Linear Programming Models

Prof. Marco Ghirardi

 Mathematical Programming Models
Mathematical programming is the selection of a best element (with
regard to some criteria) from some set of available alternatives.

In the simplest case, an optimization problem consists of

maximizing or minimizing a real function by systematically
choosing input values from within an allowed set:
 Mathematical Programming Models

A mathematical programming model is composed by:

• Decision Variables: representing (unknown) decisions to be

made. This is in contrast to problem data, which are values that
are either given or can be simply calculated from what is
given.

• Objective function f(x): objective to be either minimized or

maximized.

• Constraints: limiting feasible decisions.

LINEAR programming models
 Linear Programming Models
• When the objective function and the contraints of the model are a
linear function of the decision variables x, the model is a linear
programming model.

• The variable domain is continuous and non negative by default.

• Each variable can also be defined as integer (Z+) or binary (0-1). If at

least one variable is integer (or boolean) the model is called Integer
Programming Model (ILP).

• Special computer programs (called ‘solvers’) implementing

mathematical algorithms are able to handle those model efficiently.
• The case of continuous variable LP is the easier. Problem may become
more complicated (depending on the number of variables) in the
integer case.

• Some solvers also handle non-linear problems, with fairly good results
on special cases (quadratic models).
 Introductory Example
• SilComputers makes quarterly decisions about their
product mix. While their full product line includes hundreds
of products, we will consider a simpler problem with just
two products: notebook computers and desktop
computers.

• SilComputers would like to know how many of each

product to produce in order to maximize the profit for the
quarter.

• There are a number of limits on what SilComputers can

produce. The major constraints are as follows:
 Introductory Example
• Each computer requires a CPU. Due to tightness in the market, our
supplier has allocated 10,000 such chips to us.
• Each computer requires memory chips. A notebook needs 1 chip set
while a desktop requires 2 chip sets. We have a stock of 15,000 chip
sets to use over the next quarter.
• Each computer requires assembly time. Due to tight tolerances, a
notebook computer takes more time to assemble: 4 minutes versus 3
minutes for a desktop. There are 25,000 minutes of assembly time
available in the next quarter.
• Each notebook computer produced generates $750 profit, and each
desktop produces $1000 profit.
 Introductory Example

𝑴𝒂𝒙𝒊𝒎𝒊𝒛𝒆 𝟕𝟓𝟎. 𝟎𝟎𝟎𝒙𝟏 + 𝟏. 𝟎𝟎𝟎. 𝟎𝟎𝟎𝒙𝟐 Total profit

𝒙𝟏 + 𝒙𝟐 ≤ 𝟏𝟎 CPU availability
𝒙𝟏 + 𝟐𝒙𝟐 ≤ 𝟏𝟓 Memory availability
𝟒𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟐𝟓 Production time
Thousands of produced
𝒙𝟏 ≥ 𝟎, 𝒙𝟐 ≥ 𝟎
computers
 Linear Programming Models
• Models are useful in their own right: they
allow for a formal definition of a problem and
can be useful in thinking about a problem.
• Linear programming models are particularly
useful because we have efficient solvers able
to handle them for obtaining the optimal
decision values.
• In our example (2 variables), it is possible to
solve the problem graphically.
Constraints and feasible region
Finding the optimal profit
We include the objective function on the diagram
by drawing iso-profit lines: lines along which the
profit is the same.
Since our goal is to maximize the profit, we can
push the line out until moving it any further would
result in no feasible point. This happens at point X.
The optimal solution is found at the intersection of the
two constraints:

𝒙 + 𝟐𝒙𝟐 ≤ 𝟏𝟓
ቊ 𝟏
𝟒𝒙𝟏 + 𝟑𝒙𝟐 ≤ 𝟐𝟓

The solution is x1 = 1 and x2 = 7.

Hence, the optimal decision is to produce 1,000 notebooks and

7,000 desktops, for a total profit of $7.750.000
 The knapsack problem
• A bunch of friends is organizing an excursion and decides to put
everything in a single knapsack with capacity 10 Kilos. The friends assigned
a score from 1 to 100 to each product. They want to choose how much of
each product to load in the knapsack, in order to maximize the total score.

• The knapsack may be filled with:

– Chocolate (500g boxes) – 10 points
– Fruit juices (1 l. bottle) – 30 points
– Beer (0.33 l cans) – 6 points
– Sandwiches (100 g each) – 3 points
– Mineral water (1 l. bottle) – 20 points
– Cookies (500 g. boxes) – 8 points

• It was also decided to guarantee a minimum amount of each product,

namely: 2 boxes of Chocolate, 2 bottles of fruit juices, 6 cans of beer, 10
sandwiches, 1 bottle of mineral water and 2 boxes of cookies.
The knapsack problem
 The KP0/1 problem
• The most famous version of this problem is
the Knapsack 0/1 problem, where each
product can be loaded or not in the knapsack.
• We have a set of N of n items, where item i
has weight wi and utility ui.
 A plant problem
• A plant has to fill an order of 1000 tons of a product requiring
manganese (at least 1%), chrome (at least 18%) and molybdenum
(at least 2%).

• The suppliers sell these three metals in three different kinds of

packings. The first packing contains 2 Kilos of manganese, 2 Kilos of
chrome and 1 Kilo of molybdenum and costs 20 EUROS. The
second packing contains 2 Kilos of manganese, 3 Kilos of chrome
and 1 Kilo of molybdenum and costs 30 EUROS. The third packing
contains 1 Kilo of manganese, 2 Kilos of chrome and 5 Kilos of
molybdenum and costs 40 EUROS.

• Formulate a Linear Programming model of this problem minimizing

the total cost.
A plant problem
 A publisher problem
• A publisher publishes a newspaper whose distribution is handled through
four main distribution centers D1, D2, D3 e D4 that require daily 100000,
150000, 50000 and 75000 copies of the newspaper respectively .

• The newspaper is printed in three printing houses P1, P2 and P3 that print
daily 125000, 180000 and 70000 copies of the newspaper respectively .

• It is known that the shipping costs are 0.1 EUROS/Km (for a single copy of
the newspaper). The distances between printing houses and distribution
centers are 20, 25, 15 and 5 Km. respectively for
• the first printing houses, 12, 14, 18 and 30 Km. respectively for the
• second printing house and 19, 11, 40 e 12 Km. respectively for the third
printing house.

• Formulate a Linear Programming model of this problem minimizing the

total cost.
A publisher problem
 A publisher problem

What if I use equalities constraints?

 A publisher problem
• The total number of newspapers requested from the
publishers is 100K+150K+50K+75K = 375K
• The total number of newspapers available from the printing
houses is 125K+180K+70K = 375K

• Hence, all 375K available papers will be delivered, in the

exact required quantities. The constraints can be written as
equalities.

• In case we had more newspapers available or requested we

clearly have newspaper capacity not used, or requested
papers not delivered (in this case we need inequalities).
 A publisher problem

Can the model be written in a compact form

defining some data vectors/matrices?
 A publisher problem
• Yes! Defining:
– Number of sources (S=3) and destinations (D=4)
– Matrix C where cij is the unitary cost of delivery from i to j (note:
– Vector A where Ai are the newspapers available at source i
– Vector B where Bj are the newspaper requested at destination j
𝑆 𝐷

min ෍ ෍ 𝑐𝑖𝑗 𝑥𝑖𝑗

𝑖=1 𝑗=1
𝑛

෍ 𝑥𝑖𝑗 ≥ 𝐵𝑗 ∀𝑗 = 1, . . , 𝐷
𝑖=1
𝑛

෍ 𝑥𝑖𝑗 ≤ 𝐴𝑖 ∀𝑖 = 1, . . , 𝑆
𝑗=1
 Trasportation problem

𝑆 𝐷
In this general form, this is
min ෍ ෍ 𝑐𝑖𝑗 𝑥𝑖𝑗 called the
𝑖=1 𝑗=1
𝑛
TRANSPORTATION PROBLEM
෍ 𝑥𝑖𝑗 ≥ 𝐵𝑗 ∀𝑗 = 1, . . , 𝐷
𝑖=1 Another quite famous
𝑛
problem in the combinatorial
෍ 𝑥𝑖𝑗 ≤ 𝐴𝑖 ∀𝑖 = 1, . . , 𝑆
optimization field.
𝑗=1
A motel problem
A motel problem
 A multi-period production problem
A firm is planning the production for the next four months.
It is known the there is currently no inventory and it is required to have
no inventory at the end of the four months.
Formulate a Linear Programming model of this problem minimizing the
total production and inventory cost and satisfying sales forecast and
production capacity constraints.
Variables 𝑥𝑖 ∈ 𝑍 +
xi: production quantity - month i
zi: inventory level at the end of month i 𝑧𝑖 ∈ 𝑍 +

Min z = (34 x1 + 36 x2 + 32 x3 + 38 x4 ) + (2 z1 + 3z2 + 2 z3 )

subject to

z1 = x1 − 20
z 2 = x2 + z1 − 30
z3 = x3 + z 2 − 50
0 = x4 + z3 − 40
x1  40 x2  50 x3  30 x4  50
• Note that the constraints which impose
that the production is enough for the
demand at each months should have been:
x1  20
x2 + z1  30
x3 + z 2  50
x4 + z3 = 40

• However, they are unuseful, because the

constraints on the non-negativity of the
inventory level already imply those
constraints.
 A multi-period production problem
Consider the same problem, where the sale forecasts are increased to
30, 50, 60, 50 (months 1,2,3,4).

The available production capacity is now not enough for serving the
demand (previous model would result in an infeasible solution result).

The company has the option to activate overtime work. The cost for the
activation is 500, and in case of activation, 10 additional item are
produced in the month.
Variables 𝑥𝑖 ∈ 𝑍 +
xi: production quantity - month i
yi: overtime production activated (1) or not (0). 𝑧𝑖 ∈ 𝑍 +
zi: inventory level at the end of month i 𝑦𝑖 ∈ {0,1}

Min z = (34 x1 + 36 x2 + 32 x3 + 38 x4 ) + (2 z1 + 3z2 + 2 z3 ) +

+500(𝑦1 + 𝑦2 + 𝑦3 + 𝑦4 )

subject to

𝑧1 = 𝑥1 + 10𝑦1 − 30
𝑧2 = 𝑥2 + 10𝑦2 + 𝑧1 − 50
𝑧3 = 𝑥3 + 10𝑦3 + 𝑧2 − 60
0 = 𝑥4 + 10𝑦4 + 𝑧3 − 50

x1  40 x2  50 x3  30 x4  50
 A multi-period production problem
Consider the following additional constraints to the previous model:

• The overtime work cannot be activated for two consecutive months.

• An additional constraint for the company is to avoid non-uniform

production (months with no production). The constraint is that for
each month, the production (excluding overtime) must be at least the
10% of the total production over the 4 months.
 A multi-period production problem
• The overtime work cannot be activated for two consecutive months.
𝑦1 + 𝑦2 ≤ 1
𝑦2 + 𝑦3 ≤ 1
𝑦3 + 𝑦4 ≤ 1

• For each month, the production must be at least the 10% of the total
production over the 4 months.
x1  0.1( x1 + 10 y1 + x2 + 10 y2 + x3 + 10 y3 + x4 + 10 y4 )
x2  0.1( x1 + 10 y1 + x2 + 10 y2 + x3 + 10 y3 + x4 + 10 y4 )
x3  0.1( x1 + 10 y1 + x2 + 10 y2 + x3 + 10 y3 + x4 + 10 y4 )
x4  0.1( x1 + 10 y1 + x2 + 10 y2 + x3 + 10 y3 + x4 + 10 y4 )
 A production planning problem
• A given product is composed by three parts that can be processed
on four different production lines. Each production line has a
limited capacity in terms of available production hours.
• The following table indicates the number of parts produced per
hour by each production line with respect to each part together
with the capacity (in hours) of each production line.
• Formulate a Linear Programming model of this problem maximizing
the number of product units.
A production planning problem
 10 x11 + 15 x21 + 20 x31 + 10 x41 ,
 
max min  15 x12 + 10 x 22 + 5 x 32 + 15 x 42 
,
 5 x + 5 x + 10 x + 20 x 
 13 23 33 43 

x11 + x12 + x13  100

x21 + x22 + x23  150 xij   +
x31 + x32 + x33  80
x41 + x42 + x43  200

xij represents the hours dedicated to production of part j on line i.

The objective function written in this way is non linear.
To have an equivalent linear formulation, it is sufficient to introduce a
new variable y and modify correspondingly the cost function as follows
max y

10 x11 + 15 x21 + 20 x31 + 10 x41  y

15 x12 + 10 x22 + 5 x32 + 15 x42  y
5 x13 + 5 x23 + 10 x33 + 20 x43  y

x11 + x12 + x13  100

x21 + x22 + x23  150 xij   +
y0
x31 + x32 + x33  80
x41 + x42 + x43  200
 Max-Min
• Consider a generic PL model in which the objective function is the
following form (e1,..en are linear expressions of the control
variables):

max min{𝑒1 , … , 𝑒𝑛 }

• To linearize the model, add a variable y to be maximized, and the

following constraints:

max 𝑦

𝑦 ≤ 𝑒1
𝑦 ≤ 𝑒2
…
𝑦 ≤ 𝑒𝑛
 Min-Max
• Consider a generic PL model in which the objective function is the
following form (e1,..en are linear expressions of the control
variables):

min max{𝑒1 , … , 𝑒𝑛 }

• To linearize the model, add a variable y to be minimized, and the

following constraints:

min 𝑦

𝑦 ≥ 𝑒1
𝑦 ≥ 𝑒2
…
𝑦 ≥ 𝑒𝑛
 Min-Abs
• Consider a generic PL model in which the objective function
is the following form (e is a linear expressions of the control
variables):

min |𝑒|

• Note that |e|=max{e,-e}.

• Then, the min-max approach can be used. Result is:

min 𝑦

𝑦≥𝑒
𝑦 ≥ −𝑒
 A production planning problem
Actually, the model could have been simpler adding
the following assumption:

«There is always an optimal solution in which the 3

parts are produced in the same quantity y»

Proof: if there’s an optimal solution where more

than y parts of type t are produced, then it is
enough to remove the exceeding parts from
production from any production line. The new
solution has the above property and the same
objective function of the previous one.
max y

y = 10 x11 + 15 x21 + 20 x31 + 10 x41 =

= 15 x12 + 10 x22 + 5 x32 + 15 x42 = 5 x13 + 5 x23 + 10 x33 + 20 x43

x11 + x12 + x13  100

x21 + x22 + x23  150 xij   +
x31 + x32 + x33  80
x41 + x42 + x43  200

xij is the production of part j on line i.

 Just in time production
• A factory must plan the production of 5 lots on a
machine.
• The processing duration of the lots are
(5,7,4,7,10) minutes (lots 1,2,3,4,5).
• The production sequence is given and must be 1-
2-3-4-5.
• Every lot has a desired delivery time. Those times
are: (10.32, 10.38, 10.42, 10.52, 10.57)
• Define the starting time for each operation in
order to minimize the deviation between the
desired and achieved delivery time.
 Just in time production
𝑆𝑖 = 𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑓𝑜𝑟 𝑙𝑜𝑡 𝑖 (𝑖𝑛 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 𝑎𝑓𝑡𝑒𝑟 0.00. 𝐹𝑜𝑟 𝑒𝑥𝑎𝑚𝑝𝑙𝑒, 10.00 𝑖𝑠 600)

min 𝑆1 + 5 − 632 + 𝑆2 + 7 − 638 + 𝑆3 + 4 − 642 + 𝑆4 + 7 − 652 + 𝑆5 + 10 − 657

𝑆2 ≥ 𝑆1 + 5
𝑆3 ≥ 𝑆2 + 7
𝑆4 ≥ 𝑆3 + 4
𝑆5 ≥ 𝑆4 + 7
𝑆𝑖 ≥ 0 ∀𝑖

In order to linearize the objective function, we can introduce 5 variables, one for each
absolute value, and, accordingly, the additional constraints.
 Just in time production
𝑆𝑖 = 𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑓𝑜𝑟 𝑙𝑜𝑡 𝑖 (𝑖𝑛 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 𝑎𝑓𝑡𝑒𝑟 0.00. 𝐹𝑜𝑟 𝑒𝑥𝑎𝑚𝑝𝑙𝑒, 10.00 𝑖𝑠 600)

min 𝑦1 + 𝑦2 + 𝑦3 + 𝑦4 + 𝑦5

𝑆2 ≥ 𝑆1 + 5 𝑦1 ≥ (𝑆1 +5 − 632); 𝑦1 ≥ −(𝑆1 + 5 − 632)

𝑆3 ≥ 𝑆2 + 7 𝑦2 ≥ (𝑆2 +7 − 638); 𝑦2 ≥ −(𝑆2 + 7 − 638)
𝑆4 ≥ 𝑆3 + 4 𝑦3 ≥ (𝑆3 +4 − 642); 𝑦3 ≥ −(𝑆3 + 4 − 642)
𝑆5 ≥ 𝑆4 + 7 𝑦4 ≥ (𝑆4 +7 − 652); 𝑦4 ≥ −(𝑆4 + 7 − 652)
𝑆𝑖 ≥ 0 , 𝑦𝑖 ≥ 0 ∀𝑖 𝑦5 ≥ (𝑆5 +10 − 657); 𝑦5 ≥ −(𝑆5 + 10 − 657)
 Wedding problem
• A florist is arranging the adornment for a wedding. The following types of flowers
are available: roses, daisies, gerania and lilies.

• Roses, daisies and gerania are available in small, medium and large size, whilst
there are available 250 lilies only all in small size.

• The cost of a rose is 2, 3 and 4 EUROS respectively (small, medium and large size).
• The cost of a daisy is 1.5, 2 and 2.5 EUROS (small, medium, large size).
• The cost of a geranium is 1.5, 2.5 and 3 EUROS (small, medium, large size).
• The cost of a small lily is 2.5 EUROS.

• It is required to have at least 1000 and no more than 1300 flowers.

• It is required to have the same percentage of small, medium and large size flowers.
• It is requires that the quantity of roses (independently from the size) is at least
twice the quantity of lilies.
• Gerania and daisies are exclusive between each other (it is not possible to use
both of them in the flower mix)

• Formulate the linear programming model minimizing the total cost.

 Wedding problem
• Roses, daisies and gerania are available in small, medium and large
size, whilst there are available 250 lilies only all in small size.

• Costs:
– Rose: 2, 3 and 4 EUROS respectively (small, medium and large size).
– Daisy: 1.5, 2 and 2.5 EUROS (small, medium, large size).
– Geranium: 1.5, 2.5 and 3 EUROS (small, medium, large size).
– Small lily: 2.5 EUROS.

𝑥𝑖𝑗 = 𝑓𝑙𝑜𝑤𝑒𝑟𝑠 𝑜𝑓 𝑡𝑦𝑝𝑒 𝑖 ∈ 𝑅, 𝐷, 𝐺, 𝐿 𝑜𝑓 𝑠𝑖𝑧𝑒 𝑗 ∈ {𝑆, 𝑀, 𝐿}

𝑥𝑖𝑗 ∈ 𝑍 +
 𝑥𝑖𝑗 = 𝑓𝑙𝑜𝑤𝑒𝑟𝑠 𝑜𝑓 𝑡𝑦𝑝𝑒 𝑖 ∈ 𝑅, 𝐷, 𝐺, 𝐿 𝑜𝑓 𝑠𝑖𝑧𝑒 𝑗 ∈ {𝑆, 𝑀, 𝐿}

𝑥𝑖𝑗 ∈ 𝑍 + , 𝑦𝐺 ∈ 0,1 , 𝑦𝐷 ∈ {0,1}

min 2𝑥𝑅𝑆 + 3𝑥𝑅𝑀 + 4𝑥𝑅𝐿 + 1.5𝑥𝐷𝑆 + 2𝑥𝐷𝑀 + 2.5𝑥𝐷𝐿 +1.5𝑥𝐺𝑆 + 2.5𝑥𝐺𝑀 + 3𝑥𝐺𝐿 +2.5𝑥𝐿𝑆

𝑥𝐿𝑆 ≤ 250

σ𝑖,𝑗 𝑥𝑖𝑗 ≥1000 σ𝑖,𝑗 𝑥𝑖𝑗 ≤1300

𝑥𝑅𝑆 + 𝑥𝐷𝑆 + 𝑥𝐺𝑆 + 𝑥𝐿𝑆 = 𝑥𝑅𝑀 + 𝑥𝐷𝑀 + 𝑥𝐺𝑀 = 𝑥𝑅𝐿 + 𝑥𝐷𝐿 + 𝑥𝐺𝐿

𝑥𝑅𝑆 + 𝑥𝑅𝑀 + 𝑥𝑅𝐿 ≥ 2𝑥𝐿𝑆

𝑥𝐺𝑆 + 𝑥𝐺𝑀 + 𝑥𝐺𝐿 ≤ 1300𝑦𝐺

𝑥𝐷𝑆 + 𝑥𝐷𝑀 + 𝑥𝐷𝐿 ≤ 1300𝑦𝐷

𝑦𝐺 + 𝑦𝐷 ≤ 1
 «Big M» constraints
• A «Big M» constraint is useful for defining a
relationship between a continuous or integer variable
x and the binary variable y similar to the following:

1 𝑖𝑓 𝑥 > 0
𝑦=ቊ
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
• The constraint is:

𝑥 ≤ 𝑀𝑦
where M is a constant with a value large enough to
take any possible value of x.
 «Big M» constraints
• Big-M formulations are relatively straightforward,
but the value of the M term needs to be chosen
carefully. If M is smaller than the upper bound of
x, this situation may cut off valid solutions. If M is
too large, the model may become numerically
difficult.
• In the case of the previous exercise, 1300 is a
good value of M, because it’s, for the problem
description, the maximum value the variable x
can take.
 «Big M» constraints
X=0 X>0
𝑥 ≤ 𝑀𝑦 Y=0 Feasible NO
Y=1 Feasible Feasible

• Note that above constraint does not work EXACTLY like defined 2 slides back.

• The situation which can cause problems is the one in which y=1 and still x=0.

• Usually, variable y will represent some kind of cost in the objective function.

• In this cases, in the optimal solution, y will be set only if there is the real need of
activating variable x.

• In order to impose a double-side property (x=0 y=0 and x>0 y=1), the
following constraint must be added to the above one, where m is a small number
(in case x is integer, m can be set to 1).

𝑚𝑦 ≤ 𝑥
 Small-m constraints
• As an example, consider x as the variable defining how
much we produce of something, and that we have a setup
cost s for activating the production of that product.

• In this case we’ll model as:

– Objective function: ……… + sy
– Constraint: x≤My

• If y=0, x=0. No problem!

• If y=1, x can be both 0 or positive.
– It has no sense to pay the setup cost s for producing nothing
(y=1,x=0).
– But, in order to avoid the cost s in objective function, the solver
will never output this combination, and y will be set to 0.
• We don’t need a small-m constraint.
 Small-m constraints
• Consider x as the variable defining how much we produce of
something, and that we have a penalty cost p in the case we
produce nothing of it.

• In this case we’ll model as:

– Objective function: ……… + p(1-y)
– Constraint: x≤My

• If y=0, x=0. No problem!

• If y=1, x can be both 0 or positive.

– Problem! If we produce nothing (x=0) here the model can set y=1 and
avoid paying the penalty cost p!
– Hence, we need to avoid this possibility, with a small-m constraint:

y≤x
 Logic constraints
• When using binary variables, it is often
possible to write logic operators in form of
a linear inequality:
Another production planning example
• 4 products (A,B,C,D).
• Production can be activated or not. If it is activated, there’s a fixed
activation cost, and the factory must produce a minimum quantity.
• Each product require a certain workforce (total available is 200.000)
and provide a profit.
• In the case the production of A is activated, the factory must
activate at least one between C and D.

Product Minimum Fixed cost Workforce Profit

quantity
A 1000 14500 10 50
B 1500 10000 15 60
C 3000 8000 5 55
D 2000 9000 14 80
Variables:

xA,xB,xC,xD = number of produced items of type A,B,C and D.

yA,yB,yC,yD = boolean variables associated to the xi ones.

Question: can I write this objective
function?

min 50𝑥𝑎 − 14500 𝑦1 + (… )

Question: can I write this objective
function?

It is a valid
mathematical
formulation, but it
is
NON LINEAR (x
and y variables
min 50𝑥𝑎 − 14500 𝑦1 + (… )
cannot be
multiplied
between each
other)!