LIG 5%
left shows the gelation of the printing solution as a result of physical bonding. The
plot of shear moduli versus time on the right shows the gelation point occurring at C
~5800 s. By convention, the gelation point is defined as the point in time at which 90
Tensile strength (MPa)

the storage modulus G′ becomes greater than loss modulus G″ after the initial dis-
solution of HEC species. (B) A plot showing the dependence of gelation time as a 60
function of the pH of the printing solutions. The pH was adjusted by the addition
of varying amounts of CA. Error bars denote the SD of three trials. The plot reveals
30
that the highest gelation point, and therefore the longest printing time window,
occurs at a pH of ~3.0.
0
HEC 10% HEC 10% HEC 10% HEC 10% HEC 10%
+ + + +
of the printing solution was equally important to achieving stiffness MCC 5% MFC 3% CA 4% LIG 5%

gradients, as the viscosity determined the cross section of the deposited D

filaments, providing stiffness modulation through geometric variation. 6
Theoretical
Figure 3D shows the attained viscosity range, as well as the substantial
Log (viscosity, ) (Pa·s)

fabrication
shear thinning properties that are activated by the choice of HEC at 4 shear rate
the calculated fabrication shear rate of 28.7 s−1. Solutions of high
[10 weight % (wt %)] bulk HEC content with added microfibrillated 2
cellulose (MFC; a high-viscosity additive) served as the most viscous
Most viscous
mixture with a viscosity of 14.0 Pa·s that resulted in maximal filament 0 Least viscous
cross section, while minimizing the weight ratio of bulk HEC to
2 wt % without any additives produced more fluidic filaments with −2
a viscosity of 1.9 Pa·s. The exact methods of achieving stiffness −3 −1 1 3
gradients through these printing solutions modified in terms of E Log (shear rate, ) (1/s)
modulus and/or viscosity are expounded in the “Stiffness gradient Fig. 3. Modulation of the printing material’s mechanical and rheological
patterning and applications” section. properties. (A) A stress-strain plot showing the mechanical behavior of the base
HEC material (10 wt %) as well as with various amounts of different constituent
Design-to-fabrication workflow additives (denoted after the additive abbreviation as in table S1). The weight
The design-to-fabrication workflow (Fig. 4A and fig. S3) allows users concentration of each additive was selected by maximizing the additive concentration
to combine geometric models with gradient data to create FGM data while still ensuring printability of the solution and testability of the samples. (B) A plot
comparing the E modulus of the base material and the various modified solutions.
and to generate fabrication code. Grasshopper, a visual programing
Error bars show SDs of three samples. These test results show that CA and lignin
interface inside the 3D modeling software Rhinoceros 3D (fig. S4),
had the most impact on stiffness and could be used in the production of stiffness
serves as the platform for this workflow. Implementing the workflow gradients. (C) A plot comparing the tensile modulus of the base material and the
within a heuristic software environment creates a seamless data flow modified solutions. Error bars show SDs of three samples. (D) A plot demonstrating
between design and fabrication, allowing fabrication parameters to the shear thinning properties of the most and the least viscous solutions that can
be advantageously exploited during the design process to create be printed. The calculated shear rate of 28.7 s−1 that the printing solutions experience
diverse gradient properties. In particular, the fabrication parameters during extrusion is marked by a dotted line.

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 3 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

A Design-to-fabrication workflow B Gradient by superposing layers C Constant deposition rate

Geometric model + Gradient data Design data Fabrication code
N1 G1 X0 Y0 F50
N2 G1 X0 Y4 F75
.
.
.

Varying deposition rate

or or

Min Max M1 M2

Superposing layers Deposition rate Mixing ratio

D Representing material E Material mixing F Targeted Simulated distribution G Samples immediately Samples after
mixing ratio in grayscale without and with turbulator Nonoptimized method material distribution due to desynchronization after printing drying

Mixing ratio

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

100% M1 100% M2
Gradient-optimized method

Representative point:
Coordinates: 190, 15
Ratio: 68% M1, 32% M2

Fig. 4. Design-to-fabrication workflow of graded samples. (A) Schematics of the design-to-fabrication workflow. In addition to the geometry, designers have control
over material distribution and material composition, enabling the generation of gradient design by three different means: combining multiple layers of varied materials
(bottom left), tuning deposition rate (ml/mm) (bottom center), and tuning material ratio (bottom right). (B) Pictures of the fabrication of a multilayer material gradient. A
printing solution with low viscosity (top) is deposited first and creates a continuous material layer. A second printing solution with high viscosity (bottom) is extruded on
top of the first layer, creating stiffening ribs. The diffusion process between the different layers creates seamless gradients after sample drying. (C) Photos of samples
printed with different material deposition rates. Extrusion rate (mL/s) and printer speed (mm/s) parameters are embedded in the design data and tuned to control the
material deposition rate (mL/mm). Samples with the same deposition path vary in geometry due to a constant (top) or varying (bottom) deposition rate. (D) Representing
material gradients with a grayscale image. A grayscale image is evaluated at different locations according to its brightness and is discretized into a series of points
containing information on fabrication parameters such as position and material ratio. (E) Pictures of an extruding tip without (left) and with (right) a turbulator. The
turbulator enhances the mixing of the two incoming solutions and increases the homogeneity of deposited filaments. (F) Diagrams showing expected material distribution
in a synchronized (left column) and desynchronized (right column) fabrication system through a nonoptimized approach (top row) and gradient-optimized approach
(bottom row) to deposition path generation. The gradient-optimized method balances path continuity with material variation rate and makes the printed gradient less
sensitive to the desynchronizations between positioning and extrusion systems. The tradeoff is increased path discontinuities, shown in red. (G) Printed samples with
continuous material gradients right after deposition (left column) and final gradient result (right column). The samples with nonoptimized deposition paths (top row)
show less material contrast than the samples with gradient-optimized deposition paths (bottom row) (photo credit: Pedro A. G. S. Giachini, University of Stuttgart).

of (i) deposition sequence, (ii) deposition rate, and (iii) deposition The second strategy, the deposition rate strategy (Fig. 4C), generated
ratio are used to correspondingly enable the creation of graded objects stiffness gradients by varying the rate of the deposition, which resulted
through (i) superposing layers, (ii) varying material amount, and (iii) in varying amounts of materials along the deposition paths. The
varying material composition. deposition rate (ml/mm) depends on the extrusion rate of the syringe
First, deposition sequence is explored to develop a superposing pumps (ml/s) and the printer’s nozzle speed (mm/s). These fabrication
layer strategy (Fig. 4B) for stiffness gradient fabrication. In this process, parameters can be embedded in the geometric data through design
geometric data are associated with layers representing sequential layers or functions that determine deposition rate values and behaviors.
deposition procedures with contrasting mechanical properties. The These data are then translated into fabrication commands that
fluidic characteristic of materials with lower viscosity provides object coordinate the distribution of material, exploring material flow as
continuity, while more viscous mixtures are used to discretely alter a mean to efficiently permeate infill regions and to enable equal
stiffness. Diffusion between contrasting materials guaranteed interlayer deposition paths to manufacture objects with varied geometric stiffness
continuity and enabled the production of continuous and pliable sheets (see movie S2).
of material with patterned reinforcements (see “Stiffness gradient The third strategy generated gradients by varying material mixing
patterning” section). ratios (Fig. 4, D to G). An additional data layer is used to store the

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 4 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

mixing ratio of the materials being deposited. In this data layer, local material distribution, local cross-sectional height, and local Young’s
composition is defined either by material distribution functions (25) modulus, influencing how an object deforms (see movie S5). As an
or a graphical method that uses grayscale to represent mixing ratios example, Fig. 5D exhibits cellulose sheets of similar appearance but
(Fig. 4D). These design data of mixing ratios are translated into with varying stiffness gradients being submitted to external forces
fabrication codes that modified the extrusion rate of the corresponding (stiffness gradient measured in fig. S5). As shown, by distributing
syringe pumps. stiffness along specific directions or patterns, it is possible to achieve
In this scenario, where materials are dynamically mixed during distinct deformation behaviors.
the printing process, a turbulator is used at the merging point of the Figure 6 shows how stiffness gradients can be used to guide an
two delivery tubes to disrupt the laminar flow that results in the object’s deformation. By exploring the Young’s modulus range of
extrusion of heterogeneous mixtures (Fig. 4E and movie S3). This the different cellulose mixtures (Fig. 6A), it is possible to print visually
approach, although enhancing fluidic mixing, results in an increase similar samples with the same cross-sectional thickness that exhibit
in materials’ travel time in the tube before reaching the deposition distinct deformation behavior and end geometry due to their unique
nozzle and brings both digital and physical challenges to the process. programmed stiffness patterns (Fig. 6, B to E). This approach of
On the digital side, it creates a temporal discrepancy between the using an external force to generate the final shape of an initially flat
positioning and the extrusion systems because the increase in materials’ object allows designers to leverage simplified 2D manufacturing
travel time needs to be accounted for in the fabrication code in order strategies, avoiding more complex 3D processes. Applications range
for the positioning system to synchronize with the extrusion system from industrial design products (33) to architectural design systems
(see Material and Methods). On the physical side, diffusion between that explore elastic bending of planar objects to achieve form and

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

sequential mixtures occurs in the turbulator and reduces material structural integrity (34, 35).
contrast, an effect proportional to the dissimilarity between sequential In addition to guiding deformation processes, stiffness patterns
materials’ mixing ratio. In addition, changes in the extrusion rate of also have applications in the development of compliant mechanisms
a syringe pump require a nonzero time step to coincide with the and soft robotics (36, 37). In this scenario, stiffness gradients are
extrusion rate of the solution at the other end of the delivery tube. designed to create objects that may exhibit complex deformation
For example, in a fabrication scenario where the extrusion speed was behavior and redistribution of forces in response to actuation forces.
abruptly changed from 0 to 0.2 ml/s, delays as long as 8.1 s between Exemplifying this application, Fig. 6 (F and G) shows the deformation
the positioning and extrusion systems were recorded; this physical of cellulose samples exhibiting similar stiffness gradients that were
delay has a direct impact on the material composition accuracy. patterned using different fabrication strategies. The sequence shown
To address these challenges, we developed a computational in Fig. 6F exhibits a sample with geometric stiffness differentiation
strategy to optimize the deposition path (Fig. 4F). This strategy is through directional placement of reinforcing ribs of higher cross section,
based on the intuition that minimizing the dissimilarity between while the sequence shown in Fig. 6G exhibits a sample with E modulus
sequential material mixing ratios during deposition would reduce differentiation achieved by using a CA concentration gradient that
errors introduced by the delay due to the materials' travel time and lowers stiffness in the regions that were meant to fold. Because of
by the diffusion between adjacent material mixtures in the extrusion lower stiffness regions that resemble common curve-fold hinges, the
tube (Fig. 4F). It analyzed the position and the material data of all samples present a curved deformation pattern perpendicular to the
deposition points and created a gradient-optimized deposition path direction of the applied force. A finite element method (FEM)
that balances between path continuity and mixing ratio continuity simulation was performed to verify the observed deformation behavior
(see movie S4). Figure 4G compares the same material gradient of the sample exhibited in Fig. 6G (Fig. 6H). The sample’s fabrication
printed through the nonoptimized and gradient-optimized deposition stiffness gradient input was used to create a mesh with a fine Young’s
path methods. As expected, the sample fabricated using the gradient-­ modulus gradient that approximates the continuous stiffness gradient
optimized path method exhibits higher material contrast immediately of the physical object. As observed, the simulated mesh showed similar
after deposition and after the postdeposition diffusion that occurs deformation behavior under the application of forces that mirrored
during the drying process. the handling of the physical prototype, providing feedback on the
distribution of stresses in the deformed sample (Fig. 6I).
Stiffness gradient patterning and applications
As described in the previous sections, combining materials engineering
and manufacturing developments enables the fabrication of systems DISCUSSION
that present custom stiffness gradation by controlling both the geometric In summary, this work leveraged a combined approach of materials
and material aspects of the printed object. Through the developed engineering and digital processing to control material mixing and
strategies, gradients can be tuned at the local and global scales (Fig. 5). deposition and to enable extrusion-based MMAM of tunable viscoelastic
At the local scale, geometric stiffness can be tuned by material materials with continuous, high-contrast, and multidirectional stiffness
parameters such as viscosity and solid weight content. Figure 5A gradients. A method was established to engineer a base solution into
shows that solutions with a higher solid weight content dry into a catalog of fluidic, cellulose-based materials with distinct mechanical
comparative higher cross sections, since a larger amount of material and rheological properties. This provided the physical foundation
is left behind after solvent evaporation. Similarly, as shown in Fig. 5B, for achieving stiffness gradients. In parallel, a digital workflow was
solutions with higher viscosities result in higher cross sections after developed that allowed the embedding of gradient information
the drying process, as higher viscosities reduce the deposited material’s into design models, with integrated deposition path planning that
spread. In addition to the geometric stiffness approaches, local stiffness considered mechanical property profiles to better distribute materials.
can also be tuned according to the material mixture’s Young’s modulus The deposited filaments, being in solution state, were then able
(Fig. 5C). These parameters are combined at the global scale to control to seamlessly blend and diffuse, creating a single, coherent graded

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 5 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Local stiffness variation

Cross section
A Solid weight content (sw) B Viscosity
Wet Dry 4.1 mm
16.5 mm
H 0.05*H 0.22*H 0.9 mm 3.7 mm

2% sw 12% sw 1.9 Pa·s 14.0 Pa·s

2% solid content 12% solid content 5 mm
Wet

5 mm
Cross section width Cross section height
Dry

18 4
3.5

Dimension (mm)
15
3
12
Cross section width Cross section height 2.5
12 2 9 2
6 1.5
10
Dimension (mm)

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

1.5 1
8 3
0.5
6 1 0 0
1.9 Pa·s 14.0 Pa·s 1.9 Pa·s 14.0 Pa·s
4
0.5 Wet Dry
2
Material
0 0
2% sw 12% sw 2% sw 12% sw C E modulus
Wet Dry

250 ± 15 MPa 1800 ± 400 MPa

D Global stiffness patterns

Representative pattern 1 Representative pattern 2

Mixing ratio Mixing ratio

M1 M2 M1 M2

Mix 1: 6% sw; E = 250 ± 15 MPa Mix 2: 10% sw; E = 1430 ± 160 MPa

Fig. 5. Strategies for varying stiffness at the local scale and patterning stiffness at the global scale. (A) Diagrams and photos of the cross sections of the extruded
filaments showing the effect of solid-weight content after solvent evaporation. Printing solutions with lower cellulose concentrations produce filaments with thinner
cross sections (left) than filaments printed with the same volume of a high-concentration printing solution (right). (B) Diagrams and photos of the cross sections of the
extruded filaments from mixtures of varied viscosity. Printing solutions with lower viscosity produce layers of thinner cross sections (left), whereas printing solutions with
higher viscosity results in filaments with thicker cross sections (right). Viscosity values corresponds to the shear rate of 25.1 rad/s. (C) Range of material’s E modulus
achieved by altering the cellulose solutions with additives. (D) Patterning local material compositions and cross-sectional characteristics is used to achieve samples with
anisotropic stiffness at the global scale. Directionality was patterned by aligning less stiff regions of material, resulting in samples that show bending resistance in one
direction and programmable hinge-like behavior in another direction (left). Similarly, this effect was reproduced in multiple directions by patterning more rigid patches
of material varying in distribution and shape (right) (photo credit: Pedro A. G. S. Giachini, University of Stuttgart).

object upon solidification and drying. The payoff of integrating these materials or materials that are not chemically compatible or immiscible
physical and digital tools is the ability to achieve the same stiffness in a single solution state would not be suitable for this method. Second,
gradient in an object in multiple ways, opening design possibilities our setup currently only allowed for a maximum of two materials to
that were previously limited by the rigid coupling of material and be simultaneously mixed and extruded; a more elaborate fluid delivery
geometry. In addition, the flexibility of the developed setup and system with a higher number of extrusion channels would be required
software provides a scalable and adaptable process that can be to add an increased number of material combinations. Third, the
applied to other gradient fabrication scenarios. long drying time and low viscosity of the initially as-deposited filament
The developed method has several limitations. First, this method so far restricted the extension of this method to multilayer printing
generally relies on physical bonding of chemically compatible and of above two layers. Last, the maximum force of the syringe pump
miscible materials to achieve a solid object, meaning that dissimilar used in this work posed a limitation on the maximum viscosity of

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 6 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

A B C
99% mix 1
1% mix 2

Mixture 1
E = 1430 ± 160 MPa

D E
Mixture 2
E = 250 ± 15 MPa

1% mix 1
99% mix 2

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

G

H I
1433 MPa

248 MPa

Fig. 6. Samples showing programmable deformation due to patterned stiffness variation. (A) A schematic showing stiffness gradients represented by a grayscale
image. Mixture 1 has 10 wt % base mixture, and mixture 2 has 10 wt % base mixture and 4 wt % CA. (B to E) Thin cellulose strips of identical size were printed with various
stiffness gradient profiles along their length and exhibit different curvature profiles when subjected to the same external displacement. (B) Photograph showing that
homogenous stiffness (no gradient) resulted in a symmetric curvature profile. (C) Photograph showing that discrete regions of the weaker mixture resulted in hinge-like
behavior that distorted this strip’s curvature profile from the nongraded symmetric curve. (D) Photograph showing a gradual sinusoidal gradient resulted in a curvature
profile approximating a circle. (E) Photograph showing that the desired closed end form of this strip was achieved through the printed stiffness gradient, where the
weaker regions bent more readily than the stiffer regions. (F and G) Series of photographs of flexible graded sheets that transferred load in the transversal direction to
large body deformation in the longitudinal direction for a programmed folding effect. These identical programmed behaviors were achieved by different methods as
described in the “Stiffness gradient patterning and applications” section. (F) Sample exhibiting geometric stiffness differentiation through directional placement of
reinforcing ribs of higher cross section. (G) Sample exhibiting E modulus differentiation achieved by using a mixture of higher CA content in the regions that were meant
to fold. (H) A digital mesh model of the sheet featured in part (G). A fine gradation of E modulus values was applied to approximate the sheet’s continuous gradient.
(I) Series of snapshots of the FEM simulation that was developed to predict the programmed folding behavior under a pair of applied forces, represented by the blue block
arrows (photo credit: Sachin S. Gupta, University of Stuttgart).

the solutions that could be mixed and extruded, which, in turn, materials, the range of material stiffness can be further extended,
affected the range of the property gradients. For example, the Young’s and alternative viscous materials could be used, such as other polymer
modulus range of 248 to 1786 MPa was primarily determined by how gels, resins, or even materials used in large-scale industries, assuming
the responsible additives affected the printability of the polymer an appropriate hardware setup is introduced. Third, the main concept
solution, with the lignin substantially increasing the viscosity of the of this work could be extended to current large-scale printing
solution and the CA considerably increasing the gelation rate at technologies (38, 39) to create buildings in which geometry and
high concentrations. material continuously change to balance between structural and
In the light of these limitations, this work offers the following architectural requirements or even to create adaptable mechanisms
opportunities for further research. First, for digital processing workflow, (22) to introduce soft hinges or directional bending. Fourth, at
optimization of the path planning could be further developed by the other end of the size spectrum, it would be interesting to test
fine-tuning penalties related to path discontinuities and large differences the effectiveness of the established method at small scales, where the
in material variation rates, potentially increasing the production natural diffusion of the material may overtake the distribution specified
efficiency of this printing process. Second, for the development of by the fabrication paths, which would be crucial for applications sensitive

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 7 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

to small changes in material properties, such as tissue engineering. parameters were conducted to determine which parameters minimized
Fifth, this work can be extended to gradients other than stiffness the gelation rate. Where gelation point was not directly measured
gradients, such as optical properties, thermal conductivity, electrical using a rheometer, the quantity “thickening time” used to indicate
conductivity, or even cell nutrient concentration. Last, the implemen- the gelation rate was defined as the time it took for the stir bar to be
tation of fast-curing liquids, molds, or viscous support materials unable to spin continuously due to the solution’s increase in viscosity
(40) could even push the current work’s capabilities of printing 2D during gelation. HEC of molecular weights 90,000, 250,000, 720,000,
or 2.5D objects to create fully fledged 3D objects with these internal and 1,300,000 g/mol were all tested by preparing base 10 wt % solutions
functional property gradients. of various pH and allowing them to mix using a stir bar at ambient
temperature. Figure S1A shows the general trend that the highest
molecular weight (1,300,000 g/mol) led to the slowest thickening
MATERIALS AND METHODS time and was therefore the molecular weight of choice for all solution
Printing material preparation and testing preparations. It was hypothesized that longer average chain lengths
HEC (molecular weights of 90,000, 250,000, 720,000, and 1,300,000 g/mol), of HEC meant more potential binding sites at which the polymer
lignin, microcrystalline cellulose, CA, HCl, and sodium hydroxide chains could cross-link before the entire chain movement was restricted
were purchased from Sigma-Aldrich. MFC was graciously donated and a solid gel was formed. The same procedure was conducted
by Daicel FineChem and Weidmann Fiber Technology. to test the effect of HEC weight content on gelation rate (fig. S1B),
The multistep procedure for preparing the printing materials is with concentrations of 2, 5, and 10 wt % at a range of pH values.
as follows. First, a small quantity (~0.5 to 2.5 wt %) of HEC and/or It shows that the lower the weight concentration, the slower the

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

MFC was added to distilled water and mixed thoroughly using an thickening time. Decrease in weight content reduces the number of
overhead laboratory mixer to attain the approximate desired viscosity HEC polymer chains in solution, which slows the ability of a cross-
of the printing solution. Next, a predetermined quantity of CA linked polymer network to form throughout the volume. Last, the
(~0.1 wt %) was added to and dissolved in the solution to make the test of choice of acidifying agent between HCl and CA (fig. S1C) in
pH of the solution around 3.0. The third step was the addition of creating 10 wt % solutions of various pH was conducted, with HCl
lignin (up to ~5.0 wt %) or a second addition of CA (up to ~4.0 wt %), proving less effective than CA in increasing the gelation point. CA
which was then thoroughly mixed. The final step was the addition of was more effective because it is a weak acid and acts as a buffer at
a quantity of HEC that makes up the bulk solid content of the printing around pH 3 (pKa1 = 3.1).
material (~10.0 wt %). After a last round of mixing, the finalized One important material property that was noted during this
printing material was then loaded into a syringe by pouring and was research was the ability of the cellulose samples to absorb water
ready for fabrication. due to their hygroscopic properties. It was apparent that printed
Flow shear rate sweep rheological measurements were conducted samples of cellulose transitioned from stiff to soft and rubber-like
in a shear rate range of 10−3 to 103 s−1 with 20-mm 2.013° cone upon exposure to humidity, absorbing the water and swelling. The
plate geometry. One microliter of samples of the polymer solution use of a rheometer equipped with a humidity and temperature
was scooped using a spoon and placed on the test stage with a control chamber enabled the monitoring of the stiffness property of
50-m gap. Oscillation time sweep rheological measurements were the material from its high moisture content condition to dry state
conducted at 4% strain and a frequency of 10 rad/s and 40-mm (fig. S3A). In this couple of explorative tests, samples in their humid
parallel plate geometry. One microliter of samples of the polymer state were tested while varying the environment’s relative humidity
solution was scooped using a spoon and placed on the test stage from 95 to 0%. This was possible due to an adaptive behavior of the
with a 50-m gap. equipment during the dynamic mechanical analysis that adapted
Tensile tests were performed to characterize the mechanical the gap between samples as the material dried and shrank. The
properties of the engineered materials. The following methods were range of stiffness values measured from the humid to the dry state
carried out to identify three main parameters: Young’s modulus of the material span two orders of magnitude, and, interestingly,
(E modulus), tensile strength, and elongation at break. All samples results show that the stiffness of the material does not seem to vary
were prepared by pouring 25 ml of material at room temperature linearly between the humid and dry states, but rather the stiffness
on dogbone-shaped molds that were laser cut out of Plexiglass. All curve exhibits two steep regions where the material experiences a
samples were tested under constant ambient temperature (around rapid change in stiffness between regions of slow variation. In addition,
20°C) and in their dry state after being stored in a dry chamber until the increases in size (fig. S3B) and weight (fig. S3C) were also
their weight approached the initial solid content weight of their measured. To standardize sample thickness and exposed surface,
mix. Initial tests with strain rates varying across two orders of samples were prepared with a total initial weight of 60 g and poured
magnitude (5 to 500 mm/min) were performed to identify an appropriate into large petri dishes. A closed container was used to control
test parameter, and a strain rate of 40 mm/min was chosen for environmental conditions. Temperature was kept constant at around
characterization since it provided similar stiffness results when 20°C, while relative humidity varied according to the intended drying
compared with the strain rate of 500 mm/min while allowing for or swelling cycle. Drying was stimulated by providing a constant
longer sample observation during the test. Sample width and thickness flow of pressurized dry air with lower than 3% relative humidity.
were measured across three different sections for each sample, and mean Humidity was provided by a potassium sulfate salt solution placed
values were used as test input. Following literature recommendations inside the container, which stabilizes ambient relative humidity at
for tensile testing of polymers, stiffness values were characterized as approximately 94%. After preparation, samples were placed in the
the Secant E modulus measured at 2% strain. container under dry conditions. Two orthogonal lines were marked
As stated in the “Creating printing material with tunable rheological along the diameter of each sample, indicating the position where
and mechanical properties” section, studies of various solution measurements were to be taken. Weight and dimensional variations

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 8 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

during the drying process were measured until values became printing nozzle depends on the required print duration and resolution,
constant. The swelling cycle was stimulated next by shutting off the as well as the ability of combining multiple materials or not. A personal
dry air supply to the chamber and introducing a container with a computer was used as the input and control system, using USB
potassium sulfate solution. Again, variations in both weight and (universal serial bus) communication ports for the coordination of
dimension were measured until values became constant. both the extruding and positioning systems.
An extension to the hygroscopic properties is the ability of the A custom fabrication code consisting of a commands list for
printed samples to swell to the point of returning to a weakened gel both printer and syringe pumps was structured in accordance to the
state when saturated with water. Currently, the printed samples do hardware’s communication protocols. Tasks of both systems were
not undergo any chemical cross-linking process to prevent this grouped in a single command line, which was parsed accordingly
regression to the gel state. However, dissolution of these swollen, to each protocol at the time of execution. To establish continuous
gel-like samples in water actually enables the creation of a new movements, G-code was sent to be buffered in the 3D printer at
polymer solution with identical solid constituents to the dissolved a faster rate than the fabrication speed. Because the syringe pumps
samples, albeit with less homogenous distribution of solids as in were not controlled by the 3D printer’s controller board, the Repetier
the original printing mixture, according to qualitative observations. open-source firmware installed in the 3D printer was modified to
Preliminary tests have demonstrated that this new polymer solution supply current action feedback when requested. The feedback
can indeed be used to print once again. Thus, in addition to being consisted of adding supplementary G-code functionalities that
biodegradable due to its cellulosic material composition (29), the triggered the printer to send a feedback signal indicating that it had
printing material can be recycled (see movie S1). finished processing the last positioning command sent, thus allowing

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

synchronization between printer and extrusion system.
Geometric input and gradient patterning The printing of multiple samples with varied patterned gradients
Conventional 3D printing procedures use the CAD software to and material properties allowed us to identify that parameters such
create the desired geometry, which is then exported to a computer-­ as material viscosity and gelation time also have an impact on print
aided manufacturing (CAM) software, where the geometry gets accuracy. Through observation during printing, we identified that
rationalized into fabrication code (G-code). Because of the experimental low-viscosity materials were more likely to flow outward of their
nature of the project, where new processes had to be developed, it deposited position, mixing with neighboring materials and decreasing
was chosen that the design and fabrication workflow would be print precision, an issue that can be minimized by assigning a
developed within Grasshopper, a visual programming plugin for deposition rate coherent with the spacing between deposited filaments.
Rhinoceros 3D modeling software that allows the development of In relation to the gelation time, image comparison of recently
custom tools through programming. concluded prints and images taken during the gelation process and
To enable the fabrication of material graded samples, and since of the final sample shows that the color contrast and sharpness
traditional CAD software does not support assigning heterogeneous between neighboring materials are lower during the drying process
material properties to geometries, alternative strategies were investigated due to diffusion between the materials with differing concentrations
to store material data in the virtual geometry for fabrication. The in their viscous state.
simplest strategy consists of (i) discretizing every geometry input
into a series of points and evaluating these points according to a Printing process
function, allowing the material ratio of each point to be assigned Printing of samples started by preparing the ink solution with appro-
according to, i.e., its position in space or within the curve domain. priate cellulose and additive concentrations and tuning its viscosity.
This method, despite being convenient for assigning complex material The different inks were then loaded in the extrusion system, with
pattern data, requires from the user the ability to translate design their delivery tubes connected to a custom extruding tip for printing
needs into functions and does not enable an instinctive design process. graded samples that contains a turbulator to enhance the mixing
A second strategy included (ii) visual feedback during the design of both materials. The cellulose solutions were extruded on top of
phase by adopting a graphic design software to create grayscale a polyvinyl chloride (PVC) plate fixed to the printer’s printing bed.
images with brightness gradients corresponding to the desired After the print was concluded, the PVC plate was removed from
material ratio. The brightness evaluation returns a number between the printer, and the print material was left to complete gelation and
0 and 1 that determines the concentration proportion of two different air dry.
materials for each point of a 2D surface. To account for a higher Extrusion nozzle diameters were chosen according to ink viscosity,
number of materials, other evaluating methods based on image data, desired resolution, and print duration. Extrusion nozzles with inner
such as RGB (red, green, and blue) values, could be implemented. diameter ranging from 0.55-mm needles to 4-mm tubes were
tested, which could print solutions within 1.9 and 14.0 Pa·s viscosity.
Fabrication setup, communication, and synchronization Manufacturing with the printing material was possible due to its shear
An extrusion-based method was chosen due to its hardware flexibility, thinning behavior during extrusion. These theoretical viscosity values
which allows for easy adaptation of low-cost 3D printers and vast were calculated using conditions similar to the ones experienced
material possibility, since the method can use liquid- or paste-like during fabrication, with a flow rate of 0.18 ml/s in a 3-mm inner
materials at different scales. The positioning system consists of an diameter tube. A series of initial experiments varying material
adapted fused deposition modeling 3D printer. The usual extrusion viscosity, extrusion rate (ml/s), and printing speed (mm/s) were
nozzle was replaced by a 3D-printed piece capable of holding studied to determine the possible width range of printed filaments.
multiple delivery tubes. Two syringe pumps that control the printing The procedure consisted of extruding ink solutions with varying
material delivery were used as the extrusion system. Varied printing viscosities at a series of deposition rates (ml/mm) and recording the
nozzles were used for material deposition. The choice for a specific resulting filament width.

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 9 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Stiffness gradient measurement 8. M. Rüggeberg, T. Speck, O. Paris, C. Lapierre, B. Pollet, G. Koch, I. Burgert, Stiffness
gradients in vascular bundles of the palm Washingtonia robusta. Proc. Biol. Sci. 275,
The measurements to demonstrate gradient properties of a sample
2221–2229 (2008).
sheet were performed on a customized setup (fig. S2). The setup was 9. D. Fabris, J. C. M. Souza, F. S. Silva, M. Fredel, J. Mesquita-Guimarães, Y. Zhang,
built on an inverted optical microscope (Axio Observer A1, Zeiss). B. Henriques, The bending stress distribution in bilayered and graded zirconia-based
The indentation force was measured by a sensitive load cell (GSO-25 dental ceramics. Ceram. Int. 42, 11025–11031 (2016).
and -1 K, Transducer Techniques) mounted on a computer-controlled 10. T. A. Enab, Stress concentration analysis in functionally graded plates with elliptic holes
under biaxial loadings. Ain Shams Eng. J. 5, 839–850 (2014).
high-precision piezo motion stage (LPS-65 2″, Physik Instrumente 11. K. Kumar, J. Liu, C. Christianson, M. Ali, M. T. Tolley, J. Aizenberg, D. E. Ingber, J. C. Weaver,
GmbH & Co. KG) in the z direction, with a resolution of 5 nm and K. Bertoldi, A biologically inspired, functionally graded end effector for soft robotics
a maximum velocity of 10 mm·s−1. Fine positioning in the x and y applications. Soft Robot. 4, 317–323 (2017).
directions was done by a manual xy stage (NFP-2462CC, Position- 12. M. McEvoy, N. Correll, in Experimental Robotics, M. Ani Hsieh, O. Khatib, V. Kumar, Eds.
(Springer Tracts in Advanced Robotics, 2016), pp. 893–907.
ierungstechnik Dr. Meierling), and tilt correction was adjusted by
13. D. Rus, M. T. Tolley, Design, fabrication and control of soft robots. Nature 521, 467–475
two goniometers (M-GON65-U, Newport). The motion control of (2015).
the piezo stages and the data acquisition were performed by a 14. W. Hu, G. Z. Lum, M. Mastrangeli, M. Sitti, Small-scale soft-bodied robot with multimodal
customized Linux code (Ubuntu, Canonical Ltd.). The program locomotion. Nature 554, 81–85 (2018).
allowed control over preload, velocity, displacement in the x and z 15. L. Hines, K. Petersen, G. Z. Lum, M. Sitti, Soft actuators for small-scale robotics.
Adv. Mater. 29, 1603483 (2017).
directions, and contact time. The load cell was linked to the computer 16. S. Sant, M. J. Hancock, J. P. Donnelly, D. Iyer, A. Khademhosseini, Biomimetic gradient
via a signal conditioner (BNC-2110, National Instruments), and the hydrogels for tissue engineering. Can. J. Chem. Eng. 88, 899–911 (2010).
voltage signal from the force measurement was transferred through 17. M. Vaezi, S. Chianrabutra, B. Mellor, S. Yang, Multiple material additive manufacturing – Part 1:

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

a data acquisition board (PCIe-6259, National Instruments). The A review. Virtual Phys. Prototyp. 8, 19–50 (2013).
18. F. Pati, D. H. Ha, J. Jang, H. H. Han, J.-W. Rhie, D.-W. Cho, Biomimetic 3D tissue printing
samples were intentionally curved in an arch form, and the measurements
for soft tissue regeneration. Biomaterials 62, 164–175 (2015).
were done along the top of the arch to avoid any substrate effect 19. A. Huang, in ACADIA 2016: POSTHUMAN FRONTIERS: Data, Designers, and Cognitive
(fig. S2C). For each data point, indentation stopped when the force Machines (2016), pp. 318–325.
of 0.02 N was reached. Young’s moduli were estimated according 20. A. Menges, Integrative design computation: Integrating material behaviour and robotic
the Hertz theory manufacturing processes in computation design for performative wood construction, in
Proceedings of the 31st Annual Conference of the Association for Computer Aided Design in

​ 3 ​ ​─
F = ​─ E ​ ​R​​  _2​1​​ ​d​​  ​3_2​​ ​ Architecture (ACADIA), Banff, Canada (2011), pp. 72–81.
4 1 − ​​​  2​ 21. E. Greenberg, A. Körner, in Sustainable Design & Manufacturing, R. Setchi, R. J. Howlett,
M. Naim, H. Seinz, Eds. (Cardiff, 2014), pp. 50–66.
where F is the force, E is the Young’s modulus, R is the radius of the 22. A. Körner, L. Born, A. Mader, R. Sachse, S. Saffarian, A. S. Westermeier, S. Poppinga,
M. Bischoff, G. T. Gresser, M. Milwich, T. Speck, J. Knippers, Flectofold — A biomimetic
force probe, and d is the indentation distance.
compliant shading device for complex free form facades. Smart Mater. Struct. 27,
017001 (2017).
SUPPLEMENTARY MATERIALS 23. J. O. Hardin, T. J. Ober, A. D. Valentine, J. A. Lewis, Microfluidic printheads
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/ for multimaterial 3D printing of viscoelastic inks. Adv. Mater. 27, 3279–3284
content/full/6/8/eaay0929/DC1 (2015).
Fig. S1. Effect of different mixture parameters on rheological properties. 24. L. Mogas-Soldevila, J. Duro-Royo, N. Oxman, Water-based robotic fabrication: Large-scale
Fig. S2. Hygroscopic properties on HEC samples. additive manufacturing of functionally graded hydrogel composites via multichamber
Fig. S3. Digital fabrication workflow. extrusion. 3D Print. Addit. Manuf. 1, 141–151 (2014).
Fig. S4. Visual programming interface. 25. B. Zhang, P. Jaiswal, R. Rai, S. Nelaturi, Additive manufacturing of functionally graded
Fig. S5. Measurements of gradient mechanical properties. material objects: A review. J. Comput. Inf. Sci. Eng. 18, 041002 (2018).
Table S1. Printing materials and additives. 26. P. Michalatos, A. O. Payne, Working with multi-scale material distribution, in Acadia 2013:
Movie S1. Dissolution in water. Adaptive Architecture (2013), pp. 43–50.
Movie S2. Tunable deposition rate. 27. J. Duro-Royo, L. Mogas-Soldevila, N. Oxman, Flow-based fabrication: An integrated
Movie S3. Material mixing. computational workflow for design and digital additive manufacturing
Movie S4. Gradient-optimized print. of multifunctional heterogeneously structured objects. CAD Comput. Aided Des. 69,
Movie S5. Programmable folding prototypes. 143–154 (2015).
28. D. Saha, S. Bhattacharya, Hydrocolloids as thickening and gelling agents in food: A critical
review. J. Food Sci. Technol. 47, 587–597 (2010).
REFERENCES AND NOTES 29. J. M. G. Vasquez, T. P. Tumolva, Synthesis and characterization of a self-assembling
1. Y. Miyamoto, W. A. Kaysser, B. H. Rabin, A. Kawasaki, R. G. Ford, Functionally Graded hydrogel from water-soluble cellulose derivatives and sodium hydroxide/thiourea
Materials: Design, Processing and Application (Springer, 1999). solution. Am. J. Chem. 5, 60–65 (2015).
2. J. Aizenberg, J. C. Weaver, M. S. Thanawala, V. C. Sundar, D. E. Morse, P. Fratzl, Skeleton 30. T. T. Teeri, H. Brumer, G. Daniel, P. Gatenholm, Biomimetic engineering of cellulose-based
of Euplectella sp.: Structural hierarchy from the nanoscale to the macroscale. Science 309, materials. Trends Biotechnol. 25, 299–306 (2007).
275–278 (2005). 31. K. Markstedt, J. Sundberg, P. Gatenholm, 3D bioprinting of cellulose structures
3. A. Miserez, T. Schneberk, C. Sun, F. W. Zok, J. H. Waite, The transition from stiff from an ionic liquid. 3D Print. Addit. Manuf. 1, 115–121 (2014).
to compliant materials in squid beaks. Science 319, 1816–1819 (2008). 32. D. Correa, A. Papadopoulou, C. Guberan, N. Jhaveri, S. Reichert, A. Menges, S. Tibbits,
4. K. Kokini, J. DeJonge, S. Rangaraj, B. Beardsley, Thermal shock of functionally graded 3D-printed wood: Programming hygroscopic material transformations.
thermal barrier coatings with similar thermal resistance. Surf. Coat. Technol. 154, 223–231 3D Print. Addit. Manuf. 2, 106–116 (2015).
(2002). 33. “Reebok introduces new liquid factory,” Reebok, 20 October 2016; https://news.reebok.
5. F. Watari, A. Yokoyama, M. Omori, T. Hirai, H. Kondo, M. Uo, T. Kawasaki, Biocompatibility com/global/latest-news/
of materials and development to functionally graded implant for bio-medical application. reebok-introduces-new-liquid-factory/s/8a87d7f7-8a93-49d2-9ddd-efee2d588b76.
Compos. Sci. Technol. 64, 893–908 (2004). 34. J. Lienhard, H. Alpermann, C. Gengnagel, J. Knippers, Active bending, a review
6. D. Lin, Q. Li, W. Li, M. V. Swain, Functionally graded dental implant and its effect on bone on structures where bending is used as a self-formation process. Int. J. Space Struct. 28,
remodeling. Multi Funct. Mater. Struct. 47–50, 1035–1038 (2008). 187–196 (2013).
7. M. Herrmann, W. Sobek, Functionally graded concrete: Numerical design methods 35. E. Panagoulia, S. Schleicher, Bending-active Structures—A case study for an office chaise
and experimental tests of mass-optimized structural components. Struct. Concr. 18, lounge, in Proceedings of the 34th eCAADe Conference, A. Herneoja, T. Österlund,
54–66 (2017). P. Markkane, Eds. (University of Oulu, 2016), pp. 621–630.

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 10 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

36. A. Firouzeh, J. Paik, Robogami: A fully integrated low-profile robotic origami. Strategy – EXC 2120/1 – 390831618. Author contributions: P.A.G.S.G., S.S.G., W.W., D.W., and
J. Mech. Robot. 7, 021009 (2015). E.B. performed the conceptual development, experiment design, and data analysis. P.A.G.S.G.
37. Z. Zhakypov, J. Paik, Design methodology for constructing multimaterial origami robots and S.S.G. performed the experiments and collected the data. M.Y. and W.W. performed
and machines. IEEE Trans. Robot. 34, 151–165 (2018). gradient stiffness measurement and analysis. M.S. and A.M. supervised the research. All
38. D. Cam, M. Desylias, Ai Build; https://ai-build.com, accessed in 2019. authors contributed to the manuscript writing. Competing interests: The authors
39. P. Boyd, C. W. Weller, Branch Technology; www.branch.technology, accessed in 2019. declare that they have no competing interests. Data and materials availability: All data
40. K. Hajash, B. Sparrman, C. Guberan, J. Laucks, S. Tibbits, Large-scale rapid liquid printing. needed to evaluate the conclusions in the paper are present in the paper and/or the
3D Print. Addit. Manuf. 4, 123–132 (2017). Supplementary Materials. Additional data, including raw videos, can be requested from
D.W. and A.M.

Acknowledgments: We thank the following for their support and feedback during the course
of the research: the Integrative Technologies & Architectural Design Research (ITECH) program Submitted 20 May 2019
at the University of Stuttgart, researchers at the Institute for Computational Design & Accepted 3 December 2019
Construction (ICD) and Institute for Building Structures and Structural Design (ITKE), Published 21 February 2020
J. Knippers (ITKE), researchers and staff at the Physical Intelligence department at the 10.1126/sciadv.aay0929
Max-Planck Institute for Intelligent Systems (MPI), and researchers at the Materials Physics
group in the University of Stuttgart. Funding: The research is funded by the Max-Plank Society Citation: P. A. G. S. Giachini, S. S. Gupta, W. Wang, D. Wood, M. Yunusa, E. Baharlou, M. Sitti,
and the University of Stuttgart. The research has been partially supported by Deutsche A. Menges, Additive manufacturing of cellulose-based materials with continuous, multidirectional
Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence stiffness gradients. Sci. Adv. 6, eaay0929 (2020).

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

Giachini et al., Sci. Adv. 2020; 6 : eaay0929 21 February 2020 11 of 11

Additive manufacturing of cellulose-based materials with continuous, multidirectional
stiffness gradients
P. A. G. S. Giachini, S. S. Gupta, W. Wang, D. Wood, M. Yunusa, E. Baharlou, M. Sitti and A. Menges

Sci Adv 6 (8), eaay0929.

DOI: 10.1126/sciadv.aay0929

Downloaded from http://advances.sciencemag.org/ on February 26, 2020

ARTICLE TOOLS http://advances.sciencemag.org/content/6/8/eaay0929

SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2020/02/14/6.8.eaay0929.DC1
MATERIALS

REFERENCES This article cites 30 articles, 2 of which you can access for free
http://advances.sciencemag.org/content/6/8/eaay0929#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science Advances (ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 New
York Avenue NW, Washington, DC 20005. The title Science Advances is a registered trademark of AAAS.
Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial
License 4.0 (CC BY-NC).