Simultaneously improving fabrication accuracy and interfacial bonding strength of multi-material projection stereolithography by multi-step exposure

ABSTRACT Multi-material additive manufacturing (MMAM) takes full advantage of the ability to arbitrarily place materials of additive manufacturing technology, enabling immense design freedom and functional print capabilities. Among MMAM technologies, projection stereolithography (PSL) exhibits a great balance of high resolution and fast printing speed. However, fabrication accuracy of multi-material PSL is hindered by large overcure used to strengthen interfacial bonding weakened by chemical affinity and material-exchange process. We present a novel multi-step exposure method for multi-material PSL process to overcome this shortcoming. Firstly, the whole layer is moderately exposed producing overcure of single-material PSL level to generate geometries. Then weakened interfaces are strengthened individually with additional steps of exposure. The multi-step exposure is integrated into the already efficient materials printing order of multi-material PSL process. Curing depth and overcure of photocurable resins are modeled and characterized. Exposure required to achieve sufficient interfacial bonding of single-material interfaces built through material-exchange process and multi-material interfaces with altering materials printing order is determined with tensile tests. Microfluidic channels are used to compare fabrication accuracy of traditional single-step exposure and our multi-step exposure method. This method can be widely applied in multi-material PSL to improve fabrication accuracy in a variety of applications including microfluidic devices. Graphical abstract


Introduction
Multi-material additive manufacturing (MMAM) drastically extends the design space and functionality of manufactured parts by being able to place materials arbitrarily [1].This capability is essential for fabricating smart structures with functionalities like self-sensing [2] and shape morphing [3].Among MMAM processes, material extrusion [4], material jetting [5], and vat photopolymerization [6] have demonstrated most manufacturing and application versatility so far.Multi-material vat polymerization achieves higher resolution and more complex geometry across multiple scales compared to the other two processes [7].These advantages are mainly enabled by projection stereolithography (PSL) process, which projects 2D pattern to selectively cure liquid photocurable resin [8].Being able to cure a full layer with millions of pixels, PSL offers a great balance between resolution and speed compared to other vat photopolymerization processes including stereolithography (SLA) and two-photon polymerization (TPP) [9].As a result, most multi-material vat photopolymerization processes are PSL based [10][11][12].Enabled with these strengths, multi-material PSL is applied in various applications, such as 4D printing [13,14], metamaterials [15,16], and sensors [17,18].
Interfacial bonding is a critical challenge for MMAM processes, as premature failure may occur in multi-material interfaces caused by lower chemical affinity [19].Multi-material interfaces should be stronger than the weaker material to not compromise the strength of the multi-material structure.Interfacial bonding strength is found to be affected by materials combination, materials printing order, process parameters, and printing orientation in multi-material material jetting [20][21][22] and material extrusion [23][24][25] processes.However, studies of multimaterial interfaces of multi-material PSL are still over simplified.Most studies simply concluded that multi-material interfaces were stronger than the weaker material [26][27][28].In the meantime, overcure used in multi-material PSL is usually very high.Overcure is depth cured beyond required layer thickness per layer [29].Overcure of around 10 times layer thickness [26] and 3 times layer thickness [30] can be found in multi-material PSL processes, in contrast to overcure of 10%-35% layer thickness typically used in single-material stereolithography processes [31,32].Although overcure is necessary to bond layers together [33], large overcure will reduce fabrication accuracy both in the vertical and lateral direction [34] and block internal voids [35].As a result, fabrication accuracy in multi-material PSL process is hindered by high overcure, and little is known about bonding strength of multimaterial interfaces at lower overcure.Influence of materials printing order, which proved to be relevant in material extrusion process [25], has also not been explored.Different from multi-material material jetting and material extrusion processes, multi-material interfaces and part of single-material interfaces in multimaterial PSL process are formed through material-exchange process involving cleaning of the interfaces before bonding.Material-exchange process like rinsing or sonication in organic solvent may also lead to reduced interfacial bonding strength by removing oxygen-inhibited tacky layer surface [36].Interfaces composed of the same material built during material-exchange process could also be weakened and should be studied.
In this work, we presented a novel method to selectively add extra overcure to weakened multi-material and single-material interfaces in multi-material PSL to achieve sufficient interfacial bonding, while improving fabrication accuracy by minimizing overcure outside these interfaces to the level found in single-material stereolithography process.Multi-step exposure [37] was utilized and modified to deliver targeted exposure at different areas with minimal added printing time.Influence of overcure on weakened interfaces and materials printing order were determined by tensile tests.Different overcure necessary to produce sufficient interfacial bonding in different types of interfaces was determined.Fabrication accuracy of traditional single-step exposure and our multistep exposure method under these parameters was compared by printing a series of microfluidic channels.

Multi-material PSL system
A customized multi-material PSL system was used.Specific components of this system were described in our previous work [38].A 405 nm UV projector with resolution of 1920 × 1080 was used to project pixels of 6.04 μm with optics, resulting in a 11.6 × 6.5 mm 2 projection plane.Oxygen-permeable membrane Teflon AF2400X was used as the bottom window for vats to generate dead zone to promote separation of cured resin and vat bottom [39].

Multi-material printing
Modeling and slicing of multi-material model were also described in our previous work [38].A customized Qt C++ GUI program was used to control multi-step exposure and multi-material printing.Open-Source Computer Vision Library (OpenCV) was used in the program to calculate intersection of slice images to determine interfaces.In this way, interface with arbitrary geometries is compatible with our multi-step exposure method.Light intensity at projection plane was set to 1.3 mW cm −2 by setting LED current of the UV projector.Time duration for sonication and drying was set to 15 s and 90 s.

Multi-step exposure method for multi-material PSL
The general printing process of multi-material PSL is shown in Figure 1(a).Compared to single-material PSL, material-exchange process is added in multi-material PSL.It is used to control material cross-contamination by removing residue liquid resin on print platform and parts being printed.Multiple variations of material-exchange methods are available for multi-material PSL, including sonication in organic solvents [10], rinsing in organic solvents [15], blowing with air jet [26], dynamic flow control [30], and centrifugal removal [12].Among these material-exchange methods, we selected sonication in our custombuilt multi-material PSL system.Its main benefit is the capability to print multi-material parts with extreme low material cross-contamination, which is critical to preserve resin properties such as water solubility [38].After sonification, a blower fan is used to dry parts and the print platform by volatilizing solvent.Alcohol was used as organic solvent in this work because it has good solubility of photocurable resins, high volatility, and low toxicity.After the material-exchange process, vat position is altered to print the next material.
Although material-exchange is necessary for multi-material PSL process, it can weaken interfacial bonding.The mechanism could be the process removed oxygen-inhibited tacky layer surface required to form strong interfaces by chemical cross-linking [36].Not all interfaces in multi-material PSL are subjected to material-exchange process though: (1) No material-exchange process takes place when multiple adjacent layers are composed of the same material, mimicking single-material PSL process; (2) Process planning is commonly used in multi-material PSL process, resulting in a 'A n-1 B n-1 B n A n ' like printing order, in which n represents layer number [10].When one layer is composed of multiple materials, material-exchange process within that layer is inevitable.However, the 'A n-1 B n-1 B n A n ' printing order can effectively cut most inter-layer material-exchange processes by matching the last material printed in layer n-1 and the first material in the layer n.For instance, B n-1 B n is cut free from material-exchange process with the 'A n-1 B n-1 B n A n ' printing order.Nevertheless, process planning cannot erase all inter-layer material exchange because not all layers contain both materials.As an illustration, inter-layer material exchange cannot be avoided in B n A n+1 interface or A n-1 B n interface in a 'A n-1 A n B n A n+1 ' materials combination.Also, single-material interface weakened by material-exchange process may exist, e.g. a A n-1 A n interface in 'A n-1 B n-1 B n A n ' printing order if area of A n-1 A n is larger than zero.(3) All of multi-material interfaces go through material-exchange process to control material cross-contamination.Interfacial bonding of multi-material interfaces is further weakened by lower chemical affinity compared to single-material interfaces.
To compensate for interfaces weakened by material-exchange process and low chemical affinity, large overcure is usually used in multi-material PSL process to ensure reliable interfacial bonding.Fabrication accuracy is reduced in the entire part since large overcure is applied to the whole layer with the traditional single-step exposure method.As shown in Figure 1(b), if an insufficient t 1 exposure time was applied with single-step exposure, weakened interfaces would show poor bonding and break in or after printing.If a sufficient exposure time t 1 + t 2 was applied to whole layer with singlestep exposure, as is usually the case with multi-material PSL process, unnecessary large overcure would take place outside weakened interfaces.These areas include noninterfacial areas and interfacial areas formed without material-exchange process.They are formed under the same condition of non-interfacial areas and single-material interfaces in single-material process, so they do not require large overcure.On the contrary, unnecessary large overcure in these areas will reduce fabrication accuracy and block internal voids.We adapted a different approach to mitigate this problem by switching from single-step exposure to multi-step exposure.Our multi-step exposure method modified for multi-material PSL firstly cures the entire slice with an exposure time of t 1 , which matches exposure used in single-material PSL process to ensure small overcure.
Then multi-material interfaces and single-material interfaces formed during material exchange are found by cross-checking multi-material printing order and intersections of slice images.An additional t 2 exposure time is applied within these interfacial areas to ensure adequate interfacial bonding by increasing exposure selectively.This additional exposure time varies with material combinations and materials printing order, and it is determined later with mechanical testing.A two-step exposure is used as example here, but a maximum step number of three is possible, as will be discussed later.Fabrication accuracy can be improved outside these weakened interfacial areas, which usually composes of a large proportion of the printed part.By applying this multi-step exposure method, interfacial bonding and fabrication accuracy can be simultaneously improved for multi-material PSL process.
Specific details of the multi-step exposure method are shown in flow chart in Figure 2. The 'A n-1 B n-1 B n A n ' like printing order in multi-material PSL process is unaltered to maintain process efficiency.Instead, single-step exposure of each material is substituted with multistep exposure, adding minimal extra printing time by only including a few more exposure steps.Assume material X (A or B) in layer n is being printed.The slice of X n is exposed to get geometries printed and interfaces to layer n − 1 connected at first.Exposure time follows process parameter set for material X corresponding to a small overcure in the range used in single-material PSL.Then, interfaces going through material exchange are calculated and exposed again at the order of A n-1 X n and B n-1 X n .The combinations can be A n-1 A n , B n-1 A n or A n-1 B n , B n-1 B n depending on the material X being printed.If A or B is not found in layer n − 1, the possibility of existence of corresponding interfaces can be ruled out.If such combination does exist, then whether it went through material-exchange process needs to be determined.As discussed before, multi-material interfaces A n-1 B n and B n-1 A n will always go through material exchange.So if such materials combination exists in the materials printing order, it should be strengthened with an extra exposure step.Whether single-material interfaces A n-1 A n and B n-1 B n went through material exchange is determined by determining whether A n-1 , A n or B n-1 , B n is adjacent in materials printing order.This is done by recording each material as last (printed) material when it is finished, and comparing material X to material being printed before X.If A n-1 , A n or B n-1 , B n was adjacent, X would be the same material as last material, ruling out the need to strengthen that interface.If not, that single-material interface also requires further exposure step because it has gone through material-exchange process and has been weakened.When combination of A n-1 X n or B n-1 X n fits before-mentioned criteria, actual interfacial areas may still not exist as slices of these two materials do not overlap.Intersection of slices of material A/B in layer n − 1 and material X in layer n and area of that intersection is calculated with image processing.If area of interface A n-1 X n or B n-1 X n is non-zero, that intersection contains interfaces require higher overcure and is exposed for time according to process parameter set for that specific material combination.
A detailed example of multi-step exposure is shown in Figure 3. Assume material printing order is 'A n-1 B n-1 B n A n .'Calculation results of different types of interfacial areas through image processing is shown in Figure 3(a).Whether a certain type of interface (A n-1 B n , B n-1 B n , A n-1 A n , and B n-1 A n ) is selectively overcured is depended on conditions listed in Figure 2. In this exact example, B n-1 B n is not overcured because it is not built through material-exchange process.B n-1 A n is not overcured because its area is zero.Then a twostep exposure takes place as shown in Figure 3(b).A three-step exposure would take place when curing material A if area of B n-1 A n is non-zero.It is worth mentioning that this multistep exposure method is not limited to print two materials.It can be easily expanded to scenarios in which three or more materials are printed within the same part following the same principle.

Modeling of overcure
To produce suitable overcure for different interfacial and non-interfacial areas, relation between cure depth, exposure time and light intensity should be determined.The curing characteristics of photocurable resin can be described with empirical equation called working curve equation [40]: where C d is the curing depth, E total is the total dose projected at resin surface, E c is the critical dose required to cure the resin material, D p is penetration depth of the resin material.Both E c and D p are intrinsic to the resin material and can be calculated from experimental measurements by fitting the working curve equation [30].In multi-step exposure, E total can be calculated as linear addition of exposure in each step: where n is the step number of multi-step exposure, as discussed with Figure 2, H m is light intensity used in mth exposure step and t m is exposure time used in mth exposure step.Single-step exposure can be seen as a special case of multi-step exposure when n = 1: Another important process parameter beside E total is the layer thickness h.If C d < h, print would fail as layers are not interconnected.If C d = h, the part usually appears too weak and too soft as resin is just beyond gel point at this exposure.In most of the time an exposure at which C d > h is used to get better mechanical property and interfacial bonding.But a C d which is too high brings large overcure and severely downgrades fabrication accuracy.Overcure can be calculated as follows.
where OC is the overcure at given process parameters.It can be concluded that high total exposure dose E total corresponds to higher overcure OC, and OC grows logarithmically.As shown in Equation (2) and Equation ( 4), single-step and multi-step exposure produce the same C d and OC if E total and h is the same.Thus, E c and D p fitted with single-step exposure is applicable to multi-step with an arbitrary step number n.Two kinds of photocurable resin commonly in stereolithography, TMPTA and PEGDA, are used in this work for the validation of the multi-step exposure method.Figure 4 shows results of C d tested by single-step exposure with a variety of E total of the two materials.By fitting these data, E c of TMPTA and PEGDA resin were 1.14 mW cm −2 and 0.421 mW cm −2 , and D p of these two resins were 53.1 μm and 38.4 μm.Layer thickness in this work was set as 100 μm.By determining these parameters, C d and OC of these two resins at different E total can be calculated with Equation (1) and Equation (4).
In our software implantation, different steps of exposure are set to the same exposure intensity, so exposure time along is used to regulate exposure between different steps.This implantation simplifies variation of total exposure in each type of areas by changing total exposure time along: where H 0 is the light intensity used across steps, H 0 was set to 1.3 mW cm −2 in this work, t total is the total exposure time of each type of area.Cure depth and overcure under a set of total exposure times is calculated by Equation ( 1), (4), and (5) and listed in Table 1.To achieve high fabrication accuracy out of the weakened interfacial areas, a low overcure around 20% of the layer thickness was selected for both materials.This overcure is at the same level as single-material PSL process.A total exposure time of 8 s roughly suffice an overcure of 20% layer thickness for both resins, so 8 s was selected as base exposure in the first exposure step for both resins.

Characterization of interfacial bonding strength
To obtain sufficient exposure required to form strong interfaces, uniaxial tensile test was used to compare strength of weakened interfaces with base materials.As there is no current test standard for material interfaces of additive manufacturing, dog-bone shape tensile specimen shape was adapted from ISO 527-2 standard.Geometries of the tested specimen was adapted according to the small type 1BB in the standard, as shown in Figure S1(a).Interface formed during material exchange was designated to the center as same as previous works [12,14,26,41].Different from previous studies, materials printing order and single-material interfaces weaken by material-exchange process were taken into consideration.Specific types of tensile specimens used in this work are shown in Figure 5(a).PE stands for PEGDA resin specimen and TM stands for TMPTA resin specimen.
Both PE and TM were printed without material exchange, so they stand for tensile strength of the base materials under certain process parameters and were tested only at 8 s total exposure time.Uniaxial tensile test was done at room temperature with INSTRON 5944 machine at 0.2 mm min −1 speed.Clamping of test specimens is shown in Figure S1(b).PE and TP specimens yielded tensile strength of 0.471 MPa and 3.240 MPa, putting PEGDA the weaker one of these two resins.Four different types of possible weakened interfaces, PEGDA-TMPTA (PT, PEGDA was printed first), TMPTA-PEGDA (TP), PEGDA-PEGDA (PP), TMPTA-TMPTA (TT) are presented in four kinds of specimens.
Although PP and TT specimens do not contain multi-material interface, one time of material-exchange process was forced in the middle of the specimen to test influence of material exchange on strength of single-material interface.Total exposure time of 8 s, 11 s, and 14 s were tested for these weakened interfaces.As multi-step exposure method in this work only adds extra exposure steps to weakened interfaces, the rest of these specimens was also printed with 8 s total exposure time in a single exposure step.As a result, mechanical and other material properties are unaltered in these areas.This capability can be useful if local material properties, such as modus and glass transition temperature, are of interest [42].
Test results of PT and TP specimens are shown in Figure 5(b).Tensile test was unperformable at the base 8 s exposure for PT specimen, because either print of specimens were failed due to debonding of the multi-material interface, or the multi-material interface failed in transportation despite gentle treatment.On the contrary, TP specimen at 8 s exposure showed strong interfacial bonding, yielding a tensile strength of 0.468 MPa (99.4% of PEGDA resin).Instead of failing at the multi-material interface, they all failed at the weaker PEGDA material.Different rupture positions of specimens are shown in Figure S2.This indicates that materials printing order, which was ignored by previous studies, plays a key role in interfacial bonding of multi-material PSL process.We believe this asymmetry in behavior of PT and TP specimens was introduced by asymmetry of materials processing in multi-material PSL process.In PT specimen the PEGDA side of multi-material interface was processed in sonication because it was printed first, while TMPTA side of TP specimen was processed in sonication.These two resins exhibited different sensitivity to material exchange, which will be further proved by test results later.To strengthen the weakened PT interface, higher interface exposure time than 8 s should be used.PT specimen was printed successfully at interface exposure time of 11 s, although 100% specimens in this group still failed at multi-material interface.A tensile strength of 0.309 MPa, which was still short of PEGDA resin (65.6%), was produced.PT specimen of 14 s interface exposure time yielded a tensile strength of 0.464 MPa, which is at the same level (98.5%) of PEGDA resin.Despite having good tensile strength, 60% of specimen in this group still failed at multi-material interface.This could be explained with stress concentration led by an abrupt change in elastic modus in the multi-material interface [43].Also added exposure affected rupture pattern of weakened interfaces, as shown in Figure S3.This can explain the added tensile strength for specimens still failed at weakened interfaces.It can be concluded that an interface exposure time of 14 s is sufficient to strengthen PT interfaces, and no additional exposure is required for TP interfaces.
Test results of PP specimen is shown in Figure 5(c).With a base interface exposure time of 8 s, tensile strength was reduced to 0.225 MPa (47.8% of PEGDA resin) with an interfacial failure rate of 100%.This proves that material exchange can indeed make single-material interfaces significantly weaker.An interface exposure time of 11 s yielded a tensile strength of 0.347 MPa (73.7% of PEGDA resin) and an interfacial failure rate of 80%.When interface exposure time reaches 14 s, the specimens produced a tensile strength of 0.455 MPa, which is near the level of PEGDA resin (96.6%).Still 60% of specimens in this group failed at the interface.Although PP interface is not composed of two materials with different modus, this behavior can still be attributed to stress concentration caused by contrasting modus.It is known in PSL that modus of cured structure varies with exposure [44], which is exactly the case in the center of the specimen because the weakened interface received additional exposure compared to adjacent layers.Therefore, an interface exposure time of 14 s was required to strengthen the PP interfaces weakened by material-exchange process.Figure 5(d) shows test results of TT specimens.An interface exposure time of 8 s yielded a tensile strength of 2.14 MPa (66.0% of TMPTA resin).An interfacial failure rate of 100% indicates the single-material interface was surely weakened like the one in PP specimens.However, an interface exposure time of 11 s yielded a tensile strength of 3.13 MPa (96.6% of TMPTA resin) and an interfacial failure rate of 80%, which showed faster recovery than PT and PP specimens.It can be concluded that TMPTA resin displayed lower sensitivity to material exchange compared to PEGDA resin.An interface exposure time of 14 s did not improve interfacial bonding much further with a tensile strength of 3.24 MPa (100% of TMPTA resin) and interfacial failure rate of 40%.Therefore, an interface exposure time of 11 s was enough to strengthen TT interfaces went through material exchange.

Comparison of fabrication accuracy of multi-step exposure and single-step exposure method
Interface exposure time required to strengthen different types of weakened interfaces built through material-exchange was concluded with previous testing.Interface exposure time sufficient to strengthen weakened interface is listed in Table 2.Additional exposure time for each type of interface can be calculated and used as process parameters for multi-step exposure method.To compare fabrication accuracy of single-step and multistep exposure, exposure time selected for each resin in single-step exposure should be minimal while producing interfacial bonding strength at least equal to multi-step exposure.Then structures printed with these two methods should have similar strength because strength in fabricated structure is determined by the weakest part of it, namely the weakened interfaces in single-step exposure.Therefore, exposure time of 14 s was selected for both PEGDA and TMPTA resin in single-step exposure.It can be concluded from Table 1 that overcure outside of weakened interfaces was reduced from 44-47% to 17-23%.Higher fabrication accuracy can be expected from multi-step exposure method in these areas.
Vertical and lateral fabrication accuracy of single-step and multi-step exposure were compared by flow channels used in microfluidic devices [35].Channels fabricated to compare vertical fabrication accuracy of single-step and multi-step exposure method are shown in Figure 6.Channels of four different types and varied height of 100, 200, and 300 μm were printed with both methods.Types of these channel are determined their position, including in PEGDA resin, in TMPTA resin, at PT interface, and at TP interface.Sufficient width of 500 μm was used to ensure these channels are not blocked laterally.Vertical fabrication accuracy of channels within a single material is determined by overcure used for that material.For channels within PEGDA resin, channel with height of 100 μm was blocked in both exposure method.However, multi-step exposure method produced significant less dimensional error in channels with height of 200 μm and 300 μm due to reduced overcure, as shown in Table 3. Actual fabricated height of these channels was measured with a Keyence VHX-900F digital microscope.Multi-step exposure also achieved lower minimal channel height for channels in TMPTA resin as 200 μm channel was blocked with singlestep exposure method but open with multi-step exposure.Similar reduction of overcure and error also appeared in channel with height of 300 μm.For channels located at multi-material interface, fabrication accuracy is determined by overcure used for material at the top of the channel.Multi-step exposure is also effective to reduce overcure for this type of feature as its top is not selectively overcured as  a non-interfacial area.Instead, interfacial areas outside these channels were selectively overcured to produce good bonding strength.While multi-step exposure did not produce lower minimal channel height, error is reduced across the board like in single-material groups.This is also contributed by reduced overcure of multi-step exposure method.
Microfluidic channels were fabricated to compare lateral fabrication accuracy are shown in Figure 7.According to minimal channel height test with previous section, channel height is set as 300 μm for channels in TMPTA resin and 200 μm for channels in PEGDA resin.Channels within TMPTA resin were printed with a set of channel width of 100-200 μm at the interval of 10 μm, as shown in Figure 7(a).Because these channels are fully embedded in the inner material, fabrication accuracy of them is determined by exposure time of the inner material.Exposure time of both PEGDA and TMPTA resin outside of weakened interfacial areas was reduced from 14 s to 8 s with multi-step exposure method while maintaining interfacial bonding strength.Channels with design width less than 150 μm were blocked with large overcure produced by single-step exposure method.On the contrast, channels with designed width down to 120 μm were successfully fabricated with multi-step exposure  method.Lateral fabrication error was also reduced with multi-step exposure method, as shown in Table 4. Channels fabricated within PEGDA resin with single-step and multi-step exposure method are shown in Figure 7(b).Minimal fabricated designed channel width was down from 130 μm to 110 μm.Lateral fabrication accuracy was also significantly improved like channels fabricated within TMPTA resin.For channels within both resins, error of single-step method was reduced to the same level of multi-step method when the designed width become larger.This could be caused by larger space for the reactants to diffuse within wider channels offset the benefits of smaller overcure.However, the improvement of fabrication accuracy is still valid for these channels as multi-step exposure method still provides higher vertical accuracy.

Limitations
This multi-step exposure method for multi-material PSL process has shown its effectiveness of simultaneously improving fabrication accuracy and interfacial bonding through experiments above.However, there are still many limitations which should be covered with future research.
(1) Effects of printing orientation, which was found to be an important factor for interfacial bonding of material jetting process [20][21][22], were not covered in this work.This is mainly due to the limited printing envelop (11.6 × 6.5 mm 2 ) of our customized-built multi-material PSL printer.We were only able to fit the smallest type 1BB tensile specimen in ISO 527-2 standard vertically.Printer with a large enough printing envelop should be used to reveal pattern of interfacial bonding and fabrication accuracy in other printing orientations.(2) Fatigue performance of weakened interfaces was untouched.Abrupt change of modulus introduced by multi-step exposure may be detrimental for fatigue performance.A gradient distribution of modulus could be preferred.(3) Our research only covered limited mechanical performance under short, enclosed storage conditions.Effects of long-term storage conditions and post-processing, which are important for end-product usage, are not covered.

Conclusions
This paper presented a novel multi-step exposure method for multi-material PSL process to simultaneously improve interfacial bonding and fabrication accuracy by differentially overcure different types of interfacial and non-interfacial areas.Various types of single-material and multi-material interfaces weakened by material exchange process and chemical affinity were discussed.Multi-step exposure process to enhance interfacial bonding of these interfaces was incorporated into multi-material PSL without altering its already efficient material printing order.Two commonly used photocurable resin, PEGDA and TMPTA, were selected to test the effectiveness of the multi-step exposure method.Curing depth and overcure were modeled and characterized for PEGDA and TMPTA resin.Base exposure times for both resins were selected to produce an overcure around 20% layer thickness, which is within range used in single-material PSL process to achieve high accuracy.Tensile tests were conducted for various types of interfaces to find sufficient additional exposure time required to achieve sufficient interfacial bonding.It was also revealed from tensile tests that material printing order influenced interfacial bonding strength, and singlematerial interfaces went through material exchange were also weakened.Both phenomena were not revealed by previous studies.Fabrication accuracy of single-step and multi-step exposure method was compared under exposure parameters to achieve the same interfacial bonding strength.A set of microfluidic channels were used to benchmark these two exposure methods.Overcure was reduced from 44-47% to 17-23% layer thickness outside weakened interfacial areas with multi-step exposure method by differentially cure different types of areas.As a result, higher fabrication accuracy in both vertical and lateral directions were achieved by multi-step exposure method in microfluidic channels.In the future, this method can be widely used in multi-material PSL process to achieve better fabrication accuracy without sacrificing interfacial bonding in various types of applications including manufacturing microfluidic devices.

Figure 1 .
Figure 1.(a) Multi-material PSL process.(b) A two-step example of multi-step exposure method for multi-material PSL process.

Figure 2 .
Figure 2. Flow chart of the multi-step exposure method.

Figure 3 .
Figure 3. Multi-step exposure in layer n.(a) Calculation of interfacial areas.(b) Multi-step exposure.

Figure 5 .
Figure 5. A) Different types of tensile specimens.PE stands for PEGDA resin specimen, TM stands for TMPTA resin specimen, PT stands for specimen with PEGDA-TMPTA interface (PEGDA was printed first), TP stands for specimen with TMPTA-PEGDA interface (TMPTA was printed first), PP stands for specimen with weakened PEGDA-PEGDA interface (went through material-exchange process), TT stands for specimen with weakened TMPTA-TMPTA interface.b) Effect of interface exposure time on tensile strength of PT and TP specimens.c) effect of interface exposure time on tensile strength of PP specimens.d) effect of interface exposure time on tensile strength of TT specimens.For (a), (b), (c), and (d): N = 5.Error bars represent standard deviation.

Figure 6 .
Figure 6.Comparison of vertical accuracy between single-step exposure and multi-step exposure method.Scale bars represented 200 μm.

Figure 7 .
Figure 7.Comparison of lateral fabrication accuracy between single-step exposure and multi-step exposure method.a) microfluidic channels fabricated within TMPTA resin.b) microfluidic channels fabricated within PEGDA resin.Scale bars represented 200 μm.

Table 1 .
Effect of total exposure time on cure depth and overcure on PEGDA and TMPTA resin.

Table 2 .
Additional exposure time used for multi-step exposure.

Table 3 .
Fabricated height and error of single-step and multi-step exposure method.

Table 4 .
Fabricated width and error of single-step and multi-step exposure method.