1

Duality Theorems
Prof. V.S. Verma
Department of Mathematics and Statistics
D.D.U. Gorakhpur University, Gorakhpur

Since dual of a linear programming problem is itself a linear

programming problem, therefore, the dual of the dual problem can also be
constructed.
Here, we shall give a number of fundamental theorems to describe
the relationships between the primal and the dual. These relationships
continue to be useful in the development of mathematical programming.
Theorem 1.The dual of the dual of a given primal is the primal itself.
Proof : Consider the following linear programming problem (primal) in
the symmetric form :
Primal : Find the values of x1, x2, ..., xn which will maximize Zx = c1 x1 +
c2 x2 + ... + cn xn
subject to a11x1 + a12 x2 + ... + a1n xnb1
a21x2 + a22x2 + ... + a2n xnb2
: : : : ...(i)
am1x1 + am2 x2 + ... + amn xnbm
and satisfying x1 0, x2 0, ..., xn 0
The dual of the above primal (i) is given as follows :
Dual. Find the values of w1, w2, ...,wm which will minimize Zw = b1w1 +
b2w2 + ... + bmwm
subject to a11w1 + a21w2 + ... + am1wmc1
a21w1 + a22w2 + ... + am2wmc2
: : : : ...(ii)
a1nw1 + a2nw2 + ... + amnwmcn
and satisfying w1 0, w2 0, ..., wm 0
 2

Now, our aim is to construct the dual of the dual (ii). For this, we shall
first express the above dual problem (ii) in the symmetric form of the
primal by multiplying throughout by – 1. Thus, we obtain the following
LPP :
Maximize (– Zw) = – b1w1 – b2w2 – b3w3 ... – bmwm
subject to – a11w1 – a21w1 – ... – am1wm – c1
– a12w1 – a22w2 – ... – am2wm – c2
: : : : …(iii)
– a1nw1 – a2nw2 – ... – amnwm – cn
and satisfying w1 0, w2 0, ..., wm 0
Dual of Dual. Now, constructing the dual of (iii), we obtain the following LPP :
Minimize Zv = – c1v1 – c2v2 – ... – cnvn
subject to – a11v1 – a12v2 – ... – a1nv1 – b1
– a21v1 – a22v2 – ... – a2nvn – b2
: : : :...(iv)
– am1v1 – am2v2 – ... – amnvn – bm
and satisfying v1 0, v2 0, ..., vn 0
Now, multiplying (iv) throughout by – 1, we get the following LPP:
Maximize (Zv) = c1v1 + c2v2 + ... + cnvn where Zv = – Zv
subject toa11v1 + a12v2 + ... + a1nvnb1
a21v1 + a22v2 + ... + a2nvnb2...(v)
: : : :
am1v1 + am2v2 + ... + amnvnbm
and satisfying v1 0, v2 0, ..., vn 0
which is identical to the primal (i).
This completes the proof of the theorem.
Note 1. Of course, the variables v1, v2, ...,vn in (v) will be equal to x1, x2,
..., xn respectively.
 3

2.From the above theorem, we conclude that it is immaterial which

problem is called the primal and which one is called the dual. We can
consider the problem (ii) to be the primal and (i) to be its dual. Therefore,
sometimes, we call these two problems [(i) and (ii)] as pair of primal dual
problems.
Alternative Proof :
The above theorem can also be proved using the matrix form as follows :
Consider the following LPP (primal) in matrix form :
Maximize Zx = CX
subject to AX  b ...(vi)
and X0
The dual of the primal (vi) is given as follows :
Minimize Zw = bTW
subject to ATW  CT ...(vii)
and W0
This dual problem (vii) can be written as :
Maximize (– Zw) = – bTW
subject to – ATW  – CT ...(viii) and
W0
Now, the dual of (viii) is given as follows :
Minimize Zv = (– CT)TV = – CV
subject to (– AT)TV  (– bT)T ...(ix) and
V0
The problem (ix) can also be written as :
Maximize Zv = CV
subject to AV  b …(x)
and V0
which is identical to the primal (vi).
The completes the proof of the theorem.
 4

Theorem 2. If X is any feasible solution to the primal problem : Max. Z =

CX, subject to AX  b, X  0 and W is any feasible solution to the dual
problem : Min. Zw = bTW, subject to ATW  CT, W  0, then CX  bTW
orZx Zw.
Alternative Statement : Given the following linear programming
problems :
AX b, X  0, max. Zx = CX ...(i)
and ATW  CT, W  0, min. Zw = bTW ...(ii)
Then, for any feasible solution X to the problem (i) and any
feasiblesolution W to the problem (ii), we have CXbTW, i.e. ZxZw.
Proof : To prove this theorem, let us rewrite the pair of primal dual
problems (i) and (ii) in the form of equations by introducing the slack
variables xn + 1, xn + 2, ... xn + m in (i) and the surplus variables wm + 1, wm + 2,
..., wm + n in (ii). Thus, we obtain the following linear programming
problems:

c1x1 + c2 x2 + ... + cn xn= Zx (max.)

a11x1 + a12x2 + ... + a1n xn + xn + 1= b1
a21x1 + a22x2 + ... + a2nxn+ xn + 2= b2 ...(iii)
: : : : :
am1x1 + am2x2 + ... + amn xn + xn + m= bm
satisfying x1 0, x2 0, ... xn 0 xn+1  0, xn+2  0, ..., xn + m 0.
and𝑏1𝑤1 + 𝑏2𝑤2 + . . . + 𝑏𝑚𝑤𝑚 = Z𝑤 (min)
𝑎11𝑤1 + 𝑎21𝑤2 + . . . + 𝑎𝑚1𝑤𝑚– 𝑤𝑚 = 𝑐1
+ 1
𝑎12𝑤1 + 𝑎21𝑤2 + . . . + 𝑎𝑚1𝑤𝑚 – 𝑤𝑚 1 = 𝑐2 ...(iv)
⋮ ⋮ ⋮ ⋮ + ⋮
𝑎1𝑛𝑤1 + 𝑎2𝑛𝑤2 + . . . + 𝑎𝑚𝑛𝑤𝑚 – 𝑤𝑚 = 𝑐𝑛}
+ 𝑛
satisfyingw1 0, w2 0, ... wm 0, wm+1 0, wm+2 0, ..., wm + n 0
Let us now multiply the second equation of (iii) by w1, the third equation
of (iii) by w2, etc., and subtract successively the resulting expressions from
the first equation (objective function) of (iii). Thus, we obtain
[c1 – (a11w1 + a21w2 + ... + am1wm)] x1 + ... + [cn – (a1nw1 + a2nw2 + ... +
amnwm)] xn
 5

– (xn + 1w1 + xn + 2w2 + ... + xn + m wm) = Zx – (b1w1 + ... + bmwm) ...(v)

We now observe from (iv) that the summation within the bracket ( ) on the
right-hand side of (v) is equal to Zw and that the quantities within brackets
on the left-hand side are the values of– wm + 1, – wm + 2, ..., – wm + n,
respecitvely, so that (v) becomes
– (wm + 1x1 + wm + 2x2 + ... + xm + n xn) – (xn + 1w1 + xn + 2w2 + ... + xn + m wm)
= Zx – Zw...(vi)
If the solutions (x1, x2, ...,xn) and (w1, w2, ..., wm) constitute feasible
solutions (not necessarily basic), then all the terms on the left-hand side of
(vi) are negative, so that the right-hand side value also becomes negative.
This implies further that
Zx – Zw 0 or Zx Zw.
This completes the proof of the theorem.
Alternative Proof : The above theorem can be more easily proved as
follows :
Let X = (x1, x2, ...,xn) and W = (w1, w2, ..., wm) be feasible solutions to the
primal problem : Max. Zx = CX, subject to AX b, X  0 and the dual
problems : Min. Zw = bTW, subject to ATW  CT, W  0 respectively.
Then, we have
CX = ∑𝑛𝑗=1 𝑐𝑗 𝑥𝑗 ≤ ∑𝑛𝑗=1(∑𝑚 𝑚
𝑖=1 𝑎𝑖𝑗 𝑤𝑖 ) 𝑥𝑗 ,since𝑐𝑗 ≤ ∑𝑖=1 𝑎𝑖𝑗 𝑤𝑖 is
jthconstraint of the dual problem.
− ∑𝑚 𝑚 𝑚 𝑇 𝑚
𝑖=1 𝑤𝑖 (∑𝑖=1 𝑎𝑖𝑗 𝑤𝑖 ) ≤ ∑𝑖=1 𝑤𝑖 𝑏𝑖 𝑏 𝑊, since ∑𝑗=1 𝑎𝑖𝑗 𝑥𝑗 ≤ 𝑏𝑖 is the ith
constraint of the primal.
 CX  bTW or Zx Zw.
This completes the proof of the theorem.
 6

Theorem 3.If 𝑋̂ is a feasible solution to a primal LPP and 𝑊 ̂ is a feasible

solution to its dual such that, 𝐶𝑋̂ = 𝑏 𝑇 𝑊
̂ then 𝑋̂ is an optimal solution to
the primal LPP and 𝑊 ̂ is an optimal solution to its dual.
Proof : If 𝑋̂ is a feasible solution to the primal LPP and 𝑊 ̂ is a feasible
solution to its dual problem, then
𝐶𝑋̂ = 𝑏 𝑇 𝑊
̂
But, for any feasible solution X to the primal problem, we have proved the
result CXbTW in above theorem. Therefore, we have
CXbTW orCX 𝐶𝑋̂  using 𝐶𝑋̂ = 𝑏 𝑇 𝑊 ̂
From this result, we conclude thatis an optimal solution to the
maximization linear programming problem (i.e. primal problem).
Similarly, for any feasible solution W, to the dual problem, we have
𝐶𝑋̂ = 𝑏 𝑇 𝑊 ̂ bTW , using 𝐶𝑋̂ = 𝑏 𝑇 𝑊
̂ or 𝑏 𝑇 𝑊 ̂
and therefore, we colclude that𝑊 ̂ is an optimal solution to the
minimization linear programming problem (i.e. dual problem).
This completes the proof of the theorem.
Note :Sometimes, the Theorems 2 and 3 are stated in a combined form.
The students are advised to give complete proof of Theorem 2 before
starting to prove Theorem 3.
Theorem 4.Basic Duality Theorem. If X0 (W0) is an optimal solution to
the primal (dual), then there exists a feasible solution W0 (X0) to the dual
(primal), such that C X0 = bT W0.
Proof : The primal problem can be written as : Max. Z = CX, subject to
AX + I Xs = b,
where Xs Rm is the slack vector and I is the associated m × n identity
matrix.
Let X0 = [XB, O] be an optimal solution to the primal problem,
where XB Rm is the optimal basic feasible solution given by XB = B–1 b,
B being the optimal basis of A. Then, the optimal primal objective
function is given as follows :
Z = CX0 = CB XB
where CB is the cost vector associated with XB. Now, net evaluations in
the optimal simplex table are given as follows :
 7

𝐶𝐵 𝐵−1 𝑎𝑗 − 𝑐𝑗 , for all𝑎𝑗 ∈ A

Zj – cj = CB Xj – cj = {
𝐶𝐵 𝐵−1 𝑒𝑗 − 0 , for all𝑒𝑗 ∈ 1
Since XB is optimal, we must have Zj – cj 0 for all j. This yields the
following result :
CBB–1ajcj and CBB–1ej> 0 for all j
or CBB–1 A  C and CBB–1 O (Matrix Form)
or ATB–1 C and B–1 O.
Now, if we let B–1 = W0 Rm, then the above result becomes ATW0 C
and W00.
This implies that W0 is a feasible solution to the dual problem.
Again, the corresponding dual objective function value is given as
follows:
bTW0 = b = CBB–1b = CBXB = CX0
Thus, for a given optimal solution X0 to a primal problem, there exists
afeasible solution W0 to its dual problem such that CX0 = bTW0.
Similarly, starting with W0, the existence of X0 can be proved.
Corollary : If X0 is an optimal solution to the primal, then an optimal
solution to the dual is given by W0 = B–1 CB, where B is the primal
optimal basis.
Note : We observe that CBB–1 represent optimum value ofj's (where j =
Zj – cj) under slack variable columns.
We, now proceed to prove the following important theorem, usually called
the fundamental duality theorem that provides the basis of the foundation
of duality in linear programming :
Fundamental Duality Theorem
The following theorem is called fundamental duality theorem :
Theorem 5(a). If either the primal or the dual problem has a finite optimal
solution, then the other problem also has a finite optimal solution.
Furthermore, the optimal values of the objective function in both the
problems are the same, i.e. Max. Zx = Min. Zw.
(b) If either problem has an unbounded optimal solution, then the other
problem has no feasible solution at all.
 8

(c) Both problems may be infeasible, i.e. if the primal has no solution,
then the dual will also have no solution.
Proof (a) : We first assume that the primal is feasible and that a maximum
feasible solution has been obtained by simplex method.
The primal dual pair of problems in matrix form may be taken as
follows :
Primal : Max. Zx = CX, subject to AX b, X  0.
Dual : Min. Zw = Wb, subject to WA  C, W  0.
Now, let the primal problem has a finite optimal solution given by
XB = B–1b. ...(i)
Then, the corresponding value of primal objective function is given by
Zx* = CBXB = CB (B–1b) ...(ii)
and the accompanying conditions for optimality are given by
{Zj – cj}  0 or {CBB–1 A – C, CBB–1}  0 or CBB–1 A – C  0, CBB–1> 0 ...(iii)
Now, to verify that the associated optimal dual solution is given by
WB = CBB–1 ...(iv)
we must only observe that the condition (iii) then, demonstrates the
feasibility condition or in other words, the dual problem is satisfied.
To show Max. Zx = Min. Zw :
For the optimal dual solution (iv) i.e. WB = CBB–1, the
corresponding value of the dual objective function is given by
Zw* = WBb = (CBB–1) b = CB (B–1b) = CBXB = Max. Zx
Now, it only remains to show that Zw* is also the optimal value for
the dual objective function.
Since, by Theorem 2, we have Zx Zw for all feasible X and W, extremes
of primal and dual are related by max. Zx min. Zw.
Using (iii), we have Zw*  min. Zw.
 9

But Zw* can never be less than min. Zw. Hence, we have max. Zx= min Zw.
Similarly, starting with the finite optimal solution to the dual problem, if it
exists, we can prove that the solution to the primal is feasible and max. Z x
= min. Zw holds. To do so, we shall use the fact that 'dual of a dual is the
original primal'.
This completes the first part of the theorem.
Note : It is also important to note that if the primary constraints are
equalities, then (iv) continues to hold, but WB, no longer has to be non-
negative.
(b) To prove second part of the theorem, we suppose that for any value of
the primal objective function, no matter however large, there exists a
feasible solution yielding this value. Then, the dual problem can have no
feasible solution (W) whose components are all finite. If it did,
acontradiction would arise. By the result [(max. Zx Zw) or ( Wb)], the
dual solution would provide an upper bound on the value of the primal
objective function for any feasible primal solution. But, we have assumed
that the primal objective function can be made arbitrarily large.
A similar argument will show that the primal has no solution, when the
dual has an unbounded solution.
Note : It must be carefully noted that "if the primal has no solution then,
the dual has an unbounded solution" is never true. But, it will be true that
if the primal has no solution, then the dual will also have no solution. This
point has been explained by an example.
(c) The following example will be sufficient to reach the desired result of
third part of fundamental duality theorem :
Consider the following linear programming problem as a primal
problem :
Max. Zx = 3x1 + 4x3, subject to the constraints x1 – x2 1, – x1 + x2
0 and x1 0, x2 0.
This primal has no feasible solution, since the constraints
are inconsistent as seen from Fig.1.
 10

Similarly, the following dual of the above primal problem has

no solution as seen from Fig. 2.

𝑥2 𝑤2
𝑥1
−𝑤1
− 𝑥2
−𝑥1 − 𝑤2
≤ −1
+ 𝑥2 ≥4
≤ −1

𝑤1
− 𝑤2
≥3

𝑥1 𝑤1
O O

Fig. 1 Fig. 2

Min. Zw = – w1, subject to the constraints w1 – w2 3, – w1 + w2 4

and w1 0, w2 0.
Theorem 6. If the kth constraint of a primal LPP is an equality, then the
dual variable wk is unrestricted in sign.
Proof : As pointed out earlier, if the kth constraint is an equality, then we
can write the primal in the following form :
Max. 𝑍𝑥 = 𝑐1𝑥1 + 𝑐2𝑥2 + . . . + 𝑐𝑛𝑥𝑛
subject to 𝑎11𝑥1 + 𝑎12𝑥2 + . . . + 𝑎1𝑛𝑥𝑛 ≤ 𝑏1
𝑎21𝑥1 + 𝑎22𝑥2 + . . . + 𝑎2𝑛𝑥𝑛 ≤ 𝑏2
: ∶ ∶ ∶ …(i)
𝑎𝑘1𝑥1 + 𝑎𝑘2𝑥2 + . . . + 𝑎𝑘𝑛𝑥𝑛 ≤ 𝑏𝑘
– 𝑎𝑘1𝑥1 – 𝑎𝑘2𝑥2 – . . . – 𝑎𝑘𝑛𝑥𝑛 ≤ – 𝑏𝑘
∶ ∶ ∶ ∶
𝑎𝑚1𝑥1 + 𝑎𝑚2𝑥2 + . . . + 𝑎𝑚𝑛𝑥𝑛 ≤ 𝑏𝑚 }
so that its dual is given by
 11

Min. Z𝑤 = 𝑏1𝑤1 + 𝑏2𝑤2 + . . . + 𝑏𝑘 (𝑤 𝑘 – 𝑤 𝑘) + . . . + 𝑏𝑚𝑤𝑚

subject to𝑎11𝑤1 + 𝑎21𝑤2 + . . . + 𝑎𝑘1 (𝑤 𝑘 – 𝑤 𝑘 ) + . . . + 𝑎𝑚1𝑤𝑚 𝑐1
𝑎12𝑤1 + 𝑎22𝑤2 + . . . + 𝑎𝑘2 (𝑤 𝑘 – 𝑤 𝑘 ) + . . . + 𝑎𝑚2𝑤𝑚 𝑐2 …(ii)
: ∶ ∶ ∶ ∶
𝑎1𝑛𝑤1 + 𝑎2𝑛𝑤2 + . . . + 𝑎𝑘𝑛 (𝑤 𝑘 – 𝑤 𝑘 ) + . . . + 𝑎𝑚𝑛𝑤𝑚 𝑐𝑛 }
If we replace (wk – wk) by a new variable wk, then we observe
that the system (ii) becomes equivalent to the primal problem except that
wk is unrestricted in sign.
If, in particular, all constraints of the primal are equalities, then the
dual will be given by (4.2), except that all dual decision variables will be
unrestricted in sign.
This completes the proof of the theorem.
Theorem 7. If the pth variable of a primal LPP is unrestricted in sign, then
the pth constraint of the dual is an equality.
Proof : Consider the following linear programming problem as a primal
problem :
Max. Zx = c1 x1 + c2 x2 + ... + cp xp + ... + cn xn
subject toa11 x1 + a12 x2 + ... + a1p xp + ... + a1n xnb1
a21x1 + a22x2 + ... + a2p xp + ... + a2n xnb2
: : : : :
am1x1 + am2 x2 + ... + ampxp + ... + amnxnbm
wherex1 0, x2 0, ..., xp + 1 0, ..., xn 0, but xp is unrestricted in sign.
Since, the variable xp (unrestricted in sign) can always be expressed
as the difference of two non-negative variables, say xp and xp, therefore,
xp can be replaced by (xp – xp) in the above primal problem.
Thus, we have the new primal of the following form :
Max. Zx = c1 x1 + c2 x2 + ... + cp (xp – xp) + ... + cn xn
subject toa11 x1 + a12 x2 + ... + a1p (xp – xp) + ... + a1n xnb1
a21x1 + a22x2 + ... + a2p (xp – xp) + ... + a2n xnb2
: : : :
am1x1 + am2 x2 + ... + amp (xp – xp) + ... + amnxnbm
where x1 0, x2 0, ..., xp 0, xp 0, xp+1 0. ..., xn 0.
 12

The dual of this problem is given by,

Min. Zw = b1w1 + b2w2 + ... + bmwm
subject toa11w1 + a21w2 + ... + am1wmc1
a12w1 + a21w2 + ... + am2wmc2
: : : :

𝑎1𝑝𝑤1 + 𝑎2𝑝𝑤2 + . . . + 𝑎𝑚𝑝𝑤𝑚 ≥ 𝑐𝑝

{ }
– 𝑎1𝑝𝑤1 – 𝑎2𝑝𝑤2 – . . . – 𝑎𝑚𝑝𝑤𝑚 ≥ – 𝑐𝑝
: : : :
a1nw1 + a2nw2 + ... + amnwm cn
where the two constraints under the bracket { } in the above dual can be
written as one equality constraint, which shows that the pth constraint of
the dual is an equality constraint.
This completes the proof of the theorem.
Existence Theorem
We all know that mathematics is inevitable concerned with the
existence of solutions and the field of linear programming is not
exception. The existence theorems, which we shall prove justnow, are
extremely important and have far reaching implications in understanding
the relationship between primal and dual solutions.
Theorem 8. (Existence Theorem 1) There exists a bounded (finite)
optimal solution to an L.P.P. if and only if, there exists a feasible solution
to both, the primal and its dual.
Proof : Let the primal problem be : Maximize Zx = CX, subject to AX b,
X  0 and its dual be : Minimize Zw = bTW, subject to ATW  CT, W  0.
Let us suppose first that there exists an optimal feasible solution to the
primal problem. Then, it follows from fundamental theorem of duality that
the dual problem has at least one feasible solution.
Conversely, let us suppose that there exist feasible solutions X* and W* to
the primal and its dual respectively. Obviously, CX* and bTW* are finite.
Itfollows from Theorem 2 thatbTW* serves as an upper bound on CX*
although not necessarily the least upper bound. Thus, the primal problem
must have finite optimal solution.
 13

Theorem 9. (Existence Theorem 2) If there does not exist any feasible

solution to the dual (primal) problem but there does exist at least one to
the primal (dual) problem, then there does not exist any finite optimal
solution to the primal (dual) problem.
Proof : Let the primal problem be : Maximize Zx = CX, subject to AX b, X
 0 and its dual be : Minimize Zw = bTW, subject to ATW  CT, W  0.
Let us first consider the case when the dual problem does not possess any
feasible solution but the primal problem does. Then, any feasible solution
X* to the primal will provide the objective function value CX*. Now, we
claim that X* cannot be an optimal solution to the primal. Contrary to this,
we suppose that X* is an optimal solution to the primal. The fundamental
theorem of duality then gives that there must exist a feasible solution to
the dual problem. Thus, we arrive at a contradiction and hence our
supposition is worng. It follows, therefore, that no feasible solution to the
primal can be optimal so that there exists an unbounded solution to the
primal. Thus, there does not exist any finite optimal solution to the primal
problem.
Similarly, by assuming that the primal problem does not possess any
feasible solution, but the dual problem does, it can be proved that there
does not exist any finite solution to the dual problem
Theorem 10. (Existence Theorem 3) If, there does not exist any finite
optimal solution to the primal (dual) problem, then there does not exist
any feasible solution to the dual (primal) problem.
Proof :Let us first consider the case when there does not exist any finite
optimal solution to the primal. Without any loss of generality, we assume
that the primal does admit a feasible solution. We claim that there does not
exist any feasible solution to the dual. Contrary to this, we assume that the
dual problem admits a feasible solution. It then follows from Existence
Theorem 1 that the primal problem must admit a finite optimal solution.
Thus, we arrive at a contradiction and hence there does not exist any
feasible solution to the dual problem.
 14

Similarly, other case can be proved.

The results of Existence Theorems 1-3 can be summarized as follows :

Primal Problem
Dual Problem
Feasible solution No feasible solution
Feasible solution Optimal solution Dual unbounded
No feasible solution Primal unbounded May occur
Complementary Slackness Theorem
In this section, we shall make a further effort to relate the primal and the
dual formulations of linear programming problem. The following theorem
(called the complementary slackenss theorem) shows the intimate relation
between the primal and the dual problems :
Theorem 11.(Complementary Slackness Theorem) For the optimal
feasible solutions of the primal and the dual system, we have
(i)if the inequality occurs in the ith relation of either system (the
corresponding slack or surplus variable xn + i is positive), then the ith
variable wi of its dual is zero, i.e. wi = 0 and
(ii)if the jth variable xj is positive in either system, then the jth dual
constraint holds as a strict equality (i.e. the corresponding slack or surplus
variable wm + j is zero).
Proof : The primal objective function can be written as :
c1x1 + c2 x2 + ... + bm wm = Zw (max.) ...(i)
and the dual objective function can be written as :
b1w1 + b2w2 + ... + bmwm = Zw (min.) ...(ii)
Also, the primal-inequalities can be written as equations after
introducing the non-negative slack variables xn + 1, xn + 2, ...,xn + m as
follows :
𝑎11 𝑥1 + 𝑎12 𝑥2 + . . . + 𝑎1𝑛 𝑥𝑛 + 𝑥𝑛 + 1 = 𝑏1
𝑎21 𝑥1 + 𝑎22 𝑥2 + . . . + 𝑎2𝑛 𝑥𝑛 + 𝑥𝑛 + 2 = b2 }...(iii)
∶ ∶ ∶ ∶
𝑎𝑚1 𝑥1 + 𝑎𝑚2 𝑥2 + . . . + 𝑎𝑚𝑛 𝑥𝑛 + 𝑥𝑛 + 𝑚 = 𝑏𝑚
 15

Similarly, the dual-inequalities can be written as equations after

introducing the non-negative surplus variables wm + 1, wm + 2, ...,wm + n as
follows :
𝑎11𝑤1 + 𝑎21𝑤2 + . . . + 𝑎𝑚1𝑤𝑚 – 𝑤𝑚 – 1 = 𝑐1
𝑎12𝑤1 + 𝑎22𝑤2 + . . . + 𝑎𝑚𝑛𝑤𝑚 – 𝑤𝑚 + 2 = 𝑐2}...(iv)
: ∶ ∶ ∶
𝑎1𝑛𝑤1 + 𝑎2𝑛𝑤2 + . . . + 𝑎𝑚𝑛𝑤𝑚– 𝑤𝑚 + 𝑛 = 𝑐𝑛
Now, multiplying the equations in (iii) by w1, w2, ...,wm respectively and
then adding the resulting equations together, we obtain
𝑥1 ∑𝑚 𝑚 𝑚
𝑖=1 𝑎𝑖1 𝑤𝑖 + 𝑥2 ∑𝑖=1 𝑎𝑖2 𝑤𝑖 + ⋯ + 𝑥𝑛 ∑𝑖=1 𝑎𝑖𝑛 𝑤𝑖 𝑥𝑛+1 𝑤1 𝑥𝑛+2 𝑤2 +
𝑚
⋯ + 𝑥𝑛+𝑚 𝑤𝑚 = ∑𝑖=1 𝑤𝑖 𝑏𝑖 …(v)
Next, we subtract (v) from the corresponding sides of (i) and obtain
the following :
𝑚 𝑚 𝑚

(𝑐1 ∑ 𝑎𝑖1 𝑤𝑖 ) 𝑥1 + (𝑐2 ∑ 𝑎𝑖2 𝑤𝑖 ) 𝑥2 + ⋯ + (𝑐𝑛 ∑ 𝑎𝑖𝑛 𝑤𝑖 ) 𝑥𝑛

𝑖=1 𝑖=1 𝑖=1
𝑤1 𝑥𝑛+1 − 𝑤𝑛+2 −. . . −𝑤𝑚 − 𝑥𝑛+𝑚
𝑚
(𝑍𝑥 − ∑𝑖=1 𝑤𝑖 𝑏𝑖 )…(vi)
But, from (ii), we have 𝑍𝑥 − ∑𝑚 𝑖=1 𝑤𝑖 𝑏𝑖 …(vii)
and from (iv), we have
– wm + j = cj – (aij w1 + a2jw2 + ... + amjwm) or – wm + j = cj –∑𝑚
𝑖=1 𝑎𝑖𝑗 𝑤𝑖 , j
= 1, 2, .., n ...(viii)
Now, using the results (vii) and (viii), eqn. (vi) becomes
– (wm +1x1 + wm + 2x2 + ... + wm + n xn) – (w1xn + 1 + w2 xn + 2 + ... + wm xn +
m) = Zx – Zw ...(ix)
If X* = (x1*, x2*, ...,xn*) and W* = (w1*, w2*, ..., wm*) be the
optimal solutions to the primal and the dual problems respectively, then
we have
Zw* = Zw* ...(x)
as claimed by the duality theorem

Thus, for these optimal solutions and corresponding slack and

surplus variables
 16

x*n + i 0 (i = 1, 2, ..., m) and w*m + j 0 ( j = 1, 2, ..., n), the

equation (ix) becomes
– (w*m + 1x1* + w*m + 2x*2 + ... + w*m + n x*n) – (w*1x*n + 1 + ... + w*m x*n
+ m) = 0 [from (x)] ...(xi)
Since all the variables in (xi) are restricted to be non-negative, all the
product terms of (xi) are also non-negative and for the validity of (xi),
each term must individually be equal to zero.
Thus, we have w*m + jx*j = 0 and w*ix*n + i = 0, for all i and j.
(i)Now, if w*m + j> 0, we must have x*j = 0 and if x*n + i> 0, then w*i = 0,
which proves the first part of the theorem.
(ii) Again, if x*j> 0, then w*m + j = 0, i.e., the jth constraint of the dual
problem is strict equality and if w*i> 0, then x*n + i = 0, i.e., the ith primal
constraint is strict equality, which proves the second part of the theorem.
Hence, the theorem is now completely established.
Alternative Statement : A necessary and sufficient condition for
any pair of feasible solutions to the primal and the dual to be optimal is :
xixn + i = 0; i = 1, 2, ..., m, where xn + i is the ith slack variable in the
primal problem
and xjwm + j = 0; j = 1, 2, ..., n, where wm + j is the jth surplus variable in
the dual problem.
Remarks : As the name complementary slackness implies, the theorem
indicates that at optimality, if a primal constraint holds a strict inequality,
then its complementary dual constraint holds an equality. Conversely, if a
dual constraint holds a strict inequality, then its complementary primal
constraint holds an equality.

Primal and Dual Correspondence :

 17

With the help of duality theorems we note the following correspondence

between the primal and its dual. This can be presented in the following
tabular form :

Primal Problem Dual Problem

1. Objective function is to maximize Zp or Zx 1. Objective function is to minimize Zd or Zw

2. Requirement vector 2. Price vector

3. Coefficient matrix A 3. Transpose of coefficient matrix AT
4. Constraints with  sign 4. Constraints with  sign
5. Relation 5. Variable
6. ith inequality 6. ith variable wi 0
7. ith equality constraint
7. ith unrestricted variable wi
8. Variable
8. Relation
9. ith variable xi> 0
9. ith relation : a strict equality
10. ith unrestricted variable xi
10. ith constraint : a strict equality
11. ith slack variable positive 11. ith variable zero
12. ith variable zero 12. ith surplus variable positive
13. Finite optimal solution 13.Finite optimal solution with equaloptimal
value of objective function
14. Unbounded solution 14. No solution or an unbounded solution.
Duality and Simplex Method or Dual Simplex Method
The final simplex table giving optimal solution of the primal also contains
optimal solution of its dual in itself and conversely. This is based on the
'Fundamental Duality Theorem' which is restated as follows :
(a) If either the primal or the dual problem has a finite optimal solution, then the
other problem also has a finite optimal solution. Furthermore, the optimal values
of the objective function in both the problems are the same, i.e., Max. Z x = Min.
Zw.
(b) If either problem has an unbounded optimal solution, then the other problem
has no feasible solution at all.
(c) Both the problems may be infeasible i.e. if the primal has no solution, then the
dual will also have no solution.

Comparison of Solutions to the Primal and its Dual

 18

Consider the following pair of primal dual problems :

Primal Problem Dual Problem
Max. Zx = 40x1 + 50x2, Min. Zw = 3w1 + 5w2
subject to 2x1 + 3x2 3, subject to 2w1 + 8w2 40,
8x1 + 4x2 5, 3w1 + 4w2 50,

The primal problem in standard simplex form is:
Max. Zx = 40x1 + 50x2 + 0x3 + 0x4
subject to the constraints
2x1 + 3x2 + x3 + 0x4 = 3
5x1 + 4x2 + 0x3 + x4 = 5
andx10, x20, x30, x40.
The dual problem in standard simplex form is:
Max Zw = – 3w1 – 5w2 + 0w3 + 0w4 + 0w5 + 0w6
where Zw = – Zw subject to the constraints
2w1 + 8w2 – w3 + 0w4 + w5 + 0w6 = 40
3w1 + 4w2 + 0w3 – w4 + 0w5 + w6 = 50
andw1 0, w2 0, w3 0, w4 0, w5 0, w6 0.

x1 0, x2 0. w1 0, w2 0.

The solutions of above problems are obtained by using simplex method as
follows :
Solution of Primal Problem Solution of Dual Problem
Since the net evaluations 1, 2, 3, 4 are  hence the optimal
solution of the primal is given by x1 = 3/16, x2 = 7/8 and Zx* = (3/16) ×
(40) + (7/8) × 50 = 205/4 = 51.25.
Again, since the net evaluations 1, 2, 3, 4 are  0 hence the
optimal solution of the dual is given by w1 = 15, w2 = 5/4 and Zw* = 15 ×
3 + 5/4 × 5 = 205/4 = 51.25.
 19

Conclusion :From above comparison, it is concluded that the solution to a

Basic CB XB X1 X2 X3 Min. Ratio PHASE 1
variables X4 XB/Xk
WB W1 W2 W3 W4 W5 W6
 x3 0 3 2 3 1 0  3/3min.
x4 0 5 8 4 0 1 5/4 40 2 8 –1 1 0 0

x1=x2=0 Zx =CBXB=0 – 40 – 50 0 0 j 50 3 4 0 –1 0 1

20/3 0 16/3 –1 2/3 1 – 2/3
→ x2 50 1 2/3 1 1/3 0 1
2/3 50/3 1 4/3 0 1/3 0 1/3
 x4 0 1 16/3 0 – 4/3 1 1
← 𝑚𝑖𝑛.
2/3 5/4 0 1 – 3/6 1/8 3/4 ×

x1=x3=0 Zx = 50 – 20/3 0 50/3 0  j 15 1 0 1/4 – 1/2 –1/4 ×

PHASE 2
Basic CB XB X1 X2 X3 X4 Basic CB WB W1 W2 W3 W4
Variables Variable
x2 50 7/8 0 1 1/2 1/8 w2 –5 5/4 3 1
0 1 −
16 8
→ x1 40 3/16 1 0 – 1/4 – 3/4 w1 –3 15 1 1
1 0 −
4 2
x3=x4=0 Zx = 205/4
0 0 15 5/4 w3=w4= 0 Zw = – 205/4 0 0 3/16 7/8

(1) (2) (3) (4) (1) (2) ( 3) (4)

primal linear programming problem can provide a solution to its dual. The
relationship described by arrows should be carefully checked and grasped.
Rules for Obtaining Dual Optimal Solution from That of Primal and
Conversely
The following rules are useful for obtaining the dual optimal solution from
that of primal and conversely :
Rule 1. The optimal value of primal objective function is equal to the
optimal value of the dual objective function, i.e. max. Zx = min. Zw.
Proof : This rule has already been proved in Duality Theorem.
Rule 2. Thej's(wherej = Zj – cj) for the slack vector in optimal (final)
simplex table for the primal are the values of the corresponding optimal
dual variables in the final simplex table for the dual.
 20

Proof : Corresponding to slack and artificial variables, we have j Zj – cj =

Zj – 0 = CBB–1aj.
But, for the ith slack variable, ai becomes an identity vector ei = (0, 0, ..., 1,
0, ..., 0), with unity at the ith place.
Hence we havei= CBB–1ai = CBB–1ei = eiW = wi, where wi is the ith dual
variable.
If the primal contains greater than or equal to () constraints, then
assumeone artificial variable is introduced in each such constraint. For the
ith artificial variable, ai is an identity vector ei. In the same way, it is easy
toshow that thej'sfor the surplus vectors in an optimal solution to the
primal are the corresponding optimal variables.
As seen from the final simplex table for primal and dual problems
of above example, we have :
3 = Z3 – c3 = 15 = w1, 4 = Z4 – c4 = 5/4 = w2 and 3 = 3/16 = x1, 4 =
7/8 = x2.
Thus, we have observed that it is possible to read an optimal solution of
either problem directly from the final simplex table of the other.
Rule 3. If either problem (primal or dual) has unbounded solution, then
the other will have no feasible solution at all [see Duality Theorem part
(b)].
Thus, it is always advantageous to apply simplex method directly to
whichever problem (the primal or the dual) having lesser number of
constraints than the other, the solution to the other can be immediately
read off according to above rules.
 21

Shadow Prices or Marginal Prices in Linear Programming

Let us consider the primal problem : Maximize Z = CX, subject to the
constraints AX = b ; X  0, where X and C  Rn, b Rm and A is an m × n
real matrix. Also, let XB = B–1bbe an optimal basic feasible solution to this
primal problem, where B is the optimal basis matrix. If CB denotesthe cost
vector associated with basic variables, then the optimal value of the
objective function will be Z* = CBXB = CB (B–1b).
Now, we may define the shadow (or implicit or marginal) price pi
of the ith resource bi (right hand side value) to be the achievable rate of
increase in resource i, as follows :
𝑑𝑍 ∗
𝑑𝑏𝑖
= cBi B–1 X*i ,

where X* = CBB–1 = being the kth column vector of B–1.

Thus, X*i is the rate of change of the optimal objective function value
with respect to bi and is called the optimal dual variable or simplex
multipliers.
For several practical purposes, the calculation of shadow prices is more
important than the solution of the problem, because it permits the user to
ensure whether certain potential changes in the model requirements might
actually increase the objective function.
For example, if one of the resources of a company represent the current
production capacity of a particular plant and the shadow price for this
resource is greater than the actual unit cost increasing the capacity of the
plant, then the company could increase its profit by doing so.
 22

Advantages of Duality
The knowledge of the dual of a primal LPP is important for the following
main reasons :
(i) As seen from above worked examples, the solution may be easier to
obtain through its dual than through the original (primal) problem. This is
true of cases in which the number of original variables in the primal
problem is considerably less than the number of slack or surplus variables.
(ii)Duality is not only restricted to linear programming problems but
frequently occurs in Economics, Physics and other fields of Mathematics
also.
(a) In Economics, it is used in the formulation of the input and output
systems. The economic interpretation of the dual is found useful in
making future decisions in the activities being programmed.
(b) In Physics, it is used in the parallel circuit and series circuit theory.
(iii)In game theory having two players A and B, it is used to find the
optimal strategies of the player B when he minimizes his losses. Then, by
duality, we can change the player A's problem into player B's problem and
vice-versa. If we solve the problem for one player, the solution to the other
can be easily read-off by using duality properties.
 23

ILLUSTRATIVE EXAMPLES
Example 1. Solve the following primal LPP and its dual by simplex
method :
Max. Zx = 5x1 + 12x2 + 4x3, subject to the following constraints :
x1 + 2x2 + x3 5, 2x1 – x2 + 3x3 = 2 and x1, x2, x3 0.
Verify that the solution of primal can be read from the optimal table of the
dual and vice versa. Also, state the several steps involved in this process.
Solution.Optimal Solution of Primal Problem : By introducing the
slack variable x4 and the artificial variable x5 in the constraints of the
above problem and then applying Big-M method, we get the following
optimal primal simplex table :
Optimal Primal Simplex Table
cj→ 5 12 4 0 –M
Basic Variables CB XB X1 X2 X3 X4 X5

x2 12 8/5 0 1 – 1/5 2/5 – 1/5

x1 5 9/5 1 0 7/5 1/5 2/5
1
x3 = x4 = x5 = 0 Zx = CBXB = -28 0 0 3/5 29/5 – 2/5 + M
2

Here, we note that the starting basic variables are the slack and artificial variables
respectively, while both the optimal basic variables are original variables.
Now, the dual of given primal LPP is :
Minimize Zw = 5w1 + 2w2, subject to the following constraints :
w1 + 2w2 5, 2w1 – w2 12, w1 + 3w2 4, and w1 0, w2 is unrestricted.
Optimal Solution of Dual Problem : Since w2 is unrestricted, therefore,we
replace w2 byw2 – w2, where w2 0, w2 0. Introducing the surplus variables
w3, w4, w5 and the artificial variables w6, w7, w8 in the dual constraints, we shall
arrive at the following optimal dual simplex table :
 24

Optimal Simplex Dual Table

cj→ –5 –2 2 0 0 0 –M –M –M

Basic Variables CB XB W1W2 W2W3 W4W5W6 W7W8

w5 0 3/5 0 0 0 – 7/5 1/5 1 7/5 – 1/5 –1

w2 2 2/5 1 –1 1 2/5 – 1/5 0 – 2/5 1/5 0
w1 –5 59/5 1 0 0 – 1/5 – 2/5 0 1/5 2/5 0

w2 = w3 = w4 = Zw = CBWB 0 0 0 9/5 8/5 0 –9/5 + M – 8/5 + M M

1
w6 = w7 = w8 = 0 = – 28
5

An inspection of the optimal simplex table for the primal and dual reveals
the following interesting results :
(i)The basic variables in the starting solution of the primal are x4 and x5.
(ii)The dual variables corresponding to the primal constraint equations
containing x4 and x5 are w1 and w2 respectively.
(iii) Consider the net evaluations (j's) corresponding to x4 and x5 in the
optimal primal simplex table as follows :
Basic variables in starting primal solution : x4 x5
Corresponding optimal net evaluations : 4 = 29/5 5 = – 2/5 + M
Corresponding dual decision variables : w1 – 0 = w1w2 – (– M) = w2 + M

Then w1 – 0 = 29/5 and w2 + M = – 2/5 + M respectively gives the optimal

value of w1 as 29/5 and that of w2 as – 2/5 which is the same as obtained
by solving the dual independently.
It is interesting to note that the above result holds in general. For
verification, a similar investigation of the starting solution for basic
variables w6, w7 and w8 of the dual give the following information :
Basic variables in starting dual solution : w6 w7 w8
Negative ofCorresponding optimal net evaluations : 9 8
–M –M 0–M
5 5
Corresponding dual decision variables : x1 – M x2 – M x3 – M
 25

Now, the optimal solution of the primal problem is directly given by

9 8
x1 – M = 5 – M, x2 – M =5 – M and x3 – M = 0 – M
or x1 = 9/5, x2 = 8/5 and x3 = 0.
which was obtained in optimal primal simplex table independently.
Discussion of this example suggests the following important conclusion :
An optimal solution to the primal (dual) gives directly an optimal
solution to the dual (primal) linear programming problem.
The general steps for obtaining an optimal solution to one of the problems
from an optimal solution to the other may be outlined as below :
Step 1. First locate the starting solution in primal (dual) variables. These
variables are corresponding to the primal (dual) basic variables in the
optimal solution.
Step 2. Then, read the net evaluations(j's) corresponding to the above
located starting variables in the optimal dual (primal) simplex table. The
optimal solution to the primal (dual) problem is given by the optimal dual
(primal) net evaluations directly. If the dual (primal) is a minimization
problem, then negative ofj'swill give the optimal values.
The above primal-dual relationship in the optimal simplex table may be
summarized as :
Primal (maximization) Problem Dual (minimization) Problem
Basic variable values – (Zj – cj) under starting solution columns
(Zj – cj) under starting solution columns Basic variable values

Example 2. Write down the dual of the following primal LPP and then
solve it :
Max. Zx = 4x1 + 2x2, subject to the following constraints :
– x1 – x2 – 3, – x1 + x2 – 2 and x1 0, x2 0
Hence or otherwise, write down the solution of the primal LPP.
 26

Solution. The dual of the given primal problem is :

Min. Zw = – 3w1 – 2w2, subject to the following constraints :
– w1 – w2 4, – w1 + w2 2 and w1 0, w2 0.
Introducing the surplus variables s1 0, s2 0 and the artificial variables
a1 0, a2 0, the above dual problem becomes :
Max. Zw = 3w1 – 2w2 + 0s1 + 0s2 – Ma1 – Ma2, where Zw = – Zw
subject to the constraints : – w1 – w2 – s1 + 0s2 + a1 + 0a2 = 4
– w1 + w2 + 0s1 – s2 + 0a1 + a2 = 2
and w1 0, w2 0, s1 0, s2 0, a1 0, a2 0.
Now, to apply simplex method as usual, we construct the simplex table as
follows :

cj→ 3 2 0 0 –M –M Min Ratio

Basic
CB W B W1W2 S1 S2A1A2 WB/Wk,Wk>0
Variables
a1 –M 4 –1 –1 – 1 0 1 0 4/(– 1), neg

a2 –M 2 –1 1 0 –1 0 1  2/1, min.

w1 = w2 = s1 Zw = – 6M 2M–3 – 2↑M M 0 0↓ j=CBWB–cj

= s2 = 0
a1 –M 6 –2 0 –1 –1 1 1 No need to
→w2 2 2 –1 1 0 –1 0 1 calculate

w1 = s1 = s2 Zw = – 6M + 4 2M – 5 0 M M–2 0 2 j=CBWB–cj

= a2 = 0

It is now, clear form the optimal table that the coefficient of M in eachjis
non-negative and a non-zero artificial vector appears in the basis.
Therefore, the dual problem does not possess any optimal basic feasible
solution. Consequently, there exists no finite optimal solution to the given
problem.
 27

Example 3. Apply the principle of duality to solve the following LPP :

Max. Zx = 3x1 – 2x2, subject to the following constraints :
x1 + x2 5, x1 + 0x2 4, 0x1 + x2 6, 0x1 – x2 – 1 and x1 0, x2 0.
Solution.The dual of the given problem is :
Min. Zw = 5w1 + 4w3 + 6w3 – w4, subject to the following
constraints :
w1 + w2 + 0w3 + 0w4 3
w1 + 0w2 + w3 – w4 – 2
and w1 0, w2 0, w3 0, w4 0.
Changing the objective function into maximization form by
introducing the surplus variable s1 0 and slack variable s2 0, the dual
problem becomes :
Max. Zw = – 5w1 – 4w2 – 6w3 + w4 + 0s1 + 0s2, subject to the
following constraitns :
1w1 + 1w2 + 0w3 + 0w4 – 1s1 + 0s2 = 3
– 1w1 + 0w2 – 1w3 + 1w4 + 0s1 + 1s2 = 2
and w1 0, w2 0, w3 0, w4 0, s1 0, s2 0.

Here, w2 can be treated as artificial variable so that there is no need

to introduce artificial variable to obtain the identity vector.(10)
Now, apply simplex method to obtain the following simplex table :
cj→ –5 –4 –6 1 0 0 Min Ratio
Basic
CB W B W1W2W3W4 S 1S 2 WB/Wk,Wk>0
Variables
w2 –4 3 1 1 0 0 –1 0 No need to calcu-

w4 1 2 –1 0 –1 1 0 1 late as all j 0

w1=w3=s1 Zw = – 10 0 0 5 0 4 1  j=CBWB– cj

= s2 = 0

Since all j 0, an optimal basic feasible solution to the dual

problem is attained. Thus the optimal solution is :w2 = 3, w4 = 2, w1 = 0,
w3 = 0 and min Zw = 10.
 28

Here, it is observed that the primal variables x1 and x2 respectively

correspond to the slack and the surplus variables s1 and s2 of the dual
problem. As seen from above table, the net evaluations j corresponding
to the variables s1 and s2 are 4 and 1 respectively.
Hence, optimal solution to the original primal will be :x1 = 4, x2 = 1
and max. Zx = 10.
Example 4. A company makes three products X, Y, Z out of three raw
materials A, B and C. The number of units of raw materials required to
produce one unit of product is as given in following table :
Number of Units of Raw Material Required to
Raw Materials Produce One Unit of Product
X Y Z
A 1 2 1
B 2 1 4
C 2 5 1
The unit profit contribution of the products X, Y, and Z are Rs. 40, Rs. 25
and Rs. 50 respectively. The number of units of raw materials available
are 36, 60 and 45 respectively.
(i) Determine the product mix that will maximize the total profit and
(ii) Through the final simplex table, write the solution to the dual problem
and give the economic interpretation.
Sol. Let the company produce x, y and z units of the products X, Y and Z
respectively. Then, according to the given conditions, the LPP model
formulaton is given by :
Maximize P = 40x + 25y + 50z, where P is the profit function
subject to material constraints : x + 2y + z  36
2x + y + 4z  60
2x + 5y + z  45
and x  0, y  0, z  0.
 29

We solve this linear programming problem by simplex method and

so introducing the slack variables s1, s2, and s3 in the constraints, we get
Max. P = 40x + 25y + 50z + 0s1 + 0s2 + 0s3,
subject to the constraints : x + 2y + z + 1s1 + 0s2 + 0s3 = 36
2x + y + 4z + 0s1 + 1s2 + 0s3 = 60
2x + 5y + z + 0s1 + 0s2 + 1s3 = 45
and x 0, y 0, z 0, s1 0, s2 0, s3 0.
Let the initial solution be x = 0, y = 0, z = 0, s1 = 36, s2 = 60 and s3 = 45 so
that P = 0. Now, we try to improve this solution by simplex method by
introducting the simplex tables as given below :

Basic cj→ 40 25 50 0 0 0 Min. Ratio

Table No.
Simplex

Variables CB XB X Y Z S1S2S3 XB/Xk, Xk> 0

s1 0 36 1 2 1 1 0 0 36
Simplex Table I

s2 0 60 2 1 4 0 1 0  15, min.

s3 0 45 2 5 1 0 0 1 45
x=y=z=0 P = CBXB = 0 – 40 – 25 – 50↑ 0 0↓0 j=CBXB–cj

s1 0 21 1/2 7/4 0 1 – 1/4 0 42

Simplex Table II

→z 50 15 1/2 1/4 1 0 1/4 0 30

s3 0 30 3/2 19/40 0 1/4 1  20, min.

x = y = s2 = 0 P = CBXB = 750 – 15↑ – 25/2 0 0 25/2 0↓ j=CBXB–cj

s1 0 11 0 1/60 1 – 1/6 – 1/3 No need

Simplex Table III

z 50 5 0 – 4/3 1 0 1/3– 1/3 to calculate

→x 40 20 1 19/6 0 0 – 1/6 2/3 as all j 0
y = s2 = s3 = 0 Max. P = 1050 0 35 0 0 10 10 j=CBXB–cj

Since all j are 0, therefore, the final simplex table will give the
optimalsolution :x = 20, y = 0, z = 5 and max. P = Rs. 1050.
 30

The Dual Problem : The dual of the problem is given below :

Min. Q = 36a + 60b + 45c, subject to the following constraints :
a + 2b + 2c 40, 2a + b + 5c 25, a + 4b + c 50, and a 0, b 0, c 0.
From the final simplex table of the primal problem, the solution of
the dual problem can be obtained. The solution is obtained from the net
evaluation row elements (j) under slack variable columns (S1, S2, S3).
Thus, the solution of dual problem is as under :
a = 0, b = 10, c = 10 and min. Q = Rs. 1050.
Economic Interpretation : Suppose the manager of the company wishes
to sell three raw materials A, B and C instead of using them for production
of the products X, Y, and Z. Suppose the selling prices were Rs. a, Rs. b
and Rs. c per unit of raw material A, B and C respectively. The cost to the
purchaser of all thre raw materials would be Rs 36a + 60b + 45c. Of
course, the purchaser would like to set the selling prices in such a way so
as to minimize the total cost paid. Hence objective function becomes :
Min. Q = 36a + 60b + 45c.
The final simplex table of the primal problem indicates that the marginal
value of raw material A is Rs. zero, for B is Rs. 10 per unit and for C is
Rs. 10 per unit. Thus if the manager sells the raw material A, B and C at
price Rs. 0, Rs. 10. and Rs. 10 per unit respectively, he will get the same
contribution of Rs. 1050 which he is going to fetch in case he utilizes
these resource for production of the three products X, Y and Z.
Example 5. Apply simplex method to solve the following linear
programming problem :
Max. Zx = 30x1 + 23x2 + 29x3, subject to the following constraints :
6x1 + 5x2 + 3x3 26, 4x1 + 2x2 + 6x3 7 and all xj 0.
Also, read the solution to the dual of the above problem from the
final simplex table.
 31

Solution. Introducing the slack variables s1 and s2, the given problem
becomes :
Max. Zx = 30x1 + 23x2 + 29x3 + 0s1 + 0s2,
subject to the following constraints : 6x1 + 5x2 + 3x3 + 1s1 + 0s2 = 26
4x1 + 2x2 + 5x3+ 0s1 + 1s2 = 7
and satisfying x1 0, x2 0, x3 0, s1 0, s2 0.
We now give below the successive simplex tables for the solution of this
problem :

Simplex Table
Basic cj→ 30 23 29 0 0 Min Ratio
Variables XB/Xk,Xk > 0
CB XB X1 X2 X3S1S2

s1 0 26 6 5 3 1 0 26/6

s2 0 7 4 2 5 0 1  7/4, min.

x1 = x2 = x3= 0 Zx = CBXB = 0 – 30↑ – 23 – 29 0 0↓  j=CBXB–cj

s1 0 31/2 0 2 – 9/2 1 – 3/2 31/4

→x1 30 7/4 1 1/2 5/4 0 1/4  7/2, min.

x2 = x3 = s2 = 0 Zx = CBXB = 105/2 0 –8 47/2 0 15/2  j=CBXB–cj

s1 0 17/2 –4 0 – 19/2 1 – 5/2 No need to

x2 23 7/2 2 1 5/2 0 1/2 calculate
x1 = x3 = s2 = 0 Zx = CBXB= 161/2 16 0 57/2 0 23/2  j=CBXB–cj

Thus, the optimal solution from the final simplex table is given by
x1 = 0, x2 = 7/2, x3 = 0 and Z*x = 161/2 i.e. max. Zx = 161/2.

\
 32

Reading of solution to the dual of the given primal problem from the
final simplex table :
According to the rules, the optimal solution to the dual of the given
primal problem will be
w1 = 4 = 0, w2 = 5 = 23/2 and Z*x = Z*w = 161/2.
In this way, we can find the solution to the dual without actually
solving it.
Example 6. Use duality to solve the linear programming problem : Min
Zx= x1 – x2, subject to the constraints 2x1 + x2 2, – x1 – x2 1 and x1 0, x2 0.

Solution. The dual of given linear programming problem is :

Max. Zw = 2w1 + w2 , subject to 2w1 – w2 1, w1 – w2 – 1
and w1 0, w2 0.
We shall have lesser computational efforts in solving the dual
problem rather than solving the original one. So, first of all, we shall
express the dual problem in following standard form :
Max. Zw = 2w1 + w2 + 0s1 + 0s2 + 0a1, subject to the constraints :
2w1 – w2 + 0s1 + 1s2 + 0a1 = 1
– w1 + w2 – s1 + a1 + 0s2 = 1
and satisfyingw1 0, w2 0, s1 0, s2 0, a1 0.
We now apply two-phase simplex method to solve it as follows :
Phase 1.First of all, construct the artificial objective function Z* by
assigning – 1 cost to the artificial variable a1 and zero cost to all other
variables. Thus, we have the following artificial objective function :
Max. Z*w = 0w1 + 0w2 + 0s1 + 0s2 – a1.
Therefore, we have the following auxiliary LPP :
Max. Z*w = 0w1 + 0w2 + 0s1 + 0s2 – a1
subject to 2w1 – w2 + 0s1 + s2 + 0a1 = 1
– w1 + w2 – s1 + 0s2 + a1 = 1
and w1 0, w2 0, s1 0, s2 0, a1 0.
 33

We now construct the simplex table for solution of the auxiliary LPP in
phase I as follows :
Since all j 0 in the last simplex table and no artificial variable appears in
the basis matrix, therefore, an optimal basic feasible solution to the
auxiliary LPP has been attained. Hence, we proceed to phase II to get an
optimal basic feasible solution to the above dual problem.
Phase II. To find optimal solution of the Dual of Original LPP

We shall obtain the following simplex table for the solution during the
simplex procedure.

Basic cj→ 0 0 0 0 –1 Min. Ratio

Variables CB WB W1W2S1 S2 A1  =WB/Wk,Wk > 0

s2 0 1 2 –1 0 1 0 1/(–1), neg.

a1 –1 1 –1 1 –1 0 1  1/1, min.

w1 = w2 = s 1 = 0 Z*w = CBXB = –1 1 – 1↑1 0 0↓  j=CBWj–cj

s2 0 2 1 0 –1 1 1 No need to calcu-

→ w2 0 1 –1 1 –1 0 1 late as all j 0.

w1 = s1 = 0 Z*w = CBWB= 0 0 00 0 1  j=CBXB–cj

 34

From the above table, we see that 3 is negative and all elements of vector
S1 (not in the basis) are negative, this indicates that the problem under
Basic cj→ 2 1 0 0 Min. Ratio
Variables CB WB W1W2S1 S2 WB/Wk,Wk > 0

s2 0 2 1 0 1 1  2/1, min.

w2 1 1 –1 1 –1 0 1/(– 1), neg.

w1 = s1 = 0 Zw = CBXB = 1 –3 0 –1 0↓  j=CBWB–cj

→w1 2 2 1 0 –1 1 No need to calcu-

w2 1 3 0 1 –2 1 late as all j 0.

s1 = s2 = 0 Z*w = CBWB= 4 0 0 –4 3  j=CBXB–cj

consideration has unbounded solution. Consequently, by duality theorem,

the original primal problem will have no feasible solution.
Example 7. Use duality to solve the following LPP :
Min. Zx = 3x1 + x2, subject to x1 + x2 1, 2x1 + 3x2 2 and x1  0,
x2 0.
Solution. The dual of given linear programming problem is :
Max. Zw = 1w1 + 2w2, subject to w1 + 2w2 3, w1 + 3w2 1 and
w1 0, w2 0.
This problem can be easily solved by simplex method. Introducing
the slack
Variables s1 and s2 and proceeding in the usual simplex routine, we get
the following successive simplex table :
 35

Simplex Table
Basic cj→ 1 2 0 0 Min. Ratio
Variables CB WB WB/Wk,Wk > 0
W1 W2 S1 S2
S1 0 3 1 2 1 0 3/2
s2 0 1 1 3 0 1  1/3, min.
w1 = w2 = 0 Zw = CBWB = 0 –1 – 2↑ 0 0↓  j=CBWj–cj

s1 0 7/3 1/3 0 1 – 2/3 7/1 

→w2 2 1/3 1/3 1 0 1/3  1, min.

w1 = s2 = 0 Zw = CBWB = 2/3 – 1/3↑ 0↓ 0 2/3  j=CBXj–cj

s1 0 2 0 –1 1 –1 No need to calcul-

w1 1 1 1 3 0 1 late as all j 0

w2 = s2 = 0 Zw* = CBWB = 1 0 1 0 1  j = CBWj – cj

Hence, the optimal solution of the dual problem comes out to be w1 = 1,

w2 = 0 with Z*w = 1 i.e. max. Zw = 1.
Reading of solution of the original primal problem from the final dual
simplex table :From the above dual simplex table, we findx1 = 3 = 0, x2
= 4 = 1 and Z*x = Z*w = 1.
Example 8. A firm makes two products A and B. Each product requires
production on each of the two machines :
Machine Product Total Time Available
A B (in hours)
M1 6 4 60
M2 1 2 22

Total time available is 60 hours and 22 hours on machines M 1 and M2

respectively. Product A and B contribute Rs. 3 and Rs. 4 per unit respectively.
Determine the optimum product mix. Write the dual of this problem and
give its economic interpretation.
 36

Solution. Formulation of primal and dual problems are given below :

Primal Problem : Maximize Zx = 3x1 + 4x2, subject to the constraints :
6x1 + 4x2 60, x1 + 2x2 22 and x1 0, x2 0,
wherex1 is the number of units of product A and x2 is the number of units of
product B.
Dual Problem : Minimize Zw = 60w1 + 22w2, subject to the constraints :
6w1 + w2 3, 4w1 + 2w2 4 and w1 0, w2 0,
wherew1 is the cost per hour on machine M1 and w2 is the cost per hour on
machine M2.
Optimum Solution to the Primal Problem :
By usual simplex method, the optimal simplex table is given below :
cj→ 3 4 0 0 Min. Ratio
Basic
CB WB X1 X2 X3 X4  = XB/Xk, Xk> 0
Variables

x1 3 4 1 0 1/4 – 1/2 No need to Calculate

x2 4 9 0 1 – 1/8 – 3/4 as all j 0.

x3 = x4 = 0 Zx = CBXB = 48 0 0 1/4 3/2  j=CBXB–cj

From the above simplex table, we see that all j 0, therefore, we
have arrived at the optimal solution.
Thus, the optimal solution is :x1 = 4, x2 = 9 and max. profit = Rs. 48.
Using duality, the optimal solution to the dual problem can be read
directly from the optimal simplex table as given by :w1 = Rs. 1/4 or Rs.
0.25 per hour, w2 = Rs. 3/2 or Rs. 1.50 per hour and min. cost = Rs. 48.
Economic Interpretation of Dual Problem :
As discussed above, shadow prices are the opportunity costs that indicate
the potential profit that is lost by not having an additional unit of the
respective right hand side (resource) assuming that all right hand side
values are used optimally. Thus w1 = 0.25 and w2 = 1.50 means that
additional processing hours on machine M1 and M2 will increase the profit
by Rs. 0.25 and Rs. 1.50, respectively.
 37

Similarly, in the primal problem, if we increase the total available

hours on machine M1 from 60 hours to 61 hours, the new set of constraints
will be : 6x1 + 4x2 61 and x1 + 2x2 22.
Solving the primal problem with new set of constraints, the
optimum solution becomes :
x1 = 4.25 and x2 = 8.875 with max Zx = Rs. 48.25
This is exactly Rs. (0.25)/1 = Rs. 0.25 more than the earlier value of Zx
when only 60 hours of machine M1 were available.
EXERCISE
Use duality to solve the following linear programming problems :
1.Max. Z = 3x1 + 2x2, subject to 2x1 + x2 5, x1 + x2 3 and x1 0, x2 0
2.Max. Z = x1 + 6x2, subject to x1 + x2x1 + 3x2 3 and x1 0, x2 0.
3. Min. Z = 2x1 + 2x2, subject to 2x1 + 4x21, x1 + 2x21, 2x1 + x21 and
x10, x20.
4. Max. Z = 3x1 + 4x2, subject to x1 – x21, x1 + x24, x1, 3x23, x1<x20.
5.Max. Z = 2x1 + x2, subject to x1 + 2x2 10, x1 + x2 6, x1 – x2 2, x1 –
2x2 1 and x1  0, x2 0.
6.(a) Use simplex method to maximize Z = 5x – 2y + 3z, subject to the
constraints :
2x + 2y – z 2, 3x – 4y 3, y + 3z 5, where x, y, z are non-negative
variables.
(b) Verify your solution using the dual of the problem given above.
7.Apply simplex method to solve the following LPP :
Max. Z = 30x1 + 23x2 + 29x3, subject to 6x1 + 5x2 + 3x3 26; 4x1 + 2x2 +
5x3 7 and xj 0.
Also read the solution to the dual of the above problem from the final
simplex table.
 38

8.Apply the principle of duality to solve the following linear programming

problem :
Max. Z = 3x1 + 2x2, subject to the constraints : x1 + x21, x1 + x27, x1
+2x210, x23 and x10, x20.
9. A diet conscious house wife wishes to ensure certain minimum intake
of vitamins A, B and C for the family. The minimum daily (quantity)
needs of the vitamins A, B and C for the family are respectively, 30, 20
and 16 units. For the supply of these minimum vitamin requirements, the
house-wife relies on two fresh foods. The first one privides 7, 5, 2 units of
the three vitamins per gram respectively, and the second one provides 2, 4,
8 units of the same three vitamines per gram of the food stuff respectively.
The first food stuff costs Rs. 3 per gram and the second Rs. 2 per gram.
The problem is how many grams of each food stuff should the house-wife
buy every day to keep her food bill as low as possible ?
(i) Formulate the underlying linear programming problem (ii) Write the
dual problem, (iii) Solve the dual problem by using simplex method, (iv)
Solve the primal problem graphically, (v) Interpret the dual problem and
its solution.
ANSWERS
1.x1 = 2, x2 = 1 and max. Z = 8; w1 = 1, w2 = 2 and min. Zw = 8.
2.x1 = 3/2, x2 = 1/2 and max. Z = 9/2; w1 = 3/2, w2 = 5/2 and min. Zw =
9/2.
3.x1 = 1/3, x2 = 1/3 and max. Z = 4/3.
4. The dual problem does not possess any optimal basic feasible solution,
hence there exists an unbounded solution to the primal problem.
5.x1 = 4, x2 = 2 and max. Z = 10.
6.x = 23/3, y = 5, z = 0 and max. Z = 85/3.
7.x1 = 0, x2 = 7/2, x3 = 0 and max. Z = 161/2; w1 = 0, w2 = 23/2 and min.
Zw = 161/2.
 39

8.For dual :w1 = 0, w2 = 3, w3 = 0, w4 = 0 and min. Zw = 21; For primal :

x1 = 7, x2 = 0 and max. Z = 21.
9.(i) Min Zx = 3x1 + 2x2, subject to 7x1 + 2x230, 5x1 + 4x220,2x1 +
8x216 and x10, x20, x30.
(ii) Max. Zw = 30w1 + 20w2 + 16w3, subject to 7w1 + 5w2 + 2w33, 2w1
+4w2 + 8w32 and w10, w20, w30.
(iii) w1 = 5/13, w2 = 0, w3 = 2/13 and max. Zw = 14.
(iv) x1 = 4, x2 = 1 and min. Zx = 14.