ction

The flow stress, or the stress at which metallic materials flow and display permanent
deformation, is a key parameter in the

evaluation of their mechanical behavior and it is a subject of great interest in any

attempt to understand the role played by structural

features within the materials. A common research procedure includes assessing the
flow stress of pure metal single crystals with low

densities of crystalline defects and then assessing the effect of defects such as
impurities, dislocations, second phase particles and grain

boundaries. These studies also generally include estimates of the effects of temperature
and strain rate.

The influence of each crystalline defect, such as the type and amount of impurity
atoms in solid solution, the dislocation density and

size and the morphology and amount of second phase particles, are relatively well
predicted and there are accepted phenomenological

explanations for these predictions. However, an understanding of the effect of grain

boundaries on the flow stress is less well

developed. The main parameter associated with the grain boundaries may be the
polycrystalline average grain size which provides the

density of grain boundaries in the material. It is not easy to estimate the precise origins
of research on this topic but major breakthroughs took place in the early 50s, about
seventy years ago.

1.1. The Hall-Petch relationship or “low temperature grain refinement hardening”

Hall [1] published a paper in 1951 on the deformation of mild steels and made the
following statement:

“It is known that if a dislocation in a matrix approaches a region, such as a grain-

boundary film, which has a higher yield stress than the
matrix itself, then this region constitutes an effective potential barrier for the passage
of the dislocation, In other words, the dislocation

will experience a repulsion near the grain boundary interface, and dislocations will pile
up along the glide plane behind the grain

boundary film, until the stress concentrations around the tip of the slip band cause the
film to yield. Deformation is then transferred to the

next grain” [1].

Hall then suggested that the difference between the flow stress of a polycrystal, σ, and
the flow stress of a single crystal, σ0, is

proportional to the inverse of the square root of the grain size, d, such that [1]

(σ − σ0)∝1/d1/2 (1)

This relationship was further expanded in 1953 by Petch [2] to account for the
cleavage strength of polycrystals and Petch also

introduced the assumption of a Frank-Read source in the center of a grain as in the

following extract:

“The dislocations initially present within the crystal can probably begin to move at
small stresses, lower than the conventional yield point,

but they cannot pass beyond the confines of the grain boundary. On the glide planes
that contain a Frank-Read source, there will be

dislocation multiplication, and this will continue until the accumulation of the
dislocations on the glide plane and the interaction between

the dislocations within the accumulation stop the multiplication for that particular
applied stress” [2].

Petch suggested a quantitative explanation to the contribution of grain size to the

cleavage strength, reported experimental data of

the flow stress in agreement with Hall and then presented an equation that relates the
flow stress and the grain size in the format which

has become widely established within the materials science community:

σ = σ0 + Kd− 1/2 (2)

where K is a constant which later became known as the Hall-Petch constant.

Thereafter, it has become common practice to plot experimental data of flow stress as
a function of the inverse of the square root of

the grain size and it is found that this relation is valid for most, if not all, metals, at
least over a range of reasonable grain sizes.

However, the original explanation for the relationship was based on the concept of
dislocations emitted from a Frank-Read source in

the center of a grain piling-up at the nearest grain boundary and activating slip in the
neighboring grain and this does not agree with a

broad set of data for many materials. Also, different slopes have been reported in the
plots of σ vs d− 1/2 for different grain size ranges in

different materials. Accordingly, many different mechanisms have been suggested to

address these issues but to date there has been no

acceptable prevailing model.

The lack of consensus on this topic appears to have various reasons but a detailed
discussion on this topic is outside of the scope of

this review. Nevertheless, it is important to note that a control of the grain size for
experimental analysis may be more complicated

than simply controlling the impurities, dislocation density and presence of second
phase particles. Any production of samples with a

broad range of grain sizes is not an easy task and the early studies on this subject made
use of samples in which the minimum grain sizes

were in the range of a few micrometers. The inherent limitation on grain refinement is
of particular importance since the rule is based

on the assumption that the flow stress is inversely proportional to the square root of the
grain size and this means that large flow
stresses, which are required for any accurate evaluation of trends, depend upon the
presence of very small grain sizes. Therefore, it is

reasonable to anticipate that the lower bound for grain size in these earlier studies will
almost certainly affect the overall accuracy of

R.B. Figueiredo et al.

Progress in Materials Science 137 (2023) 101131

the predictions. Although there is the occasional early report of the production of
nanocrystalline materials, such as a grain size of 12

nm in nickel in 1986 [3], the extent of available data for bulk samples having ultrafine
grains has only increased significantly over the

last two decades. This is illustrated in Fig. 1 where data for the flow stress, σ, of
aluminum is plotted as a function of the inverse of the

square root of the grain size, d− 1/2

. These data were collected from multiple sources [4–25] and the data from papers
published in the

last century are highlighted. It is readily apparent that the range of grain sizes in
experiments will play a significant role in determining

the accuracy of any trend. Thus, early studies reported values for the Hall-Petch
constant, K, in the range of 10 ~ 65 MPa μm1/2 [26]

whereas data reported in this century show this constant, as estimated from the slope
of the plot in Fig. 1, is probably larger than ~ 100

MPa μm1/2. In fact, there may be two stages in the σ vs d− 1/2 plots for aluminum so
that coarse-grained samples display a lower slope

and finer grains display a larger slope.

The development of methods to produce nanocrystalline materials led to a significant

step forward in this research but the early
experiments were limited to very small samples which may also contain other defects.
As a consequence, there was a great deal of

dispersion in both the data and the trends. It is only very recently that methods were
developed to increase the size and homogeneity of

samples with very small grain sizes, thereby providing the possibility of accurately
evaluating the trends.

1.2. High temperature grain refinement softening and “the inverse Hall-Petch
behavior”

If the low temperature grain refinement hardening predicted by Eq. (2) is widely
known and accepted, it is also widely accepted

that the opposite trend takes place at high temperatures. This means that smaller grain
sizes produce lower flow stresses at high

temperatures. At this point it is important to clarify that the notion of “high”

temperature is not absolute and depends on the melting

point of the material. In the interests of simplicity, the present work considers “low”
temperatures as T < 0.3 Tm, “moderate” temperatures as 0.3 Tm < T < 0.5 Tm and
“high” temperatures as T > 0.5 Tm, where Tm is the absolute melting temperature of
the material.

High temperature mechanical behavior is usually described by using creep

mechanisms for which there are well-established ratecontrolling equations. The
general creep equation can be written in the form [27].

ε˙ = ADGb

kT

(σ

)n

(b

d
)p

(3)

where ε ˙ is the deformation rate, A is a constant, D is the relevant diffusion

coefficient, G is the shear modulus, b is the Burgers vector, k

is the Boltzmann’s constant, T is the absolute temperature and p and n are constants.
The type of diffusion coefficient and the constants

A, n and p vary depending on the creep mechanism. Although Eq. (3) is mostly
considered in any analysis of high temperature

deformation, for conditions of steady-state deformation a similar equation has been

suggested [28] to explain the low temperature

deformation of ultrafine grained (UFG) metals which do not display work-hardening

and also deform under steady-state conditions.

The suggested equation considered n = 8 and p = -4 which in practice predicts a Hall-

Petch relationship. Some creep mechanisms such

as dislocation viscous glide or dislocation climb occur intragranularly so that the

exponent p is equal to zero and therefore the grain size

has no effect on the creep rate or flow stress. But other mechanisms, such as Coble
diffusion creep [29] and Nabarro-Herring diffusion

creep [30,31] and grain boundary sliding [32], occur intergranularly and have a
positive exponent p which means that, for constant

temperature and strain rate, the flow stress decreases with decreasing grain size.

For example, the strain rate for superplasticity, which is controlled by grain boundary
sliding, may be estimated from the relationship [33]

Fig. 1. Flow stress of aluminum [4–25] plotted as a function of the inverse of the
square root of the grain size. The data reported in the last century

are highlighted.

R.B. Figueiredo et al.

Progress in Materials Science 137 (2023) 101131

4

ε˙ = AδDgbG

kT

(σ

)2

(b

)2

(4)

where A is a constant equal to ~ 5, δ is the grain boundary width which is usually

taken as 2b and Dgb is the coefficient for grain

boundary diffusion. This equation can be rearranged to separate the flow stress and the
grain size so that

(AδDgb

ε˙GkT )1/2

=d

b (5)

The validity of this relationship is illustrated in Fig. 2 which shows the flow stress
normalized by the strain rate and temperature

plotted as a function of the grain size normalized by the Burgers vector for a
magnesium alloy AZ31 for which there are multiple reports

in the literature [34–43] on superplasticity and grain boundary sliding. It is apparent

that the stress increases linearly with increasing

grain size and the data agree fairly well with the prediction.
The first paper reporting a true superplastic behavior was published almost ninety
years ago in 1934 [44]. The understanding of the

deformation mechanism during superplasticity was developed later but it is now well
established that grain boundary sliding is the

rate-controlling mechanism [45]. Thus, the trends reported for low temperature flow
(Hall-Petch) and for high temperature superplasticity are contrary and this is illustrated
in Fig. 3 in which the flow stress and the grain size are plotted in arbitrary units using
a

logarithmic scale. At low temperatures the flow stress is proportional to d− 1/2 and at
high temperatures the flow stress in fine-grained

samples is proportional to d under conditions in which the material undergoes

superplasticity.

It is interesting to note that the different mechanisms that were proposed [33,46,47] to
explain high temperature grain boundary

sliding consider the existence of dislocation pile-ups within the grains. Both Hall [1]
and Petch [2] also considered the existence of

dislocation pile-ups and made reference to a paper that had been recently published in
1951 on the equilibrium of linear arrays of

dislocations [48]. Hall suggested that the pile-up would generate deformation in the
neighboring grain and Petch suggested the pile-up

could initiate fracture. Ball and Hutchison [46] attributed to Friedel [49] a suggestion
that the climb of dislocations into and along

grain boundaries could be caused by the concentrated stress at the head of dislocation
pile-ups. Thus, both the low and thmetric peaks [12]. The grain size of coarse grains
(CG) is

often several micrometers to tens of micrometers, or even larger.

Coarse-grained (CG) regions exhibit good ductility but lower strength

[16,17], while ultrafine-grained (UFG) regions have a strength 1–5

times higher than CG regions [18–20], though at the cost of reduced

ductility [21–23]. This forms a heterogeneous microstructure with

alternating soft (CG) and hard (UFG) regions [24,25]. During tensile

deformation, due to the different rheological behaviors and plastic incompatibility

between hard and soft regions, uneven deformation occurs, creating strain gradients at
their interfaces [26]. To maintain

deformation continuity, geometrically necessary dislocations (GNDs)

accumulate near the interfaces, inducing heterogeneous

deformation-induced (HDI) stress. Specifically, back stress develops in

soft regions and forward stress in hard regions, collectively generating

HDI strengthening and HDI strain hardening. HDI strengthening increases yield
strength, while HDI strain hardening helps maintain and

improve ductility [27–29].

The methods for fabricating heterostructured materials include heat

treatment (HT) [30], severe cold rolling (CR) combined with HT [31,

32], and dynamic plastic deformation (DPD) [33]. These approaches

generally regulate recrystallization fractions through thermomechanical

activation, enabling the integration of fine-grained strength with

coarse-grained ductility, thereby developing novel strategies for

designing high-strength-ductility austenitic steels [9,34]. Consequently,

precise control of recrystallization fractions is critical. Currently, such

control predominantly relies on the traditional empirical paradigm of

"trial-and-error process-performance testing." Key factors influencing

recrystallization fractions include intrinsic material properties (e.g.,

activation energy), deformation conditions, annealing parameters, and

microstructural states [35]. To address this, constructing empirical

models that accurately reflect the relationship between processing parameters and
recrystallization fractions holds significant engineering

value. Although the Johnson-Mehl-Avrami-Kolmogorov (JMAK) model

provides a classical kinetic framework for recrystallization [36,37], it

neglects microstructural features (dislocation/twin density) and fails to

establish clear nucleation rate-recrystallization correlations. This limitation

underscores the urgent need for a unified model linking processing parameters-heat
treatment- recrystallization fraction.

This study selects high-manganese austenitic steel (HMAS) from the

austenitic steels stem as the research object and successfully constructs

HMAS with a bimodal grain heterogeneous structure through cold

rolling and aging annealing processes. The research focuses on analyzing

the microstructural evolution, strengthening mechanisms, and mechanical behavior of

HMAS after cold rolling and annealing. Based on

the JMAK model, a modified recrystallization model (JMAK-XIA) is

established by incorporating dislocation density and twin density as

driving forces. This model systematically addresses the limitations of the

classical JMAK model, enabling precise prediction for material design

and process optimization. It provides a theoretical basis for the

controlled preparation and mechanical design of austenitic steels with

high-performance heterogeneous microstructures.

2. Experimental

2.1. Materials and processing

The material used in this study was a high manganese austenitic

steel, with a chemical composition of 0.6 C, 18 Mn, 1.5 Al, 0.3 Si (wt%),

balanced by Fe and unavoidable impurities. The molten steel was cast at

1550–1600◦C and then held at 1200◦C for 2–4 h for hot pressing. The

ingot was subsequently hot-rolled (HR) from 30 mm to 4 mm thickness

at 1150◦C. To investigate the effects of cold rolling reduction and

holding time on the heterogeneous microstructure and mechanical

behavior of HMAS, the hot-rolled sheets were cold-rolled 50 % (2 mm

thickness) and 70 % (1.2 mm thickness) reductions, respectively. Aging

annealing treatments were performed in a box resistance furnace (SX4–10) at 600◦C

with varying holding times (5 min, 15 min, 1 h, 2 h). The

heating rate was maintained at 10 ◦C/min, and timing commenced once

the target temperature was reached. After annealing, the samples were

water-quenched. Detailed processing parameters are summarized in

Table 1.

2.2. Tensile testing

Tensile tests were conducted at room temperature using a universal

testing machine (Instron 5582) with a 30 kN load capacity, at a strain

rate of 10⁻³ s⁻¹ . The tensile specimens, machined from the HMAS sheets

along the rolling direction, had standardized dimensions of 80 mm

× 1 mm × 2 mm and 80 mm × 1 mm × 1.2 mm, respectively. To ensure

surface quality, all specimen surfaces were mechanically polished using

SiC abrasive papers. Each test group was repeated three times, and the

results were averaged to ensure reproducibility.

2.3. Microstructural characterization

Microstructural analysis was performed using electron backscatter

diffraction (EBSD, Crossbeam 550). EBSD measurements were conducted at an

accelerating voltage of 20 kV with a step size of 0.128 μm,

and the acquired data were processed using Aztec Crystal software.
Specimens were sequentially ground with SiC papers of grit sizes 400,

800, 1500, 2000, and 5000, followed by polishing with 0.25 μm diamond paste and
0.02 μm SiO₂ colloidal suspension. This preparation

eliminated residual stresses and ensured high-quality surfaces for accurate EBSD
characterization.

3. Tensile properties and microstructural characterization

3.1. Initial microstructure

The microstructures of HR, CR50 %, and CR70 % samples were

characterized by electron backscatter diffraction (EBSD), as shown in

Fig. 1. Inverse pole figure (IPF) maps reveal that the HR sample exhibits

equiaxed grains with random orientation distribution (d = 26.7 μm,

Fig. 1d), attributed to dynamic recrystallization during hot rolling. After

CR50 %, grains are elongated along the rolling direction (RD), forming a

fibrous structure with banded orientations in IPF maps, where grain

orientations begin to align preferentially along the RD < 111 > direction. Cold
deformation induces significant dislocation multiplication,

Table 1

Detailed processing parameters.

Sample Cold Rolling

Reduction

Annealing

Temperature (◦C)

Holding

Time

HR / /

CR50 % 50 % / /
CR50 %− 600℃−

5 min

50 % 600℃ 5 min

CR50 %− 600℃−

15 min

50 % 600℃ 15 min

CR50 %− 600℃−

1h

50 % 600℃ 1 h

CR50 %− 600℃−

2h

50 % 600℃ 2 h

CR70 % 70 % / /

CR70 %− 600℃−

5 min

70 % 600℃ 5 min

CR70 %− 600℃−

15 min

70 % 600℃ 15 min

CR70 %− 600℃−

1h

70 % 600℃ 1 h

CR70 %− 600℃−

2h

70 % 600℃ 2. h

X. Hu et al. Materials Today Communications 49 (2025) 113732

2

elevating stored energy and promoting grain fragmentation and

refinement, resulting in a markedly reduced average grain size (d = 24.7

μm). With increased deformation (CR70 %), grains are further elongated, exhibiting
enhanced fiberization and minimal additional refinement (d = 10.6 μm), approaching
the grain size saturation limit. Severe

plastic deformation during cold rolling drives dislocation slip, forming

dislocation cells and subgrain boundaries (Figs. 1b2 and 1c2). Concurrently, strain
localization in high-strain regions generates deformation

twins (Σ3 boundaries, marked by red lines in Figs. 1a2–c2) and stacking

faults due to dislocation slip obstruction.

The evolution of GND density was further investigated through

geometrically necessary dislocation (GND) mapping (Figs. 1a3–c3). The

results indicate that GND density increases significantly with higher cold

rolling reductions. In the hot-rolled state, recovery, and recrystallization

processes promote dislocation rearrangement, suppressing excessive

GND accumulation, resulting in a low average GND density of

2.86 × 10 ¹ ⁴ m⁻². During cold rolling, intensified plastic deformation

hinders dislocation motion at grain boundaries, leading to substantial

GND multiplication. After CR50 %, the average GND density increases

to 17.65 × 10 ¹ ⁴ m⁻². When the cold rolling reduction reaches 70 %,

severe deformation causes massive dislocation pile-ups, further

elevating the average GND density to 18.87 × 10 ¹ ⁴ m⁻². Analysis of

grain boundary misorientation distributions (Fig. 1e) shows that HR

samples are dominated by high-angle grain boundaries (HAGBs). After

cold rolling, the proportion of low-angle grain boundaries (LAGBs) increases while
HAGBs decrease. This is because sub-grain boundaries

formed by dislocation accumulation during cold rolling typically manifest as LAGBs,

and with increasing deformation degree, sub-grain

boundaries further increase.

3.2. Post-annealing microstructure

After CR50 %, HMAS samples were subjected to aging annealing at

600◦C with holding times of 5 min, 15 min, 1 h, and 2 h to investigate

Fig. 1. EBSD inverse pole figure (IPF) maps, grain boundary (GB) maps, and
geometrically necessary dislocation (GND) maps for HR (a1, a2, a3), CR50 % (b1, b2,
b3),

and CR70 % (c1, c2, c3) samples; (d) grain size distribution; (e) grain boundary
statistics. In GB maps, blue represents low-angle grain boundaries (2◦~10◦), black

denotes high-angle grain boundaries (>10◦), and red indicates special grain boundaries
(Σ3 60◦<111 >).

X. Hu et al. Materials Today Communications 49 (2025) 113732

the effect of holding time on recrystallization fraction, with microstructural

characterization shown in Fig. 2. For the CR50 %-5 min

sample, the IPF map reveals a retained fibrous structure, but localized

regions exhibit nascent fine equiaxed grains. Dislocation rearrangement

through slip and climb partially eliminates subgrain boundaries. The

proportion of LAGBs begins to decrease (LAGBs = 68.8 %), while

recrystallization nucleation initiates, leading to an increase in the proportion of

HAGBs (HAGBs = 31.2 %). Partial twins degrade due to

dislocation rearrangement, decreasing twin density, though deformation twins remain

dominant (Figs. 2a1–a3). At this stage, the material

undergoes recovery, where dislocation rearrangement forms clearer

subgrain structures, while full recrystallization has not yet been initiated. After 15 min
of annealing, fine equiaxed grains nucleate in highstrain-energy regions. Subgrain
boundaries further diminish with

ongoing dynamic recovery, forming a continuous HAGBs network.

Recrystallization disrupts the original twin structure, causing twin

Fig. 2. EBSD IPF maps, GB maps, and LAGBs/HAGBs proportion distributions of

CR50 %-600℃-5 min (a1, a2, a3), CR50 %-600℃-15 min (b1, b2, b3), CR50 %-

600℃-1h (c1, c2, c3), and CR50 %-600℃-2h (d1, d2, d3) samples.

X. Hu et al. Materials Today Communications 49 (2025) 113732

density to decrease abruptly (to 49.4 %).

Both CR50 %-600℃-1h and CR50 %-600℃-2h samples exhibit

numerous fine recrystallized grains, composed of CG (indicated by black

dashed lines) and UFG (marked by white dashed lines) regions. After 1 h

of annealing, sufficient thermal activation energy enables the nucleation

of recrystallized grains (UFG) in localized areas, marking the initial

recrystallization stage. These recrystallized grains adopt an equiaxed

morphology with lamellar distribution, forming a heterogeneous

microstructure where deformed grains (CG) coexist with recrystallized

grains (UFG). The proportion of LAGBs significantly decreases (LAGBs =

44.1 %), while recrystallized boundaries dominate with large misorientation angles
between grains. Extending the annealing time to 2 h

enhances recrystallization nucleation and growth, as further stored energy release

promotes additional recrystalliza