Rheological and Mechanical Gradient Properties of Polyurethane Elastomers for 3D-Printing with Reactive Additives

Polyurethane (PU) elastomers with their broad range of strength and elasticity are ideal materials for additive manufacturing of shapes with gradients of mechanical properties. By adjusting the mixing ratio of different polyurethane reactants during 3D-printing it is possible to change the mechanical properties. However, to guarantee intraand inter-layer adhesion, it is essential to know the reaction kinetics of the polyurethane reaction, and to be able to influence the reaction speed in a wide range. In this study, the effect of adding three different catalysts and two inhibitors to the reaction of polyurethane elastomers were studied by comparing the time of crossover points between storage and loss modulus G′ and G′′ from time sweep tests of small amplitude oscillatory shear at 30∘C. The time of crossover points is reducedwith the increasing amount of catalysts, but only the reaction timewith one inhibitor is significantly delayed. The reaction time of 90% NCO group conversion calculated from the FTIR-spectrum also demonstrates the kinetics of samples with different catalysts. In addition, the relation between the conversion as determined from FTIR spectroscopy and the mechanical properties of the materials was established. Based on these results, it is possible to select optimized catalysts and inhibitors for polyurethane 3D-printing of materials with gradients of mechanical properties.


Introduction
Polyurethane (PU) elastomers with their broad range of strength and elasticity have found many applications in the past few decades, such as resins for automotive, medical, and aerospace industry. On the other hand, additive manufacturing is one of the fastest developing technologies and is challenging traditional manufacturing. In recent years, many different techniques of additive manufacturing have been developed for different polymers. For example, ABS, PLA and PC are commonly used materials for fused deposition modeling (FDM), while PA and PEEK are typical polymers for selective laser sintering (SLS) [1]. However, there are few researches on additive manufacturing of polyurethane based elastomers because only very few commercial devices can print polyurethane.
Elsner [2] has set up an inkjet printing prototype for additive manufacturing with piezoelectric nozzles, which can precisely adjust the ratio of two or more pre-polymer components by ejecting tiny amount of liquid. Müller et al. [3] studied the printability of polyisocyanate and polyols with this 3D-printing prototype. They demonstrated that by adjusting the mixing ratio of the polyurethane reactants it is possible to change the mechanical properties of each voxel, and established a technique of printing gradient materials. Uhlmann et al. [4,5] investigated the printability of polyurethane further by optimizing the printing parameters, and managed to print small polyurethane cylinders by use of this prototype 3D-printing device (Figure 1).
Polyurethane materials are mainly obtained from a polyaddition reaction between di-isocyanates and polyols [6][7][8]. The basic reaction equation is shown as follows: In this reaction, the hydroxyl groups of polyol lose a hydro-Eckart Uhlmann, Georg Gerlitzky: Department of Machine Tools and Factory Management, Technische Universität (TU) Berlin, Pascalstraße 8-9, 10587 Berlin, Germany Figure 1: Prototype of 3D-printer and specimen of a 3D-printed test shape gen atom and form an oxygen anion to connect with the carbon of isocyanate groups. The main effect of catalysts in this reaction is to help hydroxyl groups lose hydrogen ions. Common catalysts are amine-based 1,4-Di-Aza-Bi-Cyclo-(2,2,2)-Octane (DABCO), and tin-based Di-Butyl-Tin-Di-Laurate (DBTDL) [9]. Many studies [10][11][12][13][14][15][16] have already investigated the effects of these two types of catalysts onto the polyurethane formation reaction. The catalytic mechanism of DABCO is shown as follows for example: The function of an inhibitor is opposite to the action of catalysts, relying mostly on Lewis acids to hinder the proton transfer to the isocyanate groups. Commonly used inhibitors include Benzoyl-chloride and para-Toluene Sulfonic Acid (p-TSA) [17,18], and these materials play an essential role in the pre-polymer preparation of polyurethane. However, it has also been noticed that the reaction can be accelerated by adding an excess of strong organic acids [19].
For the printing of polyurethanes, it is important to know the reaction time when the soft material is able to maintain a stable shape during the crosslinking reaction, because it determines when the next layer can be printed on the previous layer. Therefore, rheological characterization is an essential technique. By measuring the crossover point of storage modulus G ′ and loss modulus G ′′ with constant amplitude and oscillatory frequency, this time can be determined and guaranteed precisely. Winter and Chambon et al. [20][21][22][23][24][25] have made great efforts in characterizing the gel point by rheological measurements for crosslinking polymers, including polyurethane. Moreover, Hu et al. [26] applied this method to bismaleimide/cyanate ester. They demonstrated that this method is practicable to compare the kinetics of a crosslinking reaction.
To characterize the overall polyaddition process of polyurethane, many studies have applied Fouriertransform infrared (FTIR) spectroscopy [27][28][29][30]. Particularly with the help of in-situ FTIR, one can observe the delay of the peak of the isocyanate group, and by comparison of isocyanate peak area and reference peak area, the conversion a of the NCO group can be calculated as shown in the following equation [31,32]: A 1 (0) and A 1 (t) are the peak areas of the isocyanate group at the beginning of the reaction and at time t, respectively; A 2 (0) and A 2 (t) are the reference peak areas at the beginning and at time t, respectively. In this work, we investigate the effects of various reactive additives to the polyaddition reaction of polyurethane elastomers. For 3D-printing of reactive polymers like such polyurethanes, it is essential to be able to adjust the reaction rate, because the possible printing speed, especially in the vertical z-direction, relies on the reaction time until high viscous polymer is formed. By conducting timedependent measurements with small-amplitude oscillatory shear (SAOS), it is possible to evaluate a processing time window from the dynamic mechanical response. Furthermore, the reaction process can be monitored also from conversion of chemical groups by FTIR spectroscopy. In addition, the mechanical properties of polyurethanes at different conversion stages are reported.

Materials
Two polyols (Lupranol L1100 and Lupranol L3300, BASF Polyurethanes GmbH) and one polyisocyanate (Lupranat M20S, BASF Polyurethanes GmbH) were used as the main reactants and diluted by anhydrous toluene (Sigma-Aldrich). Lupranol L1100 is a difunctional polyol, which is used for production of foams and elastomers, and Lupranol L3300 is a trifunctional polyol based on glycerine, which is used for the production of rigid foams, but also for non-foamed polyurethane. Lupranat M20S is a solventfree polyisocyanate, containing oligomers and isomers, which is mainly used for high-density rigid foams as well as for packing and casting materials. Lupranat M20W was used in a previous study [3], but is not commercially available any more, and therefore, Lupranat M20S is used as replacement with comparable properties. Table 1 presents the typical properties of the three main reactants.
The molar ratio (MR) between the NCO group and the OH group as well as the polyol ratio (PR) has a significant effect on the reaction rate, and MR and PR are defined as where α NCO and α OH are number of moles of NCO and OH, respectively; where m L1100 and m L3300 are mass of L1100 and L3300, respectively.
In this work, the molar ratio was chosen to be 105, which is higher than 100, in order to release the rich of an excess of isocyanate to react with water in air, forming allophonates [3]. From the results of tests conducted by time sweep tests in the rheometer for different polyol ratios, PR with 0.33 was chosen as the main polyol ratio in order to compare the effects of different reactive additives. All the reactants and tools were thoroughly dried in advance in a vacuum oven at a temperature of 50 ∘ C for 24 hours before preparation of samples. All compositions of reacting compounds investigated are shown in Table 2.
To prepare polyurethane reaction compounds, which can be printed with piezoelectric nozzles, Lupranol L1100, Lupranol L3300 and Lupranat M20S were firstly diluted by Toluene of 30 wt.%, 32 wt.% and 25 wt.%, respectively [3]. Then Lupranol L1100 and Lupranol L3300 were dropped in a neat beaker according to the mass fraction intended, and finally mixed. For rheological measurements, Lupranat M20S was poured into a disposable cup used under curing conditions. The polyol mixture was dropped into the polyisocyanate within 1 min. The reactants were stirred for 1 min and then transferred to the instrument. This is defined as the pure PU sample without catalysts or inhibitors.
For the sample preparation with additives, the catalysts were added after mixing of the three reactants, and then stirred for another 1 min before transfer to the instrument. When inhibitors were added, a certain amount of inhibitor needed to be dissolved in the polyol mixture in advance to form an acid liquid, and then it was added with the polyols to the reaction.

Rheological testing
The storage and loss modulus G ′ and G ′′ as well as the crossover time were measured by using an Anton Paar MCR-301 rheometer with a CTD-450 oven chamber.
The time sweep tests were performed by plate and plate geometry with a diameter of 25 mm at a temperature of 30 ∘ C. To investigate the appropriate strain amplitude in linear viscoelastic region, amplitude sweep tests were first conducted for a strain range between 1% and 100% at an angular frequency of 0.5 rad/s. Time sweep tests were performed in controlled strain mode with 1% strain and at a frequency of 1 Hz. Rheological measurements were stopped much later than the crossover times, and repeated at least twice for every type of sample.

Fourier-Transform Infrared (FT-IT) Spectroscopy
The FTIR measurements were conducted on Thermo Nicolet 380 spectrometer with ATR mode to follow the conversion reaction. To maintain a reaction temperature of 30 ∘ C, a small oven was placed on the top of the ATR unit. After sufficient preheating for thermal equilibrium, the reaction liquid was dropped onto the diamond window measurements repeated with a certain time interval. This method mimics an in-situ measurement.
For the FTIR measurements, the samples were prepared with similar procedures as for the rheological testing. After the samples were stirred homogeneously, they were transferred with a pipette to the ATR crystal, and measured every 2 min for the first 10 min, then every 10 min for the first hour, every 20 min for the second hour, every 30 min for the third hour, every 60 min for the following time period until the absorbance of the NCO peak is lower than a relative intensity of 0.05.

Mechanical testing
The samples for the mechanical tests were prepared in a polytetrafluoroethylene (PTFE) mold, which was designed based on the international standard ISO 527-2 ( Figure 2). After homogenously mixed, the reactants without additives were slowly casted into the mold to avoid the formation of bubbles. Then the mold with liquid reactants was shifted into a freezer with a temperature of about 5 ∘ C and kept there for 24 hours. This step slows down the crosslinking reaction and decreases the formation of bubbles. When taken out from the freezer, the samples have already become high viscous liquids and were subsequently stored in an oven at a temperature of 30 ∘ C.
When the samples had obtained a stable shape, they were demolded and kept at 30 ∘ C for further reaction. Before the tensile testing, three different points on the specimen were tested by FTIR to confirm its homogeneity in conversion (Figure 2), and the Shore A hardness was measured by a durometer. Afterwards, the samples were longitudinally stretched at a rate of 1 mm/min for the determination of the tensile modulus, and then by 100 mm/min for the remaining stress-strain diagram. The relationship between reaction conversion and corresponding mechanical properties was studied.  Figure 3 shows for PU with different PRs the complex viscosities and dynamic moduli with changes in the G ′ and G" crossover points. The curves show some characteristic regions and onset times, at which they grow with different rates. In Figure 3, the viscosities decreases with increasing amount of L1100 in the reaction, because L1100 has a low viscosity and can react with isocyanate, forming longer linear chains and networks with a low density of crosslinks. L1100 dilutes the reaction system and hinders the reaction between L3300 and isocyanate to form a 3-dimensional network structure. Therefore the sample with PR=0.66 needed the longest time to reach the crossover point. The compound with PR=0 has the highest growth rate of the three samples both for the viscosity and the moduli, because there are no diols in the reaction, which means that the resulting structure has more crosslinks than the other compounds. However, the resulting high viscosity slows down the reaction. Therefore, PR=0 needed more time to reach the crossover point than PR=0.33. By adding 33 wt.% of L1100, isocyanate forms more linear chains than PR=0 resulting in a lower viscosity, which in turn results in faster crosslinking with L3300. PR=0.33 is therefore the compound which reached the crossover point after the shortest reaction time. Since the compound with PR=0.33 has an earlier crossover point than other two compounds, it was chosen as the test compound in order to compare the effects of different catalysts and inhibitors. In the primary test for samples with different PRs, the pure reactants were used in bulk polymerization. As a result, the reaction heat was difficult to control, and the reaction heat accelerated the reaction between isocyanate and water in the air, forming a few bubbles. To decrease the viscosity for printing, the three reactants were diluted by toluene. In addition, toluene can absorb part of the heat produced by the reaction and isolates the water in the air from the reaction. Therefore, fewer bubbles were produced by the diluted reaction, and a higher G ′ /G" crossover point was obtained, but the time to reach the crossover point was extended to 612 min (black curve in Figure 4a-c) because of the decreased concentration of the reactants. Measured data of G ′ and G ′′ at the start of the reaction were not reliable due to the extremely low viscosity, and they are not shown here.

Crosslinking behaviour from time sweep tests
From the diagrams, it can be seen that all the G ′ curves have a similar shape and reach the level of 10 6 Pa, which means the crosslinking density of all samples are almost  Comparing the time to reach the crossover point in Figure 5, it is obvious that the time of samples with different amounts of catalyst L and A obey a power law, and L is more effective than A. This is because catalyst A is an amine-based catalyst, which not only improves the reaction between isocyanate and hydroxyl, but also the isocyanate-water reaction [19]. Stannous octoate (catalyst O) is a Tin-based complex compound. The results reveal that catalyst O has the lowest efficiency when a small amount (0.01 wt.%) is added, but is much more effective at higher concentrations. For the sample with 0.2 wt.% of O, the lowest crossover time is obtained.
Considering the value of G ′ and G ′′ at the crossover point, the trend is first a decrease and then a plateau for catalyst L and O, while catalyst A still follows a power law. The decrease of the modulus values at the crossover point with increasing catalyst concentration is due to the fact that the catalysts create an efficient three-dimensional network.
To delay the curing time of this polyurethane, as typical inhibitors, Benzoyl chloride (inhibitor C) and para-Toluene-Sulfonic Acid (inhibitor S), were applied and their effects on the dynamic modulus are shown in Figure 6. Benzoyl chloride reacts with polyol of the system to form HCl. From the results, it is obvious that the time to reach the crossover point for the sample with 0.1 wt.% Benzoyl chloride is almost the same as for the sample without inhibitor, which means that such a small amount has only little effect on the reaction. If the amount of inhibitor is increased to 2.5 wt.%, it can delay the reaction more significantly. However, only poor reproducible results can be obtained for this amount because the reaction between Benzoyl chloride and polyol seems difficult to control.
Para-Toluene-Sulfonic Acid can be added directly to the reaction to vary the pH value. From Figure 6, it is seen that 0.05 wt.% has little effect as the time to the crossover point is nearly the same as without inhibitor. The most effective inhibitor amount is 0.1 wt.%, which results in a time of 634 min. However, when the amount is increased to 0.2 wt.%, the inhibitor may work as a catalyst, which reduces the time strongly to 312 min [19].  Figure 7a shows the FTIR spectrum of the pure polyurethane PR=0.33 sample. The NCO group absorbance occurs between 2200-2350 cm −1 with a peak at 2280 cm −1 and as the reaction proceeds, the absorbance gradually drops. Figure 7b is the 3-dimensional spectrum transformed from Figure 7a. From this plot, it is obvious that the reaction rate of NCO is fast during the first 200 min, and then is reduced progressively because of the spatially hindering network structure and the high viscosity limiting the further reaction between functional groups. With the help of the reference peak at 2960 cm −1 for C-H stretching vibration in the CH 2 group of a polyether, the decrease of the peak area can be used to calculate the conversion of NCO group by use of equation 1 (Figure 8). From Figure 8, it can be seen that the conversion of NCO group for polyurethanes with different reactive additives increases also sharply at the beginning of the reaction and then enters a plateau, which is similar to the pure PU as shown in Figure 7b. Compared to the starting point of the pure PU sample without catalyst, the starting points of PU with catalysts all have already a higher conversion, and with increasing amount of catalyst, the conversion is higher at comparable reaction times. This is because there is a certain loading time between adding the catalyst to the reaction compounds and the start of the measurements. This indicates that the catalysts have an obvious catalytic function during this loading time.

Degree of conversion from FTIR spectroscopy
The catalysts L and O are both tin-based, but the efficiency is obviously different when compared at the same mass fractions. When 0.01 wt.% of L is added, the reaction time decreases to 815 min, which means that the reaction rate almost doubles. However, when the same amount of catalyst O is added, the reaction time is 1568 min, which is almost similar or even slightly longer than the time for the pure PU sample. This reveals that catalyst L is more effective than O at low mass fractions. Combined with the results from Figure 4b, it can be said that 0.01 wt.% of O is effective at the beginning of the reaction, but has no significant effects later during the reaction.
When the mass fraction is increased from 0.1 wt.% to 0.2 wt.%, the reaction time with addition of L is only reduced from 466 min to 341 min, which indicates that it is nearly saturated for this reaction. In contrast, when the mass fraction of catalyst O is increased to 0.1 wt.%, the reaction time decreases more than 10 times to 124 min, and when it is increased to 0.2 wt.%, the reaction time decreases almost another 10 times to 19 min. That means, for mass fractions above 0.1 wt.%, catalyst O is more effective than L. This may be explained in terms of the molecular structure, because O is sterically smaller than L, and therefore may diffuse much easier through the PU network to catalyze functional groups.
The addition of catalyst A with 0.01 wt.% results only in a small acceleration of the curing reaction, whereas with 0.1 wt.% and 0.2 wt.%, the reaction times are reduced to 136 min and 100 min, and possibly with 0.2 wt.% a satura-

Tensile properties and shore hardness
The changes of tensile properties of polyurethane elastomers are shown in Figure 9. It is found that the pure PU sample at crossover time of 612 min corresponds to a NCO group conversion of approximately 0.7. Thus, the first mechanical tests started at the conversion of 0.7. It can be seen that the elastomer with the lowest conversion 0.7 is extremely soft and weak. However, the stress-strain slope increases with higher degrees of conversion, while the strain at break has a trend to firstly increase and then to decrease. Two samples, which were post-cured for one week and two weeks, respectively, have significantly increased stresses at break (22 MPa and 36 MPa), but decreased strains at break (8% and 6%) compared to not fully reacted samples shown in Figure 9a. Figure 9b shows the increase of tensile modulus for materials with increasing conversion. The increase is only small until a conversion of 0.85 is reached, but then increases significantly. The two post-cured samples have a much higher tensile modulus of 300 MPa and 396 MPa, respectively. This may also be largely due to the evaporation of toluene, because the characteristic peak of toluene in the FTIR spectrum (Figure 9c) decreases and finally disappears with continued post curing.
The Shore hardness of polyurethanes was measured at different degrees of conversion before tensile testing. However, hardness can be reliably measured only after passing the crossover point when the elastic response is dominating, i.e. when the conversion is larger than about 0.7. It is obvious that polyurethane turns harder subsequently because more crosslinks are created when the reaction proceeds. The growth trend can be fitted to a first approxima-tion with a linear function (Figure 9d). If the function is extrapolated to full conversion of NCO, the result agrees relatively well with the Shore A hardness value of about 89 from previous work [3].

Conclusions
In this work, the effect of adding different reactive additives to the polyaddition reaction of polyurethane elastomers were studied by monitoring the crossover time of G ′ and G ′′ in time sweep tests at 30°C. The crossover time is typically reduced with an increasing amount of accelerating catalyst, whereas the reaction time of some inhibitors is found to be delayed. The reaction progress was studied by FTIR spectroscopy for the NCO group conversion. In addition, the relation between the 90% NCO conversion as determined from the FTIR spectrum and the mechanical properties of the materials was established.
Concerning acceleration of the reaction rate, all three accelerating catalysts have good catalytic effects, which rise with increasing amount of catalysts. The two tin-based catalysts L and O are more efficient than the amine-based catalyst A. For mass fractions higher than 0.2 wt.% the catalyst O is the most effective additive.
The two inhibitors C and S have only a limited effect on delaying the reaction. Only S extends the crossover time to 634 min at a mass fraction of 0.1%. It seems that the acidic environment is able to inhibit the curing reaction of polyurethane to some extent.
Based on these results, it is possible to select optimized recipes with catalysts or inhibitors for 3D-printing with polyurethanes within large limits of reaction rate. With a certain ratio of reactants and additives, the printed layers can be adjusted to the printing time with improved or guaranteed intra-layer adhesion as well as to the time interval for inter-layer adhesion. After printing several layers, the load carrying capacity of the support layers also needs to be considered. Moreover, when the printing procedure is finished, the conversion diagram can be used to determine the time for removing the printed object.
With increasing conversion, characteristic mechanical properties such as hardness and tensile modulus of the polyurethane materials increased. These data are essential to set up process parameters for 3D-printing of polyurethane shapes, especially also with gradients in mechanical properties.
Thus, it is possible to optimize the 3D-printing process with respect to processing conditions as well as end-use properties. Yet, further inhibitors and recipes should be investigated in further detail in the future.