Analytic Framework For Students' Use of Mathematics in Upper-Division Physics

PHYSICAL REVIEW SPECIAL TOPICS - PHYSICS EDUCATION RESEARCH 9, 020119 (2013)
Analytic framework for students’ use of mathematics in upper-division physics

Bethany R. Wilcox, Marcos D. Caballero,* Daniel A. Rehn, and Steven J. Pollock
Department of Physics, University of Colorado, Boulder, Colorado 80309, USA
(Received 3 July 2013; published 13 November 2013)
Many students in upper-division physics courses struggle with the mathematically sophisticated tools
and techniques that are required for advanced physics content. We have developed an analytical frame-
work to assist instructors and researchers in characterizing students’ difficulties with specific mathemati-
cal tools when solving the long and complex problems that are characteristic of upper division. In this
paper, we present this framework, including its motivation and development. We also describe an
application of the framework to investigations of student difficulties with direct integration in electricity
and magnetism (i.e., Coulomb’s law) and approximation methods in classical mechanics (i.e., Taylor
series). These investigations provide examples of the types of difficulties encountered by advanced
physics students, as well as the utility of the framework for both researchers and instructors.
DOI: 10.1103/PhysRevSTPER.9.020119 PACS numbers: 01.40.Fk, 01.40.Ha
mathematical tools. It is often an explicit goal for advanced

I. INTRODUCTION
courses to develop students’ ability to connect mathe-
Previous research has identified a considerable number matical expressions to physics concepts. For example, con-
of students’ conceptual and mathematical difficulties, sensus learning goals for upper-division courses at the
particularly at the introductory level (see Ref. [1] for a University of Colorado Boulder (CU) [16] include
review). Substantial work has also been done to character- ‘‘Students should be able to translate a physical description
ize student problem solving in introductory physics [2]. In of an upper-division physics problem to a mathematical
addition to the significant work at the introductory level, equation necessary to solve it,’’ and ‘‘to achieve physical
researchers have recently begun to characterize students’ insight through the mathematics of a problem.’’ To improve
conceptual knowledge in more advanced physics courses student learning in advanced physics courses, we find it
[3–10]. Furthermore, a small but growing body of research necessary to move away from merely noting students’ con-
suggests that upper-division students continue to struggle ceptual difficulties towards systematically investigating
to make sense of the mathematics necessary to solve how students integrate mathematics with their conceptual
problems in physics [11–13]. knowledge to solve complex physics problems.
Upper-division physics content requires students to In order to address the issues that arise when solving
manipulate sophisticated mathematical tools (e.g., multi- physics problems that rely on sophisticated mathematical
variable integration, approximation methods, special tech- tools, we must first understand how students access and
niques for solving partial differential equations, etc.). coordinate their mathematical and conceptual resources.
Students are taught these tools in their mathematics However, canonical problems in upper-division courses are
courses and use them to solve numerous abstract mathe- often long and complex, and students’ reasoning is simi-
matical exercises. Yet, many students still struggle to apply larly long and complex. Making sense of the difficulties
mathematical tools to problems in physics. This is not that arise requires a well-articulated framework for analyz-
necessarily surprising given that physicists use mathemat- ing students’ synthesis of conceptual knowledge and
ics quite differently than mathematicians (i.e., to make mathematical tools. We use the term framework to refer
inferences about physical systems) [14,15]. However, per- to a structure of guiding principles and assumptions about
sistent mathematical difficulties can undermine attempts to the underlying relationship between a physical concept
build on prior knowledge as our physics majors advance and the mathematics necessary to describe it. At the
through the curriculum. Upper-division instructors face upper-division level, in particular, this relationship can be
significant pressure to cover new content, a task made strongly dependent on the particular concept in question,
more difficult by constantly having to review the relevant suggesting that a useful framework needs to be adaptable
to a wide variety of physical concepts and mathematical
*Current address: Department of Physics and Astronomy, tools.
Michigan State University, East Lansing, Michigan 48824, USA. We first encountered the need for such a framework
while investigating students’ understanding of approxima-
Published by the American Physical Society under the terms of
the Creative Commons Attribution 3.0 License. Further distri- tion methods (i.e., Taylor series) in a middle-division
bution of this work must maintain attribution to the author(s) and classical mechanics course [17] and with integration of
the published article’s title, journal citation, and DOI. continuous charge distributions (i.e., Coulomb’s law) in
1554-9178=13=9(2)=020119(17) 020119-1 Published by the American Physical Society

WILCOX et al. PHYS. REV. ST PHYS. EDUC. RES. 9, 020119 (2013)
an upper-division electrostatics course [18]. Our initial researcher dealing with mathematical difficulties in the
analysis focused on identifying emergent themes in stu- physics classroom.
dents’ work. We quickly identified a multitude of common It has been well documented that students do not
difficulties, but, beyond producing a laundry list of errors, approach physics problems in a manner consistent with
we struggled to organize these issues in a productive way. Redish’s model [14]. In fact, students often approach phys-
This lack of coherence made it challenging to identify ics problems in a way that seems haphazard and inefficient
relationships between the difficulties and to produce to experts [19]. Some attempts have been made to address
actionable implications for instruction or further research. this at the introductory level by explicitly teaching students
To provide a suitable organizational structure, we a problem-solving strategy that is more aligned with the
developed a framework to address students’ activation of expert approach. Wright and Williams [20] incorporated a
mathematical tools, construction of mathematical models, problem-solving strategy into their introductory physics
execution of the mathematics, and reflection on the results course that involved four steps: (1) what’s happening?,
(ACER). The ACER framework is a tool designed to aid (2) isolate the unknown, (3) substitute, and (4) evaluation
both instructors and researchers in exploring when and (WISE). The WISE strategy was designed as a heuristic
how students employ particular mathematical tools to that physics students could use to become more efficient
solve canonical problems from upper-division physics and accurate problem solvers.
courses. Our goal is to provide a scaffold for describing Similarly, Heller et al. [21] developed a strategy to help
student learning that is explicitly grounded in theories of their introductory students integrate the conceptual and
learning but can still be leveraged by instructors who are procedural aspects of problem solving. This strategy
not thoroughly versed in such theories. included five steps: (1) visualize the problem, (2) physics
This paper serves the dual purpose of describing the description, (3) plan the solution, (4) execute the plan,
theoretical grounding and development of the ACER and (5) check and evaluate. Docktor [22] modified and
framework (Secs. II and III) as well as presenting the extended this strategy to develop a validated physics
methods and findings of two investigations of student problem-solving assessment rubric. With the goal of pro-
difficulties at the upper-division level employing this viding consistent and reliable scores on problem-solving
framework (Sec. IV). It then closes with a discussion of
tasks, this rubric is scored based on five general processes:
limitations and implications for future work (Sec. V).
useful description, physics approach, specific application
of physics, mathematical procedures, and logical progres-
II. PROBLEM-SOLVING STRATEGIES AND sion. Useful description is the process of summarizing a
THEORETICAL FRAMEWORKS problem statement by assigning symbols and/or sketching.
Physics approach and specific application of physics
There are two common aspects to understanding the
problems students encounter when utilizing mathematics represent the process of selecting and linking the appro-
in physics. The first is to characterize physicists’ use of priate physics concepts to the specifics of the problem.
mathematics; such a characterization helps produce Mathematical procedures refers to the mathematical op-
instructional and analytical tools to align students’ problem erations needed to produce a solution, and logical progres-
solving with experts’ problem solving. The second is to sion looks at the focus and consistency of the overall
describe what the students are actually doing, not just in solution.
terms of how it does not make sense to physicists, but in The strategies presented above suggest considerable
terms of how it does make sense to the students. Here, agreement as to the general structure of expert problem
we review some of the previous research using these two solving as well as some indication that this structure can be
approaches. used as a guide to assess student work at the introductory
The first of these two aspects seeks to better understand level. The prescriptive nature of these problem-solving
the crossroads between physics and mathematics. Redish strategies lends itself well to the kinds of problems encoun-
[14] has developed an idealized model of how physicists tered in introductory physics. However, upper-division
use math to describe physical systems. He identifies four problems are more complex and less likely to respond to
steps that guide this process: (1) map the physical struc- a prescriptive approach. Additionally, problem-solving
tures to mathematical ones, (2) transform the initial mathe- strategies are intentionally independent of specific content
matical structures, (3) interpret the results in terms of the so as to be generally applicable, and on their own offer
physical system, and (4) evaluate the validity of the results. limited insight into the nature of students’ difficulties with
This iterative model makes it clear that the source of specific mathematical tools.
students’ difficulties may not be as simple as not knowing The other aspect of understanding the problems students
the necessary mathematical formalisms. While the inten- encounter when utilizing mathematics in physics focuses
tionally broad nature of the model makes it widely appli- on explaining why students solve problems in a particular
cable, we found it challenging to utilize it to identify way. Tuminaro [23] used videotaped problem-solving ses-
concrete, actionable implications for the instructor or sions with introductory students to develop a theoretical
020119-2
ANALYTIC FRAMEWORK FOR STUDENTS’ . . . PHYS. REV. ST PHYS. EDUC. RES. 9, 020119 (2013)
framework describing students’ use of mathematics in students often make multiple mistakes or take unnecessary
physics. This model of student thinking blends three theo- steps which must then be tracked through the solution. This
retical constructs: mathematical resources [24], epistemic undermines attempts to pinpoint the fundamental difficul-
games [25], and frames [26]. Mathematical resources are ties that cause the students to struggle or to identify rela-
the abstract knowledge elements that are involved in tionships between these difficulties. The ACER framework
mathematical thinking. Tuminaro [23] includes in the cate- provides an organizing structure that focuses on important
gory of mathematical resources a student’s intuitive mathe- nodes in students’ solutions. This removes some of the
matics knowledge and sense of physical mechanism, ‘‘noise’’ in students’ work that can obscure what is going
their understanding of mathematical symbolism, and the on. This section provides a general overview of the frame-
strategies they use to extract information from equations. work and its development before demonstrating its appli-
Epistemic games are coherent patterns of activities cation to specific mathematical tools.
observed during problem solving. Each game is character-
ized by different sequences of moves and types of resour- A. Overview
ces used by the student. The game that a student chooses to
play is governed by the frame they are operating in, which ACER was developed in conjunction with research into
is determined by their tacit expectations for what kind of student learning of two topics in upper-division physics:
activity they are engaged in. Taylor series [17] and direct integration [18]. Direct inte-
The framework presented by Tuminaro [23] was devel- gration and Taylor series were selected because they are
oped for introductory students and relies on students’ representative of the kinds of mathematical tools that
explicit discussion of the details of their work. Upper- upper-division physics students are expected to use.
division students, on the other hand, tend to work more Additionally, previous work from both math and physics
quickly and externalize less of their specific steps. To education suggest that these two topics are challenging for
address this, Bing [27] leveraged the theoretical constructs students [28–33]. The results of applying the framework to
of mathematical resources and epistemic framing to ana- these specific topics will be discussed in detail in Sec. IV;
lyze upper-level students’ use of mathematics. Epistemic here, we present the general development and form of
framing is the students’ unconscious answer to the question ACER. The ACER framework, like the frameworks pre-
‘‘What kind of activity is this?’’ Bing argues that a stu- sented by Tuminaro [23] and Bing [27], is fundamentally
dent’s framing can be identified by examining the types of cognitive and assumes a resource view on the nature of
justifications and proof that they offer to support their knowledge [24].
mathematical claims, rather than the specific ‘‘moves’’ In order to better understand students’ difficulties, we
they make. performed a modified version of task analysis [34,35] on
There are several limitations to the theoretical frame- canonical problems relating to each topic. Task analysis is a
works from Tuminaro [23] and Bing [27]. To understand method used to uncover the tacit knowledge used by experts
student work in terms of epistemic games or epistemic when solving complex problems. Our modified use of task
framing, one must have data on the students’ real-time analysis is described in greater detail in Sec. III B; however,
reasoning. This largely restricts the potential data sources the general process requires a content expert to work
to video and audio data, eliminating students’ written through the problem while documenting and reflecting on
work. Additionally, effective application of either frame- all elements of a complete solution. These elements are then
work requires considerable familiarity with the underlying discussed with several other content experts to reach con-
theoretical constructs in physics education research (PER). sensus that all important aspects of the solution have been
In practice this will prevent many instructors, particularly identified. After several iterations, we found that these
at the upper-division level, from productively utilizing the various problem-specific elements could be organized into
frameworks. four components that appeared consistently in the solutions
Describing experts’ use of mathematics and characteriz- to a number of content-rich problems utilizing sophisticated
ing students’ problem solving are complementary aspects mathematical tools. These four components are activation
of understanding mathematical difficulties in physics. of the tool, construction of the model, execution of the
The ACER framework leverages ideas from both in order mathematics, and reflection on the result. Each component
to target students’ use of mathematics in upper-division is described in greater detail below.
courses. In order to solve the back-of-the-book or exam-type
problems that ACER targets, one must determine which
mathematical tool is appropriate (activation) and construct
III. ACER FRAMEWORK
a mathematical model by mapping the particular physical
ACER is an analytical framework designed to guide and system onto appropriate mathematical tools (construction).
structure investigations of students’ difficulties with the Once the mathematical model is complete, there is often
sophisticated mathematical tools used in their physics a series of mathematical steps that must be executed in
classes. When solving upper-division physics problems, order to reduce the solution into a form that can be readily
020119-3
Activation Construction Execution of the mathematics: In order to arrive at a

of the tool of the model
solution, it is usually necessary to transform the math
structures produced in the construction component (e.g.,
Execution of the
unevaluated integrals) into mathematical expressions that
mathematics can be more easily interpreted (e.g., evaluated integrals).
Each mathematical tool requires specific background
Reflection on the
knowledge and base mathematical skills (e.g., how to take
results derivatives or integrals). The mathematical manipulations
performed in this component are not necessarily context-
FIG. 1. A visual representation of the ACER framework. free. When employing these base mathematical skills, an
expert maintains an awareness of the physical meaning of
interpreted (execution). This final solution must then be each symbol in the expression (e.g., which symbols are
interpreted and checked to ensure that it is consistent with constants when taking derivatives or integrals) [14].
known or expected results (reflection). The four general Reflection on the result: Solutions to problems in upper-
components are emergent from experts’ problem solving division physics usually result in expressions that are not
and are consistent with previous literature on problem- merely superficial manipulations of formulas provided in
solving strategies (see Sec. II). Though the framework the textbook or notes. Instead, they are new entities that
suggests a certain logical flow, we are not suggesting that offer meaningful insight into explaining or predicting the
all experts or students solve problems in a clearly organ- behavior of physical systems. Reflecting on these expres-
ized, linear fashion. sions is a crucial part of understanding the system and gain-
A convenient visualization of ACER is given in Fig. 1. ing confidence in the calculation performed (e.g., how do we
The framework provides a researcher-guided outline that know an expression is the correct one?). At the most basic
organizes key elements of a well-articulated, complete level, reflection involves checking expressions for errors
solution. The framework does not assign value by provid- (e.g., checking units) or comparing predictions to established
ing an ideal solution path towards which the students or expected results (e.g., checking limiting behavior). This
should strive. ACER is also not designed to be general kind of reflection can help to identify mistakes that
enough to be applied to open-ended problems; however,
occurred in the other components of the framework.
its targeted focus means it can be operationalized for a
The theoretical constructs that ground the frameworks
variety of mathematical tools used in context-rich prob-
presented by Tuminaro [23] and Bing [27] are commensu-
lems. Section III B will provide several examples of how
rate with the implicit theoretical constructs that ground
ACER is operationalized for specific tools and topics.
ACER. For example, a problem solver accesses different,
ACER is a tool for understanding and characterizing the
possibly overlapping, networks of resources depending on
difficulties seen in students’ work, but its structure is not
the component of the framework in which they are work-
meant to approximate students’ actual solutions. Instead,
the general structure of ACER was developed to accom- ing. Similarly, certain epistemic frames would be more
modate the complex and often iterative solution patterns useful than others when operating in different components.
characteristic of upper-division problems. Bing identifies four epistemic frames used by upper-
Activation of the tool: A problem statement contains a division students: invoking authority, physical mapping,
number of explicit and/or implicit cues that prime or calculation, and math consistency [27]. Invoking authority
activate different resources (or networks of resources) can be a valuable frame while in the activation component.
associated with any number of mathematical tools [24]. For instance, appealing to authority (e.g., the book or
These cues can include the goal of the problem (e.g., notes) is often a good way to identify which mathematical
calculate the potential) as well as the language and sym- tool to use. In the construction component, when trying to
bols used. The resources students activate depend on the map that tool to a specific problem, relying on authority
individual student and their perception of the nature of the (e.g., depending on a similar problem in the book) can
task (i.e., their epistemic framing [27]). easily sidetrack the unwary student. However, a physical
Construction of the model: In physics, mathematics are mapping frame would likely be productive for both
often used to express a simplified picture (i.e., a model) the construction and activation components. While we
of a real system. These mathematical models are typically acknowledge the value of leveraging theoretical constructs
necessary to solve physics problems. Mathematical models like resources and epistemic frames in conjunction with
are generally written
R in a remarkably compact form ACER, we have intentionally avoided explicit identification
(e.g., ¼ GdM=r) where each symbol has a spe- of specific resources or frames as part of the framework. In
cific physical meaning, which may be context dependent. this way, it is not necessary to have a strong background in
Different representations (e.g., diagrammatic or graphical) theories of learning in order to utilize ACER.
are sometimes necessary to construct or map the elements The general components of ACER were created by
of the model [14]. identifying broad themes that emerged from the modified
020119-4
task analysis of problems described in the next section. the modified task analysis remains flexible to modification
These components are consistent with Redish’s idealized based on emergent analysis of student work.
model for the way physicists utilize mathematics [14], as The following sections provide two examples of the
well as the steps in the problem-solving strategies pre- operationalized framework from upper-division electro-
sented for introductory physics [19–22]. Yet, ACER goes statics and middle-division classical mechanics.
beyond these broad descriptions by providing a mechanism
to target specific topics and mathematical tools. This 1. Example from electrostatics
mechanism is described in the following section.
Determining the electric potential or electric field from a
continuous charge distribution using the integral form of
B. Operationalizing ACER Coulomb’s law is one of the first topics that upper-division
The utility of ACER as a framework for understanding students encounter in junior-level electrostatics. For the
students’ use of mathematics in physics comes when remainder of the paper, we use Coulomb’s law to refer
it is operationalized for a specific mathematical tool. to the integral equation allowing for direct calculation of
Operationalization is the process by which a particular the electric field or potential from a continuous charge
problem or set of problems that exploit the targeted tool distribution:
are mapped onto the framework. This involves identifying * 1 Z dq
Eð r Þ ¼ r^ ; (1)
important elements in each component that together result 40 V j* r j2
in what an expert or instructor would consider a complete
and correct solution. * 1 Z dq
Vð r Þ ¼ : (2)
We used a modified form of task analysis to operation- 40 V j* rj
alize the framework. Formally, task analysis [34,35] is *
Here, dq represents the differential charge element and r is
accomplished by having a subject matter expert (SME) * *0
solve problems while explaining their steps and reasoning the difference vector r r between the source and the
to a knowledge extraction expert (KEE) who keeps a observation location (i.e., Griffiths’ script r) [37]. In this
record. This method for uncovering the tacit knowledge case, the ‘‘tool’’ we refer to is integration, and we describe
used by experts has been exploited to produce example its application to problems determining the potential or
electric field from an arbitrary, static charge distribution
solutions designed to improve students’ ability to solve
via Coulomb’s law. We will focus here only on charge
novel problems [36].
distributions that cannot easily be dealt with using
Our modified task analysis does not include a KEE. This
Gauss’s law. The element codes below are for labeling
was done because such an expert was not readily available
purposes only and are not mean to suggest a particular order
to us, nor did we want the need for a KEE to prevent other
nor are all elements always involved for any given problem.
researchers or instructors from utilizing the framework.
Activation of the tool: The first component of the frame-
Instead, the SME works through the problems, document-
work involves the selection of a solution method. The
ing their reasoning and mapping the vital elements of their
modified task analysis identified four elements that are
solution onto the components of ACER. This record is then
involved in the activation of resources identifying direct
shared with several other SMEs to ensure that all important
integration (i.e., Coulomb’s law) as the appropriate tool:
aspects of the solution are accounted for. Additionally, these
experts come to a consensus in classifying each element CA1: The problem asks for the potential or electric
into a specific component (i.e., activation, construction, field.
execution, or reflection). These preliminary elements are CA2: The problem gives a charge distribution.
then applied to student work and the operationalized frame- CA3: The charge distribution does not have appropri-
work is refined to accommodate patterns of student reason- ate symmetry to productively use Gauss’s law.
ing not present in the SMEs solutions. CA4: Direct calculation of the potential is more effi-
Our motivation for removing the KEE was entirely cient than starting with the electric field.
practical in origin; however, not utilizing a KEE may
have implications for the theoretical foundations of our Elements CA1–CA3 are cues typically present in the
modified task analysis. The KEE, as a content novice, helps problem statement. Element CA4 is specific to problems
to force the SME to fully and clearly justify their steps even asking for the electric potential and is included to account for
when they include decisions based on procedural and the possibility of solving for potential by first calculating the
declarative details the SME no longer thinks about [34]. electric field. This method is valid but often more difficult.
Removing the KEE from the task analysis process makes it Construction of the model: Here, mathematical resour-
more difficult to ensure that the important elements identi- ces are used to map the specific physical situation onto the
fied in the solution are complete from the point of view of a general mathematical expression for Coulomb’s law. The
novice as well as a SME. For this reason it is important that resulting integral expression should be in a form that could,
the operationalized ACER framework which is produced by in principle, be solved with no knowledge of the physics of
020119-5
this specific problem. We identify four key elements that develop insight into a newly constructed solution are
must be completed in this mapping: ubiquitous practices in physics. At CU, physics students
typically first encounter Taylor series from a formal,
CC1: Use the geometry of the charge distribution to mathematical perspective as freshman in calculus and
select a coordinate system. then again as sophomores in their middle-division classical
CC2: Express the differential charge element (dq) in mechanics course from an applied physics perspective. We
the selected coordinates. use Taylor series to refer to the general series approxima-
CC3: Select integration limits consistent with the tion of continuous functions:
differential charge element and the extent of X
1
1 ðnÞ
the physical system. fðxÞ ¼ f ðx0 Þðx x0 Þn
n¼0 n!
*
CC4: Express the difference vector r in the selected
coordinates. 1
¼ fðx0 Þ þ f0 ðx0 Þðx x0 Þ þ f00 ðx0 Þðx x0 Þ2 þ :
2
Elements CC2 and CC4 can be accomplished in multiple (3)
ways, often involving several smaller steps. In order to
express the differential charge element, the student must Here, fðxÞ represents some continuous function with con-
combine the charge density and differential to produce tinuous derivatives over the domain of interest. We will
an expression with the dimensions of charge (e.g., refer to x x0 as the expansion parameter, to x as the
dq ¼ dA). Construction of the difference vector often expansion variable, and to x0 as the expansion point. In this
includes a diagram that identifies vectors to the source point case, the ‘‘tool’’ we refer to is Taylor series, and its use is to
*0 * describe approximations to complex expressions in order
r and field point r . to gain insight about the underlying physics. In this paper,
Execution of the mathematics: This component of the we will focus only on examples from classical mechanics
framework deals with the mathematics required to com- though the framework could be applied to Taylor series in
pute a final expression. In order to produce a formula any domain.
describing the potential or electric field, it is necessary to Activation of the tool: The first component of ACER
CE1: Maintain an awareness of which variables are involves selecting Taylor series as an appropriate tool for a
being integrated over (e.g., r0 vs r). given problem. Our modified task analysis identified three
CE2: Execute (multivariable) integrals in the elements that are likely to activate resources (or a network
selected coordinate system. of resources) associated with Taylor series:
CE3: Manipulate the resulting algebraic expressions TA1: The problem asks for a Taylor approximation
into a form that can be readily interpreted. directly.
Reflection on the result: The final component of the TA2: The problem asks for an approximate expres-
framework involves verifying that the expression is con- sion to a complex function.
sistent with expectations. While many different techniques TA3: The problem uses language and/or symbols that
can be used to reflect on the result, these two checks are imply one physical quantity is much smaller than
particularly common: some other physical quantity (e.g., ‘‘small,’’
‘‘near,’’ ‘‘close,’’ or ).
CR1: Verify that the units are correct.
CR2: Check the limiting behavior to ensure it is We include TA1 because Taylor series are often referred
consistent with the total charge and geometry to explicitly in middle-division classical mechanics prob-
of the charge distribution. lems. The physical quantities that are compared in TA3
must have the same units, and the ratio of these quantities
Element CR2 is especially useful when the student must be less than 1.
already has some intuition for how the potential or electric Construction of the model: In this component, mathe-
field should behave in the limits. However, if they do not matical resources are used to map particular physical quan-
come in with this intuition, reflection on the results of this tities onto the general expression for Taylor series [Eq. (3)].
type of problem is a vital part of developing it. After the mapping is complete, the approximation could in
In Sec. IV B, we will apply this operationalization of principle be completed with no additional knowledge of the
ACER to investigate student work on a canonical electro- physics of the problem. For Taylor series, we identify four
statics problem (Fig. 2). key elements to complete this mapping:
TC1: Identify the physical quantities for which the

2. Example from classical mechanics problem explicitly states or implicitly suggests
Using Taylor series to construct an analytically tractable a comparison of scale (e.g., length scale, mass
problem, to approximate a complex expression, or to scale, time scale).
020119-6
TC2: Determine about which point the comparison is In Sec. IV C, we will apply this operationalization of
being made (i.e., expansion point). ACER to investigate student work on several Taylor series
TC3: Express the comparison explicitly by con- problems.
structing a dimensionless ratio of physical
quantities (i.e., expansion variable). IV. APPLICATION OF ACER
TC4: Recast the expression to be expanded in terms
of the expansion variable. To demonstrate the utility and versatility of ACER, we
present findings from two investigations of student diffi-
Element TC2 is often neglected because many approx- culties in the advanced physics courses at CU: direct
imations in physics are computed about zero (i.e., a integration of continuous charge distributions and Taylor
Maclaurin series). In most problems, there are several series as an approximation method. These investigations
combinations of physical quantities that could be used to were conducted independently as part of broader trans-
construct a dimensionless ratio, but one must identify formation efforts associated with CU’s upper-division
only those for which a comparison of scale is implied. Principles of Electricity and Magnetism 1 (E&M 1) course
Determining the appropriate expansion variable can be [38,39] and middle-division Classical Mechanics and
aided by sketching the physical situation and identifying Mathematical Methods 1 (CM 1) course [40]. Data for
the relative scales of physical quantities in the problem. these studies come from analysis of student solutions to
Execution of the mathematics: This component of the traditional exam questions and formal, think-aloud inter-
framework is concerned with employing mathematics to views. In both cases, initial data collection and analysis
compute a possible solution. Once the appropriate model began prior to the development of the ACER framework.
has been constructed, the expansion can be computed. Application of the framework to initial data motivated a
Strictly speaking, executing a Taylor series requires one to second round of interviews for both topics. This section
presents the methods and findings of these two investiga-
TE1: Maintain an awareness of the meaning of each tions with particular emphasis on how ACER contributed
symbol in the expression (e.g., which symbols to the analysis.
are constants when taking derivatives).
TE2: Compute derivatives of functions. A. Background
TE3: Evaluate the derivatives of nontrivial functions Data for these studies were collected in association with
at the expansion point. the E&M 1 and CM 1 courses at CU. Below, we provide
TE4: Manipulate the resulting algebraic expressions additional details on the methods for our direct integration
into a form that can be readily interpreted. of Coulomb’s law (Sec. IV B 1) and Taylor series
(Sec. IV C 1) studies. E&M 1 typically covers the first six
Alternatively, one might neglect elements TE2 and TE3,
chapters of Griffiths [37], which includes both electrostat-
if one has knowledge of common ‘‘expansion templates’’
ics and magnetostatics. CM 1 uses Boas [41] along with
(e.g., sinx x x3 =3!) and how to adapt these templates
Taylor [42] and covers up to but not including calculus
to the mathematical models developed previously. Hence,
of variations. The student population for both courses is
there are two pathways to execute a Taylor series: a formal
composed of physics, engineering physics, and astrophys-
method involving all elements and an abbreviated method
ics majors, with a typical class sizes of 30–70 students.
that shortcuts TE2 and TE3. The abbreviated method itself
These courses have been transformed to include a number
includes substeps, the details of which are beyond the
of research-based teaching practices including peer
scope of this study and thus have not been articulated here.
instruction [43] using clickers and tutorials [38,39].
Reflection on the result: The final component describes
In order to determine the types of difficulties students
how to verify that the approximate expression is consistent
have with Coulomb’s law integrals and Taylor series, we
with expectations. The expressions that result from per-
analyzed student solutions to canonical exam problems on
forming a Taylor series are often novel entities, not super-
continuous charge distributions (N ¼ 172) and approxima-
ficial manipulations of formula from textbooks or notes,
tion methods (N ¼ 116) and conducted two sets of think-
and these expressions must be checked:
aloud interviews (total N ¼ 18) to further probe student
TR1: Verify that the units are correct. understanding. The specific details of each exam problem
TR2: Check the behavior in the regime where the are described in greater detail below. Interviews were
approximation applies to ensure it is consistent videotaped and students’ written work was captured with
with prior knowledge or intuition about the embedded audio. Interviewees were paid volunteers who
physical system. responded to an Email request for research participants. All
interviewees had successfully completed E&M 1 or CM 1
This component is particularly important for Taylor one or two semesters prior. Participants in both studies
series because such approximations are used to check or demonstrated a wide range of abilities and received course
make sense of solutions to many other problems. scores ranging from A to D.
020119-7
Exams were analyzed by identifying each of the key by hand. Both non-PER faculty asked for the total charge
elements from the framework that appeared in the students’ on the disk first and then for the potential on the z axis.
solutions. Each element was then coded to identify the Interview data came from two sets of think-aloud inter-
types of steps made by students. These codes represented views (N ¼ 10), performed approximately 1 year apart on
emergent themes in the students’ work around each ele- different sets of students. The first set of interviews was
ment and were not predetermined by the framework. This structured to probe the preliminary difficulties identified in
coding helped to ensure that the expert-guided framework the student exams. The students were asked to calculate the
did not miss important but unanticipated aspects of student potential from two parallel disks of charge by direct
solutions. The interviews were similarly analyzed by clas- integration, and they were provided with a diagram of
sifying each of the student’s major moves into one of the the charge distribution and Eqs. (1) and (2). In terms of the
four components of the framework. Exams provided quan- ACER framework, this prompt completely bypassed the
titative data identifying common difficulties and interviews activation component. Also, while the first interview proto-
offered deeper insight into the nature of those difficulties. col offered important insight into how students spontane-
ously reflect (or not) on their solutions, it provided no
B. Coulomb’s law explicit probe of the reflection component. The second
interview protocol specifically targeted activation by asking
1. Methods students to find the potential along the z axis outside a
Our E&M 1 students are exposed to the Coulomb’s law spherical shell with nonuniform charge density ðÞ without
integral for the electric field [Eq. (1)] before the analogous providing a diagram or prompting them to solve the problem
expression for the electric potential [Eq. (2)]. However, in any specific manner. An additional question targeted
the vector nature of the electric field makes Eq. (1) sig- reflection by asking students to determine which of three
nificantly more challenging to calculate, and historically expressions could represent the potential from a static, lo-
instructors at CU tend to ask students to compute the calized charge distribution with total charge Q (see Fig. 3).
potential on exams. The exam problem examined here
asked students to calculate the electric potential along an 2. Results
axis of symmetry from a disk with charge density ðÞ
This section presents the identification and analysis of
(Fig. 2). We selected this problem because it is a recogniz- common student difficulties with Coulomb’s law integrals
able Coulomb’s law question which requires integration organized by component and element of the operational-
and has been asked on the first midterm exam for multiple ized ACER framework (see Sec. III B 1).
semesters. Activation of the tool: Roughly three-quarters of our
Exams were collected from four semesters of the course students (73% of 172) correctly approached the exam
(N ¼ 172), each taught by a different instructor. Two of question using Eq. (2). The remaining students (27% of
these instructors were physics education researchers *
involved in developing the transformed materials and two 172) attempted to calculate the potential by determining E,
were traditional research faculty. All four semesters uti- by either Gauss’s law or Eq. (1), and then taking the line
lized some or all of the available transformed materials. integral (i.e., missing elements CA3 and CA4). Rather than
The exact details of the disk question, while similar, were stemming primarily from a failure to recall Eq. (2), we
not identical from semester to semester. One of the PER argue below that this difficulty likely originated from a
faculty asked the students to sketch the charge distribution failure to reject these other methods.
and then to calculate an expression for the potential on the Identifying evidence of activation in the exam solutions
z axis (as in Fig. 2). The other PER faculty asked the was challenging because students did not typically write
students to calculate the total charge on the disk but only out their thought process as they began the problem. In
required them to set up the expression for the potential on particular, there was rarely explicit evidence that the
the x axis as the resulting integral cannot be solved easily
z
P
σ(φ)
a
x
FIG. 2. An example of the canonical exam problem on con- FIG. 3. Three equations presented in the second interview set to
tinuous charge distributions. target reflection. Students must determine the units of a, b, and c.
020119-8
students attended specifically to CA1 and CA2 (i.e., the appeared in the construction component, particularly
prompt asked for potential and provided information on the when expressing the differential charge element and dif-
charge distribution). However, we did not see students ference vector (elements CC2 and CC4). These difficulties
attempting to calculate quantities unrelated to the potential cannot be explained purely by students failing to conceptu-
or attempting to utilize methods inconsistent with the alize the integral or lacking the mathematical skills to set
information provided. up integrals over surfaces and perform vector subtractions.
More easily identified was element CA3, which elimi- Rather, students had trouble keeping track of the relation-
nates Gauss’s law as a valid approach. Approximately a tenth ships between various quantities as they adapted the decep-
of our students (11% of 172) attempted to employ Gauss’ law tively simple general formula [Eq. (2)] to a specific
*
to solve for E and then to calculate V by taking line integral. physical system.
These students often justified their answers with comments Almost all of the exams (97% of 172, N ¼ 166) con-
such as, ‘‘Since we want the voltage at a point outside the tained elements from the construction component (i.e., the
disk, the E-field we use will appear to be that of a point student did more than just write down the equation). Of
charge at the origin.’’ This inappropriate use of Gauss’s law is these students, only two did not use the appropriate coor-
consistent with previous research at the junior level [11]. dinates (i.e., cylindrical), indicating that students at this
Interestingly, none of the students in the single semester level are adept at selecting appropriate coordinate systems
(N ¼ 25) that were asked to sketch the charge distribution in highly symmetric problems (element CC1). Similarly,
rather than to calculate total charge attempted to use Gauss’s only one of the interview participants started with an
law. This suggests that calculation of the total charge likely inappropriate coordinate system, and this student eventu-
activated resources associated with Gauss’s law. ally switched after attempting the problem in Cartesian
The misapplication of Gauss’s law was also the primary coordinates. This finding is somewhat surprising given
issue observed in the interviews. Even when the students prior research indicating that even middle-division physics
were explicitly prompted to use direct integration, one of students often have a strong preference for Cartesian coor-
five students still attempted to use Gauss’s law. Two stu- dinates [13].
dents in the second set of interviews explicitly considered The remaining elements of construction proved more
using Coulomb’s law but rejected it in favor of using challenging. Nearly half the students (42% of 166) had
Gauss’s law or the expression for E from a point charge. difficulty expressing the differential charge element (ele-
ACER states that there are a number of cues (elements ment CC2) and some (14% of 166) failed to provide limits
CA1–CA3) embedded in the prompt of a physics problem of integration or gave limits that were inconsistent with
that can guide a student to the appropriate solution method. their differential (element CC3). The most common errors
For example, if the prompt provides a boundary condition made while expressing the differential charge element (dq)
rather than a charge distribution, this is likely to cue the were (see Table I) performing the integration over a region
student to use separation of variables or method of images. of space with zero charge density, using a differential with
Elements CA1 and CA2 are identical for questions that can the wrong units, and plugging in total charge instead of
be solved by Gauss’s law and Coulomb’s law [i.e., it asks charge density.
*0 Initially, we interpreted difficulties with dq as a failure
for V or E and provides ð r Þ]. However, our students tend
to be more comfortable with Gauss’s law (i.e., their to conceptualize Eq. (2) as a sum over each little ‘‘bit’’ of
Gauss’s law resources are easily activated); therefore, charge. Previous research on student difficulties with the
they must first reject Gauss’s law as appropriate before concept of accumulation as it applies to definite integrals
they will attempt to use Coulomb’s law. supports this interpretation [32]. However, the interviews
Even without Gauss’s law, it is still possible to solve for suggest that the problem was more subtle than that. Even
V by first calculating E using Eq. (1), but this calculation
requires considerably more work (element CA4). Indeed, TABLE I. Difficulties expressing the differential charge ele-
of the students who attempted this method (15% of 172) ment (dq). Percentages are of just the students who had difficulty
only a few (N ¼ 3) completed the exam problem success- with dq (42% of 166, N ¼ 69). Codes are not exhaustive or
fully. One virtue of the electric potential in electrostatics exclusive but represent the most common themes; thus, the total
N in the table need not sum to 69.
is to allow for easier calculation of the electric field via
* *
E ¼ rV. However, the students may have jumped to Difficulty N Percent
*
calculating V from E because they were exposed to E first Not integrating only over charges, 37 54
e.g., dq ¼ drdzrd
and resources associated with the electric field were more
Differential with the wrong units, 23 33
easily activated. This difficulty was not observed in the
e.g., dq ¼ drd
interviews. Total charge instead of charge density, 10 14
Construction of the model: For Coulomb’s law integrals, e.g., dq ¼ Qtot drrd
the largest number of common student difficulties
020119-9
those interviewees who failed to produce an appropriate Six of the eight interview participants who used
expression for dq made statements or gestures indicating Coulomb’s law also spontaneously drew the difference
they understood the integral to be a sum over the charge vector, and a seventh drew the vector but did not explicitly
*
distribution. Additionally, post-test data from the classical identify it as r . However, even those students who were
mechanics course at CU shows that more than 80% of our able to articulate the difference vector as the distance
students can correctly determine the differential area ele- between the source and field point struggled to produce a
ment for a cylindrical shell one semester prior to taking useful expression for it. Only one interview participant
E&M. Thus the problem appeared to be neither that the arrived at a correct expression for the difference vector
students were not conceptualizing the integral as a sum *
while the others were either unable to express j r j or treated
over the charges nor that they could not construct a differ-
it as a single variable like r or r0 . The greater degree of
ential area element. Instead, the difficulties appeared when *
students were asked to apply these two ideas simulta- difficulty with r observed in the interviews may be due to
neously to produce an expression for dq consistent with the time delay between the participants completing the
a specific charge distribution. course and sitting for the interview.
* *0
*
The magnitude of the difference vector j r j must also be Using Griffith’s ‘‘script-r’’ notation, rather than r r ,
expressed such that it is consistent with the specific charge has a number of advantages including making Coulomb’s
distribution (element CC4), and most students (86% of law for continuous charge distributions look very similar to
172, N ¼ 148) attempted to do so. About half of these Coulomb’s law for a point charge. However, it may be that
*
(47% of 148) were unable to produce a correct formula for this notation also encourages students to look at r as a
*
j r j. The most common errors included (see Table II) using separate entity that they must remember rather than a
a magnitude appropriate for a ring of charge, setting the quantity they construct. In fact, most students made com-
magnitude equal to the distance to the source point (r0 ), ments in the interviews about not remembering the formula
*
setting the magnitude equal to the distance to the field point for r or which direction it pointed, and few even attempted
(r), and never expressing the magnitude in terms of given to use the source and field point vectors to answer these
variables or quantities. It was difficult to distinguish questions. Only three of the eight interviewees spontane-
between the middle two difficulties because students’ no- * *0
ously drew r and r , suggesting that the script-r notation
tation rarely distinguished clearly between the source and obscured the importance of these two vectors. Failure to
field variables; these issues are combined in Table II. The properly distinguish between r, r, and r0 often resulted in
remaining students were distributed over a variety of dis- improper cancellations in the execution component.
tinct, but not widely represented issues. Execution of the mathematics: Given the high pressure
Students’ spontaneous use of diagrammatic representa- and individual nature of both exams and interviews, we
tion may be an additional aspect of the construction com- expected that many students would make mathematical
* *0 *
ponent. For example, drawing the vectors r , r , and r is a errors particularly with element CE3. Yet our data offer no
*
helpful step towards a correct expression for j r j. We found evidence that mathematical errors with either integrals or
that about two-thirds of our students (66% of 148, N ¼ 98) algebraic manipulations (elements CE2 or CE3) were spe-
drew one or more of these vectors on the exams; however, cific to solving Coulomb’s law problems nor that they rep-
only half of these students (50% of 98) made explicit use of resented the primary barrier to student success on these
this diagram in their solution. It may be that our students problems. More than half the student exam solutions (60%
have seen enough of these types of problems to know that of 172) contained elements from the execution component.
they should draw a diagram but have not internalized how The significant reduction in number was due primarily to the
to use it productively. one of the four classes (N ¼ 55) that was only asked to set up
the integral for V. Additionally, not all students progressed
far enough in their solutions to actually evaluate integrals.
TABLE II. Difficulties expressing the magnitude of the differ- We were not able to produce a quantitative measure
ence vector (r). Percentages are of just the students who had of student difficulties with element CE1 from the exams
difficulties with r (47% of 148, N ¼ 69). Codes are not exhaus- because the majority of students did not consistently dis-
tive but represent the most common themes; thus, the total N in tinguish between source and field variables (i.e., r vs r0 ).
the table need not sum to 69. However, of the four interview participants who made a
Difficulty N Percent distinction between the source and field point, none con-
sistently used the primed notation. Two of these students
Ring of charge, pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 27 39
* ended up integrating over the r variable as if it were r0 .
i.e., j r j ¼ a2 þ r02
Overall, half the students’ exams containing elements
Distance to source or field point, 17 25
* * of execution (51% of 103, N ¼ 53) made various mathe-
i.e., j r j ¼ r or j r j ¼ r0
No expression for j r j
*
8 12 matical errors while solving integrals or simplifying
their expression algebraically (elements CE2 and CE3).
020119-10
Roughly half of the students with mathematical errors at large r (i.e., expression 1 of Fig. 3). Only one of the five
(49% of 53) made only slight mathematical errors, such students recognized the need for an expansion without
as dropping a factor of 2 or plugging in limits incorrectly. prompting. Another three argued that the expression
The remaining students (51% of 53) made various signifi- clearly did not fall off like a point charge. However,
cant mathematical errors, such as pulling integration vari- when directed to Taylor expand, all three were able to
ables outside of integrals or not completing one or more manipulate the expression in order to isolate the small
integrals. Similar trends were observed with the seven quantity and determine the leading term in the series. A
interview participants who attempted to complete one or more detailed discussion of student difficulties with Taylor
more calculations. Four students made significant mathe- series through the lens of ACER is given in Sec. IV C.
matical errors, two made only slight mathematical errors,
and one made no errors. 3. Summary and implications
Reflection on the result: In many cases, mistakes in the We found that our junior-level students tended to
construction or execution component resulted in expres- encounter two broad difficulties which inhibited them
sions for the potential which had the wrong units and/or from successfully solving for the potential from a continu-
limiting behavior (elements CR1 and CR2). While our ous charge distribution using Coulomb’s law. First, stu-
students were able to identify these checks as valuable dents struggled to activate direct integration via Coulomb’s
when explicitly prompted, we found that they rarely spon- law as the appropriate solution method. In particular, some
taneously check these properties to gain confidence in their students tried to calculate the potential by first calculating
solutions. the electric field by Gauss’s law or Coulomb’s law. For
Only a small number of students (8% of 172) made instructors, this suggests that presentation of Eq. (2) should
explicit attempts to check their work on exams and almost be accompanied by explicit emphasis on when and why
exclusively by checking limiting behavior. While it is Gauss’s law cannot be used as well as the utility of calcu-
possible that a greater number of students did perform lating the electric potential rather than the electric field.
one or more checks (i.e., elements CR1 and CR2) but The latter should be aimed at helping students to develop
simply did not write them out, the interviews suggest this strong connections between the conceptual idea of the
is less likely. When they had not been prompted to check or potential and various mathematical formulas which allow
reflect on their solutions, half of the interview participants *
them to calculate Vð r Þ. Second, students had difficulty
made no attempt to do so. Two of the remaining students coordinating their mathematical and physical resources
only made superficial comments about being uncertain to construct an integral expression for the potential which
if their solution was correct. One stated that her answer was consistent with the particular physical situation, spe-
did not makes sense but was not able to leverage this cifically when expressing the differential charge element
realization to correct her earlier work. The final two stu- *
dq and difference vector r . Instructors may be able to
dents both mentioned checking the units of their solutions,
help by highlighting the relationships between these
though not recalling the units of o prevented one of them quantities to encourage students to view Eq. (2) as a
from actually doing so. coherent whole rather than a conglomeration of discon-
The second set of interviews explicitly targeted reflec- nected pieces. We also found that while our juniors were
tion by directly asking the students to determine if three capable of correct and meaningful reflection when
formulas (Fig. 3) could represent the potential from a explicitly prompted, very few executed these reflections
static, localized charge distribution with positive total spontaneously. We consider the ability to translate
charge Q. All five students suggested checking the units between physical and mathematical descriptions of a
of these expressions, yet all but one had difficulty doing so problem and to meaningfully reflect on or interpret the
because they did not recall the units of o . This may be part results as two defining characteristics of a physicist, yet
of why units checks were not more common in the exam these are areas where our students struggled most when
solutions as well. Eventually, all the students were able to manipulating Coulomb’s law integrals.
execute a units check once shown a method for getting
around the units of o by considering the formula for the
C. Taylor series
potential of a point charge. Additionally, all five students
suggested checking that in the limit as r ! 1 the potential 1. Methods
went to zero. Only two students spontaneously argued that Our students are formally exposed to Taylor series
V would need to fall off as 1=r. The other three made this expansions [Eq. (3)] in mathematics courses taken prior
argument when their attention was specifically drawn to to CM 1. In CM 1, students learn to use Taylor series in
the fact that the charge distribution was localized and had problems with physical context. Here, we examine two
positive total charge. exam questions that represent typical problems asked of
One of the three expressions for V required an appro- our sophomore students with different contexts: motion
priate Taylor expansion in order to determine its behavior and energy. The first problem [Fig. 4(a)] was given prior
020119-11
member’s second exam. In the first part of this problem,

students were asked to check the units of the expression for
the potential energy. Then, they computed the approximate
expression for the potential energy in part ii.
Interview data came from two sets of think-aloud inter-
views (N ¼ 8), performed approximately one semester
apart. Both studies asked students to solve a number of
Taylor series problems, which included formal math and
physics questions (see Fig. 4). The first study was per-
formed prior to the development of ACER and asked formal
math questions first. Physics questions, which were asked at
the end of the interview, explicitly cued students to use a
Taylor series (e.g., ‘‘perform a Taylor expansion’’) and
included parts i and ii of the drag question [Fig. 4(a)].
After the development of ACER, it was clear that the first
study limited the possibility of observing attempts to pro-
cess implicit cues. In the second study, formal math prob-
lems were moved to the end of the interview and the physics
questions contained only implicit cueing (e.g., ‘‘find an
approximate expression’’). Part ii of the energy question
in Fig. 4(b) appeared as part of this study.
2. Results
This section presents the analysis of student work and
the identification student difficulties with Taylor series
organized by component and element of the operational-
FIG. 4 (color online). Students’ solutions to these Taylor series ized ACER framework (see Sec. III B 2).
exam problems were analyzed using the ACER framework.
Activation of the tool: TA1–TA3 are cues embedded in
(a) Motion problem developed prior to ACER. (b) Energy prob-
lem developed after ACER and used as part of interview studies. the problem statement that can lead a student to activate
Italicized text did not appear on interview documents. resources associated with Taylor expansions, and in some
sense they are organized by the likelihood that they will
do so. The first exam study and think-aloud interview
to the development of ACER. It explicitly asks students to study [Fig. 4(a)] targeted students’ responses to explicit
perform a Taylor series expansion on an expression for cueing. Almost all students in the exam study (93% of 45)
the 1D position of a particle moving under linear drag. The attempted a Taylor series on part i of the problem, and most
second exam problem [Fig. 4(b)] was written after the students (87% of 45) did so again on part iii. Those
development of ACER to directly target aspects of activa- students who did not attempt a Taylor expansion used
tion and reflection. Students must find an approximate some inappropriate form of the binomial expansion [e.g.,
expression for the gravitational potential energy of a ða þ bÞn rather than ð1 þ Þn ] or skipped part iii. We saw
bead sliding inside a frictionless cylinder. similar success in the first interview study, where no stu-
Exams were collected from two semesters of the course dent failed to start the problem with a Taylor expansion
(N ¼ 116), each taught by a different instructor. One when explicitly prompted.
instructor was traditional research faculty and the other From the point of view of ACER, the first exam and
was physics education research faculty involved in the interview studies limited investigations of activation to
development of transformed course materials. Both element TA1. The second exam study [Fig. 4(b)] was
instructors made use of these transformed materials. In initially designed to target element TA2 by asking students
the first exam study, students (N ¼ 45) solved the linear to ‘‘[f]ind an approximate expression’’ in part ii. However,
drag problem [Fig. 4(a)] on the traditional faculty member’s the instructor felt this cueing was too vague, so additional
first exam. In part i, students were asked to compute the first wording was added to the problem statement [the italicized
two terms of a canonical Taylor expansion about t ¼ 0. text in Fig. 4(b)]. In this study, most students (87% of 71)
Students needed to clearly state the significance of these attempted a Taylor expansion. Those who did not typically
two terms in part ii. Finally, students needed to perform a misconstrued the problem by constructing some sort of
Taylor expansion of the same function around t ¼ m=b. For differential equation (6% of 71) or left the problem unan-
the second exam study, students (N ¼ 71) were asked to swered (6% of 71). This indicates that students have little
solve the energy problem [Fig. 4(b)] on the PER faculty trouble activating Taylor series when cued explicitly or
020119-12
implicitly; however, we suspect that the addition of the TABLE III. Difficulties constructing an expansion around a
italicized text in Fig. 4(b) made the cueing more explicit nonzero expansion point–part iii in Fig. 4(b). Percentages are
of the students who had difficulty with the nonzero expansion
than originally intended.
point (65% of 40, N ¼ 26). Codes are not exhaustive but
The second interview study offered a clearer view of represent the most common themes; thus, the total N in the
students’ responses to implicit cueing (TA2) with a ques- table need not sum to 26.
tion nearly identical to the problem in Fig. 4(b), but with-
out the italicized text. Two of the four interviewees Difficulty N Percent
immediately plugged in the given value [i.e., ¼ 0 in Used answer to part i, 16 62
Fig. 4(b)] to determine the approximate expression [e.g., i.e., xðtÞ vx0 t þ vx0 mb t2
UðÞ 0]. Later in the interview, after working through Incorrect functional dependence, 7 27
the formal math problems, both participants asked to return i.e., xðtÞ vx0bm ð1 1eÞ þ vex0 t vem
x0 b 2
t
to the physics problems and solved them again using
Taylor approximations. This suggests that these formal
math problems primed the student’s resources associated of 40, N ¼ 21) had coefficients in their Taylor expansion
with Taylor series expansions allowing them to connect that were consistent with evaluating the function and its
these resources back to the physics. A recent study of derivatives at t ¼ m=b. Two-thirds of these students (67%
students’ use of Taylor series approximations in the context of 21) also had the correct functional dependence [i.e.,
of statistical mechanics also indicates that upper-division ðt m=bÞn ]. The remaining one-third (33% of 21) used
students have difficulty knowing when to use a Taylor the form for an expansion around zero (i.e., tn ). Difficulties
expansion when not explicitly prompted to do so [29]. with constructing an expansion around a nonzero expan-
Construction of the model: In both the exam and inter- sion point are summarized in Table III.
view studies, the mathematical representation of the physi- Given the specific questions used, our exam studies
cal model was constructed for the students [i.e., xðtÞ and provided little insight into how students compare the scales
UðÞ in Fig. 4]. However, to compute the Taylor expansion of physical quantities (TC3) or how students recast expres-
of each function, the physical quantities in each equation sions (TC4). All students in the first exam study who
had to be mapped onto the general expression for Taylor attempted a Taylor expansion maintained the already-
series [Eq. (3)]. Identifying elements TC1–TC4 in a stu- constructed dimensionless ratio (i.e., bt=m) throughout
dents’ written solution was a challenge because students their work. The expression in Fig. 4(a) was constructed
rarely documented their thought process while performing such that the dimensionless ratio appeared in the exponen-
this mapping. When coding for these elements, we focused tial. In the second study, the expansion variable can be
on how students treated the symbols appearing in each compared to a number directly because it is technically
problem. The analysis was holistic, taking into account dimensionless. However, follow-up questioning of inter-
the full solution that students provided. From this view, viewees provided evidence that students do not have a
nearly all of the exams (study 1, part i—93% of 45, strong grasp of comparative scales. Only one student in
N ¼ 42; study 1, part iii—89% of 45, N ¼ 40; study eight clearly articulated that for an expansion to be ‘‘good’’
2—89% of 71, N ¼ 63) contained elements from the it must be performed over dimensionless variables that are
construction component (i.e., the student did more than smaller than 1. The other seven students believed their
superficially manipulate the expressions). expansion was a ‘‘good’’ approximation to the original
In all studies, every student who attempted a Taylor expressions if the variable (e.g., t) was ‘‘small compared
expansion identified the appropriate symbol as the expan- to 1’’ regardless of the expression under consideration or
sion variable (TC1). Moreover, when determining the ex- the presence of a natural comparative scale. Mathematics
pansion point (TC2), most students in the first (93% of 42) education researchers have also observed that some stu-
and second (97% of 63) exam studies had no trouble when dents struggle to identify the range in which an approxi-
this point was zero (i.e., a Maclaurin series). Students often mation is ‘‘good,’’ even in purely mathematical problems
demonstrated their identification of the variable and expan- with no inherent comparative scale [28].
sion point through mathematical manipulations (e.g., taking Execution of the mathematics: Elements TE2–TE4 pro-
derivatives and constructing functions) or their use of ca- vide opportunities to capture the type and nature of the
nonical symbolic forms [e.g., xðtÞ a þ bt þ ct2 ][44]. mathematical errors made while computing Taylor expan-
While most students correctly identified the expansion sions. While the data presented below provide evidence
around zero, a substantial fraction (47% of 40, N ¼ 19) of that students made a number of mathematical mistakes, we
students failed to properly identify nonzero expansion did not find that such mistakes were the primary barrier to
points [i.e., t m=b in part iii of Fig. 4(a)]. Most of these student success.
students (84% of 16) simply responded to part iii with their Almost all students in the exam studies (study 1—93%
answer for the expansion around t ¼ 0 (part i). Students of 45, N ¼ 42; study 2—89% of 71, N ¼ 63) performed
who correctly identified m=b as the expansion point (53% some mathematical manipulation captured by the
020119-13
execution component (TE1–TE4). Identifying constants (29% of 28) computed the derivative of the associated
and variables (TE1) in the given expressions was only a functions incorrectly. The remaining students struggled to
significant barrier to one student in part i of the first exam perform any of the necessary mathematics. Mathematical
study and three students in the second exam study. Students errors were more prevalent in our interview studies, but
typically demonstrated an awareness of the nature of each few were serious. Of the eight participants, seven made
symbol by taking derivatives or using an expansion tem- some mathematical mistake, but only one participant com-
plate with the appropriate variable [e.g., t in Fig. 4(a)]. puted derivatives incorrectly.
Interview participants often explicitly pointed to symbols Once the computation is complete, it is typical to organ-
and clearly identified them as constants. ize terms in increasing order (TE4). This practice makes
Students computed Taylor expansions through both the interpretation of the solution somewhat simpler because
formal and abbreviated methods [e.g., working through terms with similar orders are grouped together and their
Eq. (3) versus using ‘‘expansion templates’’ like cosx effect can be discussed together. Most students successfully
1 þ x2 =2!]. Those who used expansion templates shortcut organized their solution in this way (study 1, part i—83% of
elements TE2 and TE3 even though these templates stem 42; study 1, part iii—70% of 40; study 2—97% of 63).
from taking derivatives of the associated function and Similarly, all interview participants spontaneously organ-
evaluating those derivatives around the expansion point. ized their solutions in order of increasing power. However,
Of students who showed evidence of the execution com- the practice of organizing solutions did not mean students
ponent, a significant fraction used expansion templates could readily interpret their solution. As discussed below,
when the expansion was around zero (study 1, part i— many students struggled to make meaningful statements
67% of 42; study 2—90% of 63). The remaining students about the physics of their proposed solutions.
computed derivatives of the associated functions. Reflection on the result: Once a solution has been
For part iii of the first exam study, fewer students overall constructed, it should be checked for errors and an inter-
were coded in TE2 and TE3 (57% of 42, N ¼ 24) because pretation should be made. As we discuss below, students
a substantial fraction (Table III) used their answer for rarely offered checks or spontaneously interpreted their
the expansion around t ¼ 0 (part i). Of the remaining solution. When prompted in the second exam study, stu-
students, more than two-thirds (71% of 24) used formal dents checked the units of a solution successfully, but in
methods to compute their Taylor expansion. This suggests the first exam study students struggled to interpret solu-
that students are more familiar with templates of Maclaurin tions meaningfully.
expansions. We observed similar trends in our interviews. In the first exam study, no student spontaneously
When confronted with simple functions or expressions, checked their solution to part i or part iii for errors. A
interviewees overwhelmingly elected to use or ask for check of the units (TR1) would have helped a small frac-
expansion templates. When simple functions were tion of students on part i (10% of 42), but on part iii it could
embedded in more complicated expressions, seven of eight have clued more than a third of students (33% of 40) that
interviewees employed formal methods; only one student something was incorrect about their solution. Part ii of the
used an expansion template. first exam study [Fig. 4(a)] forced students to interpret their
The broad ACER framework (Sec. III B 2) does not solution and to connect it to their prior knowledge about
capture all the nuances of students’ mathematical errors; motion (TR2). Most students offered little substance in
hence, we found it constructive to create a number of their interpretation. Common responses for the linear
subcodes to capture more details. Considering all coded term included ‘‘it’s the initial v times t’’ and ‘‘it’s the
instances of execution, about one-third (34% of 147, velocity.’’ Only a quarter of students (25% of 40) men-
N ¼ 50) contained some mathematical error. Some of tioned something similar to ‘‘the distance covered in vac-
these students made only slight algebraic manipulation uum.’’ For the quadratic term, the same fraction of students
errors (44% of 50) such as forgetting a minus sign or mentioned that it was the ‘‘drag term’’ or the ‘‘correction,’’
dropping numerical factors. More than half of the students but not a single student mentioned the sign difference
with mathematical errors (56% of 50, N ¼ 28) made more between the linear and quadratic terms. In our interviews,
serious mistakes, which occurred primarily in part iii of the no student clearly connected a solution to this problem to
first exam study. More than half of these students (54% of the underlying physics.
28) made serious expansion mistakes, such as appending In the second exam study, students were prompted to
variables to ‘‘patch up’’ their solutions. That is, students check the units of the expression prior to starting the
would produce a solution that did not depend on t [e.g., problem [Fig. 4(b)] and most students did this correctly
xðtÞ ¼ a0 þ a1 m=b þ ] and in the next line append a t (80% of 71). Students in the first exam study were not
[e.g., xðtÞ ¼ a0 þ a1 m=bt þ ]. This was not observed asked to reflect on their solution directly. Eventually, all
in the interviews, so it is unclear if patching up an expres- four participants in the second think-aloud study produced
sion represents an error in construction or execution, or, a solution to the problem that depended on 2 . They were
possibly, a ‘‘success’’ in reflection. About a quarter then asked to discuss any physics that could help them
020119-14
interpret their solution. Only one of the four students made and structure investigations of student difficulties with
an interpretation of the solution. This student suggested Coulomb’s law and Taylor series. This has allowed us to
that the system ‘‘looks like a harmonic oscillator,’’ and more clearly identify prevalent difficulties our students
gestured to indicate the oscillation around the bottom of demonstrated with each of these topics and to paint a
the cylinder. Even with additional prompting by the inter- more coherent picture of how these difficulties are
viewer, the other three participants expressed only super- interrelated.
ficial reflections, ‘‘yeah, that looks different [from the As with any expert-guided description, it should not be
original expression].’’ assumed a priori that the operationalized ACER frame-
work will span the space of all relevant aspects of actual
3. Summary and implications student problem solving. It is intended to provide a scaffold
We found that sophomore-level students encountered from which researchers and instructors who are less
several challenges when solving Taylor approximation familiar with qualitative analysis can ground an analysis
problems. These challenges limited the production of com- of what students actually do when solving mathematically
plete, well-articulated solutions. First, knowing when to demanding physics problems. However, additional
use Taylor approximations is challenging to students when research comparing the operationalized framework, as
prompts are less explicit. This difficulty is likely under- produced by the expert task analysis, to interviews and
represented in our data because we have not explored how group problem-solving sessions will be necessary to
students grapple with minimal cueing (TA3). Processing explore the limitations of ACER in terms of capturing
implicit cues is a skill that will follow students throughout emergent aspects of students’ work.
their physics careers. Instructors should be aware of what There are several important limitations to the ACER
cues they include in problems and how these cues impact framework. The framework was designed to target the
student success on Taylor approximations. Second, while intersection between mathematics and physics in upper-
students are relatively adept at performing expansions division physics courses, and it is not well suited to
around zero (i.e., Maclaurin series), they struggle to per- describing student reasoning around purely conceptual or
form Taylor expansions around nonzero expansion points. open-ended problems. Additionally, the framework inher-
Difficulties here ranged from failing to demonstrate under- ently incorporates some aspects of representation because
standing of expansions around points other than zero to the translation between verbal, mathematical, graphical,
expanding around appropriate points but not producing the and/or pictorial representations is almost always required
correct functional form (Table III). Not all Taylor expan- to solve physics problems; however, the exact placement
sions in physics occur around zero, and students must be of multiple representations within the framework is likely
prepared to solve general expansion problems. Third, soph- to be highly content dependent. Furthermore, we have not
omore students (like juniors) rarely reflect spontaneously commented on the integration of prediction and metacog-
on their solutions. We have commonly observed this chal- nition into the framework, in part because we rarely
lenge for students in all upper-division courses. Checking observe our students showing explicit signs of either with-
solutions for errors and constructing meaningful interpre- out prompting. Application of ACER to additional topics
tations are practices that are equally important to using and tools will clarify how the framework can shed light on
mathematics. Yet, these practices are underemphasized in these aspects of problem solving.
our current upper-division courses. Problems and activities Ongoing projects with ACER include its use to frame
should be designed to develop students’ skills with reflec- investigations of upper-division students’ difficulties with
tive practices. delta functions and complex exponentials. Future work will
include analysis of students’ difficulties with separation of
V. SUMMARY AND DISCUSSION variables in the context of Laplace’s equation. Each of these
We have presented an analytic framework, ACER, that is projects will facilitate further validation and refinement of
specifically targeted towards characterizing student diffi- ACER as a tool for understanding student difficulties. Future
culties with mathematics in upper-division physics. The work will also involve leveraging ACER to investigate the
ACER framework provides an organizing structure that evolution of students’ difficulties with specific mathematical
focuses on important nodes in students’ solutions to com- tools over time. Specifically, Newton’s law of gravity for
plex problems by providing a researcher-guided outline extended bodies is mathematically very similar to the use of
that lays out the key elements of a well-articulated, com- Coulomb’s law for continuous charge distributions but is
plete solution. To account for the complex and highly typically encountered in sophomore physics. By identifying
context-dependent nature of problem solving in advanced students’ difficulties with gravitation in sophomore classical
undergraduate physics, ACER is designed to be operation- mechanics and comparing them to difficulties with direct
alized for specific mathematical tools in different physics integration in junior electrostatics, we will be able to inves-
contexts rather than as a general description. We have tigate how these difficulties change (or not) as students
utilized the operationalized ACER framework to inform advance through the curriculum.
020119-15
The ACER framework was designed to be a tool not only and complete assessment of all aspects of students’ prob-
for researchers but instructors as well. We have already lem solving.
discussed a number of suggestions for instructors that
may help students avoid or overcome the difficulties we ACKNOWLEDGMENTS
identified. However, ACER can also be used to critique and
The authors gratefully acknowledge the generous contri-
design problems. Examining the prompt of a question can
butions of CU faculty members: A Becker, M. Dubson, E.
identify which components of the framework the problem
Kinney, A. Marino, and T. Schibli. This work was funded by
targets and which ones it might short-circuit (e.g., bypass-
NSF-CCLI Grant No. DUE-1023028, the Science Education
ing activation by instructing the student to use a Taylor
Initiative, and a National Science Foundation Graduate
series to approximate a function). This can help instructors
Research Fellowship under Award No. DGE 1144083.
to produce homework sets and exams that offer a balanced
[1] J. L. Docktor and J. P. Mestre, A Synthesis of Discipline- [14] E. F. Redish, arXiv:physics/0608268.

Based Education Research in Physics (National Research [15] J. Wagner, C. Manogue, and J. Thompson, in Proceedings
Council, Washington, DC, 2010). of the Physics Education Research Conference, Omaha,
[2] L. Hsu, E. Brewe, T. Foster, and K. Harper, Resource letter NE, 2011 (American Institute of Physics, College Park,
RPS-1: Research in problem solving, Am. J. Phys. 72, MD, 2012), Vol. 1413, pp. 89–92.
1147 (2004). [16] R. Pepper, S. Chasteen, S. Pollock, and K. Perkins, in
[3] D. E. Meltzer and R. K. Thornton, Resource letter ALIP-1: Proceedings of the Physics Education Research
Active-learning instruction in physics, Am. J. Phys. 80, Conference, Omaha, NE, 2011 (Ref. [15]), pp. 291–294.
478 (2012). [17] M. Caballero, B. Wilcox, R. Pepper, and S. Pollock, in
[4] R. Pepper, S. Chasteen, S. Pollock, and K. Perkins, in Proceedings of the Physics Education Research
Proceedings of the Physics Education Research Conference, Conference, Philadelphia, PA, 2012 (American Institute
Portland, OR, 2010 (American Institute of Physics, of Physics, College Park, MD, 2013), Vol. 1513, pp. 90–93.
College Mark, MD, 2011), Vol. 1289, pp. 245–248. [18] B. Wilcox, M. Caballero, R. Pepper, and S. Pollock, in
[5] C. S. Wallace and S. V. Chasteen, Upper-division students’ Proceedings of the Physics Education Research
difficulties with Ampère’s law, Phys. Rev. ST Phys. Educ. Conference, Philadelphia, PA, 2012 (Ref. [17]), pp. 418–421.
Res. 6, 020115 (2010). [19] F. Reif, J. H. Larkin, and G. C. Brackett, Teaching general
[6] C. Singh, M. Belloni, and W. Christian, Improving stu- learning and problem-solving skills, Am. J. Phys. 44, 212
dents’ understanding of quantum mechanics, Phys. Today (1976).
59, No. 8, 43 (2006). [20] D. S. Wright and C. D. Williams, A wise strategy for
[7] T. I. Smith, J. Thompson, and D. Mountcastle, in introductory physics, Phys. Teach. 24, 211 (1986).
Proceedings of the Physics Education Research [21] P. Heller, R. Keith, and S. Anderson, Teaching problem
Conference, Portland, OR, 2010 (Ref. [4]), pp. 305–308. solving through cooperative grouping. Part 1: Group vs.
[8] L. Deslauriers and C. Wieman, Learning and retention of individual problem solving, Am. J. Phys. 60, 627 (1992).
quantum concepts with different teaching methods, Phys. [22] J. Docktor, Ph.D. thesis, University of Minnesota, 2009.
Rev. ST Phys. Educ. Res. 7, 010101 (2011). [23] J. Tuminaro, Ph.D. thesis, University of Maryland, 2004.
[9] M. Ayene, J. Kriek, and B. Damtie, Wave-particle duality [24] D. Hammer, Student resources for learning introductory
and uncertainty principle: Phenomenographic categories physics, Am. J. Phys. 68, S52 (2000).
of description of tertiary physics students’ depictions, [25] A. Collins and W. Ferguson, Epistemic forms and epis-
Phys. Rev. ST Phys. Educ. Res. 7, 020113 (2011). temic games: Structures and strategies to guide inquiry,
[10] G. Zhu and C. Singh, Surveying students’ understanding Educ. Psychol. 28, 25 (1993).
of quantum mechanics in one spatial dimension, Am. J. [26] P. Deborah Tannen, Framing in Discourse (Oxford
Phys. 80, 252 (2012). University Press, Oxford, England, 1993), ISBN .
[11] R. E. Pepper, S. V. Chasteen, S. J. Pollock, and K. K. [27] T. J. Bing, Ph.D. thesis, University of Maryland, 2008.
Perkins, Observations on student difficulties with mathe- [28] D. Champney and E. Kuo, in Proceedings of the
matics in upper-division electricity and magnetism, Phys. Conference on Research in Undergraduate Mathematics
Rev. ST Phys. Educ. Res. 8, 010111 (2012). Education, Portland, OR, 2012 (SIGMAA, Washington,
[12] W. M. Christensen and J. R. Thompson, Investigating DC, 2012), Vol. 15, pp. 94–107.
graphical representations of slope and derivative without [29] T. I. Smith, J. R. Thompson, and D. B. Mountcastle,
a physics context, Phys. Rev. ST Phys. Educ. Res. 8, Student understanding of Taylor series expansions in
023101 (2012). statistical mechanics, Phys. Rev. ST Phys. Educ. Res. 9,
[13] E. C. Sayre and M. C. Wittmann, Plasticity of intermediate 020110 (2013).
mechanics students’ coordinate system choice, Phys. Rev. [30] N. Khan, D. Hu, D.-H. Nguyen, and N. S. Rebello, in
ST Phys. Educ. Res. 4, 020105 (2008). Proceedings of the Physics Education Research
020119-16
Conference, Omaha, NE, 2011 (Ref. [15]), pp. 227– [37] D. J. Griffiths, Introduction to Electrodynamics (Prentice-
230. Hall, Upper Saddle River, NJ, 1999), ISBN .
[31] D. Hu and N. S. Rebello, in Proceedings of the Physics [38] S. V. Chasteen, S. J. Pollock, R. E. Pepper, and K. K.
Education Research Conference, Philadelphia, PA, 2012 Perkins, Thinking like a physicist: A multi-semester
(Ref. [17]), pp. 186–189. case study of junior-level electricity and magnetism,
[32] P. W. Thompson and J. Silverman, Making the Am. J. Phys. 80, 923 (2012).
Connection: Research and Teaching in Undergraduate [39] S. V. Chasteen, S. J. Pollock, R. E. Pepper, and K. K.
Mathematics (Mathematics Association of America, Perkins, Transforming the junior level: Outcomes from
Washington, DC, 2008), Vol. 73, p. 43. instruction and research in E&M, Phys. Rev. ST Phys.
[33] D. Kung and N. Speer, Do they really get it? Evaluating Educ. Res. 8, 020107 (2012).
evidence of student understanding of power series, [40] S. Pollock, R. Pepper, and A. Marino, in Proceedings of
PRIMUS 23, 419 (2013). the Physics Education Research Conference, Omaha, NE,
[34] R. Catrambone, in Proceedings of the 2011 Learning and 2011 (Ref. [15]), pp. 303–306.
Technology Symposium, Columbus, GA, 2011, http:// [41] M. Boas, Mathematical Methods in the Physical Sciences
cunningham.columbusstate.edu/technologysymposium/ (John Wiley & Sons, New York, 2006), ISBN .
docs/Catrambone%20white%20paper.pdf . [42] J. R. Taylor, Classical Mechanics (University Science
[35] R. Catrambone, The subgoal learning model: Creating Books, Herndon, VA, 2005), ISBN .
better examples so that student can solve novel problems, [43] E. Mazur, Peer Instruction: A User’s Manual, Series in
J. Exp. Psychol. Gen. 127, 355 (1998). Educational Innovation (Prentice Hall, Upper Saddle
[36] R. Catrambone, Generalizing solution procedures learned River, NJ, 1997).
from examples, J. Exp. Psychol. Learn. Mem. Cogn. 22, [44] B. L. Sherin, How students understand physics equations,
1020 (1996). Cognit. Instr. 19, 479 (2001).
020119-17

Analytic Framework For Students' Use of Mathematics in Upper-Division Physics

Uploaded by

Copyright:

Available Formats

Analytic Framework For Students' Use of Mathematics in Upper-Division Physics

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analytic Framework For Students' Use of Mathematics in Upper-Division Physics

Uploaded by

Copyright:

Available Formats

PHYSICAL REVIEW SPECIAL TOPICS - PHYSICS EDUCATION RESEARCH 9, 020119 (2013)

Analytic framework for students’ use of mathematics in upper-division physics

DOI: 10.1103/PhysRevSTPER.9.020119 PACS numbers: 01.40.Fk, 01.40.Ha

mathematical tools. It is often an explicit goal for advanced

1554-9178=13=9(2)=020119(17) 020119-1 Published by the American Physical Society

Activation Construction Execution of the mathematics: In order to arrive at a

TC1: Identify the physical quantities for which the

member’s second exam. In the first part of this problem,

[1] J. L. Docktor and J. P. Mestre, A Synthesis of Discipline- [14] E. F. Redish, arXiv:physics/0608268.

You might also like