Physica A 468 (2017) 307–314

Contents lists available at ScienceDirect

Physica A
journal homepage: www.elsevier.com/locate/physa

A quantum anharmonic oscillator model for the stock market

Tingting Gao, Yu Chen ∗
Department of Human and Engineered Environmental Studies, Graduate School of Frontier Sciences, The University of Tokyo, 5
Chome-1-5 Kashiwanoha, Kashiwa, Chiba Prefecture 277-8561, Japan

highlights
• A quantum anharmonic oscillator model is proposed to describe the stock price return.
• Excited states are used to explain the leptokurtic distributions of stock indices.
• The model is applicable to both liquid and illiquid markets.
• Data filtering is applied to extract price data for the trend following behavior.

article info abstract

Article history: A financially interpretable quantum model is proposed to study the probability distribu-
Received 5 July 2016 tions of the stock price return. The dynamics of a quantum particle is considered an analog
Received in revised form 6 October 2016 of the motion of stock price. Then the probability distributions of price return can be com-
Available online 2 November 2016
puted from the wave functions that evolve according to Schrodinger equation. Instead of a
harmonic oscillator in previous studies, a quantum anharmonic oscillator is applied to the
Keywords:
stock in liquid market. The leptokurtic distributions of price return can be reproduced by
Quantum anharmonic oscillator
Stock market
our quantum model with the introduction of mixed-state and multi-potential. The trend
Potential operator following dominant market, in which the price return follows a bimodal distribution, is
Probability distribution discussed as a specific case of the illiquid market.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction

As a newly developed interdisciplinary subject applying physics theories and methods into economic studies,
econophysics was first introduced in the mid-1990s, followed by a series of relevant works [1–4]. However, econophysics-
like study of finance had been started much earlier. In 1900, L. Bachelier in his doctoral thesis described a dynamic model
of option price using stochastic process [5], which was honored as the first piece of work in mathematical finance. Modern
econophysicists view this work as an application of Brownian motion, the fundamental phenomenon of statistical physics,
to the modeling of financial market. During the last two decades, different kinds of theories and methods originating from
physics have been used for stock price analysis [6,7], option pricing [8,9], portfolio management [10,11], etc. Although
statistical physics remains the major resource for methodologies, tools from other branches such as quantum mechanics
also played an important role in the field of econophysics [8,9]. For example, B. Baaquie systematically explained the path
integral method for option pricing in his book [12], while M. Schaden tried to use wave functions of cash and security to
describe the state of a financial market [13].
Concurrently, another body of econophysicists concentrated on the dynamics of stock price with the help of quantum
wave packets [14–17]. Ye et al. proposed a quantum harmonic oscillator model in 2008 to explain the persistent fluctuations

∗ Corresponding author.
E-mail address: chen@k.u-tokyo.ac.jp (Y. Chen).

http://dx.doi.org/10.1016/j.physa.2016.10.094
0378-4371/© 2016 Elsevier B.V. All rights reserved.
308 T. Gao, Y. Chen / Physica A 468 (2017) 307–314

in stock markets. The squared modulus of the ground state wave function was used as a probability measure of the stock
price in stationary market [14]. As the study focused on the dynamics of the stock price under sudden information, no
convincible financial interpretation of its Hamiltonian was addressed and one could deplore a lack of comparison between
the derived probability distributions and that of the classical models. Zhang et al. explained the financial meaning of the wave
function and gave a financial form of the uncertainty principle in 2010 [15]. Although they demonstrated the feasibility
of quantum model by proposing a cosine formed potential, the model did not show any advantage over the quantum
harmonic oscillator or the classical random walk. In this study, we propose a financially interpretable quantum Hamiltonian
for stationary market and compute the probability density functions of price return from the Schrodinger equation. The
potential consisting of a quadratic and a quartic term, which can be widely applied to different market environments, is
derived from the dynamics of order excess in the stock market.
In the following parts of this paper, basic quantum variables and the corresponding dynamic equations (Schrodinger
equation) are specified in Section 2. In Section 3, the potential operator with detailed financial interpretation for a stationary
market is proposed, followed by the adjustment of parameters for liquid market. The trend following dominant market is
discussed as a special case of illiquid market by both analyzing the real data and adjusting the parameters of the quantum
model in Section 4. Discussion and conclusion are addressed in Section 5.

2. A quantum description of the stock market

In order to study the stylized facts of the stock price, we refer to the price return as the fractional price change
p(t ) − p(t − 1t )
r (t , 1t ) = , (1)
p(t − 1t )
with p(t ) the stock price at time t and 1t the time interval [18]. Assuming that the motion of a stock price is similar to that of
a quantum particle, the price return can then be regarded as an analog to the position of the quantum particle. The position of
a quantum particle is indeterminate and follows a probability distribution, according to Copenhagen interpretation, which
is equal to the squared modulus of a wave function ψ(r , t ). Thus the stock price can analogously be described by a wave
function of price return whose dynamics is described by Schrodinger equation as
∂
ih̄ ψ(r , t ) = H ψ(r , t ), (2)
∂t
with Hamiltonian operator

h̄2 ∂ 2
H =− + V (r , t ), (3)
2m ∂ r 2
where i is the imaginary unit, h̄ is the reduced Planck constant.
The Hamiltonian operator consists of a kinetic and a potential energy part, where h̄ can be regarded as the uncertainty of
irrational transaction and m represents the intrinsic properties of the stock such as capital [15,17]. It shows that the state of a
stock particle is influenced by both its internal properties and the external potential field. Assuming that both h̄ and m are of
unit value, the potential operator remains as the variable part of the model. Focusing on the stationary market environment
in this study, we use a time independent potential to reduce Eq. (2) into
H (r )ϕ(r ) = E ϕ(r ), (4)
with ψ(r , t ) = ϕ(r )T (t ) and T (t ) = exp(−iEt /h̄), where E is a real number interpreted as the eigenenergy of the stock
price and ϕ(r ) is the corresponding eigenfunction.
Instead of the classical consideration that the price return is of a fixed value at a definite time, in the quantum description,
the price return is indeterminate before ‘‘measurement’’ (the finish of a deal) and follows a probability distribution according
to different possible values. The probability density of price return in a stationary market is obtained from

ρ(r , t ) = |ψ(r , t )|2 = |ϕ(r )|2 . (5)

3. A quantum anharmonic oscillator model

We firstly induce an equation about price return from the dynamics of order excess. The relation equation of the
instantaneous price return r (t ) and instantaneous order excess 1φ is
r (t ) = 1φ/λ, (6)
1 dp(t )
where r (t ) = , 1φ = φ+ − φ− (φ+ : instantaneous demand; φ− : instantaneous supply), and λ is the measurement
p(t ) dt
of market depth [19].
We assume that the market consists of three types of participants: market makers, contrarians and trend followers.
Market makers, working as intermediaries, absorb the existing orders by providing buy and sell quotations of the stocks
 T. Gao, Y. Chen / Physica A 468 (2017) 307–314 309

hold by themselves and other market participants. Contrarians and trend followers place buy or sell orders according to the
anticipated return of a stock. In this model, the anticipated return can be replaced by the instantaneous price return since
we ignore memory effect. The contrarians sell their holdings when the stock price increase (positive price return) and place
buy orders when the stock price decrease (negative price return). The trend followers act conversely. Both the contrarians
and trend followers are assumed risk avoiders. Based on these behavioral characteristics of the market participants, we can
construct the dynamic equations of the order excess (buy and sell orders) in the market as
dφ ± 

 = −γ φ± ; (7a)
dt MM
d1φ 

 = −(aCT − bCT r 2 )r ; (7b)
dt CT
d1φ 

 = (aTF − bTF r 2 )r . (7c)
dt  TF

In Eq. (7a) for the market makers, γ is a positive parameter measuring their collective absorbing ability, which is assumed to
be symmetric for buy and sell orders; the minus sign indicates that the market makers’s operation decreases the number of
orders by clearing the market. In Eqs. (7b) and (7c), aCT , aTF are positive parameters representing respectively the behavioral
dependencies of contrarians and trend followers on the anticipated price return, where the traders’ responses to positive
and negative returns have been assumed symmetric; the terms −bCT r 2 and −bTF r 2 show that the contrarians and the trend
followers are more hesitated to order when the volatility (r 2 is used as instantaneous volatility) of the price return becomes
larger. Moreover, the risk aversion terms should not be too large as to reverse the intended action of the contrarians or
trend followers, i.e. aCT − bCT r 2 ≥ 0 and aTF − bTF r 2 ≥ 0. All the parameters reflect the collective behaviors of the market
participants, including the information of both the population and the activeness per participant.
From Eq. (6) and the dynamics of order excess Eq. (7), we obtain the dynamic equation of r (t ) as

dr
= −(γ + c )r + kcr 3 , (8)
dt
where c = (aCT − aTF )/λ, representing the competition between contrarians and trend-followers without the concern of
risk aversion (c > 0 means more orders placed by the contrarians while c < 0 means more orders from the trend followers,
which will be described as stronger contrary trading and stronger trend following respectively in the following text); the
risk aversion effect is assumed to be identical for the contrarians and the trend followers, i.e. k = bTF /aTF = bCT /aCT with
k ∈ [0, 1/r 2 ].1
As in previous literatures [19], the Brownian price return can be described by a Langevin equation:

d2 r dr dV (r )
mr = −η − + ξ (t ), (9)
dt 2 dt dr
where mr is the mass of the fictitious particle of price return, η is the damping coefficient, V (r ) is a time-independent
potential, and ξ (t ) is a random force. The requirement of consistency between Eqs. (8) and (9) results in the overdamped
2
condition, mr ddt 2r = 0, as well as

dV (r )
− = −(γ + c )r + kcr 3 . (10)
dr
The overdamped condition implies that the mass of price return mr is ignorable in the classical model. Thus we derive a
financially interpretable potential in the form of

γ +c kc
V (r ) = r2 − r 4. (11)
2 4
Adopting this potential into Eq. (3), we can rewrite the Schrodinger equation of our quantum model

1 d2 γ +c
  
kc
− + r2 − r4 ϕ(r ) = E ϕ(r ). (12)
2 dr 2 2 4
The efficient stock market is always of high liquidity, which is ensured by the strength of market makers in our model. Thus
γ ≫ |c | is required in a liquid market, resulting in a potential operator (consisting of a positive quadratic term and a quartic
perturbation term) having the potential form of an anharmonic oscillator.

1 Risk aversion effect here is measured by the parameter k, which together with the instantaneous volatility r 2 determines the shrinkage percentage of
the new orders.
310 T. Gao, Y. Chen / Physica A 468 (2017) 307–314

Fig. 1. The probability distributions of the price return for stationary liquid market with different c values, where γ = 1 and k = 0.01 are assumed and
the distribution curves have been normalized to those of the same volatility (σ 2 = 1).

Fig. 2. (a) The dependency of kurtosis on c with γ = 1 and k = 0.01. (b) The dependency of kurtosis on k with different c /γ values.

It is known that the solutions of Schrodinger equation are not unique, but multiple with eigenvectors corresponding to
different energy levels. However, as the excited states are always unstable, we can use the wave function of the ground state
to describe the stock in a stationary market. The squared modulus of the ground state wave function of a harmonic oscillator
can be analytically obtained as a Gaussian function, indicating that a quantum harmonic oscillator model can describe the
price return of a stock as accurately as the classical random walk model. As an anharmonic quantum oscillator, our model
is expected to provide more information than Gaussian distribution.
Schrodinger equation for the anharmonic oscillator is solved with finite difference method (FDM) [20], and the obtained
probability distributions are shown in Fig. 1. The range of c is considered as c ∈ [−0.2γ , 0.2γ ] due to the liquidity
requirement. It can be seen that the introduction of the quartic perturbation term in the potential causes the deviation
of the probability distribution from Gaussian, namely, a more negative quartic term results in sharper peak and fatter tails
of the distribution.
Fig. 2 gives the kurtoses of probability distributions corresponding to different c and k respectively, illustrating that the
kurtosis is closely related to risk aversion. It indicates that a more negative quartic perturbation term (positive kc) can bring
a better modeling result because a probability distribution of price return with a sharper peak and fatter tails can emerge,
which is consistent with Fig. 1. In other words, the risk aversion is a key factor that contributes to the stylized facts of stock
price in real market, and the leptokurtic distributions of price return may partially be caused by the weaker collective risk
aversion of trend followers than that of contrarians (and also that of market makers if it is considered).
By adjusting the parameters of the potential in the quantum model, we can reproduce the probability distributions for
different market data. Given the time series of a market index with the time interval 1t, such as one day for daily data and
one month for monthly data, the index return of the jth tick, rj = (pj − pj−1 )/pj−1 , can be considered the instantaneous
return of the quantum model with dt = 1. As the influence of parameters c and k on the modeling result is small (refer
Figs. 1 and 2), we focus on the parameter γ in the adjustment. It is noted that the feasibility of adjusting γ to price series is
based on the assumptions: (1) the intrinsic property of the stock is kept invariant in the stationary market, (2) the close of
each deal is a measurement of the stock price, and (3) the statistical distribution of price return series is approximated to
the stationary probability distribution before a measurement.
Fig. 3(a) shows the modeling results of monthly and daily data of Nikkei 225 using the ground states of the quantum
oscillators. The ground state gives a reasonable modeling for the monthly data, where the imperfectness can be explained
 T. Gao, Y. Chen / Physica A 468 (2017) 307–314 311

Fig. 3. (a) Fitting of the quantum anharmonic oscillator at ground state to the monthly and daily Nikkei 225 Index from January 4, 1996 to June 1, 2016,
by adjusting γ . Other parameters have been assumed as c = 0.2γ and k = 1. (b) Fitting of the quantum anharmonic oscillator at mixed-state to the daily
Nikkei 225 Index. The dotted line is the probability distribution described by the ground state of the quantum model (γ = 1.55 × 107 ); the dashed line is a
mixed-state distribution of Eq. (13) with ω0 = 1 and ωn = 0.005, n = 1, 2, . . . , 10 (γ = 1.80 × 107 ); the solid line is the combination of two mixed-state
distributions with almost the same parameter settings except for γ1 = 4.55 × 107 and γ2 = 5.20 × 106 respectively.

Table 1
The adjusted γ with corresponding values of standard deviation σ and kurtosis κ for different market data using different modeling methods. The standard
deviations of probability distributions have been required to be consistent with the market indices. The multi-potential modeling also gives consistent
kurtoses.
Market indices Data Ground-state M. Mixed-state M. Multi-potential M.
σ κD γ κM1 γ κM2 γ1 γ2 κM3
Nikkei 225 0.0154 8.34 3.75 × 10 6
3.0003 8.50 × 10 6
6.69 3.40 × 10 7
3.88 × 10 6
8.34
SSE composite index 0.0171 7.96 2.44 × 106 3.0003 5.65 × 106 6.70 1.75 × 107 2.73 × 106 7.96
S&P 500 0.0123 10.76 9.10 × 106 3.0005 2.10 × 107 6.69 4.32 × 108 6.75 × 106 10.76

as that the data sample in fact covers a long period when the potential of the market is not rigorously time independent. On
the other hand, the ground state seems unable to model the daily data well although the data covers the same period as the
monthly one. We attribute it to the neglect of the excited states of the stock. It has been mentioned that the use of ground
state is based on the assumption that the stock at excited states is not stable, similarly to physical quantum particles, which
means that the stock would return back to the ground state in a short time if it were in a excited state. Thus the ground state
is sufficient to model the monthly data because the time interval is long enough for the stock to return back to the ground
state. But the situation would be different for the daily data or data with higher frequencies when the time intervals is too
short for the stock to return back.
In order to model the distributions of the real market data with short time intervals, we define a mixed-state probability
density function:
N
1 
|Ψ (r )|2 = ωn |ϕn (r )|2 , (13)
Ω n=0

where n represents the energy level, ωn is the weight of the nth state proportional to the time that the stock stays in the nth
n=0 ωn , and |ϕn (r )| is the probability density function of the nth energy level. Although it is not easy
N
excited state, Ω = 2

to precisely determine the weights of different states for a given data sample, we can just screen weights to improve the
modeling result. As mentioned above that the potential of the data sample is in fact changing with the time, the combination
of the mixed-state probability distributions for different potentials is supposed to be more capable of modeling the data. It
is shown in Fig. 3(b) that the mixed-state probability distribution fits the data much better in tails than the ground state
one although the weights of excited states are much smaller than the weight of ground state. Moreover, the introduction of
multi-potential makes the derived distributions closer to that of the real data. We may thus say that the stock stays at the
ground state most of the time but sometimes jumps to the excited states possibly under the influence of abrupt information.
The adjusted parameters of other market indices (SSE Composite Index and S&P 500) for the same period are listed in
Table 1. The statistics of the fitted probability distributions show the relative appropriateness of the three modeling methods
(refer Fig. 3(b)) as: Multi-Potential Modeling > Mixed-State Modeling > Ground-State Modeling.

4. The trend following dominant market

Illiquid markets can be modeled by large |c |/γ , which indicates that the strength of market makers is weak so that
the order cannot be absorbed efficiently. In this case, if the market were dominated by contrarians with a large positive c,
312 T. Gao, Y. Chen / Physica A 468 (2017) 307–314

Fig. 4. (a) The probability distributions for the markets with different strength of trend following, where c ′ = c /γ + 1. (b) The time series of daily
Nikkei225 Index from March 1, 1996 to February 29, 2016 (it is noted that only the first 500 ticks are plotted in this figure to help distinct the price and
MA line), the corresponding 60 day MA, and an example of artificial price series obtained by shuffling the price returns of the original Nikkei 225 daily
data, where the initial prices are set the same. (c) The probability distributions for trend following dominant market obtained from real data (the number
amounts to 303) and the quantum model with the same volatility (σ = 0.0190). The adjusted parameters are: γ = 2.4 × 1011 , c = −1.00047γ , k = 1.
(d) The standard deviations and kurtoses of the probability distributions for trend following dominant market, calculated from the original Index data and
100 shuffled random data series.

the probability distributions of price return would be similar to that for liquid market (see Fig. 1). On the other hand, the
probability distributions for the market with negative c of relatively large absolute values, illustrated in Fig. 4(a), shows
that when the trend following became stronger, the peak of the distribution in the middle would collapse, which causes the
unimodal distribution turning into a bimodal one. The bimodal distribution implies that if a huge number of trend followers
participated the market, the stock price would be apt to crash or boom.
We can extract the data for trend following dominant market from a consecutive price series with the help of moving
average (MA). There are several trading rules [21], named Granville’s rules, widely used by the market participants (chartists)
as a simple but efficient tool for timing buy and sell signals. For example, as shown in Fig. 4(b), it is regarded as a buy signal
when the stock price moves upward through a flat or rising moving average line, and conversely, if the stock price moves
downward though a flat or declining moving average line, it may be a good time for the trend followers to sell. Assuming that
the trend followers becomes extremely active around the crossovers of the moving average and the stock price, we collect
the price returns around those crossovers, finding that the distribution of the filtered data is consistent with that of our model
for trend following dominant market as in Fig. 4(c). The adjusted parameter c /γ , although has larger absolute value than the
liquid market, is close to −1, which means that the trend following is strong but not strong enough to overwhelm the effect
of market makers. This can be attributed to the coarseness of the data filtering process. We used the 60 day MA while in
reality the trend followers utilize MA with different length of MA terms and they also use other methods to help themselves
make decisions. The data sample we obtained by the simplified Granville’s rules inevitably includes some impurities.
In order to confirm the reliability of the filtered data as for the trend following dominant market, we shuffle the price
returns of the original price series to produce random price series (one of them is plotted in Fig. 4(b)) and filter them with
the simplified Granville’s rule illustrated above. The probability distribution of the filtered data for each random price series
is marked as a point in Fig. 4(d) according to its kurtosis and standard deviation. It is shown that the points for the random
price series cluster at a distance from that for the original data, indicating that the chartists do not trade randomly around
the crossovers.
In addition, we checked the distributions of filtered data using MA with different lengths of MA term and for other
markets such as SSE Composite Index and S&P 500 (Fig. 5). It is shown that the short-term MAs (such as those with the
 T. Gao, Y. Chen / Physica A 468 (2017) 307–314 313

Fig. 5. (a) Probability distributions of filtered Nikkei 225 daily data according to different lengths of MA term. (b) Standard deviations of the filtered data
for Nikkei 225, SSE Composite Index and S&P 500 respectively.

length of MA term shorter than 5 days) are rarely considered by the chartists. The standard deviations of the filtered data, as
a measurement of the utilization frequency corresponding to different MAs, indicate the chartists’ preference of the length
of MA term. Furthermore, long-term MAs are more preferable in Nikkei 225 and S&P 500 than in SSE Composite Index.

5. Discussion and conclusion

Starting from a market containing three types of participants, we have introduced a financially interpretable quantum
anharmonic oscillator which can be used to model the motion of stock price in a liquid market. It is obvious that a
classical oscillator with the same potential form is incapable of describing the statistics of the stock price motion. Instead
of a harmonic oscillator, the quantum anharmonic one considering the risk aversion is more reasonable in modeling the
stock in the real market. The derivation of the potential operator gives the financial interpretations of previous quantum
models based on isolated systems [14–17] or open system [22]. The modeling of liquid market told us that the leptokurtic
distributions may be partially attributed to the weaker risk aversion behavior of trend followers. The fitted probability
distribution of anharmonic oscillator at mixed-state shown that instead of only staying at the ground state, the stock in real
market sometimes may jump into excited states and returns back in a short time. Moreover, a time independent potential
is not sufficient for the movement of a stock price since the market environment changes along the time. We also analyzed
the trend following dominant market by filtering the data with the simplified Granville’s rules and succeeded in modeling
it with our quantum model. The filtered data indicated some patterns of the usage of MAs.
Although further quantitative analysis, such as the financial correspondence of energy level and the measurement of
potential is necessary, this study shows that our quantum model is better than the simple random walk model and previous
quantum models at describing the probability distributions of price return in stock market.

Acknowledgment

We acknowledge the financial support from China Scholarship Council (CSC).

References

[1] R.N. Mantegna, H.E. Stanley, An Introduction to Econophysics: Correlations and Complexity in Finance, Cambridge University Press, Cambridge, 1999.
[2] M. Takayasu, T. Mizuno, H. Takayasu, Potential force observed in market dynamics, Physica A 370 (2006) 91–97. http://dx.doi.org/10.1016/j.physa.
2006.04.041.
[3] T. Mizuno, H. Takayasu, M. Takayasu, Analysis of price diffusion in financial markets using puck model, Physica A 382 (2007) 187–192. http://dx.doi.
org/10.1016/j.physa.2007.02.049.
[4] J. Huang, Experimental econophysics: Complexity, self-organization, and emergent properties, Phys. Rep. 564 (2015) 1–55. http://dx.doi.org/10.1016/
j.physrep.2014.11.005.
[5] L. Bachelier, The Random Character of Stock Market Prices, MIT Press, Cambridge, 1964.
[6] Y. Liu, P. Gopikrishnan, P. Cizeau, M. Meyer, C.-K. Peng, H.E. Stanley, Statistical properties of the volatility of price fluctuations, Phys. Rev. E 60 (2)
(1999) 1390–1400.
[7] E. Scalas, The application of continuous-time random walks in finance and economics, Physica A 362 (2006) 225–239. http://dx.doi.org/10.1016/j.
physa.2005.11.024.
[8] V. Linetsky, The path integral approach to financial modeling and options pricing, Comput. Econ. 11 (1998) 129–163.
[9] A. Cassagnes, Y. Chen, H. Ohashi, Path integral pricing of outside barrier Asian options, Physica A 394 (2014) 266–276. http://dx.doi.org/10.1016/j.
physa.2013.09.067.
[10] S. Sharifi, M. Crane, A. Shamaie, H. Ruskin, Random matrix theory for portfolio optimization: A stability approach, Physica A 335 (2004) 629–643.
http://dx.doi.org/10.1016/j.physa.2003.12.016.
[11] J. Daly, M. Crane, H. Ruskin, Random matrix theory filters in portfolio optimisation – a stanility and risk assessment, Physica A 387 (2008) 4248–4260.
http://dx.doi.org/10.1016/j.physa.2008.02.045.
[12] B.E. Baaquie, Quantum Finance, Cambridge University Press, Cambridge, 2004.
314 T. Gao, Y. Chen / Physica A 468 (2017) 307–314

[13] M. Schaden, Quantum finance, Physica A 316 (2002) 511–538.

[14] C. Ye, J. Huang, Non-classical oscillator model for persistent fluctuations in stock markets, Physica A 387 (2008) 1255–1263. http://dx.doi.org/10.1016/
j.physa.2007.10.050.
[15] C. Zhang, L. Huang, A quantum model for the stock market, Physica A 389 (2010) 5769–5775. http://dx.doi.org/10.1016/j.physa.2010.09.008.
[16] L.-A. Cotfas, A finite-dimensional quantum model for the stock market, Physica A 392 (2013) 371–380. http://dx.doi.org/10.1016/j.physa.2012.09.010.
[17] X. Meng, J.-W. Zhang, J. Xu, H. Guo, Quantum spatial periodc harmonic model for daily price-limited stock markets, Physica A 438 (2015) 154–160.
http://dx.doi.org/10.1016/j.physa.2015.06.041.
[18] N.F. Johnson, P. Jefferies, P.M. Hui, Financial Market Complexity, Oxford University Press, New York, 2003.
[19] J.-P. Bouchaud, R. Cont, A Langevin approach to stock market fluctuations and crashes, Eur. Phys. J. B 6 (1998) 543–550.
[20] R.J. LeVeque, Finite Difference Methods for Ordinary and Partial Differential Equations: Steady-State and Time-Dependent Problems, SIAM,
Philadelphia, 2007.
[21] J.E. Granville, A Strategy of Daily Stock Market Timing for Maximum Profit, Prentice-Hall, 1960.
[22] X. Meng, J.-W. Zhang, H. Guo, Quantum Brownian motion model for the stock market, Physica A 452 (2016) 281–288. http://dx.doi.org/10.1016/j.
physa.2016.02.026.