This work focused on three polyolefin-based materials. The first material is an isotactic polypropylene produced by REPSOL, ISPLEN^® PR595C2M, which is a polypropylene random copolymer containing ∼3% ethylene with high fluidity intended for injection moulding. It is characterised by its high transparency and good organoleptic properties. Moulds are easily filled with this grade, permitting short cycle times. The resin exhibits a good balance of properties: optical transparency, stiffness, impact resistance, and organoleptic properties. This polypropylene contains a nucleating agent based on dibenzylidene sorbitol. This resin exhibits a melt flow rate (230 °C; 2.16 kg) of 45 g/10 min as evaluated by the ISO 1133 test method. This material is referred to as PPH.

We also used a low melt flow index isotactic polypropylene, namely, Molplen HP 500N from Lyndon Basell Molpen. HP500N is a homopolymer used for general-purpose injection moulding applications. It exhibits good flow and stiffness and is suitable for food contact applications. It exhibits a melt flow rate (230°C/2.16 kg) of 12 g/10 min. This material is referred to as PPL.

The third material used in this work was supplied by Lyndon Basell as Hostalen GC 7260, which is a high-density polyethylene resin used in general injection moulding applications. Hostalen GC 7260 provides good flowability and high rigidity and has been extensively used in injection moulding and other conversion processes. The manufacturers state that it is not intended for use in medical and pharmaceutical applications. It exhibits an MFI (190°C/2.16 kg) of 8.0 g/10 min. This material is referred to as PEL.

The operando X-ray scattering measurements were performed using a specially designed injection moulding system that was designed, fabricated, and tested at the Centre for Rapid and Sustainable Product Development (CDRSP). This has been described extensively in previous publications (Arioli et al., 2023; Costa et al., 2022; Costa et al., 2024). In this work, we have upgraded the injection moulding system to replace the cold runner with a single-nozzle hot runner, and as part of these changes, we relocated the injection unit to the side of the mould. The hot runner is not simply a heated cold runner, and in this work, we used a hot runner system tailored and developed by YUDO EU S.A. It consists of electrically heated components placed outside the mould due to geometrical restrictions, and it has the main function of conducting the molten plastic from the tip of the nozzle until the gate of the mould, maintaining a controlled adjustable temperature. The heated block is equipped with an inlet, a manifold, and a nozzle. The major parts of the experimental system are shown in Figure 1, which shows a schematic of the X-ray source, the mould assembly, and the small-angle X-ray scattering (SAXS) detector. The SAXS-to-sample distance was set at 6 m to ensure that the detector covers the required scattering vector range and that the zero-angle scattering can be masked by the beam stop and separated from the scattering data. This is also dependent on the beam stop size and shape and the physical dimensions of the detector. Figure 2 shows the injection moulding system mounted on the ALBA NCD-SWEET Beamline (Gonzalez et al.) in Barcelona. This shows the adjustable support for the injection unit with a free bearing support to allow the injection unit to move with the mould as it is repositioned using the translation stage on which the mould system is mounted. A part prepared using this equipment from a polypropylene, PPH, is shown in Figure 3.

The source of X-rays on the ALBA NCD-SWEET Beamline is an invacuum undulator, which is an array of permanent magnets driven by 3 GeV electrons. This provides a highly spatially and energetically collimated source. The most intense harmonic generated is at an energy of 12.4 keV corresponding to a wavelength of 1Å, delivering a flux at the sample position of 1.5·10¹² ph/s at 12.4 keV @ 150 mA (BL11 NCD). Currently, the ALBA synchrotron light source operates at 250 ma, so the current flux will be 2.4·10¹² ph/s at 12.4 keV @ 250 mA. The principal challenge in designing an operando X-ray scattering measurement system for injection moulding is the windows in the mould cavity. In this work, the mould cavity is fabricated from a standard aluminium alloy widely used for moulds, and we have prepared reasonably X-ray transparent windows using thinned areas of 0.08 mm thickness. For 12.4 keV photons, the transmission is ∼70% (Hubbell and Seltzer), and these windows are sufficiently strong to withstand the high pressure exerted by the fluid polymer during the injection phase. It could be that higher energy photons would be beneficial in increasing transmission, but, in that case, the scattered signal will be lower as the scattering factor decreases with increasing photon energy.

In this system, we can make several distinct measurements to provide feedback on the processes within the mould cavity. These measurements do not require the plastic to be transparent in the visible range. They are not restricted to measurements of the surface of the part. We now consider each of these measurement groups.

As shown in Figure 1, the X-ray beam enters the mould cavity through one of six windows in the mould cavity prepared by precision machining of the metal to leave a thinned area of a thickness of 0.08 mm. There is an equivalent window on the opposite side of the mould cavity, and the transmitted unscattered beam eventually impinges on the SAXS detector mounted more than 6 m away; the vast fraction of this path is evacuated to eliminate the air and reduce the effects of scattering by the air molecules. This zero-angle transmitted beam is not detected by the SAXS area detector as an absorbing beam stop is positioned to absorb the beam and prevent it from saturating the detector. Mounted inside the beam stop is an X-ray-sensitive photodiode that produces a signal proportional to the X-ray intensity. This signal is integrated over the same time period as the SAXS detector and can be used to obtain information about the fluid polymer in the mould cavity at the point where the X-ray beam passes through the mould cavity. The X-ray beam has a cross-section of 200 µm × 100 µm. The intensity, I(t), at time t measured at the photodiode can be written as I t = I 0   exp   − μ d , (1)

where I ₀ is the intensity of the beam with an empty cavity, µ is the linear absorption coefficient, and d is the effective sample thickness. Equation 1 can easily be rearranged to yield the value of µd. Using tabulated values of the mass absorption coefficients for C and H for various photon energies (Hubbell and Seltzer), we obtained a value for u/ρ, where ρ is the density, which leads to a value of µ of 1.138 cm⁻¹ for the 12 keV photons.

The incident X-ray beam intensity is more or less constant and depends on the current of the electrons in the “ring.” The ALBA synchrotron light source operates in the so-called “top-up” mode, in which small injections of current are made to maintain the ring current at a more or less constant value. This offers considerable advantages over the historical approach of injecting a large current, waiting for it to decay, and then refilling. The thermal load on the front end of each beamline is highly dependent on the ring current, and maintaining a near-constant current greatly increases the stability and reliability of each beamline. Data files contain the ring current at the time of the data accumulation, and the beamline includes an incident beam monitor, both of which can be used to correct for small variations in the incident beam intensity. The following experiments were performed using the second window counting from the gate, which corresponds to a distance of 30 mm.

The purpose of this measurement is to provide dynamic feedback so that the injection moulding process can be dynamically optimised. For this purpose, we need fast processes relative to the injection mould cycle, see Figure 4. In earlier work, when first testing the operando X-ray scattering measurement system (Costa et al., 2022; Costa et al., 2024), we employed a 1 s data accumulation cycle time, and Figure 5 shows a plot of the measured value of µd against time during the injection moulding cycle.

The plot shows a steady value of ∼0, which suddenly jumps to a value approaching 0.1 between successive data measurements. We can also observe a small drop in the signal when it reaches a maximum. We can also observe a drop in the value when the first pressure stage completes and the second pressure stage begins. In these experiments, the first pressure was 40 bar, and the second pressure was 30 bar. As anticipated, the “mould filling” is faster than the 1 s data cycle. With only the SAXS detector active, the fastest data cycle possible is 37 ms, which corresponds to the latency time of the SAXS detector. We repeated the moulding cycle with a data cycle time of 0.1 s, and the resultant plot is shown in Figure 6.

We can observe the transition between the empty cavity and the filling process more clearly. The step up in the value of μd now contains a few points and shows, as we might expect, that the filling of the space probed by the X-ray beam is linear in time. We note the same change in the signal prior to the step up in value of μd, as observed in Figure 6, where the limited time resolution has blurred the signal to a certain amount. This part of the signal, which we can see more clearly in Figure 6, was a puzzle at first, as it suggests that the measured signal is higher than when measured from an empty cavity. At first, a simple explanation linked the nature of the signal to the effect of the injection process on the alignment of the mould, but the effect was repeated on successive injection moulding cycles, so that explanation was discounted. We know from other available timing information that the injection process starts at the deviation of the signal from ∼0. We now see this as the consequence of the detector geometry in the beamstop, and we are working to explore if changes to the design are possible to minimize or eliminate this effect.

The latency of the SAXS detector is 37 ms, so we measured the transmission signal with a data cycle of 0.050 s. Of course, as the data cycle shortens, the noise in the data increases. We will see in a later section the influence of this reducing data cycle time on the SAXS signal. Figure 7 shows a plot of μd as a function of time for a data cycle of 0.05 s. This is broadly similar to that shown in Figure 6. The data quality, even at this short data cycle time, is good. There is a rapid step up when the plastic enters the mould cavity. We note that when comparing the curves with a 0.05 s data cycle for the two polypropylenes, there is a noticeable slowing of the rate of filling time for the polypropylene (PPL) with the lower melt flow index than that recorded for PPH.

Clearly, the use of the transmission signal is straightforward, and quantitative information is available on a time scale sufficiently fast to provide dynamic feedback to control, for example, the pressure applied to ensure complete filling. There are options to adjust the position of the sampling X-ray beam along the mould cavity length.

X-ray scattering arises from differences in electron density within the sample, and, as most industrial thermoplastics have broadly similar elemental composition, these electron density differences arise from density differences (BL11 NCD). These may arise from changes in the scale of mixing to the development of organized structures, such as occur for block copolymers and in the crystallisation of polymers to form chain-folded lamellar crystals, the morphology of which has a major impact on properties. Thus, we might expect that SAXS measurements will be most effective in terms of semi-crystalline polymers, block copolymers, and in mixtures of polymers and less effective with amorphous thermoplastics.

Figure 8 shows an SAXS pattern recorded with a data cycle time of 10 s for PPH corresponding to the time points shown graphically. At time t = 0 s, the pattern shows an elliptical nature, indicating that materials have already entered the mould cavity. In Figure 8B, we can observe a well-developed SAXS pattern typical of a semi-crystalline polymer, in which crystallisation has been driven by row nuclei, which are extended chains parallel to the flow direction in the mould. We can see two well-defined maxima on the horizontal axis that arise from stacks of chain-folded lamellar crystals. The peak position gives the long period and the repeating distance in the stack. The morphology of the chain-folded crystals is a strong determining factor of the product properties. The peak position determines the long period, which, with a knowledge of the degree of crystallinity, we can determine the crystal thickness, which is largely determined by the temperature of crystallisation. Using calibration curves determined for the same material, we can use this information to determine the temperature of the polymeric material in the mould cavity and thus determine the rate of cooling at different parts of the mould. The peak intensity gives information on the level of crystallinity. The peak height will increase until the proportion of crystalline material in the lamellar stack is 50%. After that, interpretation is more challenging. The azimuthal variation in the intensity at the same scattering angle as the maxima can be used to quantitatively determine the level of preferred orientation. Obtaining such information in a time-resolved manner gives valuable information on the properties of the product, as it allows the first material to be crystallised to be differentiated from that which takes place at a later time. Post analysis of the product using electron microscopy coupled with differential etching can reveal the complex morphologies that form and the variation in crystallinity in different parts of the product (Mitchell and Tojeira, 2016). Although such a post-analysis approach is very powerful, it cannot be performed in a time-resolved manner to give feedback to an optimisation process using digital twin technology. Of course, the SAXS data shown in Figure 8 were obtained using a data cycle time of 10 s; we now explore the level of information available with a data cycle time of 0.05 s.

Figure 9 shows the raw SAXS pattern recorded with a data cycle time of 0.05 s. We can see that there is a higher level of noise, but we can also see that there is a clear feature positioned in the same manner as that shown in Figure 8B. The SAXS detector is essentially an array of independent photo counters. Figure 10A shows a plot of the intensity as a function of pixel number for a line that passes through the centre. The noise reduction in the equivalent line is now ±10 pixels wide, in which the value plotted is the average of the 20 pixels at that horizontal position. The position of the lamellar stack peak can be seen and evaluated in the graph shown in Figure 10B. Critical to this process is that the average is taken over geometrically equivalent pixels. We note that for a typical polymer processing technology, the sample symmetry is often close to a uniaxial case, and thus, we find it useful to work in polar coordinates |Q|, the modulus of the scattering vector Q, and α, the angle between the scattering vector and the symmetry axis of the sample (Mitchell and Tojeira, 2016). The recorded scattering pattern I (|Q|, α) may be represented as a series of spherical harmonics with amplitudes I_2n(|Q|). Only the even harmonics are required, as any non-absorbing sample must exhibit an inversion centre. This process of representing the scattering as a series of spherical harmonics has the principal advantage of separating the effects of preferred orientation from the structure of the material. Indeed, the calculation of the amplitudes of the spherical harmonics naturally leads to the evaluation of the level of preferred orientation in terms of a series of orientation parameters < P₂>, <P₄>, <P₆>, <P₈>, etc. (Mitchell and Tojeira, 2016; Mitchell et al., 2005). These are closely correlated with properties. The amplitude of the spherical harmonics can be readily calculated in Equation 2. I 2 n Q _ = 4 n + 1 ∫ 0 π / 2 I Q _ , α P 2 n cos ⁡ α sin ⁡ α . d α . (2)

The orientation distribution function D(α) can be described in a similar manner using the amplitudes of a series of spherical harmonics D_2n. X-ray scattering enables the complete orientation distribution function to be obtained in comparison to the more limited information available from birefringence and spectroscopy measurements (Mitchell and Windle, 1988).

Calculating these spherical harmonics naturally involves the integration over a number of geometrically related pixels, and hence, such functions will have a much-reduced noise, enabling quantitative measurements to be made.

We describe wide-angle X-ray scattering (WAXS) data here for completeness. In our current system, it is not possible to collect the WAXS data due to the geometry of the mould system, but we are currently developing a revised system that will enable the observation and recording of WAXS data. WAXS data provide information on the crystal structure of the phases present in the product (Mitchell and Tojeira, 2016). Some polymers, such as isotactic polypropylene, exhibit several different crystal structures with different properties. In the case of iPP, these crystal structures include α, β, and γ, and rapid cooling can lead to a so-called mesomorphic phase. Each of these has very different properties. In the same way, it is possible to use the ratio of the area of the sharp peaks to the broad amorphous peaks to calculate the level of crystallinity over the complete crystallisation range. Finally, it is possible to use the azimuthal variation in intensity to evaluate the preferred orientation of the crystal planes. Each analysis procedure is quantitative and mathematically rigorous.

The largest contrast or electron density difference arises from voids. Figure 11 shows the small-angle X-ray scattering pattern of an injection moulded copolymer sample based on polyhydroxybutyarate. The horizontal streak through the zero-angle point arises from small voids that have been extended in the flow direction, which is vertical on the page. Quantitative analysis of the peak shape can yield information on the shape of the voids. In this case, the voids are highly anisotropic extended voids parallel to the flow axis. In principle, the fraction of voids can also be calculated, but this requires the intensities recorded using the SAXS detector to be placed on an absolute scale (Zhang et al., 2009), a process regrettably rarely performed.

The time-resolved measurement of the absorption factor of the material in the mould cavity is described in Section 4.3. This can be employed to evaluate the material thickness and whether this is less than or greater than the length within the mould cavity. In the event that filling of the mould cavity is confirmed by another method, the absorption measurement scan be used to evaluate the density of the material in the mould cavity.

Figure 12 shows two sequences of time-resolved SAXS patterns captured at the start of the injection cycle. Time runs from left to right. The flow direction in the mould cavity also runs from left to right. The top row marked A is for PPH, a nucleated polypropylene. The left-hand image at t = 0.00 s corresponds to an empty mould cavity. As we move to the images on the right, we can observe the emergence of a horizontal scattering feature that passes through the zero-angle position and corresponds to a row of nucleated crystals (da Silva et al., 2023), which lie normal to the flow direction. The very high level of preferred orientation is a characteristic of such crystals (da Silva et al., 2023). The last image in the sequence is for a pattern recorded after 9.50 s, and we can observe that the basic characteristics of the preceding image in terms of the anisotropy of the scattering are maintained. Typically, the first crystals form in the materials adjacent to the much cooler mould wall. Subsequent crystallisation takes place with the crystallisation front moving into the interior of the part. Clearly, in Figure 12A, the mode of crystallisation remains the same. In contrast, Figure 12B shows the crystallisation behaviour for PEH. Again, in the first few patterns starting from t = 0.00 s, we can see the emergence of the highly anisotropic horizontal feature, but as time increases, another more isotropic scattering emerges, indicating that the crystallisation mode changes as crystallisation progresses through the part. The crystal thickness of the chain-folded lamellar crystals is strongly correlated with the temperature of crystallisation. As we can determine the crystal thickness as a function of time, the crystallisation moving through the mould cavity provides a route to determining the temperature within the mould cavity rather than only in the mould wall and mould.

Of course, we are not the first to identify the value of small-angle X-ray scattering in the study of injection moulded polymer products, and here we highlight some SAXS measurements of polyethylene (Zhao et al., 2020) and those of Zhu and Edward on polypropylene (Zhu and Edward, 2004). Structure–property relationships have been established for polymers (Kantz et al., 1972). The advantage of X-ray scattering is that it is a quantitative technique. Taking preferred orientation as a specific example, as outlined in Section 4.4, we can extract all the quantitative parameters from the SAXS patterns shown in Figure 12. These procedures enable us to reconstruct the complete orientation distribution function for each pattern. We can obtain this distribution for both the crystalline and amorphous components in the case of semi-crystalline polymers. The mechanical properties of the injected part have been shown to strongly depend on the morphology of the semi-crystalline polymer developed from the processing (Kantz et al., 1972), as well as some factors that depend on the known properties of the polymer chains. We can see that the X-ray scattering measurement can give information on the strength and modulus of the product. The data corresponding to the presence of a spatially varying morphology can be used to predict the level of warpage rather than using post-processing laboratory-based data (Li et al., 2019).