Content uploaded by Alexander Curth
Author content
All content in this area was uploaded by Alexander Curth on Oct 17, 2024
Content may be subject to copyright.
1
EarthWorks: Zero Waste 3D Printed
Earthen Formwork for Shape-Optimized,
Reinforced Concrete Construction
Alexander Curtha, Natalie Pearla, Emily Wissemanna, Tim Cousina, Latifa Alkhayata, Vincent Jackowa,
Keith Leea, Oliver Moldowb, Mohamed Ismailc, Caitlin Muellera, Lawrence Sassa
aMassachusetts Institute of Technology
bUniversity of Stuttgart
cUniversity of Virginia
77 Massachusetts Avenue
Cambridge, MA 02139
USA
Abstract
Rapid global urbanization is driving governments and builders to seek paradigm-shifting technologies to
speed the construction of housing and infrastructure at a low economic and carbon cost. Here, we present
a novel method for fabricating materially efficient, shape-optimized, code-compliant, reinforced concrete
structures cast in directly recyclable 3D printed earth formwork, hereby referred to as EarthWorks. This
research demonstrates the potential of zero waste, circular formwork that can be manufactured with
construction waste soils directly on site. Methods are described for formwork design and toolpathing that
accounts for hydrostatic pressure, conventional reinforcement, high accuracy connections, and the
fabrication of complex, 3D-shaped geometry with continuous extrusion. In addition, the building design
and performance potential of the EarthWorks method are assessed and compared to existing additive
formwork technologies from a carbon perspective. Case studies are fabricated demonstrating cast-in-
place, tilt-up, and on-site prefab methods to produce bespoke columns, beams, and frames designed to
California building code.
Keywords:
Low-Carbon Construction, 3D Printed Earth, 3D Printed Architecture, 3D Printed Earth Formwork
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
2
1 Introduction
1.1 The dual challenges of rapid urbanization and climate change
As the rate of global, rapid urbanization continues to grow, researchers and practitioners in Architecture,
Engineering, and Construction (AEC) are being pushed to explore automation to accelerate the
development of new housing and infrastructure [1]. At the same time, the construction industry is
contributing 11% of global annual carbon emissions, a number projected to grow over the next 50 years.
Eight percent of this impact is a direct effect of cement production for concrete, currently the world’s
most ubiquitous building material [2]. To address the twofold challenge of rapid urbanization and the
climate impact of growing cities, novel systems must be developed to minimize the use of concrete in
new construction with globally scalable construction automation technologies. An extensive body of
research illustrates methods for material reduction in concrete structures through shape optimization;
however, fabricating shape-optimized structural elements remains a critical challenge. We present novel
methods for 3D printing earth as formwork for shaped, reinforced concrete building elements based on
previously patented research by the authors (Curth 2023, Pending). Using additive manufacturing, soil,
often considered waste on a construction site, can be transformed into infinitely recyclable formwork for
mass-customized, shape-optimized, structural elements readily deployable in industry.
1.2 Significance of formwork
Concrete is the most widely used modern building material for good reason: it is cheap and strong, its
components are globally accessible, and a broad spectrum of building typologies, from single-story
houses to 100-story skyscrapers, can be constructed using the same essential building technologies. For
many, concrete is a symbol of modernity and success. It is the material of the 21st-century city. There is a
strong likelihood that concrete usage will continue to increase steadily into the foreseeable future.
Reducing the material’s climate impact can be coalesced into three basic strategies:
1. Change the chemistry of concrete to reduce its carbon emissions,
2. Change how cement is manufactured,
3. Use less concrete.
The first strategy is the subject of extensive research with promising but ultimately limited results in
terms of CO2 savings relative to cost and accessibility [4]. Waste streams from steel manufacturing and
other industrial processes can be used to offset some of the concrete’s cement content. However, those
waste streams come from industries that are rapidly transforming to reduce their own carbon impact and
minimize the very waste driving concrete emissions reductions. Secondly, transitioning cement
production to electrochemical systems that utilize a renewable grid mix could cut a portion of emissions,
but it is expensive and has not been demonstrated at scale [5]. The third option is seemingly simple and
immediately actionable: find ways to use less concrete in new construction. Research has shown that
more than half the concrete in a typical building is structurally unnecessary and could be eliminated
through shape optimization [6]. The challenge lies in fabrication. Reinforced concrete structures are
typically made from consistent rectilinear sections to minimize the complexity of both formwork and
reinforcement. Any added complexity in a concrete building can be closely correlated to increased cost
and material waste [7]. Unique geometry requires unique formwork and highly skilled labor to fabricate
it. As a result, improving the material efficiency of concrete buildings is primarily a challenge of
efficiently producing complex formwork. The question explored in this research is whether additive
manufacturing with local materials can close the gap between shape-optimized structural design and low-
cost, high-speed formwork fabrication.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
3
1.3 Aims
The aims of this research are twofold. First, harness the capability of additive manufacturing to produce
materially efficient, shape-optimized structural elements with low-carbon, infinitely recyclable formwork.
Second, to outline an additive manufacturing formwork methodology applicable across a range of
environments and fabrication typologies, including cast-in-place, tilt-up, and pre-fabrication. By making
complex formwork more accessible, we make it possible for designers and builders to choose low-carbon
options that may previously have been prohibitively expensive and difficult to manufacture.
2 State of the art
2.1 Formwork
Since Roman times, formwork for concrete has predominately been produced from processed timber. The
size and availability of timber and the skilled labor to mill it into precise shapes have driven what could
be built, leading to the common saying that a concrete building is “built twice” first as a timber formwork
structure, then a cast concrete structure [8]. Today, more sophisticated reinforced concrete projects may
use modular steel forms or even massive slip-forming molds for simple geometries like cooling towers or
building cores. However, most housing and urban infrastructure construction still utilizes a system of flat
timber boards or plywood held in place by wooden supports. Depending on the quality of the formwork
material and the skill of the builders, it may be reused for more than one concrete pour. However, at the
end of their lifecycle, formwork materials typically go to landfill [9]. For much of the world, timber, steel,
and foam formwork are becoming increasingly costly as supplies of each material dwindle. As global
efforts continue to push to reduce climate impact in the built environment, scalable technologies that offer
both minimal carbon impact and high-speed construction are of paramount importance.
Digital fabrication and additive manufacturing, in particular, have been demonstrated to be a viable path
toward implementing performative shape-optimized geometries [10]. Examples have been fabricated at
MIT using plastic and foam 3D prints as formwork for reinforced concrete [11]. Researchers at ETH
Zurich have investigated the potential of mineral foams as formwork [12] as well as direct powder
printing (self-supporting) floor systems [13]. In addition, CNC cut molds have been produced for shaped
concrete beams[6]. Fabric molds have been used to create architectural structures since the late 1800s and
have recently been used as thin shell formwork for large spanning vaults[14], [15]. Earth has also been
directly explored as formwork for low-carbon foundations, such as materially efficient hypar shell
structures [16]. Ice and sand have even been utilized as digitally fabricated formwork for concrete [17].
Each of these examples offers compelling possibilities for specific use cases, however plywood and steel
forms are still the industry norm.
2.2 3D printed formwork
Establishing scalable, automated construction methods is a widespread challenge for the building
industry. As other manufacturing processes, including automotive and tech, have moved forward with
increasing levels of automation, construction has remained a largely manual process [18]. When Behrok
Khoshnevis first developed contour crafting in the early 2000s, one of the first applications he explored
was formwork for reinforced concrete [19]. Concrete 3D printing is growing as an industrialized method
for automation in construction, primarily as formwork for more conventional, code-compliant, cast-in-
place reinforced concrete structures. This technique can be seen in buildings produced around the world
by researchers and construction companies, including CoBod, Apis Core, and Icon. Studies have also
shown the viability of 3D-printed mortar for shape-optimized spanning systems, including waffle
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
4
slabs[10] and post-tensioned, topology-optimized beams [20]. Recent studies have detailed the advances
in 3D-printed cementitious formwork [21], illustrating the key differences between the existing typologies
of large-scale additive manufacturing (powder bed and fused deposition).
As the technology has evolved over the past two decades, similar formwork research efforts have been
made with other industrially processed, engineered materials, including thermoplastics, geopolymers, and
ceramics. Each of these approaches presents different advantages ranging from high resolution,
cantilevering geometry printed over the course of days in plastics [22] to low resolution, minimally
cantilevering geometries printed in a matter of minutes using mortar and mineral composites [12]. Printed
ceramic materials have been utilized as incremental lost formwork in a fired and unfired state, and their
accuracy/deformation during casting assessed through 3D scanning (Bruce et al. 2021; Mohamed Ismail
2023). Each of these studies has a recurring tradeoff between fabrication speed, resolution/geometric
freedom and precision, and material performance. In this study, we work to balance these tradeoffs,
creating a system that can be deployed on-site or in a prefabrication context using waste soils already
available on building sites.
In each system, the integration of reinforcement and the afterlife of the formwork material are key
sticking points. Finding methods for the use of standard reinforcement design made of readily available
rebar sizes and commonly formed shapes facilitates the broader application of these formwork methods
[25], [26]. The time required to build a sufficiently robust rebar cage can far exceed that required to
produce formwork conventionally or with additive technologies. As such, standardized, prefabricated
units are often employed in industry. To realize the potential of mass customization for material saving
and design freedom, equal attention must be paid to formwork and reinforcement. For lost formwork
strategies, reinforcing may need to be placed sequentially or in modular elements as printing proceeds,
creating the potential for added complexity and an increase in the steel used, resulting in higher carbon
impacts. In this study, we explore multiple approaches to reinforcing and methods for the mass
customization of bespoke rebar cages for integration with 3D-printed earth formwork.
Table 1: A selection of critical 3D-printed formwork studies organized chronologically and compared. Speed, Embodied Carbon,
and Reinforcement Strategy are widely discussed in the literature, yet often not characterized.
Study
Source
Material
Printing
Technology
Circularity
Embodied
Carbon
(kgCO2/kg)
Print
Resolut
ion
(mm)
Speed
(mm/s)
Cantilever
limit
(degrees)
Test
Size
Maximum
speed per M2
of formwork
(hrs/m2)
Reinforcement
strategy
Eggshell
[22]
Thermoplastic
(PP)
6-axis Robot.
CEAD pellet
extruder
N
Not reported
1-2
25-45
Not
Reported
1.8m
column
5.56hrs/m2
Rebar
Mineral foam
[12]
Mineral foam
6-axis Robot
Ram extruder
N
Not reported
7-10
9-50
5
200mm
tall
cylinder
0.56hrs/m2
Glass fiber
Fired Clay
[24]
Fired Clay
6-axis Robot
Pneumatic ram
N
Not
Reported
2
20
Not
Reported
1.25
x.25m
beam
6.94hrs/m2
Rebar
Contour
Crafting
[19]
Mortar
3-Axis gantry
Concrete
pump, Ram
Extruder
N
Not
Reported
13
20
Not
Reported
1.5x0.2
x0.6m
wall
section
1.07hrs/m2
Rebar
Mobile
Additive
Manufacturing
[27]
Clay
Mobile 6-axis
robot
Y
Not
Reported
Not
Reporte
d
Not
Reported
Not
Reported
1.3m
tall
column
Not Reported
Rebar
Programmable
Mud
[28]
Raw Earth
6-axis robot
Y
0.006
10-25
20-50
20
3 x 3 x
0.3m
0.22hrs/m2
Rebar
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
5
Progressive
Cavity Pump
wall
section
Digital
Construction
Platform
[11]
Expanding
foam
6-axis robot
Pressurized
foam sprayer
N
Not
Reported
10-25
50-100
20
15m
diamete
r, 3m
tall
dome
0.11hrs/m2
None
Cocoon
[23],
[29]
Raw Clay
6-axis robot
Ram Extruder
Y
Not
Reported
1-3
Not
Reported
Not
Reported
1.3m
tall
column
Not Reported
GFRC
Clay Robotics
[30]
Raw Clay
6-axis robot
Ram Extruder
Y
Not
Reported
1-3
Not
Reported
Not
Reported
Not
Reporte
d
Not Reported
None
Concrete
Choreography
[31]
High Strength
Mortar
6-axis robot
Progressive
Cavity Pump
N
Not
Reported
5
180
20
3m tall
column
0.3hrs/m2
Rebar
DOE for Clay
Molds
[32]
Raw Clay
6-axis robot
Ram Extruder
Y
Not
Reported
1-3
100
Not
Reported
200mm
tall,
100mm
wide
0.93hrs/m2
None
Parametric
Waffle Slabs
[10]
Mortar, Fired
Ceramic
6-axis robot
N
Not
Reported
10-15
300
30
50x50x
50cm
0.06hrs/m2
Rebar
The broad spectrum of architectural scale 3D printed experiments can be roughly compared with a time
per square meter interpretation of vertical printing rate (VPR), a metric discussed in previous studies [22].
However, such a comparison would only be meaningful if we assume that each experiment was produced
as fast as possible with the highest quality equipment; for example, the projects Concrete Choreography
and Parametric Waffle slabs employed similar robotic systems and materials. However, the latter was
manufactured at nearly twice the VPR of the former by an industrial partner using proprietary pumping
equipment and a proprietary 3D printing mortar formula, highlighting the challenges of comparing
existing formwork experiments and validating their reproducibility. Notably, no prior studies provided an
embodied carbon estimation of their formwork method. What can be compared directly is the circularity
of the processes described. Only formwork produced with unfired clay and earth can be directly recycled
back into the printing process by adding water. Ismail’s study of low-cost formwork for shaped beams is
the only research on 3D-printed formwork to date, which suggests how such molds might be used to
produce more than one cast. However, no such tests were conducted on printed experiments.
2.3 Earthen materials
Digital fabrication with earth is a growing area of research with innovations in the production of discrete
compressed earth block and rammed earth wall sections, as well as continuous or monolithic fabrication
through poured earth and additive manufacturing [33]. In the realm of additive manufacturing, the critical
challenge is rationalizing a fundamentally variable, locally sourced material into a reliable feedstock for
precision fabrication. In the earliest study of 3D-printed engineered clay materials for formwork, Wang
establishes the need for the assessment of hydrostatic pressure in the forms. Recent studies by the authors
and others detail field tests and processing strategies necessary to make earth a viable material for
consistent 3D printed results at an architectural scale. In this study, the material processing and testing
methods for printing earth are detailed in previous research and patents by the authors [3], [28]. These
methods include a hydrostatic pressure assessment and a novel buttressing strategy demonstrated in the
case studies section. Further gaps in the literature exist around water uptake and demolding strategies,
which we work to address here through post-printing coatings.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
6
2.4 Computation and shape optimization
A significant body of research details methods for the geometric optimization of structural elements to
reduce material usage, indicating that, in many cases, 50% of the material in a building’s structural
system is unnecessary. Examples of this research include shape-optimized one and two-way floor
systems, as well as approaches for massive material savings through the application of shell footings [10],
[34]. Shape optimization is a design exploration process in which a parameterized boundary
representation is constrained by strength and deflection requirements. Complex geometry is often
required to achieve maximum material savings. Such geometry may include single and double curvature
or voids of varying sizes, making the shared challenge of these existing shape-optimized structural
systems fabrication. As a result, there is a critical need for automated fabrication systems that can manage
mass-customized, generally complex geometry creation, a gap that additive manufacturing potentially
addresses cost-effectively. This study synthesizes current shape optimization with material and machine
properties to produce a novel computational workflow suitable for designing and fabricating optimal
geometry with locally sourced materials.
2.5 Life Cycle Assessment of construction processes
Life Cycle Assessment is critical, yet often lacking in studies on additive manufacturing of concrete
formwork. Without a clear understanding of the carbon cost of a given strategy at an industrial scale, it is
difficult to understand its efficacy. Estève-Bourrel has established guidelines for assessing the Life Cycle
and carbon impacts of earthen building materials in early-stage design [35]. In [28], an LCA compares
various earthen printing systems with conventional construction and mortar printing. Compared to mortar
3D printing, printing earth requires 2% the carbon to produce wall sections of equivalent structural
performance. This study compares conventional and digital fabrication techniques for concrete formwork
from a life cycle perspective.
2.6 Research gap and opportunity
Additive manufacturing for reinforced concrete formwork is a growing area of research, as is earth 3D
printing. In prior studies, full-scale printed molds are produced with lost formwork that is often made
from a high-carbon material like plastic or cement. Those molds made from circular materials,
specifically unfired clay, are comparatively small, low-strength, and slow to produce (Table 1). There is a
lack of comparison or discussion around the tradeoffs between material cost, embodied carbon, accuracy,
and speed, all of which are critical factors for industrial adoption. In the realm of earthen 3D printing,
material characterization continues to be a challenge. Recent research has trended toward engineered
earthen materials, which increase cost and carbon impact, particularly when stabilizers are added. In this
study, we analyze existing approaches to balance the factors listed above. Building on prior research, [3],
[28] we choose minimally modified, carefully sourced local earth and then develop computation methods
to maximize its performance for formwork applications, reducing the need for characterization.
This study addresses the following research gaps:
1. 3D printing locally sourced earth for formwork applications.
2. Computational methods for managing mold drying and hydrostatic pressure during casting.
3. Fabrication methods for tilt-up/prefabricated and hybrid print/cast-in-place building elements.
4. Methods for effective mold preparation and reinforcement alignment.
5. 6-axis, space-filling earth printing with continuous extrusion for 3D-shaped molds.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
7
6. Comparative life cycle assessment of paired shape-optimization and zero-waste EarthWorks
method.
3 Methodology
Additive manufacturing (AM) is a flexible method of digital fabrication. In this work, we engage AM
processes to produce both finished building elements and temporary formwork for traditional reinforced
concrete. Processing and mechanical parameters of the local soil being used are factored into early-stage
design to fully leverage the construction potential of AM. The result is building elements that are
materially efficient, highly performative, and produced with zero waste. To test and demonstrate material-
aware earth 3D printing for formwork, we produced three prototypes (Figure 1).
1. Cast-in-place: An earth wall acting as lost formwork for a cast-in-place, curving frame poured
after the printed earth reaches a dry state in 72-96 hours.
2. Tilt-up: A 2D-shape optimized portal frame cast on its side in an earth mold printed and
reinforced with optimized buttressing to resist hydrostatic pressure on the mold in a partially
dried state, reducing the time between printing and casting to 12-24 hours.
3. Pre-fab: A 3D-shape optimized beam cast in an earth mold printed to maintain even wall
thickness throughout for rapid drying to allow casting within 24-48 hours and 6-axis surfacing
toolpaths for geometric accuracy to maximize material savings.
Each case study required the implementation of novel material-driven computational design
methodologies to minimize print time and concrete use. In each case, the design process is driven by the
California Building Code structural design and concrete fabrication requirements, allowing us to directly
manufacture prototypes that could be used in multistory construction today.
Figure 1: EarthWorks is a flexible framework for manufacturing materially efficient reinforced concrete building elements.
3.1 Material sourcing and processing
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
8
A simple mixture of earth, water, and straw produces a raw earth material suitable for 3D printing. The
appropriate mixture for climate and printing needs is determined by adjusting the ratio of each
component. For the case studies outlined in this paper, two earth sources are used: raw local earth and that
same earth recycled from prior 3D mud prints. In each case, earth is sourced at an excavated construction
site in Montecito, CA, as described in [28].
Figure 2: Adapted from Curth 2023. 3D printing earth is a fundamentally circular process of direct material reuse.
After excavation, raw earth is sifted to sort out particles greater than 1cm in diameter, a constraint defined
by the pumping system (Imer Small 50 progressive cavity pump). Sifting is more easily enabled when the
earth is dry, and if necessary, it can be spread into a thin layer to sun dry prior to sifting. Earth is passed
through a sifting structure with an angled wire mesh and thus sorted to an appropriate particle size for
pumping to the 3D printing end effector. More industrial methods exist for sifting large quantities of
earth. However, these methods were beyond the scope of the study.
Once earth is sifted, it is weighed and mixed. For these case studies, 100 kg of earth batches were mixed
at a time. This is a feasible quantity to batch in a six-cubic-foot concrete mixer and transport via
wheelbarrow. To reach a final mixture with 15%-20% hydration, 20 kg of water (by weight) is added into
the mixer, followed by 50 kg of the dry earth in 10 kg increments. Once the earth and water have
homogenized (5-10 minutes mixing), 0.03 kg mulched straw with a maximum length of 10 cm is added.
Once thoroughly incorporated, the remaining 50 kg of sifted earth is added to the mixture in batches of 10
kg at a time. The mud thickens with each addition of dry earth, and when mixing is completed, the mud
body can flow but is still viscous and cohesive. In total, each batch consists of 100 kg of dry sifted earth,
20 kg of water, and 0.03 kg of chopped straw.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
9
The processing of recycled (already printed) earth varies slightly because this material has already been
sifted and modified with straw. Therefore, the mud does not need to be crushed and sifted; it can simply
be rehydrated. If the mud to be recycled contains pieces of concrete from the casting process larger than
the 1cm pump limit, it can be dried, crushed, and re-sifted. Recycled mud is more easily re-hydrated if the
surface area is increased by breaking apart the dried chunks. Recycled 3D mud prints are crushed,
weighed, and added to a wheelbarrow using a sledgehammer. Again, more industrial methods for
crushing dried soil were beyond the scope of the study. A ratio of 20% water by weight is then poured
over the dried pieces of 3D prints. A hoe or shovel is helpful for mixing and turning the mud as it
hydrates. The rehydrated chunks and straw can then be added to the concrete mixer or turned and
combined with a shovel until the desired homogenous mixture is achieved. Leaving the rehydrated
material for 12-48 hours facilitates the complete homogenization of the mix. However, it is entirely
possible to print immediately after remixing a batch of the earth so long as the mixing is thorough. Larger
scale equipment used in geotechnical engineering, such as sifting shake tables and higher volume
concrete mixers, could be used to accelerate and scale the material processing workflow.
3.1.1 Site considerations
One key advantage of additive manufacturing with the earth is the potential for not only producing
formwork on-site but also integrating earth excavated during initial construction into the formwork’s
earth mixture. Onsite fabrication of formwork with excavated earth could cut down on the transportation
costs of structural elements and disposal of excavated earth and allow real-time modification of formwork
designs. For this approach to be effective, it is critical that local and site-specific environmental
conditions, such as humidity, precipitation, and sun exposure, are taken into consideration and that soil is
tested to evaluate how it will perform as formwork. Environmental conditions could impact the ratio of
ingredients used in the mud mixture, printing time, and drying time. Slight workflow modifications, such
as shade structures, additional water, or fans, can accommodate environmental fluctuations.
3.1.2 Material sourcing
Earth for printing can be sourced directly from the construction site where the printed forms are needed.
However, there are significant variations from site to site depending on geography, depth of excavation,
and geologic history. Some indicators that local earth is good for additive manufacturing are if there is
existing earthen architecture in the region or if there is evidence of erosion and silty deposits nearby. The
color and texture of the earth can be a good indication as well. There are simple onsite field tests and
printing-specific tests detailed in the authors’ previous material studies [28] for determining a material's
cantilever and bridging potential, which can indicate if a site’s soil is adequate for earthen construction
and formwork [36]. An ideal earth mixture will use earth that is approximately 30% clay and 70% sand
and can be defined as a ‘Sandy Clay Loam.’ Field tests for evaluation are adequate for 3D printing
formwork uses; however, further geotechnical, structural, and chemical laboratory tests can be done as
well if the intention is to leave earthen formwork as part of the completed structure for structural, thermal,
or aesthetic reasons (Case Study 2). Further lab testing could include dry density, unconfined compressive
strength, particle size distribution, plastic limit, expansion index, and specific gravity.
The case studies outlined in this paper use waste soil from a nearby construction excavation in Goleta,
California. The soil is a well-graded sandy clay loam. The climate in Goleta, where these prototypes were
printed, is arid and coastal, with an average annual humidity of 50–60%.
3.2 Mold preparation
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
10
Informal experimentation by the authors determined that mold release agents are helpful in removing the
concrete from the earth molds and limiting water exchange. As a result, various mold coatings are
explored to minimize moisture exchange between printed earth and poured concrete, given the potential
to reduce concrete strength and consistency. Test coatings included ceramic wax (car wax), oil lubricants
(WD40), and spray-on rubber. 3D printed earth cylinders with an interior dimension of 20cm diameter
and 100cm height were printed and left to dry to a “leather hard state” (moisture content approximately
5%) before being coated with three layers of each material, leaving time for each coat to dry before the
next is applied. Rubber coating is removed from the experiment because it limits the recyclability of the
earth mixture, while wax does not. Three samples of wax-coated and uncoated cylinders are prepared,
along with three control cylinders of the same dimension cast in plastic molds. Concrete is then cast in the
cylinders, de-molded after one week, and load tested after one month.
3.3 Computational design methods
The computational design workflow used in this study allows the builder to factor in material, structural,
and fabrication considerations from an early stage. The design software uses Rhino3D/ Grasshopper [37],
the Design Space Explorer plugin [38], Python 3.8, and the Karamba 3D plugin [39] to integrate material
properties and structural optimization to produce 2D and 3D shape-optimized concrete elements,
including frames, beams, and foundations. The fabrication method to be used is defined by the size of the
building element and how it can be manufactured as quickly as possible; for example, instead of printing
formwork for a frame to be cast in one piece standing vertically, the model is positioned on its side, and a
tilt-up strategy is employed, reducing print time from days to minutes. As the designer works, the model
provides key metrics, including material usage in both final geometry and forms, as well as estimated
fabrication time and stability of the mold during fabrication and under the hydrostatic pressure of casting,
allowing for carbon-informed design exploration within the latent space of the shape optimization.
Adjacent studies by the authors investigate the integration of thermal performance into this design and
fabrication workflow [40]. Toolpaths for the finalized geometry are then generated with the specified
layer resolution. The toolpaths used in this study are continuous to minimize print time, requiring careful
consideration of print sequence.
Figure 3: Integrating material and fabrication analysis in early-stage design encourages the designer to minimize carbon impact
from the start.
3.3.1 Hydrostatic pressure
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
11
Hydrostatic pressure is a critical challenge in large-scale casting processes. Pressure increases linearly
with depth, leading to high lateral loads on formwork at the base of a wall or column. In a previous study,
baseline hydrostatic pressure tests were conducted for printed earth [28]. In the EarthWorks methodology,
buttressing is parametrically generated in response to a model of hydrostatic pressure after the shape
optimization of a base geometry is completed. This technique is both computationally quick and can be
applied across a wide range of geometric conditions, facilitating casting before the mold is dry. For
hybrid, lost formwork earth/concrete structures, more bespoke consideration of hydrostatic pressure is
required. Tests up to three meters were conducted on cylinders of various moisture content samples from
wet to dry printed earth with an interior diameter of 20cm, wall thickness of 40mm, and a 20mm layer
height. Cylinders were chosen to provide a uniform loading condition in 360 degrees. Concrete with a
density of 2.4g/cm3 was used, and five seconds of vibrating compaction with a vibratory rod was used
after the concrete reached a depth of 10cm and again at 20cm to properly simulate our larger scale casts.
No deformation or failure was observed in all but the highest water content tests, indicating the material’s
robustness as formwork but necessitating future studies with higher hydrostatic loads.
3.3.2 Toolpathing for 2D and 3D conditions
Within the scope of traditional additive manufacturing and for most of the printing paths outlined in this
research, a 3-axis gantry system is sufficient. However, with increased access to robotic arms, 6-axis
systems can achieve increasingly complex geometries, where the tool path is no longer relegated to
planar, vertically-oriented configurations. Non-planar configurations can better conform to 3D surface
conditions, provide more accurate surface reproduction, and limit undesirable surface finishes [41].
As desired geometry increases in size and with lower robotic mobility (fixed robotic arms), it is
imperative to develop a robotic locomotion strategy for reachability and to avoid singularities within a
print. All machine toolpaths exit the digital model with polar planar rotations to maximize the arm's reach
and allow continuous printing through typically problematic quadrant-based singularities. In this research,
the strategy was twofold: geometry positioning and localized plane orientation adjustment.
1. A tool path can be placed within the robot cell’s given work area, sometimes spanning 360
degrees, making full use of axis 1. Specific geometries are better suited for such a configuration,
for example, tilt-up frames (Case Study 2) that can wrap around a robot's mounting location.
However, positioning is not always entirely sufficient, and instead, the robot axes must respond to
the toolpath’s location at any given time. In the case of a long-spanning beam (>4m) (Case Study
3), where the entire cartesian x or y axis of a robot is needed, singularities in the wrist (between
axes 4 and 6) can occur. However, such an event depends on work height and whether or not the
robot’s configuration will render axis five at 0 degrees. As such, a secondary path-related strategy
was employed.
2. After a tool path is positioned, its position frames are calculated and reoriented so that the x-
axis points in the robot's direction in a polar fashion. Often, additional universal manual plane
rotations varying from -20 to 20 degrees can improve printability. In such a case, the plane's z-
axes have not changed their vectors.
3.3.2.1 Print surface rationalization
In some of the case studies illustrated here, the geometry to be cast exceeded the tolerance of the available
build area. The ground plane was not flat enough to maintain a submillimeter tolerance across a four-
meter beam. As such, a roughing raft pass is used to create a surface precisely parallel to the robot’s base
plane. This may only require one pass over the build area, or in more extreme scenarios (printing on-site
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
12
or in an environment where site preparation is not practical – such as the rocky Martian surface), an
interpolated leveling strategy can be employed by building up increasing planar layers of material relative
to a starting mesh conformal to the build surface. This is achieved through 3D scanning the site,
projecting the base toolpath to the scan, and interpolating layers upward to a plane from the scan surface.
In the case of the experiments outlined here, a single layer of material was sufficient to rationalize the
warehouse floor and provide a precise platform for printing.
3.3.2.2 Infill/roughing
Given that printing systems using minimally processed local materials can be time-consuming to calibrate
for retraction, we chose to employ continuous extrusion and use strategic toolpathing algorithms to avoid
the need for travel moves. This also significantly reduces the cost and complexity of both the extruder
pump and pump control system, as no pinch valves or speed control integration is necessary. For print
base surface rationalization, our model uses a simple linear parallel path or, in the case of more complex
geometry, a reaction-diffusion space-filling curve. For roughing and finishing passes in 2D and 3D
shaped molds, a square waveform of consistent frequency but variable amplitude is used to tailor a
continuous curve to produce an accurate surface. For 3D forms, a roughing pass may be printed in 2.5D
(X and Y rotation fixed, Z moving) for high-speed printing, followed by a fully 3D finish pass normal to
the model surface (Case Study 3).
3.4 Reinforcement and detailing
Designing and fabricating additive earth formwork requires careful detailing. Earth, as a traditional
building material, absorbs tolerance errors. As the material takes on a more significant role in modern
construction and as concrete formwork, it is critical to consider tolerance, margin of error, and installation
of structural elements. Detailing differs for each case study, but critical overarching considerations
include rebar clear cover that meets local building code, lifting points for modular cast elements, and the
moment connections between the foundation, wall, and beam. In each case, higher resolution 3d printed
thermoplastic jigs are used. Printed clips make it possible to hand bend rebar approximately to shape and
then additively improve the accuracy of the cage as each clip is connected, completing the cage. These
elements can also be mass customized and deployed as needed to maintain clear cover or register a
moment connection (See Figure 3).
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
13
Figure 4: Parametric rebar clips allow precise construction and positioning of a shaped, code-compliant rebar cage (Image:
Emily Wissemann).
In preparation for lifting and moving the cast-in-place mud wall (Case Study 1), the earth print was
modified to create the negative space required in the foundation for direct lifting with a forklift. In Case
Study 2, plastic is laid down so the print can contract upon itself as it dries, and lifting straps are laid
below the plastic. After the precast frame has cured, the straps can be used to lift and move the element to
the site.
Modular concrete elements rely on robust moment connections. These connection points are dictated by
the digital model and are coordinated with footing and beam connections. For foundation connections,
steel plates with holes for bolted connections, were welded to the rebar cage, producing moment
connections compliant with California building code [42, p. 19]. 3D-printed plastic plugs were inserted
into the bolt holes, preventing concrete from being cast in these zones so that bolts could be placed and
tightened when the element was lifted into position.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
14
Figure 5: Left: 3D-printed plastic blocks are placed in the bolt holes of the frame base to facilitate assembly after casting. Right:
Moment connections in the frame were built to California building code. Note that small cracks appeared during drying but did
not impact the accuracy of the cast (Images: Alexander Curth).
3.5 Life Cycle Assessment
The life cycle assessment of the earth printing systems used in this study is described in detail in Curth
2024. This cradle-to-grave model assumes that both the robotic arm and pump are operating at full power
all the time and does not account for the circularity of the material. The conservative estimate of
0.006kgCo2/kg from the prior study indicates that, unlike other additive manufacturing processes, carbon
impact is driven by print time, not material quantity. The most considerable carbon impacts for a
functional unit of printed earth come from the power draw of the pumping system. Practically, however,
limiting material usage is directly correlated to minimizing print time; the shorter the print path, the less
run time and material are required. Compared to other AM formwork methods using foam, mortar,
thermoplastics, or even fired clay, the carbon impact is so low that entirely different design considerations
are possible. Material usage is not a serious cost or carbon concern so long as overall fabrication time
does not impede production. This opens up possibilities like using earth as support material and infill for
3D-shaped molds or as buttressing to resist hydrostatic pressure without incurring a significantly elevated
carbon impact.
4 Results
4.1 Hydrostatic pressure
The hydrostatic pressure test cylinders are loaded with wet concrete to a pressure of 0.11 MPa. Across
three samples for each water content (5%, 10%, 15%, 20%), failure is only observed in the 15-20% water
content cylinders. Below 10% water content (or a “leather hard” state by feel), the cylinders were not
deformed; above 10% water content, local buckling is observed in a classic hoop stress failure, resulting
in a split in the mold approximately 5cm from the base of the cylinder. Frictional forces prevent the
failure from occurring at the bottom of the mold, where pressure is highest. Given the initial results
described here and in [28], a larger-scale test is conducted on a single cylinder at 5% water content (the
typical moisture level at which the case studies are cast). The cylinder’s height is extended with three
meters of plastic tubing. When fully loaded to a hydrostatic pressure of 0.17 MPa, no deformation is
observed by filling the tube with wet concrete. Further hydrostatic testing could include other test
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
15
geometry, higher-density concrete mixes, and consideration of the effect of fiber content on the ability of
the printed earth mold to resist lateral loads.
Notably, after these tests, the design workflow is reconfigured to assume the earth at casting moisture of
5% could resist 0.1 MPa of hydrostatic pressure. This assumption gives the model a minimum safety
factor of 1.5 based on these tests and enables prediction of where buckling would occur, particularly in
long, straight sections of the mold.
Figure 6: Simulation-driven buttressing was successfully employed in the casting of a 3 x 3 meter tilt-up frame (Image:
Alexander Curth).
4.2 Mold coating
Load tests of the cast concrete cylinders do not indicate a significant difference between wax-coated and
uncoated molds. However, water is observed leaking from the base of the uncoated mold during casting,
and the earth at the base of the mold is noticeably wetter than when casting had been initiated. After
demolding, poor concrete quality is observed in the uncoated sample, particularly near the base of the
cylinder. Neither of these phenomena are observed in the wax-coated samples, leading to the use of a
wax-coating methodology throughout the production of subsequent case studies. A set of six cast concrete
cylinders are then prepared, three from waxed earth molds and three from unwaxed molds. In unconfined
compression tests, no statistically significant difference is found in performance between the cylinders. It
is possible no negative impact on compressive strength is observed from the water exchange because
equilibrium is reached as the earth’s pore structure fills, limiting water uptake to minor exchanges with
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
16
the surface of the sample, making the poor concrete quality primarily a cosmetic issue. Future work could
further investigate other coatings, assess water exchange between the earth mold and fresh concrete, and
explore larger-scale tests to better understand its impact on strength across various geometries.
4.3 Case studies
To illustrate the generalizability of the EarthWorks method, three case studies are designed and
fabricated.
1. A cast-in-place wall and frame system,
2. a 2D shape-optimized tilt-up portal frame, and
3. a 3D shape-optimized, flanged beam.
Each modality presents unique advantages and challenges both in computational design and fabrication.
The resultant cast, reinforced concrete elements are designed and fabricated to meet California building
codes[42], making each technology immediately applicable to industry. A provisional patent has
protected the methods described to ensure wide-ranging access to this approach to building efficient
concrete structures with zero waste [3]. Ultimately, the novel methods presented in each case study
synthesize and test the described material and fabrication digital design workflow and open a path for
future investigations into other building elements and additive manufacturing-driven formwork design.
4.3.1 Cast in place wall
A cast-in-place hybrid earth and concrete wall is designed to illustrate the potential for in-situ casting of
concrete frames in nonbearing earth walls. The wall consists of a reinforced concrete footing cast into 3D
printed earth formwork, a 3D printed earth wall, four reinforced concrete columns cast into the wall, and a
reinforced concrete sill. The earth print could be removed from the sill but remains in place for the wall
and foundations after casting. Code-compliant rebar cages are bent and tied by hand using 3D-printed
plastic clips to ensure proper spacing and position. All thread rods with threaded couplings were
integrated into the base of the column cages, allowing for the placement of the vertical column rebar
sequentially in half-meter sections as the print progressed in height. While the horizontal sections of the
cage matched conventional construction, adding an all-thread rod in the columns increased the cost of the
prototype (1/2” all-thread is approximately twice the price of equivalent #4 rebar in California). The use
of rebar-specific couplings could mitigate some of this cost. Reinforcement continues to be a critical area
for future research.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
17
Figure 7: After the sill and footing are cast into printed earth, an earth wall is printed, which includes forms for moment-
connected columns. The rebar cage is coupled to threaded inserts protruding from the cast foundation, which are held in
tolerance by 3D-printed plastic jigs (Image: Saleh Jamsheer).
The fabrication of the hybrid earth wall is more time-consuming than the other case study geometries due
to drying requirements between prints of vertical wall sections. After 30cm of printing, the material needs
to lose at least 5% of its moisture content to facilitate the addition of another 30cm of wet earth without
failing under its self-weight. Drying time is entirely dependent on atmospheric conditions in the printing
space; on hot, low humidity (18-30o C) days, printing up to a meter in 3-4 printing sessions spaced three
hours apart was possible. On higher humidity and lower temperature days (8-18 o C), printing could be
limited to 50-60cm with 6 hours between 2-3 printing sessions. Future work could further quantify this
limitation and investigate methods for uniformly accelerating drying. In earlier experiments printed
outdoors, wind is found to be a remarkable drying accelerant but also the cause of differential drying and
resultant cracking.
4.3.2 Tilt-up shaped frame
A portal frame is designed using a 2D multi-objective shape optimization to minimize concrete while
meeting design and code structural requirements. The 2D shape optimization approach taken here, while
accessible and easy to use, is a numerical method that does not accurately model concrete mechanics,
instead treating the concrete as an elastic material. This is in contrast to Case Study 3, in which a 3D-
shape optimization model is used, taking into account the full mechanical behavior of reinforced concrete.
The frame is fabricated on its side, a “tilt-up” casting strategy common in the site manufacturing of
monolithic concrete walls for big-box stores and warehouses with sufficient staging areas. Here, the goal
is to produce a materially efficient structural element, maximizing the working range of a single robot
arm while keeping the fabrication time as low as possible. The formwork consists of a space-filling
pattern on the ground plane, a printed extrusion of the frame’s 2D profile, and printed buttresses to
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
18
manage the hydrostatic pressure of concrete during the casting process. The frame includes steel plates for
bolted connections at its base and connection points for beams along its top. Once cured, the 3D-printed
formwork is removed, and the frame is tilted into its upright position. Of the three methods tested, the tilt-
up frame could be considered the most successful in that the ratio of material used in both the mold and
the cast element is minimal, the rebar cage is conventional and straightforward to fabricate, and the final
product is immediately ready for use as a structural component of a code compliant building. This is the
fastest way to manufacture case studies while also being the largest single cast.
4.3.2.1 Design methodology
Figure 8: Optimal designs were explored across loading conditions ranging from 1-3 stories, assuming the forces typical of
multistory housing in Southern California.
The tilt-up frame is designed for an interior wall condition with a tributary area of 3m x 3m, or 9m2.
These loads are transferred to the frame as a distributed line load. In the computational model, the frame
was treated as a 2D mesh, with 274 points along the top of the frame, which receives the total live (1.9
kN/m2) and dead (12.94 kN/m2) loads. The frame is designed to have two fixed connections with the
ground.
Live loads, dead loads of the overlying slab, and frame weight are integrated into the model for single,
double, and three-story conditions. The cast frame has a volume of 0.237 m3 and a weight of 569.5 kg,
based on an estimated concrete density of 2,400 kg/m3. Frame reinforcement totals 28.65 m of #4 rebar
and weighs 28.57 kg, based on an estimated steel weight of 1 kg/linear meter. Based on these values, the
total weight of the tilt-up frame is approximately 675 kg or 6.62 kN. Additional levels are loaded at two
points along the vertical columns of the frame. Lateral load is calculated by using 20% of the total weight
of the frame (6.62 kN), which is 1.3 kN.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
19
Figure 9: The as-built frame balanced code constraints on rebar cover with the optimal removal of structurally unnecessary
material.
4.3.2.2 Multi-Objective Optimization
The two objectives of our optimization are reducing maximum displacement and the 2D mesh area of the
frame. The mesh area is parameterized with eight fixed and 12 adjustable control points. x1, x4, x9, and
x12 are parameterized to move horizontally, while all other parameterized control points move vertically.
Displacement is calculated from a Karamba 3D model, modeling the frame as a 2D shell with a constant
thickness. The optimization objective is to decrease the mesh area in 2D while maintaining a design
constrained by the maximum allowable displacement (4mm). A penalty is applied if the max
displacement is met and the displacement value is multiplied by a factor of four, exaggerating the penalty
to quickly limit the design space of acceptable options. This value is then added to the output mesh area,
dramatically increasing the objective value and clearly designating a poor solution. Optimization is run
using the DSE toolbox [38]. First, a global optimization (GN-ISRES) was run, followed by a local
optimization (LN-COBYLA). The optimization result is then exported and adjusted based on fabrication
constraints (Figure 8).
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
20
Figure 10: 3D-printed plastic rebar chairs remain visible to illustrate the shaping of the rebar to match the overall optimization.
Printed plastic inserts used to accurately register the moment connections at the top of the frame are removed after casting
(Images: Oliver Moldow).
4.3.3 Pre-Fabricated 3D-shaped beam
Figure 11: The unearthed beam (Images: Alexander Curth).
A four-meter, 3D-shaped beam is designed using Beam Shape Explorer (BSE) [24], a tool that optimizes
a series of the sections along the beam within a set of fabrication constraints, accurately accounting for
concrete mechanics, and then lofts the resultant sections into a complete beam. The resultant solid
geometry is then translated into a printable mold design, which includes a 3-axis roughing pass and a 6-
axis finishing pass. The roughing pass takes advantage of the maximum bridging distances possible
within two layers of printed earth at 20 cm layer heights, creating space for airflow underneath the mold
to allow for even drying and the reduction of print material, thus reducing print time. The finishing pass
utilizes space-filling waveforms with a fixed frequency and variable amplitude to effectively cover the
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
21
complex 3D surface of the optimized beam mold. The printing takes place over two days, leaving the
roughing pass overnight to dry slowly to an adequate strength for the finish pass.
The reinforcement cage consists of one piece of #4 rebar bent to an optimal shape as specified by BSE
and a grid of steel wire in the flange to provide transverse and temperature reinforcement. It is possible to
bend the single, 2D-shaped bar accurately by chalking a template curve on the floor. This can be done by
hand through measurement or by simply attaching a drawing tool to the available CNC/robot system in
use for printing. The rebar is positioned using the edge of the printed mold to support temporary
suspension points for the cage. Additionally, 3x3x3 cm concrete Dobie blocks are placed under the cage
to maintain proper clear cover during casting. These blocks are cast in reusable 3D-printed PLA molds.
While printing this case study is far more complex than the 2D-shaped tilt-up frame, the reinforcement is
significantly more straightforward to manufacture and leaves less room for human error and
compounding tolerance issues. The critical challenge of fabricating the 3D shape-optimized beam is
manufacturing an accurate, doubly curved mold surface through space-filling toolpath generation and 6-
axis extrusion oriented normal to the surface.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
22
Figure 12: The 3D beam mold consisted of four layers (left to right): roughing pass, bridging surface pass, and finish pass. After
completion, a conventionally fabricated rebar was suspended in the mold on blocks (bottom) (Images: Alexander Curth).
This case study reveals two critical challenges to the EarthWorks method. First, rationalizing a 3D geometry into a
mold toolpath containing three and 6-axis paths requires careful tolerance calibration and simulation to ensure the 6-
axis paths do not cause the printer nozzle to intersect or scrape elements of the roughing pass. Once integrated into
the parametric design model, further adjustments for similar beam geometry are straightforward. However, these
pathing issues could be different for any number of 3D geometries, leading to a need for additional work on the
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
23
generalization of the surfacing algorithm. The second challenge is surface finish. In the tilt-up frame case study, we
leverage the imprint of the print path on the cast concrete for its aesthetic and expressive quality; here, we explore
making a smooth casting surface. While areas of the resultant beam have only a subtle and subjectively beautiful
imprint of the finish pass, other regions with slight under-extrusion produced small holes in the mold surface,
leading to seam lines in the concrete cast. This challenge could be addressed by smoothing the mold surface after
printing either robotically or by hand. However, it speaks to issues of extrusion with a 25mm nozzle. Each corner of
the space-filling square wave making up the finish pass contains a 12.5mm radius fillet from the nozzle geometry,
making small gaps in the surface inevitable. Programmed over extrusion could address this challenge, as is now
common in desktop-scale plastic printing.
4.4 Dimensional validation
Each case study mold and resultant concrete geometry is scanned and measured wherever possible. Lost
formwork casts were not extracted for measurement. Various strategies were employed, each with its own
assumed accuracy and tradeoffs:
1. LiDar scanning with iPhone 14 Pro and the PolyCam app (2% declared relative accuracy).
2. Structured light scanning with an Einscan Pro HD (0.1mm accuracy).
3. Touchpoint measurements of mold using ABB 4600 (0.5mm accuracy).
4. Tape measure (2mm accuracy).
5. Weighing cast parts when possible.
Large-scale metrology is a well-known and thoroughly researched challenge. 3D scanning at the scale of
a four-meter object poses inherent accuracy issues, given that none of the geometry produced could be
scanned from a single point. Systems, such as a total station, needed to conduct accurate and repeatable
measurements of sub-millimeter dimensions on our full-scale geometry would be more expensive than the
entire existing printing system combined. A 2mm dimensional accuracy goal is chosen to comply with
conventional, stick-built structures in California. Ultimately, it is found that direct measurement of critical
dimensions, such as relative moment connection position by hand and with a pointer on the robot arm, is
the most relevant to constructing accurate building elements. The surface of the cast reflects the resolution
of the printed mold, resulting in a ridged texture. Global measurement of excess material from these
ridges or other inaccuracies in the mold is difficult to quantify without weighing the element.
Figure 13: Fitting a 3D scan to the digital model of the as-cast geometry of the frame indicates ~10-15mm of excess material on
some edges, primarily in the form of thin pieces of concrete left in the negative space between the extruded layers of earth (see
figure 10).
An advantage of LSAM is the ability to precisely adjust the resolution of a given geometry through
manipulations of layer height and print orientation. In the case of the 3D-shaped beam, the flange surfaces
are ultimately more accurate than the side walls of the beam because they could be printed normal to the
original input geometry. Doing so in the vertical walls of the beam would have required a smoothing
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
24
nozzle attachment due to limitations of reach and self-intersection. In these studies, inaccuracies primarily
occur as a result of the following factors:
1. Irregular extrusion
2. Print resolution (fillets caused by nozzle dimension, print bead shape)
3. Differential drying
Irregular extrusion is primarily a result of mix inconsistency; future work could include sensing for real-
time extrusion width adjustment or a more industrialized material preparation system to ensure
consistency. Print resolution is essentially a matter of layer height choice and print nozzle sophistication.
In Khoshnevis’s earliest Contour Crafting study, he illustrates the potential for entirely smooth, singly
curved print surfaces through active troweling. Differential drying is a more complex problem to address
[19]. Given the addition of straw to our mix and the likely alignment of clay platelets in the extrusion
process, the material is fundamentally anisotropic. Frictional forces between the print and the print
surface also influence how a print changes during drying. In these tests, most of these concerns are
addressed by simply casting when the print still retains 5-10% of its moisture, limiting measured
shrinkage to under 1%. A shrinkage factor was built into the parametric mold pipeline. Adjustments to the
print material’s shrinkage rate by adding larger aggregate and deflocculation are a likely path forward to
addressing this issue; such rheologic adjustments are the subject of extensive research in the concrete
industry. In the case of this study, a balance is struck between minimal modification to locally sourced
material and the accuracy needed to make functional, code-compliant building elements.
Table 2: Summarizing the printed formwork strategies developed in our case studies highlights the relative tradeoffs of each
approach. The cast-in-place wall is the most time-consuming approach but also the approach with the greatest potential for
direct, on-site, whole-building fabrication. Assuming 30mm/s, a 20mm layer height, a 50mm extrusion width, and a concrete
density of 2.4 g/cm3.
Case Study
Print Time
(Hrs at 30mm/s)
Mass of earth
printed (Kg)
Mass of cast
concrete (Kg)
Cast-in-place frame
16.76
3619
556
Tilt-up frame
3.37
727
569
Pre-fab beam
4.55
982
278
Cast-in-place Foundation
0.73
155
452
5 Discussion
5.1 Summary of Novel Contribution
This study presents novel additive manufacturing methods for fabricating reinforced concrete structures
with 3D-printed, locally sourced earth. The ability to print geometrically complex (or simple) molds at a
super-meter scale with directly recyclable, truly zero-waste material opens extensive opportunities for
architects and engineers working to lower the climate impact of new construction. Additionally,
EarthWorks offers new aesthetic potentials in overall form and surface consideration. The process can be
calibrated to the resolution required in different building elements through adjustments to layer height and
print orientation.
The materially informed computational design methods developed to produce the case studies in section 4
provide a reproducible framework for a wide range of potential building applications. Hybrid earthen
structures with integrated voids for cast-reinforced concrete elements and geometrically complex-shaped
concrete elements can be manufactured. Tools for rationalizing large print surfaces make these methods
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
25
accessible for on-site fabrication and in a prefabrication factory setting. The automated buttressing
algorithm facilitates the production of typically unstable print geometries (long straight sections) and
allows for stable casting without form ties. Methods for calibrating tolerance in key mold sections
through printed inserts make it possible to accurately register elements with an otherwise low-resolution
surface. Similarly, mass customized 3d-printed rebar clips allow conventional rebar fabrication
techniques to be accurately applied to shaped concrete geometry, minimizing steel usage and simplifying
the construction of code-compliant rebar cages. The combination of printing and reinforcement strategies
enables flexible and rapid application of the EarthWorks method to geometrically complex building
elements.
5.2 Challenges and future work
The key challenges to scaling the EarthWorks method to industrial applications are similar to the
limitations found in other earth printing applications: the effect of drying time on print speed, shrinkage,
and deformation under self-weight. While these challenges may seem to fall squarely in the domain of
material science, this article has shown that design offers several pathways forward. Printing to low
heights for tilt-up elements, working sequentially on multiple prints at once to allow vertical mold and
wall sections time to dry, and incorporating inherently stable geometries like waveforms more capable of
tolerating shrinkage without cracking than long straight sections. However, extensive work remains to
develop systems for earth mixtures with larger aggregate and lower water contents to mitigate shrinkage
and self-weight deformation. Compatible systems may already exist in larger industrial concrete pumping
equipment. Accelerating drying is also an open area of study; as shown in this work, directed air from
simple box fans can improve vertical build rate so long as even drying can be achieved from all sides to
avoid differential shrinkage. The development of integrated drying systems, either in printed geometry or
in printing systems, could be a compelling area of research. In summary, additive manufacturing
formwork with earth, and additive manufacturing with earth more generally, is a developing field with
significant potential for impactful future research to lower carbon impacts in new construction.
5.2.1 Reinforcement strategies
Each case study used a different reinforcing method, all code-compliant but with their own drawbacks.
Case Study 1 required the use of nonconventional and expensive steel for assembly (all-thread). Case
Study 2 entailed constructing a complex, curving rebar cage based on the digital model, which needed to
be precise to maintain a clear cover while minimizing the concrete used. Case Study 3, while the most
geometrically complex, required simple reinforcing that was easily fabricated by hand.
Fortunately, 3D printing earth formwork offers alternative workflows. For example, an approach to the
fabrication of elements with highly complex rebar cages would be to print the base of the mold, insert the
rebar cage, 3D scan and measure, input rebar mesh into the digital model, modify 3D printing toolpath as
needed for clear cover, and 3D print mud formwork with rebar cage in place. This validation-oriented
approach would be especially relevant for tilt-up scenarios where the entire cage is typically preassembled
and placed as one unit, risking tolerance issues across the complex, continuous cage. This approach would
allow consistent concrete coverage of rebar, further reducing the volume of concrete used and the margin
of error.
Rebar cages could be robotically prefabricated. However, robotic rebar fabrication is time-intensive and
requires highly specialized, expensive equipment [43]. Rebar cages could also be preordered in designs
manufactured by local rebar bending prefabricators. However, this could pose limitations to design
optimization and increase cost.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
26
An alternative method for reinforcement would be to explore prestressing with networks of cables or
fibers. The reinforcement system would be pre-woven, set in place, and pulled into tension after the
formwork is printed. Prestressing with flexible anchor point jigs would allow for the adjustment of the
reinforcement to mitigate minimal concrete cover. It would also reduce the cross sections and necessary
material use.
5.2.2 Retraction
The robotic setup used in these case studies poses some limitations to tool path design because of the lack
of data flow between the robot and the pump. The print has to be a continuous path with a constant bead
dimension from start to finish, as a mechanical system for stopping extrusion has yet to be implemented.
Such systems exist and are commercially available. Inline pinch valves hardwired to the pumping and
robotic control systems are now industry standard. Retraction would be beneficial when geometry is
nested (ex., a column within a wall) or when designing openings in a wall. It would increase flexibility for
element placement (column, window, door), untethered from the consideration of the efficient, singular,
continuous path. Conversely, designing towards a simple robotic system that only allows for continuous
extrusion does lead to the development of maximally efficient toolpathing, no print time is wasted
traveling between sections of geometry.
5.2.3 Aesthetic potential
Concrete has been rendered monotonous due to the standardization of its formwork globally. 3d printed
earth is a compelling alternative to complex timber forms for introducing textural design elements in the
surface of cast concrete. 3D-printed earth formwork would allow aesthetic diversity to be achieved again
easily through surface manipulation and ornamentation. Textures similar to that of Yale University’s
Rudolph Hall can be attained without requiring meticulous labor to place thousands of wooden elements
within the formwork [44]. Building on the existing history of Brutalist Concrete expression, entirely new
forms of ornamentation are possible and open to exploration.
Figure 14: The textured surface of Rudolph Hall was produced with handcrafted wooden forms and then chipped after casting.
The EarthWorks process creates expansive potential for the exploration of functional and aesthetic surface ornamentation.
(Images: Left – John Hill, Middle, Right – Alexander Curth)
High environmental and economic costs are often associated with non-standard or unconventional,
nonrectilinear forms. Due to the circular nature of printed earth formwork, designers can design more
freely and deliberately for efficient and expressive architecture.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
27
6 Conclusion
The significance of the EarthWorks method lies in its direct and flexible applicability to existing
structural design standards. By designing a method around how reinforced concrete buildings are
constructed today, we provide a framework for implementing impactful carbon and cost-saving designs
that were previously inaccessible due to geometric complexity. The use of local soil, a widely available
material, and reproducible computational methods that can be adapted to a given soil's mechanical
properties make 3D printing earth formwork feasible anywhere that a robotic printing system can be
deployed. As large-scale additive manufacturing continues to grow as an industrial construction system,
so will the utility of the EarthWorks methodology.
7 Contribution
Alexander Curth: Conceptualization, software development, structural design, fabrication, analysis,
writing
Natalie Pearl: Fabrication, analysis, writing
Emily Wisseman: Fabrication, writing
Tim Cousin: Structural design, fabrication
Oliver Moldow: Robotic control, fabrication
Latifa Alkhayat: Fabrication, writing
Vincent Jackow: Structural design, fabrication
Keith Lee: Structural design
Mohamed Ismail: Structural design, supervision
Caitlin Mueller: Supervision
Lawrence Sass: Supervision
8 Acknowledgments
Thank you to Clay Studio Santa Barbara, Lynda LABS, for providing research facilities and to Joey
Watson, Jared Vazales, Bruce Ohannessian for assistance with material sourcing and processing.
9 References
[1] UNEP, “2021 Global Status Report for Buildings and Construction: Towards a zero-emissions,
efficient and resilient buildings and construction sector,” Nairobi, 2021. [Online]. Available:
https://globalabc.org/resources/publications/2021-global-status-report-buildings-and-construction
[2] M. Röck et al., “Embodied GHG emissions of buildings – The hidden challenge for effective
climate change mitigation,” Applied Energy, vol. 258, p. 114107, Jan. 2020, doi:
10.1016/j.apenergy.2019.114107.
[3] A. Curth, “3D Printed Earth Formwork for Reinforced Concrete Construction,” 63/507,725
[4] S. Barbhuiya, F. Kanavaris, B. B. Das, and M. Idrees, “Decarbonising cement and concrete
production: Strategies, challenges and pathways for sustainable development,” Journal of Building
Engineering, p. 108861, Feb. 2024, doi: 10.1016/j.jobe.2024.108861.
[5] L. D. Ellis, A. F. Badel, M. L. Chiang, R. J.-Y. Park, and Y.-M. Chiang, “Toward electrochemical
synthesis of cement—An electrolyzer-based process for decarbonating CaCO3 while producing
useful gas streams,” Proceedings of the National Academy of Sciences, vol. 117, no. 23, pp. 12584–
12591, Jun. 2020, doi: 10.1073/pnas.1821673116.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
28
[6] M. A. Ismail and C. T. Mueller, “Minimizing embodied energy of reinforced concrete floor systems
in developing countries through shape optimization,” Engineering Structures, vol. 246, p. 112955,
Nov. 2021, doi: 10.1016/j.engstruct.2021.112955.
[7] B. García de Soto et al., “Productivity of digital fabrication in construction: Cost and time analysis
of a robotically built wall,” Automation in Construction, vol. 92, pp. 297–311, Aug. 2018, doi:
10.1016/j.autcon.2018.04.004.
[8] J. Fitchen, Building construction before mechanization. Mit Press, 1989.
[9] W. Li, X. Lin, D. W. Bao, and Y. Min Xie, “A review of formwork systems for modern concrete
construction,” Structures, vol. 38, pp. 52–63, Apr. 2022, doi: 10.1016/j.istruc.2022.01.089.
[10] Alexander Curth, Ashley Hartwell, Tim Brodesser, and Caitlin T. Mueller, “Parametric waffle slabs:
Optimal geometry materialized with additive construction,” in International Association for Shell
and Spatial Structures 2022, Beijing, Aug. 2022, pp. 3183–3193. Accessed: Feb. 08, 2023.
[Online]. Available:
https://scholar.google.com/citations?view_op=view_citation&hl=en&user=eXO3wG8AAAAJ&cita
tion_for_view=eXO3wG8AAAAJ:pqnbT2bcN3wC
[11] S. J. Keating, J. C. Leland, L. Cai, and N. Oxman, “Toward site-specific and self-sufficient robotic
fabrication on architectural scales,” Sci. Robot., vol. 2, no. 5, p. eaam8986, Apr. 2017, doi:
10.1126/scirobotics.aam8986.
[12] P. Bedarf, C. Calvo-Barentin, D. M. Schulte, A. Şenol, E. Jeoffroy, and B. Dillenburger, “Mineral
composites: stay-in-place formwork for concrete using foam 3D printing,” Archit. Struct. Constr.,
Feb. 2023, doi: 10.1007/s44150-023-00084-x.
[13] B. Dillenburger, R. Flatt, Joseph Schwartz, and M. A. Meibodi, “The Smart Slab - 3D-printed
formwork for a radical new concrete aesthetic,” dbt. Accessed: Dec. 13, 2018. [Online]. Available:
http://dbt.arch.ethz.ch/project/smart-slab/
[14] D. Veenendaal, M. West, and P. Block, “History and overview of fabric formwork: using fabrics for
concrete casting,” Structural Concrete, vol. 12, no. 3, pp. 164–177, Sep. 2011, doi:
10.1002/suco.201100014.
[15] M. A. Popescu, “KnitCrete: Stay-in-place knitted formworks for complex concrete structures,”
info:eu-repo/semantics/doctoralThesis, 2019. [Online]. Available:
https://search.ebscohost.com/login.aspx?direct=true&db=ddu&AN=528D67A0F8E00E5D&site=eh
ost-live&scope=site&authtype=sso&custid=s8978330
[16] K. A. Feickert, “Thin shell foundations: Embodied carbon reduction through materially efficient
geometry,” Thesis, Massachusetts Institute of Technology, 2022. Accessed: Apr. 06, 2023. [Online].
Available: https://dspace.mit.edu/handle/1721.1/144920
[17] O. Gericke, D. Kovaleva, and W. Sobek, Fabrication of Concrete Parts using a Frozen Sand
Formwork. 2016.
[18] Q. Chen, B. García de Soto, and B. T. Adey, “Construction automation: Research areas, industry
concerns and suggestions for advancement,” Automation in Construction, vol. 94, pp. 22–38, Oct.
2018, doi: 10.1016/j.autcon.2018.05.028.
[19] B. Khoshnevis, “Automated construction by contour crafting—related robotics and information
technologies,” Automation in Construction, vol. 13, no. 1, pp. 5–19, Jan. 2004, doi:
10.1016/j.autcon.2003.08.012.
[20] G. Vantyghem, W. De Corte, E. Shakour, and O. Amir, “3D printing of a post-tensioned concrete
girder designed by topology optimization,” Automation in Construction, vol. 112, p. 103084, Apr.
2020, doi: 10.1016/j.autcon.2020.103084.
[21] A. Jipa and B. Dillenburger, “3D Printed Formwork for Concrete: State-of-the-Art, Opportunities,
Challenges, and Applications,” 3D Printing and Additive Manufacturing, vol. 9, no. 2, pp. 84–107,
Apr. 2022, doi: 10.1089/3dp.2021.0024.
[22] J. Burger et al., “Eggshell: Ultra-Thin Three-Dimensional Printed Formwork for Concrete
Structures,” 3D Printing and Additive Manufacturing, vol. 7, no. 2, pp. 48–59, Apr. 2020, doi:
10.1089/3dp.2019.0197.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
29
[23] M. Bruce, G. Clune, R. Xie, S. Mozaffari, and A. Adel, Cocoon: 3D Printed Clay Formwork for
Concrete Casting. 2021.
[24] M. A. Ismail, “Reshaping concrete: Empowering development through low-carbon structural
design,” Thesis, Massachusetts Institute of Technology, 2023. Accessed: Nov. 30, 2023. [Online].
Available: https://dspace.mit.edu/handle/1721.1/150043
[25] M. Momeni, J. Relefors, A. Khatry, L. Pettersson, A. V. Papadopoulos, and T. Nolte, “Automated
fabrication of reinforcement cages using a robotized production cell,” Automation in Construction,
vol. 133, p. 103990, Jan. 2022, doi: 10.1016/j.autcon.2021.103990.
[26] D. Asprone, C. Menna, F. P. Bos, T. A. M. Salet, J. Mata-Falcón, and W. Kaufmann, “Rethinking
reinforcement for digital fabrication with concrete,” Cement and Concrete Research, vol. 112, pp.
111–121, Oct. 2018, doi: 10.1016/j.cemconres.2018.05.020.
[27] G. Dielemans et al., “Mobile Additive Manufacturing: A Case Study of Clay Formwork for
Bespoke in Situ Concrete Construction,” in Third RILEM International Conference on Concrete
and Digital Fabrication, vol. 37, R. Buswell, A. Blanco, S. Cavalaro, and P. Kinnell, Eds., in
RILEM Bookseries, vol. 37. , Cham: Springer International Publishing, 2022, pp. 15–21. doi:
10.1007/978-3-031-06116-5_3.
[28] A. Curth, N. Pearl, A. Castro-Salazar, C. Mueller, and L. Sass, “3d Printing Earth: Local, Circular
Material Processing and Life Cycle Assessment.” Rochester, NY, Dec. 05, 2023. doi:
10.2139/ssrn.4653718.
[29] S. Mozaffari, M. Bruce, G. Clune, R. Xie, W. Mcgee, and A. Adel, “Digital design and fabrication
of clay formwork for concrete casting,” Automation in Construction, vol. 154, Oct. 2023, doi:
10.1016/j.autcon.2023.104969.
[30] S. Wang, S. Dritsas, P. Morel, and K. Ho, Clay robotics: A hybrid 3D printing casting process.
2017, p. 88. doi: 10.1201/9781315198101-20.
[31] A. Anton, L. Reiter, T. Wangler, V. Frangez, R. J. Flatt, and B. Dillenburger, “A 3D concrete
printing prefabrication platform for bespoke columns,” Automation in Construction, vol. 122, p.
103467, Feb. 2021, doi: 10.1016/j.autcon.2020.103467.
[32] S. Wang, Z. Xuereb Conti, and F. Raspall, “Optimization of Clay Mould for Concrete Casting
Using Design of Experiments,” pp. 283–292, 2019, doi: 10.52842/conf.caadria.2019.2.283.
[33] M. Gomaa, W. Jabi, V. Soebarto, and Y. Xie, “Digital manufacturing for earth construction: A
critical review,” Journal of Cleaner Production, vol. 338, p. 130630, Jan. 2022, doi:
10.1016/j.jclepro.2022.130630.
[34] K. Feickert and C. Mueller, “Thin shell foundations: Quantification of embodied carbon reduction
through materially efficient geometry,” Architecture, Structures and Construction, Nov. 2023, doi:
10.1007/s44150-023-00101-z.
[35] P. Estève-Bourrel et al., “Developing an integrated BIM/LCA framework to assess the sustainability
of using earthen architecture,” IOP Conference Series: Earth and Environmental Science, vol. 1078,
p. 012100, Sep. 2022, doi: 10.1088/1755-1315/1078/1/012100.
[36] M. Delgado and I. Cañas, “The selection of soils for unstabilised earth building: A normative
review,” Construction and Building Materials, vol. 21, pp. 237–251, Feb. 2007, doi:
10.1016/j.conbuildmat.2005.08.006.
[37] Robert McNeel & Associates, “Rhino3D®and Grasshopper3D®.” 2022. [Online]. Available:
https://www.rhino3d.com
[38] N. C. Brown, V. Jusiega, and C. T. Mueller, “Implementing data-driven parametric building design
with a flexible toolbox,” Automation in Construction, vol. 118, p. 103252, Oct. 2020, doi:
10.1016/j.autcon.2020.103252.
[39] C. Preisinger and M. Heimrath, “Karamba—A Toolkit for Parametric Structural Design,” Structural
Engineering International, vol. 24, no. 2, pp. 217–221, May 2014, doi:
10.2749/101686614X13830790993483.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
30
[40] A. Curth, E. Gascon, L. Sass, L. Norford, and C. T. Mueller, “Additive Energy: 3D printing
thermally performative building elements with low carbon earthen materials.” Research Gate, Sep.
2023. doi: 10.13140/RG.2.2.14662.65600.
[41] D. Ahlers, F. Wasserfall, N. Hendrich, and J. Zhang, 3D Printing of Nonplanar Layers for Smooth
Surface Generation. 2019. doi: 10.13140/RG.2.2.34888.26881.
[42] “Chapter 19: Concrete,” in California Building Code, Title 24., vol. 1 & 2, in California Code of
Regulations, vol. 1 & 2. , International Code Council, Inc., 2022. Accessed: Apr. 17, 2024. [Online].
Available: https://codes.iccsafe.org/content/CABC2022P1/chapter-19-concrete
[43] N. Hack and W. V. Lauer, “Mesh-Mould: Robotically Fabricated Spatial Meshes as Reinforced
Concrete Formwork,” Architectural Design, vol. 84, no. 3, pp. 44–53, May 2014, doi: 10/gg55fz.
[44] W. Niebrzydowski, “From ‘As Found’ to Bush-Hammered Concrete – Material and Texture in
Brutalist Architecture,” IOP Conference Series: Materials Science and Engineering, vol. 471, p.
072016, Feb. 2019, doi: 10.1088/1757-899X/471/7/072016.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4874925
Preprint not peer reviewed
