Seeing the unseen—enhancing and evaluating undergraduate polarization experiments with interactive Mixed-Reality technology

Learning from hands-on experiments requires learners to interpret their concrete interactions with the setup in terms of abstract physical concepts. To facilitate conceptual learning and close the gap between abstract physical models and the haptic interaction with the pertinent experimental setup, we developed an interactive Mixed-Reality learning environment centred around an undergraduate lab experiment dealing with light polarization. The use of Smartglasses (Microsoft HoloLens II) enables real-time visualization of data measured in the setup and ensures a high degree of spatial and temporal contiguity between functional components and model-based representations. A pilot study with N = 73 undergraduate students was conducted in a pre/post design to evaluate the learning environment with respect to learning outcome and learners’ affection towards the experiment. The results show that students’ knowledge had significantly increased after working with the learning environment with a large effect size (t(72) = 8.50, p ≤ 0.001, d = 1.03), and the activities are perceived as interesting and enjoyable. This proves the effectiveness of the environment with regard to learning about polarization and opens the pathway for an extension of our approach to other topics in science education.


Introduction
Laboratory experiments play a fundamental role in physics as a discipline of science as well as in the individual development of professional knowledge and skills in science education [1,2].Lab courses can serve a variety of learning objectives, gradually preparing students for the handling of complex experimental setups and the appropriate preparation, interpretation, and presentation of measurement results [3].Especially, introductory lab courses are not only intended to support students' methodological skills, but additionally often comprise a set of iconic experiments intended to exemplify fundamental theoretical concepts of introductory physics [3,4].However, empirical findings suggest that the traditional 'cookbook' lab course is not well-suited to promote conceptual learning [4][5][6].More successful approaches promote an inquiry-based style and require learners to reflect upon their own actions during experimentation [1,7,8], but of course, they also confront learners with many additional challenges.This makes careful planning and design of the learning environments even more important [9].
As far as challenges of conceptual learning are considered, it is helpful to analyze the learning process during experimentation from a cognitive science perspective, for example, by applying concepts of cognitive load theory (CLT, [10]) and the cognitive theory of multimedia learning (CTML, [11]).

Cognitive challenges in lab experiments
To construct conceptual knowledge from the experiment, students need to be able to draw inferences from their experimental actions and the resulting data onto the underlying theoretical model in question.This is a challenging task for learners given that the theoretical model is not physically inherent to the experiment itself but is based on abstract and often mathematical concepts that must be mentally constructed by the learner.According to CTML, the cognitive effort associated with creating internal mental representations can be reduced by providing appropriate external representations through different sensory modalities following the so-called 'modality principle' [12] or emphasize connections between relevant elements of the learning material according to the 'signaling principle' [13].Furthermore, spatial proximity between related elements has been shown to be an important factor for effective learning with multiple representations according to the 'spatial contiguity principle' [13,14].

Mixed-Reality in hands-on experiments
Mixed-Reality (MR) technology provides new opportunities to put the theoretical implications provided by CTML into practice, since it allows to project additional visual model representations into the real experimental setup, which can also be fed by real experimental data.The term 'Mixed-Reality' is a rather broad description of various technological means that allow to enrich the physical world with additional digital elements [15].The key aspect here is that the digital elements are not merely passively displayed, but instead interactively coupled with the physical world they are embedded in [16].This provides learners with an immersive experience of interacting with a concrete hands-on experimental apparatus along with abstract model representations needed to link the measurement results with the pertinent physical concepts.Furthermore, the technology bears the potential to increase students' interest and motivation by including elements of gamification to the activities at hand [17].
Mixed-Reality has recently become a popular field of development and research in science education [18,19].In terms of empirical evaluation, their effectiveness is most widely discussed in terms of conceptual learning and learners' affection towards the learning activities [20].However, most of the applications discussed in the literature so far use the technology to simulate or substitute integral parts of real experimentation rather than augmenting it in an authentic way [20][21][22].This use-case can be advantageous as far as it allows to simulate experimentation activities which could not be implemented otherwise, because they might be dangerous or require expensive equipment [18].On the other hand, previous research has not yet considered the full potential of coupling MR technology to real hands-on experiments and use the real experimental data rather than merely simulating integral parts of the experiment or its underlying models.There are yet very few examples in the literature that follow this alternative approach [23,24].While the reports on these learning environments indicate the positive effects on learning outcome, the small number of available implementations along with different technological systems used makes it difficult to derive general conclusions regarding the conditions of effective usage of MR technology at this point.It is therefore necessary to extend the range of available experiments to cover all aspects and potential affordances of MR technology.One crucial aspect that has not yet been focused on in previous implementations is the possibility to not only provide passive visualizations of measurement data within a given setup but to account for continuous motoric interaction with the experimental setup itself in addition to adaptive real-time visualizations of the obtained results.This idea marks a key feature implemented in our learning environment and offers unprecedented ways of merging digital and real elements together in one application.The resulting MR learning environment closes the gap between concrete experiment and abstract model representations by allowing students to interact with both levels of representation during the experimentation process.

A new approach on MR experiments
In the remaining parts of this paper, we report on the implementation of a hands-on MR experiment dealing with concepts of light polarization for undergraduate students (figure 1).
Unlike previous examples for effective MR learning environments, our setup provides visualizations that not only passively display measurement data, but also actively react on motoric interactions with the experimental apparatus in real time.The Utilization of the setup's manipulable components as input devices for model-based visualizations provides a new level of interactivity along with an intuitive and natural way of working with the experimental apparatus.
The article is structured as follows: in section 2, we explain the physical concepts needed in the context of our experiment.We then briefly explain how different affordances of MR technology relate to practical issues during experimentation, which leads to a more detailed description of the features of our MR learning environment.Section 3 provides a report on how the concept is implemented within an undergraduate lab course along with an evaluation of learning outcomes and affective measures.We conclude by discussing ideas for expanding the range of potential experiments within this new approach.

A Mixed-Reality polarization experiment
The experiment presented in this paper is based on a typical setup for investigating effects of light polarization and its manipulation by optical components such as polarizers, half-wave plates and polarizing beam splitters.Controlling and preparing light polarization is of great importance for many optical experiments.Students need to understand what polarization is and how to manipulate it if they want to perform more advanced optics experiments on the graduate level and beyond.The different components used in the experiment allow to manipulate the polarization of a laser diode by means of selective absorption (linear polarizer), by inducing a relative phase shift due to birefringence (half-wave plate), or by exploiting the Brewster-condition of reflection (polarizing beam splitter).

Shortcomings of a classical setup
A typical setup without MR enhancement is shown in figure 2. A laser diode serves as a partially polarized light source, which is linearly polarized after passing an initial polarizer.Several additional components can be inserted into the setup before the light intensity is measured.The output voltage of a photodiode serves as an intensity estimate and can be read out by a multi-meter.
However, since it is only possible to measure (time-averaged) light intensities and not the amplitude of the electric field itself, there is no direct way of measuring the polarization vector or its direction.It can indirectly be inferred from measuring light intensities and relate them to the current experimental configuration, e.g. of polarizer angle.In the classical setup, there is no additional visual aid helping the students to draw that inference correctly, meaning that learning about the concept of polarization is, in fact, disjunct to performing the experiment.Furthermore, since the data are only represented as numerical values, there is no convenient way of visually assessing the results or compare multiple measurements to each other.The classical setup hence tempts the students to perform the experiment in a fashion of mechanical data recording instead of the active, inquiry-based experimentation desired to promote model-based conceptual learning [9,[25][26][27].It is typically expected from the learner to produce their own graphical representations of the data when writing their lab report after the conduction of the experiment, but this activity is decoupled from the active experimentation phase.
This short analysis shows that even in a rather simple experimental environment there is much potential for a better integration of representations and activities during experimentation.The next section shows how the setting can be improved by means of MR technology.

An enhanced MR experiment
Regardless of the various technological systems available for implementation, we consider the aspect of interactivity between real-world entities and virtual representations as the defining feature of an MR environment [16].We therefore designed the MR environment such that the virtual elements are fed by real data from the experiment.Adding virtual elements to the experiment is a two-edged endeavour: on the one hand, it can promote conceptual learning by offering additional information addressing multiple sensory channels in line with CTML's multimedia and modality principles [12,13], on the other hand, it bears the risk of shifting away the focus from the experiment itself as well as overwhelming the learners with too much information [28].Hence, interactions were designed to be centred around the physical interaction with the real experiment itself using the optical components as physically manipulable input devices (see figures 3 and 5).The technical platform to achieve these goals is provided by Microsoft's smartglass-system HoloLens II, which provides maximal immersion and allows the students to have their hands free to work on the experiment.Physical signals from the experiment (photodiode voltages and rotation angles of components) are processed by an Arduino microcontroller and sent to the lab-PC via USB.The lab-PC in turn is used as a server for streaming the data to the HoloLens for real-time visualizations.
By integrating additional model-based representations into the experimentation process, MR helps turning the lab experiment into a multi-representational learning environment (figure 4).These visualizations include a vector representation of incoming and transmitted polarization for each optical component, depending on the chosen angle of rotation (figure 5(b)).A linear polarizer, for example, only transmits the component of E 0 parallel to its transmission axis, which can mathematically be described and visualized as a vector projection:  This law's foundations in the vectorial nature of a light wave's amplitude is directly intelligible by looking at the changes in the visualization when turning the polarizer.The law of Malus itself is experimentally accessible by measuring the light intensity behind the setup for different polarizer angles.The second important visualization is hence a polar diagram displaying the photodiode voltage as a function of rotation angle (figure 5(c)).The photodiode's output voltage is displayed in real time.The currently displayed value can also be easily saved and added to the plot by using simple hand gestures to control a built-in measurement interface.The data is then displayed directly within the plot and also saved as a text file on the lab-PC for further analysis in the lab report.This functionality shifts the focus from gathering data to assessing and interpreting the data.
The vector representation and the intensity plot hence provide access to Malus' law and other concepts associated with polarization from the two different theoretical perspectives 'light wave's amplitude' and 'light intensity', which complement each other and allow students to explore how they are related.This ability to establish mental links between different representations (known as 'constructing representational coherence' [30][31][32]) is often seen as a necessity to form deeper understanding [30,33].Additional elements of the MR environment include textual elements such as short information on the optical components, which is available by simply pointing a finger at the component of interest, and a display of the current experimental task.Only a minimal amount of navigation within textual interfaces is necessary, and all of it can be done easily by using the intuitive finger-tapping gestures provided by the HoloLens system.
Another great affordance of the MR system is that all different representations can be arranged in spatial proximity to each other and directly within the experimental setup where the interaction takes place.This has more effects than just structuring representations intuitively: according to CTML's spatial contiguity principle [13,14], placing related representations in spatial proximity to each other can reduce the cognitive load needed to mentally integrate these representations and increase learning outcome.By using the HoloLens students can manipulate the experiment and watch the effects all in one gaze and in real-time instead of having to switch their focus between the setup itself and, for example, an additional computer screen.

The MR experiment in an undergraduate lab course
We designed the MR environment for a mandatory undergraduate lab course in experimental physics at a large German university, where it is part of the curriculum since 2021.Similar experiments without MR enhancements had traditionally been part of the obligatory experiments in that lab course, so the new setting could easily be integrated into the canon.Apart from offering the students a MR-enhanced experience during the experimentation phase itself, the general conditions were just like with any other experiment during the course.To evaluate whether the experiment is suitable to promote conceptual learning, we tested participants' knowledge on topics of light polarization and the functionality of optical components before and after doing the experiment.

Participants and ethics statement
In the summer terms of 2021 and 2022, we gathered data on N = 73 participants who performed the MR experiment.The course typically takes place in the fourth semester of the bachelor's degree.Most of the participants were physics majors (N = 58), but the course was also open for students of physics for teaching professions (N = 13) and of geophysics (N = 2).The general requirements and tasks during the experiment did not differ between these groups.Most of the students (N = 72) had never worked with an MR headset before.
All participants gave written informed consent to participate in the study.The research process was carried out in accordance with the Declaration of Helsinki.Published data do not provide any information which could be used to identify individuals.Approval of an ethics committee is not necessary for this kind of research in Germany.

Measures
We developed a set of eight quiz items aligned with the contents of the experiment to assess the participants' knowledge on basic concepts of light polarization and the functionality of the optical components used in the experiment.Figure 6 shows an example item of the quiz.In this case, the item asks for the correct vector representation of linearly polarized light when passing through a polarizer with its transmission axis oriented under a given angle.To find the correct answer, students must apply the concept of vector projection onto the transmission axis of the polarizer.To ensure content validity of the test items, they were reviewed by two experts in physics education and adjusted based on their recommendations.The same set of eight items was used in pre-and post-test, so that the learning gains during the experimentation phase can be evaluated.
Additional data was gathered regarding the participants' affection towards the activities during the four different parts of the experiment.For that purpose we used the 'Interest/ Enjoyment' subscale of the 'Intrinsic Motivation Inventory' [34] in a shortened and translated version [35].The scale consists of three items on a five-point likert scale which are coded from −2 (strongly disagree) to 2 strongly agree), higher levels of agreement indicating higher degrees of Interest/Enjoyment.

Data collection and analysis
One week before their participation in the experiment was scheduled, each student was asked to take a short online test (eight items) to assess their prior knowledge on the topic of polarization.After completion of the test, they were given access to some introductory information necessary for preparing themselves for the experiment.
The lab day itself started with a short briefing by the advisor after which the students completed the built-in HoloLens tutorial in order to get used to the MR headset.During the experimentation phase, each student was required to work with the experiments individually, having to complete a sequence of four experimentation activities which required setting up an appropriate setup and performing adequate measurements: • Determine the polarization profile of the laser diode and estimate its degree of linear polarization • Investigate the applicability of Malus' law on a set of two and three polarizers • Characterize the effect of a half-wave plate on the polarization state • Investigate the properties of a polarizing beam splitter and characterize its two output channels Each of the four activities was accompanied by a sequence of different tasks displayed by the MR system to foster the formulation of hypotheses and the reasoning on explanations of observed effects along with the quantitative aspect of gathering data.During the experimentation phase, the students were furthermore asked to fill out the Interest/Enjoyment questionnaire after each of the four parts of the experiment.An overview of the design of the study is given in figure 7.
Once all four activities were completed, the students could access the online post-test directly from the lab-PC.As usual in this kind of lab course, they were required to hand in a written lab report a week after the completion of the experiment.The task for the lab report was to characterize the functionality of the optical components based on the acquired data and to estimate their optical quality by an appropriate parameter of their choice (e.g. by calculating extinction or transmission ratios).
The empirical data obtained from the pre/post-test and the Interest/Enjoyment ratings were analyzed and visualized using R version 4.10.The evaluation of learning gains was done by performing paired t-tests on the pre-and post-test quiz scores and calculating the associated effect size (Cohen's d).Assessment of the Interest/Enjoyment ratings was done in a purely descriptive manner by plotting the respective ratings for each of the four experimentation tasks.

Results
3.4.1.Learning outcome.The results of the paired t-tests show that there is a significant increase in test scores after performing the experiments (t(72) = 8.50, p < .001)with a mean increase of 2.0 points in the post-test (figure 8(a)).The magnitude of this effect is d = 1.03, which is considered a large effect size [36].The results show that the newly developed MR experiment promotes student learning about topics of polarization during experimentation, which proves the applicability of the concept.
3.4.2.Interest/Enjoyment.Figure 8(b) shows the mean rating on Interest/Enjoyment for each part of the experiment as described in section 3.3.The total mean over all items and all parts of the experiment is M = 1.1, indicating that all in all students enjoyed working with the experiment.This is also backed up by positive student feedback regarding their user  experience when working with the MR environment, which was described as innovative and motivating to work with as well as helpful for understanding.

Limitations and perspectives for further development
Based on the finding that the current version of our MR learning environment was successfully implemented into an existing lab course and brought about positive learning results as well as encouraging feedback by the students taking part in the lab course, the concept seems worth to be developed further and used for additional research.It is important to perform further evaluations of the concept as well as comparing the MR-enhanced setup to the traditional one in order to clearly identify benefits and drawbacks of the MR concept in further studies.The current data still lacks a comparison to a traditional setup, which can be considered the main limitation of the present study.
For example, the beneficial effects of the learning environment presented in this article can be made plausible by a variety of hypotheses which refer to different design aspects and theoretical perspectives, and need to be tested in more specific research designs that allow comparison to a control group.An essential aspect that motivated the design of the MR environment is the unique opportunity to merge and align the real experiment with the abstract model visualizations, thereby facilitating the process of relating both kinds of representations to each other and drawing inferences on the interconnections between them.The active engagement with the measurement data in interpreting the graphs and values during the experiment might also plausibly have contributed to the positive results.
Finally, an explanation in terms of educational psychology might trace back the positive effects to the high degree of spatial contiguity between the different representations, as the MR environment allows to present the model visualizations very close to the experimental components.This line of reasoning opens up new opportunities for the investigation of learning with experiments under a multimedia learning perspective, which has not yet been an explicit focus of empirical research.Some first efforts have been made in that direction [23,24], but it will be necessary to develop more MR experiments covering a variety of topics to extend the basis on which empirical data can be gathered in studies, allowing to derive more context-and topic-independent conclusions on the design of MR learning environments.
With the products of our first development at hand it is straightforward to develop additional learning environments for a broader set of optical experiments in a similar fashion.A particularly interesting field of application for MR learning environments is the field of quantum optics, as the ability to make mathematical and abstract concepts visible will open up more possibilities to implement these complex experiments in an accessible and motivating way for students.To that end, we started developing a multi-user MR environment to perform quantum key distribution in the advanced physics lab course and hope to evaluate this new experiment soon.

Conclusion
In this article, we presented a newly developed Mixed-Reality learning environment for a set of conventional experiments on light polarization.While the experiments themselves are rather typical for undergraduate lab courses, Mixed-Reality technology provides an innovative way to redesign these experiments to promote conceptual learning during experimentation (figure 1).We tested and evaluated our concept with N = 73 students in an undergraduate physics lab course by testing participants' knowledge on polarization in a pre-/post-design and found a significant positive effect on students' knowledge with a large effect size of d = 1.03 (figure 8).
The positive results obtained when putting the new learning environment into practice in a beginners' lab course show that the experiment indeed fosters conceptual learning of the participants and provides learners with an interesting and enjoyable learning experience.We believe that our approach of providing an interface that reacts to direct motoric interaction with the experiment bears great potential from the practitioners' point of view as well as to stimulate further research into learning with multiple representations in hands-on experiments.Since the available data does not yet include a comparison to traditional experimental setups, the results presented in this paper should be seen as a starting point for further investigations to clear up the interplay of different representations in different sensory channels in learning from experimentation activities.

Figure 1 .
Figure 1.Working with the interactive MR experiment.Holograms are shown as seen through the MR headset.

Figure 2 .Figure 3 .
Figure 2. Typical setup used to perform polarization experiments in lab courses.

Figure 4 .
Figure 4. Virtual interactive elements displayed within the real experiment as seen through the MR headset.See figure 5 for a more detailed view of the model-based representations.

Figure 5 .
Figure 5. Interactive model representations.(a) Rotation angles of the real components serve as input for the visualizations.(b) Vector representation showing how an incoming polarization vector is projected onto the transmission axis of a linear polarizer.(c) Representation of present light intensity and previously measured values in a polar diagram.

Figure 6 .
Figure 6.Example item of the knowledge quiz.question text in German reads: 'vertically polarized light passes through the given setup.Which of the following depictions is a correct representation of the change of the electric field vector?'The correct answer is option (b).

Figure 7 .
Figure 7. Overview of the procedures and measures of the study.

Figure 8 .
Figure 8.(a) Comparison of participants' mean scores in pre and post-test.(b) Mean item rating on the Interest/Enjoyment scale for each of the four parts of the experiment as described in section 3.3.The ratings for each item were coded from −2 to +2, higher ratings indicating higher values of Interest/Enjoyment.